Eastern North Carolina health care atlas a resource for healthier communities. |
Previous | 1 of 4 | Next |
|
small (250x250 max)
medium (500x500 max)
Large
Extra Large
large ( > 500x500)
Full Resolution
|
This page
All
|
Introduction to the 2006 Eastern North Carolina Atlas of Mortality A tombstone in an eastern North Carolina church cemetery is inscribed: In Memory of James Bonner Foreman Who was born The 1st of December 1785 And died The 22nd of December 1807 Aged 22 Years and 21 Days Come view my Tomb as you pass by, As you are now so once was I; As I am now so must you be, Therefore prepare to follow me. Death is a personal event that we will all eventually experience. It is also something fundamentally empirical, recordable, and therefore measurable. The tradition and culture of recordkeeping varies throughout the world and in the west some countries have been compiling data on peoples’ lives for centuries either for ecclesiastical or secular purposes. One extremely important secular purpose is the amassing of individual records over time and place into part of North Carolina’s vital statistics collection. Eventually, every North Carolina resident shows up in the vital statistics registry “ book” as a single data record, an abstraction, of a life once lived. Unlike Mr. Foreman’s epigraph two centuries ago, more data and information pertaining to the circumstances of the mortal event are recorded. In addition to date of birth and death ( i. e., age at death), these include the decedent’s location at death, cause of death, race, sex, and residence. The data recording the circumstances surrounding people’s deaths can be formed into a picture about the conditions of living in their period of time and their society when aggregated at various scales and dimension. The atlas format is an appropriate means of display and description of vital events such as mortality. The present chapter is an introduction to the approach and concepts used in the current edition of the Eastern North Carolina Atlas of Mortality. Specifically addressed topics can be found using the following linked headings. Overview of the Atlas Portraying Geographic Data Data Sources Mapping with GIS Software Maps in the Atlas Time Series Charts in the Atlas Overview of the Atlas The 2007 edition of the Eastern North Carolina Atlas of Mortality is a narrated collection of such statistical pictures that describe the spatial and temporal facets— the descriptive geography-- of death in the eastern- most 41 counties of North Carolina ( ENC). Over the last three decades, this region has seen thousands of individuals dying in excess of what would be expected or experienced in other parts of the country. The underlying motivation for this work is to bring this ongoing tragedy to light and to show health professionals and policy makers where and on what problems need their attention. The information presented in this atlas will allow the reader to form a coherent image in his or her mind of the history and future of mortality in Eastern North Carolina. It is hoped that these statistical images will lead to not only an increased awareness of the conditions of life-- and death-- in ENC but that it will also stimulate thinking about hypotheses, research questions, policy, and strategies for making life better in our region. In this work, the geographical distributions of mortality from leading causes are aggregated and portrayed for the years 2000 to 2004 ( 5 years) and chronicled over a 26- year time series beginning in the year 1979 an ending in 2004. From 2004, rate projections ( linear best fit lines) are included. Figure 1.1 portrays the 100 counties of North Carolina and delineates the major regions used in this Atlas. The regional focus is the eastern- most 41 counties whose western boundary is approximated by I- 95 and extends to the coastline. ENC 41 also corresponds to the physiographic province of the Coastal Plain. The 41- county region is further divided into two sub- regions: ENC 29, comprised of the northeastern- most 29 counties of ENC 41, and a remaining southern 12- county region. ENC 29 corresponds spatially to the county service area of University Health Systems of Eastern Carolina. ENC 41 possesses North Carolina’s greatest levels of poverty and ethnic diversity, while population and economic growth lags behind the remaining western 59 counties. To contrast and compare mortality rates with the rest of the state, the remaining 59 counties are grouped into two regions corresponding to the Piedmont ( PNC) and the western mountain region ( WNC). Over the last 30 to 40 years, PNC and WNC have experienced rather different population and economic trajectories than the east and this is reflected in their more favorable mortality outcomes. The Atlas traces the spatial and temporal domains of ENC’s mortality experience with the use of maps, tables, and time series charts. These three components of the Atlas are built on measures that summarize the population’s mortality experience. Summary measures like mortality rates are calculated from several of the descriptive elements of the individual death record. The resulting rate calculations are then tabulated by county, region, and time period. In contrast to the simple table, maps are a 2- dimensional spatial ordering of mortality rates that describe a place’s mortality experience and burden. Time series charts portray the temporal order of mortality rates for regions, counties, and their constituent population groups. These charts show general parallel, convergent, or divergent 2 trends among regions and population groups. Relative and absolute mortality rate comparisons can be made from the maps, tables, and charts to determine progress toward the elimination of rate disparities and mortality burden over space and time. Portraying Geographic Data Maps are the most important feature of a geographical atlas. Along with other graphical means of communication, a wide range of topical literature has evolved that discuss the nature of maps and the geographic information and meaning that they portray from a variety of technical and philosophical both within and without the discipline of geography. A good discussion of the foregoing, which also includes Information Theory, can be found in Poore and Chrisman’s Order from Noise: Toward a Social Theory of Geographic Information ( Poore & Chrisman, 2006). The more salient and general points concerning maps and time series data found in this work are discussed below. For a more technical treatment of charts, with a strong emphasis on the proper construction of graphics that convey meaningful information from quantitative data, the reader is directed to the works of Tufte ( Tufte, 1995; Tufte, 1997; Tufte, 2001; Tufte, 2006). Pragmatically, different aspects of various techniques and perspectives necessarily come together in the development of any atlas and how they come together may distinguish one atlas’s approach from another. In this Atlas, our approach is one of description and chronicling in such a way that the reader can make meaningful geographical comparisons of the regional mortality experience. One functional definition of geography considers both space and time as referential systems. Borrowing terminology from Werlen ( Werlen, 1993), a space can be defined as a three dimensional container. This type of space orders events ( an occurrence or areas with given attributes like mortality rates) by measuring their positional relationships ( the x and y axes) and their sizes or magnitudes ( the z axis). Another dimension can be added that orders those events temporally and therefore, sequentially. The 2- dimensional or 3- dimensional static map can be stacked or sequenced along a temporal axis to form a time series of maps. As long ago as 1964, Berry ( Berry, 1964) described and operationalized a very similar concept as the geographical data matrix, where the matrix is the container of geographically referenced data— attributes/ characteristics ( or mortality rates) that are linked to places or areas. With some modifications, this prosaic and functional conceptualization describes how spatially referenced data are managed in modern Geographic Information Systems ( GISs). With a GIS, these data can be stacked or sequenced in temporal order very quickly to create a moving picture of a geographic process. Because of space constraints, only the most current 5- year maps of mortality rates are provided in this Atlas, but they are accompanied by charts that show temporal trends among regions and population groups. 3 Geographical referencing and the binding together of attribute data over points in time or sequence of time periods are a means to the comparative study of trends in mortality processes. In both spatial and temporal referential systems, there is a well- known tendency for objects within the system that are nearer to one another to be more alike than those more distant or, as stated in Waldo Tobler’s first law of geography, “… everything is related to everything else, but near things are more related than distant things.” ( Tobler, 1970) This notion of propinquity and similarity is important for understanding relationships among demographic, social, biological, and physical attributes of places. For example, a group of neighboring counties such as those found in Eastern North Carolina will tend to have similar age, race, and sex structures because they have had similar economic and demographic histories or, more generally, have experienced similar social relations and processes as well as live within similar spatial structures ( Gregory & Urry, 1985). Since age is the greatest risk factor for mortality we would also expect a group of neighboring counties that share a similar age structure to have similar mortality rates. In varying degrees, these same counties may also have similarities in other known risk factors such as certain occupations, race, housing, and poverty. Within the spatial analytical line of inquiry, this well known propensity in geography is extremely useful for constructing hypotheses, modeling, and theory testing. Maps can be thought of as models of real- world patterns and processes at a given point in time. They reduce reality to a set of graphical and geometric objects that have an a priori common meaning, which is necessary for interpretation and communication. This reality is not produced, reproduced, or experienced in exactly the same way by any two persons or reflected in individual death records but collectively similarities and patterns can emerge and be traced for population aggregates. A map as a representation allows a way for the user to apprehend a myriad of facts about places and order them both spatially and temporally into one coherent mental picture. Once geographic data have been integrated into a suitable level of coherency, assessment and analyses can begin with a certain set of well- grounded assumptions. These assumptions might include Tobler’s first law of geography ( the closer, the more similar) or considered in conjunction with certain risk factors such as age or diabetes with certain mortality outcomes. However, it should always be borne in the mind of the map user or analyst that these newly acquired understandings and cognitive models are ultimately based on a reduced reality— that is, in the time- worn phrase: the map is not the territory. Finally, maps can be used either as arguments to make a case for further study into the etiology of the causes of mortality and morbidity or they can be used as propositions ( or hypotheses) addressing potential causes of observed mortality and morbidity patterns ( Koch, 2005). To illustrate, given the range of social and structural inequalities that exist among certain demographic groups in the US and particularly in the South, the Atlas provides evidence for the argument that differences in the underlying social fabric will manifest themselves in the 4 observed patterns of mortality for Whites and Non- whites in eastern North Carolina and for all Eastern North Carolinians versus the rest of the state. The case can be made by employing maps, tables, and charts that permit comparisons among the race- sex groups at county, regional, and national scales. Maps of related demographic and socio- economic variables are either included or referenced in the Atlas as propositions about relationships underlying the observed mortality patterns. As a tool for integrating disparate data, either as argument or proposition, the Atlas can assist in developing research questions for topics on health disparities, health resources, and economic development. Representational data used in the construction of maps are of two distinct classes. The first data class is made up of a limited set of geometrical objects that are used to represent a large range of real- world features on a map. The most basic of these data is the geometric point that is located on a geometric plane. The point can represent an event, institution, or place, for example. On this same plane an additional point will define a line and a series of lines can represent features such as road networks, stream systems, or social relationships and connections. Three or more points will define a polygon and can represent real- world entities such as counties or urban areas. In some maps polyhedra or solids defined by four or more polygons can be constructed to represent specific types of features. These geometrical representations ( or features) have some measurable quality or attribute assigned to them, which provides the basis for making comparisons and discerning patterns. Points, lines, and polygons can be assigned an attribute, quality, or quantity that describes map features. This second class of data can be partitioned into three categories: nominal, ordinal, and interval/ ratio ( Earickson & Harlin, 1994). Nominal data refers to the binary presence/ absence of a quality or one or more types of a given feature, such as vegetation cover or soil. Ordinal data are ranked in ascending or descending order and can be used to describe a hierarchical system of, for example, health states or levels of care quality measured as poor, fair, good, or excellent. Finally, interval/ ratio scale ( or metric scale) data measure quantities like mortality rates, dentist to population ratios, or disease prevalence. For interval data the difference between any pair of values is always the same no matter where they are located along the metric scale. There is a small but important distinction when considering either interval or ratio data. Interval data can include values that are less than an arbitrarily defined zero, such as temperature or elevation. However, unlike elevation, one cannot speak of a temperature being twice as cold or hot as another. These data are strictly interval in nature. Ratio data are interval data that can be compared meaningfully. For example, one could make the statement that the mortality rate for female breast cancer in county A is 33% greater than the rate in county B. Interval data can be evaluated as “ twice as much,” “ half as great,” or as some percent or proportion of one value in relation to another. 5 Data Sources The predominant types of data employed in this Atlas are polygons bound or joined to interval/ ratio data attributes. Polygons are used to represent counties, which are the basic units of analysis and are the building blocks for larger multi-county regions. County- level polygon data ( i. e., boundary files) are obtainable from the geography page of the US Census website. These data are available in several formats and are ready for use with most GIS packages. Because boundary files have unique county identifiers, they are also ready to “ join” or link to attribute data. A wide variety of county- level attribute data are employed in this work. Demographic and socio- economic data can be obtained from the American FactFinder section of the US Census website and the NC State Data Center. In the Atlas, mortality rates by leading causes of death are calculated from two sources. The North Carolina source is located at the University of North Carolina’s Odum Institute, which provides the most up to date vital statistics for the state. Mortality data for the nation and other areas of the county are calculated from data found in the Compressed Mortality File ( CMF) series produced by the National Center for Health Statistics. These data tend to be 3 to 4 years behind the latest year for North Carolina. Mapping with GIS Software Today, nearly all data required for GIS and mapping exist in a digital form. Many printed tabular data sources, collected in more remote periods of time, have been archived either on paper or microforms. These data sources can be scanned or imaged into formats suitable for optical character recognition ( OCR) programs or other software tools that will transform the printed character or numeral into a digital rendition. Once obtained, the data need to be stored in some type of database. Storage can be in a large relational enterprise level database such as MS- SQL ® or Oracle ® with member tables distributed according to function anywhere on the globe or data storage can simply be in a spreadsheet “ database” residing on a desktop PC. In Microsoft’s Excel ® , one or more data ranges ( i. e., columns × rows) described in a worksheet can behave as individual database tables within a workbook. These data ranges and tables loosely correspond to Berry’s geographic data matrices. ( Berry, 1964) Using a small set of basic database functions in Excel, it is possible to link and match records ( table rows) in a way similar to what is done in a true relational database. In order to match records, there must be a field serving as an index. An index field contains rows of unique identifiers and is common to all tables that will be linked or joined. In this Atlas, we use either the unique county name within the state or the Federal Information Processing Standard ( FIPS) code that uniquely identifies any county among the more than 3,000 counties in the US. These same identifiers are used to match attribute data to county polygons prior to mapping in a GIS. 6 Map- making today is largely done using GIS software that integrates a wide variety of disparate data sources and data types. The construction of maps is actually one of many functions a modern GIS can perform. Other functions include spatial querying, spatial analyses, modeling, as well as layering and combining spatial objects and their attribute data to develop new data. For the purposes of descriptive spatial epidemiology and ultimately the comparisons that will be made, the Atlas here employs the primary and more basic functions of a GIS which manage geographically referenced data and quickly generate map layers with accompanying cartographic elements. Cartographic elements include the legend or map key derived from data and feature classification and symbology. Data in an atlas of mortality are typically rates and percentages ( interval/ ratio data). A GIS is able to partition and classify a data distribution with a choice of automated default methods ( e. g. quantile, equal interval, natural breaks, or statistical) or the user can classify the data manually. The choice of method is based on the purpose of the map ( e. g., statistical description, proposition, or argument) and the intended audience of map readers ( Wilson & Buescher, 2002). The GIS also provides color palettes for selecting a hue for each theme. A hue can be further divided into a series of graded shades with hue saturation corresponding logically to category ranges. Analysis proceeds by examining the resulting patterns of categorized rates represented as shades: do counties with more saturated shades tend to cluster together? Or are they more dispersed, demonstrating no real comprehensible pattern? Such basic analyses can yield ideas for the development of hypotheses or intervention strategies if something is known about the processes that created them. Different ideas about presentation of map data and experimentation with categories can proceed quickly with a GIS. What took several days to produce by hand as recently as twenty years ago today only takes several minutes. The maps in this Atlas were created in ESRI’s ArcGIS 9.1 and 9.2. Maps in the Atlas The Atlas is organized in a way that invites the assessment of patterns in both the spatial and temporal domains. Maps show the distribution of categorized county rates of mortality for the years 2000 to 2004. Mortality rates are, in effect, measures of density. They measure the density of events ( deaths by selected causes) in relation to the population producing those events. Both crude and age- adjusted rates are employed for making regional comparisons in those maps depicting total deaths by cause, while only age- adjusted rates are used for making county and regional comparison by race- sex groups. Crude rates are constructed by dividing the number of events ( or case mortality by cause) in a county by that county’s total population, and then multiplying the result by 100,000, which has the effect of reducing in a certain time period the number of decimal places and thereby making the rate more easily understood. A crude rate is the actual rate and is useful for measuring the burden of disease mortality 7 in an area and time period. However, making comparisons among counties with crude rates is problematic because the differences in their respective age-structures can confound interpretation. For example, knowing that increased age is the greatest risk factor for dying in a given time period, a county with a larger proportion of elderly ( e. g., retirees) will naturally produce a greater crude rate than a county where there are larger proportions of college- age students or individuals stationed on military bases. To make meaningful comparisons, a county’s age structure ( the numbers of people in each previously defined age group) must be adjusted. Essentially this adjustment is a re- weighting of a county’s population that produces an expected, as opposed to actual or observed, number of deaths for that population. The weights are based on an external or synthetic population structure known as the standard million population. Age- specific death rates based on the weights are calculated for each group in the age structure and then summed to produce an age- adjusted rate ( Buescher, 1998). An age- adjusted county rate is the rate a county would have if it had the same age structure as the external or standard million population and renders this county’s rate comparable to any other county using the same standard million population. It should be emphasized that age-adjusted rates used in making comparisons are not the actual observed rates but are the rates that would be expected if each county and region had the same age structure. The external population used in this work is the US Standard Million for the year 2000. Knowing which standard million population is used is extremely important when comparing rates calculated from mortality data from different states and time periods, otherwise the rates are simply not comparable. Time Series Charts in the Atlas Time series graphs for the years 1979 to 2004 provide a synoptic view of mortality trends for regions and race- sex groups. Age- adjusted rates are used to make comparisons among the 41 counties of ENC, the remaining 59 counties of the state ( RNC), North Carolina, and the US ( 1979 to 2002). Time series plots for four ENC race- sex groups ( male and female Whites and Nonwhites) are provided on an additional graph. Best- fit lines are incorporated into the time series plots for both regional and population charts so that the user can assess differences and trends. How well the trend and projection line fits the data is described by the coefficient of determination, R2. ( R2 is a statistic with values 0.0 to 1.0; the closer to 1.0, the better the fit.) For some leading causes of death there are Healthy People 2010 goals, which are age- adjusted target rates for the year 2010 ( U. S. Department of Health and Human Services, 2000). Where applicable, target values are included in the chart and can be used in conjunction with the projected trend lines. This permits the user to make comparisons among regions and population groups in terms of the amount of progress that is being made against a nationally recognized standard. 8 Over the course of many years, mortality rates will ebb and flow with small annual perturbations deviating from the general trend. A larger view over many decades may show gradually decreasing ( the ebb) or increasing ( the flow) trends for chronic diseases and intermittent spiking for epidemics during that period of time when communicable and infectious diseases were predominant causes of death ( see figure 1.2). Long term directional changes and pattern shifts in mortality rates are known as secular trends.∗ These trends are both responding and contributing to the underlying long term shifts in demographic, socio-economic, and environmental processes. One of the best examples is the nearly complete decline of mortality due to infectious diseases in the early part of the twentieth century. Infectious diseases tend to carry off larger proportions of susceptible young as well as those in the older age groups. Socio- economic and environmental processes such as improved access to better food and nutrition, improved sanitation, and generally better living conditions resulted in fewer deaths of the young as a result of contagion. In turn, a gradual shift in demographics occurred: more children survived into adulthood and into later life. This demographic shift— the result of more individuals now surviving into the older age groups-- is a major influence on the rise of the crude mortality rate from cardiovascular disease ( with the exception of stroke) in the early- mid twentieth century. These kinds of changes are described in Omran’s work on the epidemiologic transition ( Omran, 2005). The long term mortality trends resulting from different causes of death may not all be the same. Generally, mortality rates over the long term trace curvilinear patterns. As these patterns are examined more closely, parts of the curve begin to take on a more linear form. To simplify and give a general snapshot of recent trends, the mortality time series depicted in this Atlas models the data linearly. The benefit to this is that it provides easily understandable summary measures of mortality events occurring over three decades. However, the reader is cautioned to examine the general pattern of the entire series, giving more weight to events that have occurred later in the series than earlier. The maps, tables, and charts found in the Eastern North Carolina Atlas of Mortality form an armamentarium for understanding, integration, and synthesis of the region’s mortality burden and experience. Singly, an individual’s death, like the one found in an obscure corner of a church cemetery may appear to be a random event. However, when lone events like these are amassed into numerators and then rates, meaningful pictures about the conditions of life in a place can be created. In the end, it should always be kept in mind when gazing upon the abstract representation of the mortality map that ultimately it was the lives rather than the deaths of people that generated the observed patterns. ∗ The term secular as used here refers to a characteristic pattern for a given age or time period in population history. For example, until the First World War in the United States infectious and communicable diseases had a much more prominent role in observed mortality patterns than they do today. The last several decades of the twentieth century has seen a gradual decline for certain chronic diseases like those of the heart and some cancers. 9 The next chapter addresses general mortality. In this chapter, the leading causes of death for ENC 41 are delineated for the 5- year period, 2000 to 2004. Discussion of the spatial and temporal distributions of mortality from all causes ( i. e., general mortality) follows, including a more in- depth treatment of rates and measures in light of the observed data. Subsequent chapters address the 10 leading causes of death for the region and will generally follow the pattern of discussion found in the chapter on general mortality. References Berry, B. J. L. ( 1964). Approaches to regional analysis: A synthesis. Annals of the Association of American Geographers, 54( 1), 2- 11. Buescher, P. A. ( 1998). Age- adjusted death rates. Raleigh, North Carolina: North Carolina Center for Health Statistics. Earickson, R., & Harlin, J. M. ( 1994). Geographic measurement and quantitative analysis. New York : Macmillan ; Toronto; New York: Maxwell Macmillan Canada; Maxwell Macmillan International. Gregory, D., & Urry, J. ( 1985). Social relations and spatial structures. New York: St. Martin's Press. Koch, T. ( 2005). Cartographies of disease : Maps, mapping, and medicine ( 1st ed.). Redlands, Calif.: ESRI Press. OMRAN, A. R. ( 2005). The epidemiologic transition: A theory of the epidemiology of population change. Milbank Quarterly, 83( 4), 731- 757. Poore, B. S., & Chrisman, N. R. ( 2006). Order from noise: Toward a social theory of geographic information. Annals of the Association of American Geographers, 96( 3), 508- 523. Tobler, W. R. ( 1970). A computer movie simulating urban growth in the detroit region. Economic Geography, 46( Supplement: Proceedings. International Geographical Union. Commission on Quantitative Methods), 234- 240. Tufte, E. R. ( 2006). Beautiful evidence. Cheshire, Conn.: Graphics Press. Tufte, E. R. ( 2001). The visual display of quantitative information ( 2nd ed.). Cheshire, Conn.: Graphics Press. 10 11 Tufte, E. R. ( 1997). Visual explanations : Images and quantities, evidence and narrative. Cheshire, Conn.: Graphics Press. Tufte, E. R. ( 1995). Envisioning information ( 5th printing, August 1995 ed.). Cheshire, Conn.: Graphics Press. Werlen, B. ( 1993). Society action and space : An alternative human geography [ Gesellschaft, Handlung und Raum.] . London ; New York: Routledge. Wilson, J. L., & Buescher, P. A. ( 2002). Mapping mortality and morbidity rates. Raleigh, North Carolina: North Carolina Center for Health Statistics. Pitt Wake Hyde Duplin Bladen Bertie Pender Wilkes Moore Onslow Union Surry Ashe Beaufort Craven Halifax Robeson Nash Sampson Iredell Columbus Swain Carteret Burke Brunswick Johnston Anson Guilford Randolph Harnett Wayne Jones Chatham Macon Rowan Hoke Martin Tyrrell Dare Lee Stokes Stanly Lenoir Franklin Buncombe Warren Granville Davidson Jackson Haywood Gates Person Caldwell Wilson Forsyth Polk Caswell Cumberland Orange Pamlico Rutherford Madison Yadkin Gaston Clay Cherokee Richmond Cleveland Catawba Davie Rockingham McDowell Hertford Alamance Vance Avery Yancey Mecklenburg Northampton Edgecombe Montgomery Durham Graham Scotland Greene Watauga Henderson Washington Transylvania Mitchell Alleghany Currituck Camden Chowan Perquimans Pasquotank New Hanover Lincoln Cabarrus Alexander Western ( WNC) Piedmont ( PNC) Remaining 59- County Region ( RNC 59) Eastern North Carolina 29- County Sub- region ( ENC 29) Eastern North Carolina 12- County Sub- region Eastern North Carolina 41- County Region ( ENC 41) North Carolina County and Regional Locations Center for Health Services Research and Development East Carolina University Greenville, NC ECU, Center for Health Services Research and Development, 2007 Figure 1.1 Six Leading Causes of Mortality in the US 1900 to 2001 0 100 200 300 400 500 600 700 1900 1910 1920 1930 1940 1950 1960 1970 1980 1990 2000 Year Heart Disease Cancer ( All Types) Pneumonia & Influenza Tuberculosis ( All Forms) Diarrhea & Enteritis Three Infectious/ Communicable and Three Chronic Diseases Deaths per 100,000 Population* Sources: Leading Causes of Death, 1900- 1998 http:// www. cdc. gov/ datawh/ statab/ unpubd/ mortabs/ hist- tab. htm ( Last accessed Dec. 29, 2005) Data for 1999- 2001 from NCHS’s Compressed Mortality Files Stroke * Rates are not age- adjusted Center for Health Services Research and Development East Carolina University Greenville, NC ECU, Center for Health Services Research and Development, 2007 Figure 1.2 General Mortality in Eastern North Carolina 2000 to 2004 The chapter on general mortality is divided into several topics related to mortality from all causes for eastern North Carolina. They can be accessed directly with the following links. Introduction The Spatial Distribution of Crude and Age- Adjusted General Mortality Rates The Temporal Distribution of Age- Adjusted General Mortality Rates Mortality Burden The Spatial Distribution of Premature Mortality from All Causes The Temporal Distribution of Premature Mortality from All Causes From Empiricism to Explanation: General Mortality Disparities Conclusion Introduction General mortality includes all causes of death over a specified time interval. Causes of death are further defined and classified into internationally recognized series of grouped codes, such as the International Statistical Classification of Diseases and Disorders and Related Problems, 10th Revision or ICD- 10 ( World Health Organization, 2004). ( For the most recent revision of codes, see the electronic version at the World Health Organization’s website ( World Health Organization, 2006).) Periodically, revisions are made to incorporate changes in medical knowledge and to incorporate and facilitate improved coding rules ( see U. S. Department of Health and Human Services, Centers for Disease Control and Prevention, & National Center for Health Statistics, 2006). Standardized coding, in conjunction with using standard populations for age- adjustment, permits comparability of rates among different time periods and geographical units. Once the cause of death has been coded, each record is accumulated into a time and place- specific total number of deaths. The accumulated totals are then used to determine the relative ranking and importance of leading causes of death for a county or a region. Figure 2.1 portrays the resulting 5- year totals for the ten leading causes of death proportionally to the total number of deaths in ENC from 2000 to 2004. In this figure, two general classes of mortality causes dominate the mortality experience of ENC: Total Cardiovascular Disease ( TCVD) and Malignant Neoplasms ( All Cancers). From 2000 to 2004, more than 59% or 65,442 deaths have occurred due to these two disease categories. The remaining eight leading causes of death account for just 21.4% or 23,605 deaths during this same period. The number one leading cause of death in ENC for the study period is TCVD, which accounts for 37.0% ( 40,820) of the region’s 110,390 deaths. ( The TCVD category is based on the definitions proposed by the American Heart Association ( American Heart Association, 2005) and includes mortality due to stroke.) Death from malignant neoplasms is the second of the ten leading causes of death and 2 accounts for 22.3% ( 24,622) of all regional mortality. A distant third leading cause is attributed to Chronic Obstructive Pulmonary Disease and Chronic Lower Respiratory Disease COPD/ CLRD with 4.9% ( 5,384) of all deaths from this cause. Mortality from Diabetes Mellitus follows with 3.5% ( 3,904) of all ENC deaths. In fifth place, death from Unintentional Motor Vehicle Injuries ( UMVI), accounts for 2.8% ( 3,047) of the region’s deaths. Septicemia is the tenth ranking cause of death claiming 1.7% of all deaths. The ten leading causes of death are followed by a single category, All Other, which accounts for 19.3% ( 21,343) of General Mortality. Within this final category, 1,378 people have committed suicide, 1,193 people have died from chronic liver disease and cirrhosis, 1,104 people have been murdered, and 776 people have died from AIDS due to HIV ( Human Immuno- deficiency Virus). Regionally, deaths from specific causes in the All Other category make up very small percentages within general mortality. Nevertheless, when counties are examined separately, the seemingly insignificant causes of death at the regional scale can be important causes of death at the more local county level scale. It is therefore important to monitor at the “ basement” level so that emerging mortality trends at regional and local scales can be detected. The present chapter is organized around three general topics. The first two topics describe patterns of mortality from all causes, but using two different approaches in its portrayal. The first approach examines the spatial and temporal patterns of two density measures: crude and age- adjusted general mortality rates. These two measures describe mortality quantities in relation to population sizes and their distributions in space and time. However, density measures do not provide information about what part of a population is being affected. Mortality, whether from specific or general causes, can affect populations in a differential manner across spatial and temporal dimensions. Measuring the cumulative differences of age- at- death of individuals that occur before an accepted standard age- at- death ( say, 75 years) produces information about the level of premature mortality. Larger amounts of years of potential life lost in a population signify greater levels of mortality burden being shouldered by that population. The second topic covered in this chapter addresses the distributions of premature mortality in eastern North Carolina and the state. Finally, we move from empirical descriptions to a brief discussion of how patterns of general mortality can be explained by their relationships to other factors. The Spatial Distribution of Crude and Age- Adjusted General Mortality Rates A map of crude mortality rates will draw the map- reader’s attention to those areas that are experiencing the highest numbers of deaths relative to their local populations. The crude mortality rate measures the density of resident deaths occurring in an area in relation to the population of that area. It is a summary measure representing the proportion of a population that has died over a given interval of time. Because this proportion is frequently a very small value, it is multiplied by a larger number ( of persons) like 1,000, 10,000, 100,000, or even 3 1,000,000 for extremely rare causes of death. ( This atlas will employ the multiple of 100,000 persons when discussing and comparing density measures.) Because age is the greatest risk factor for dying, the map of crude mortality rates is also, to some degree, a map of the underlying spatial distribution of population age structures. Controlling for the effects of age variation will permit the map reader to make comparisons of mortality rates among different areas on the map. This is accomplished through the technique of age- adjustment, which adjusts the observed number of deaths to an expected number of deaths if the population under study had the same age structure as some external reference population ( Buescher, 1998). In this atlas, the US Standard Million for the year 2000 is employed ( Anderson & Rosenberg, 1998). It is extremely important that the standard population used in each case is the same when comparing age-adjusted maps from one period of time to another, or when comparing maps of age- adjusted rates from other states. Different model or standard populations will generate different age- adjusted rates even when the actual or observed distribution of deaths across the population age distribution remains constant. Figure 2.2 shows the mapped distribution of crude mortality rates from all causes for counties in the contiguous US from 2001 to 2003. The category classification is based on the extension of the classification scheme used in the North Carolina mortality maps discussed later in this chapter.∗ Higher rates of general mortality in this map are concentrated in the central part of the nation, which includes the Great Plains and Midwest, the South, and the outlying high rate counties in the Far West. There is also a significant cluster of counties centered in mountainous West Virginia and eastern Kentucky. Recall that age is the greatest risk for dying. Figure 2.3 is a map that shows the distribution of county level proportions of people 60 years of age and older for the 2000 US Census year. The distribution of the proportions of elderly is similar to the distribution of the higher crude mortality rates seen in the previous figure. Statistically, the relationship, measured as a correlation, results in an r- value of 0.81 and an R2 of 0.66, which means that 66% of the variation in county crude rates of mortality is explained by just the proportions of individuals greater than 60 years of age. Although the ∗ The North Carolina rates ( crude and age- adjusted) are based on more current numbers from the State Center for Health Statistics and State Data Center and use a five- year period such as 2000 to 2004. Because numbers for the entire US are usually available three years behind the state’s and that there are a significant number of US counties that experience small numbers of mortality events, US county rates ( crude and age- adjusted) in this work are based on three- year aggregations ( 2001 and 2003) from the National Center for Health Statistics’ Compressed Mortality File ( CMF) 1999 to 2003. The center point year or fulcrum year for the US county rate maps is 2002. That same year is the fulcrum for the period 2000 to 2004, which is the period used in the state and regional discussions throughout the Atlas. However, in the state and regional comparisons, the US value for one year is 2002, because their numbers are sufficiently large not to warrant aggregation. It should also be noted that the rates generated for North Carolina counties in the three- year US map will be slightly different than those generated for the five- year NC maps seen elsewhere in the Atlas. This is the result of using different numbers of data points ( three and five) and slight differences found in the denominators ( i. e., county populations) between the US ( CMF 1999- 2003) and NC ( state demographic estimates) data sources. 4 crude mortality rate map depicts where mortality is occurring in relation to population age structure, this map cannot be used to make meaningful comparisons among individual counties because their respective age structures are different. Figure 2.4 shows the effect that age- adjustment has on the county mortality pattern using an external standard population ( US 2000 Standard Million). The high rate counties shift and concentrate their spatial distribution to the Ozarks, Lower Mississippi Valley, the southern Coastal Plain, and the south-central Appalachian region of West Virginia and eastern Kentucky. A few outlying high rate counties are found scattered throughout the west, which generally correspond to Indian reservations. The inset map in figure 2.4 shows that ENC is a northern extension of the high rates of age- adjusted mortality found in the southern Coastal Plain. To contrast, the national age- adjusted map also shows that most of the remaining counties on North Carolina ( RNC) are part of the southern extension of the much more favorable mortality conditions of the Northeast. Maps showing the spatial distribution of crude and age- adjusted mortality rates from all causes in both NC and ENC for the years 2000 to 2004 are found in figure 2.5. Individual county and regional mortality rates are listed in table 2.1 and their locations can be found using the map in appendix A of this chapter. The state map for crude rates shows that the greatest mortality burden is experienced at both ends of the state. Western North Carolina ( WNC) has the highest crude general mortality rate of 1,080 deaths per 100,000 people. The next highest crude mortality rate is found in the northeast 29- county region of North Carolina ( ENC 29) with a rate of 967.7-- 7% higher than the 41- county ENC regional rate of 905.4. The highest county rates in ENC cluster together along a northwest- southeast axis. In this cluster, Hertford County experiences the highest rate of 1,330.6 and the second highest rate is found in its southern neighboring county, Bertie, at 1,313.3. Onslow County’s rate is the lowest in the region ( 518.4), experiencing mortality at just 39% of Hertford’s level. The greatest local impact of mortality from all causes is felt in the northern county cluster of ENC which also possesses an older aging- in- place population. This contrasts starkly to Onslow County where a significant portion of its population is made up of young, transient, military service- age people. When the county rates are age- adjusted ( figure 2.5b), the high mortality categories become more concentrated in the east and less so in the state’s mountainous west-- where the relative older ages of those county populations played a role in that region’s observed higher crude rates. The effect of age adjustment on relatively youthful county populations with low crude mortality rates can be quite dramatic. For example, when the crude rates for Cumberland and Onslow-- two counties with large military aged populations-- are age- adjusted, there is an apparent jump in mortality rates of 51% and 84%, respectively. At the regional level, age- adjustment widens the disparity for general mortality between ENC and RNC to more than 12% from the crude rate difference of 7%. Within the ENC region, counties with the highest age- adjusted rates form two centers, 5 one in the northeastern 29- county sub- region of ENC and the other in the remaining southern 12- county sub- region. In the 29- county sub- region, Edgecombe County possesses the highest rate ( 1125.6) among the eight county cluster found there. Other counties in this cluster include Halifax ( 1023.9), Northampton ( 1007.0), Hertford ( 1124.5), Gates ( 1061.1), Bertie ( 1111.2), Martin ( 1091.0), and Washington ( 1027.8).∗ The highest age- adjusted general mortality rate is found in the southern sub- region. Robeson County with 1133.3 age-adjusted deaths per 100,000 has the highest general mortality rate in the state and forms the core of the southern- center- high- rate- county- cluster. This county cluster includes Scotland ( 1063.6), Hoke ( 1015.5), Bladen ( 1101.7), and Columbus ( 1069.9) counties. A north- northeast linear series of adjacent high rate counties ( Sampson, Wayne, and Lenoir) continues from the northeastern border of Bladen County. Immediately adjacent to the east of the southern cluster is a three county cluster of the lowest age- adjusted general mortality rates found in the entire 41 county region. The counties in this cluster include New Hanover ( 832.7), Brunswick ( 842.8), and Pender ( 830.1). All three of these counties have rates that are more favorable than the RNC 59 county rate of 866.1 age- adjusted deaths per 100,000. The examination and comparison of crude and age- adjusted general mortality in North Carolina yields two conclusions. First, the higher crude rates found in the east like in the western counties, can partly be explained as a function of greater proportions of elderly. Second, when general mortality rates are adjusted for age, 16 of the 20 highest rate counties are found in ENC 41, which clearly demarcates this region as one experiencing a greater mortality burden in both an absolute and relative sense. Many of the counties in this discussion will be seen again in later sections of the Atlas when the geographic patterns of mortality from specific causes are explored. Figure 2.6 shows the contributions that race- sex specific age- adjusted general mortality make to the overall pattern of general mortality in North Carolina seen in the age- adjusted map ( fig. 2.5b). Applying the same age- adjusted rate category classification found in figure 2.5b to the rate distributions of each of the four demographic groups in figure 2.6 produces four distinct map patterns. Males of both races have higher rates ( i. e. they occupy the highest rate category: 1,004.3 to 2,107.8) of general mortality throughout the state. White males ( fig. 2.6c) have a state rate of 1,034.0 and a regional ( ENC 41) rate of 1,100.2 compared to nonwhite males whose rates are 1,336.2 and 1,406.0, respectively. Counties with larger proportions of retirement age populations, found within each of the state’s three physiographic provinces, as well as the larger metropolitan counties of the Piedmont have lower rates of death from all causes for white males. For nonwhite males, 95 of the state’s 100 counties are in the highest rate category. The remaining five counties are found in the westernmost portion of the state, and their lower rates for nonwhite males are probably the result of fewer people being in this race- sex group in the western region. The age- adjusted death rates ∗ In the three year average ( 2001 to 2003) for the 3,100 plus counties of the US, Martin County raked 15th highest in the nation at 1313 age- adjusted deaths per 100,000. 6 for females of either race are significantly less than their male counterparts. White females ( fig. 2.6b) have a state rate of 720.4 and a regional ( ENC 41) rate of 773.5 compared to nonwhite females ( fig. 2.6d) whose rates are 857.3 and 891.6, respectively. White females have rates in the lowest map category throughout the state. Eight out of eleven of this group’s highest rate counties are found in ENC. Nonwhite females have the most complex spatial distribution of mortality. A wide range of rates are observed throughout the state with the largest concentration of high rate counties found in ENC. Another large concentration of higher rate counties can be found along a north- south axis in the central Piedmont. In some counties, these high rates may be attributed to the smaller representation of this demographic group and thereby the potential effects of random variation of rates due to small numbers. Overall, there is very little geographic effect on nonwhite males with respect to the age- adjusted general mortality map patterns. White males, and females of both racial groups appear to shape or delimit the regional distribution of mortality from all causes, while the relatively greater proportion of nonwhite males in ENC further accentuates the high general mortality rates found in that region. When general mortality rates for North Carolina are age- adjusted for the years 2000 to 2004, 35 of ENC’s 41 counties ( 85%) emerge with rates above the state rate of 896.5, while 26 of RNC’s 59 counties ( 44%) do so. Partitioning the general mortality map for the total population into four separate maps based on race and sex reveals how the distribution of rates for the total population is weighted and shaped by its constituent sub- populations. Later chapters of the atlas will show the impacts of specific leading causes of death on these sub-populations and their subsequent contribution to the observed spatial patterns of general mortality. The age- adjusted general mortality map of NC and ENC represents the integration of the patterns produced by component leading causes of death. It is also the culmination of many different mortality processes that have been operating at their own characteristic scales, tempos, and modes. The next section discusses how some of these processes have affected the observed pattern of mortality in ENC over time. The Temporal Distribution of Age- Adjusted General Mortality Rates The following two figures ( 2.7 and 2.8) show how mortality has evolved over the 26- year time period from 1979 to 2004. The last five data points ( the years 2000 through 2004) in the ENC 41, RNC 59, and NC time series illustrate the amount of variation in annual rates that are subsumed into the single age- adjusted five-year ( 2000- 2004) rate seen in the preceding table and maps. A trend line, shown by dashes in the figure, is fitted for each of the time series and extended to the year 2010. The trend line is calculated based on information from the entire series of data points ( i. e., annual rates). Additional information about the trend line is also provided below the figure. This information includes the percent change in rates from the initial year to the latest year in the time series. The R2 value is a measure of how well the fitted trend line corresponds to the observed 7 series. The equation of the line, also shown, generates the trend line that allows the investigator to calculate an expected value for a given point in time. Time series trend lines can diverge, converge, or run parallel to one another. To make analysis easier, linearity of the observed data is assumed for the 26- year period in these time series graphs. However, broader temporal scales of observation show that mortality from any number of causes is generally non- linear ( see figure 1.2). With the simplifying assumption of linearity, it is possible to calculate an approximate time when two series will have the same rate ( convergence) or when two series began to separate ( divergence) from each other by setting the two equations of the line equal to one another. However this should be done only when R2 values are high ( i. e., approaching 1.00) and when making projections into the near future or more recent past. Making projections too far into the future, or past, over- extends the more limiting and linear perspective of recent mortality trends, resulting in the danger of making spurious conclusions about long- term and, most, likely non- linear processes. For example, using the equations- of- the- line in the trends description section found in figure 2.7, the age- adjusted general mortality rate for ENC 41 and RNC 59 will not be equal or converge until the year 2154! Clearly, the use of linear trend lines should only be used short term prognostication. Their utility lies in permitting the researcher to make summary assessments and examine potentially meaningful trends, emerging differences or improvements in rate disparities. Figure 2.7 illustrates four solid trends in regional declines of general mortality. The goodness- of- fit lines are all above 0.90, indicating that from 1979 to 2004 there are very tight fits to the modeled trend line and that predictions for the next several years could be reasonably and confidently made. Over this 26- year study period, age- adjusted mortality rates have declined by 16 and 17 percent for all four regions. The greatest decreasing coefficient belongs to the US (- 7.75) and the least to RNC 59 (- 6.34). This translates into an average growing disparity of age- adjusted general mortality rates of about 1.4 age- adjusted deaths per 100,000 per year over the course of the last 26 years. Although all trends are certainly favorable in absolute terms, the ENC 41 trend line stands out well above the others with the line equations demonstrating persistent relative disparity in mortality rate trends between this region and RNC 59. A closer look at the mortality experience of ENC 41 reveals substantial differences by race and gender. Figure 2.8 shows relatively flat trends ( from negligible to 7% decrease) for females, with only a slight growth in disparity by race ( see trend descriptions) over the 26- year period. White males have had the greatest amount of rate decline with a 28% decrease from 1979 to 2004. The trend for white males is very consistent over time and can probably be used reliably as an indicator of mortality scenarios in the near future. Nonwhite males follow with a more modest 16% decrease and a less confident trend line than their white counterparts. Although the trend lines for males from either racial 8 group are decreasing, the relative rate disparity between them, as measured by the equations- of- the- line, increases from 17% in 1979 to 37% in 2004. Since 1979, age- adjusted general mortality has been improving for all males in the region, while rates have remained relatively flat or changing little for regional females. The female pattern suggests that mortality rates may reach an asymptotic level for a period of time. One reason for this flattening out might be that all benefits from current health technologies, innovations, knowledge concerning care and behaviors have been nearly realized for that group over the last two to three decades. There may also be a certain amount of intra- regional “ balancing out” or counteracting of high and low rates among counties in different parts of ENC 41. The trend lines for males are converging on the trend lines for females— with white males approximating the mortality rates for nonwhite females some time around the year 2014 or 2015. It will be interesting to see if white males, and probably much later for nonwhite males, begin to approach a similar mortality asymptote as has been the case for females. It is likely that the reasons for the relatively low rates for females have yet to be completely realized for males, but the rates show that they are still in the process of responding to or adopting mortality reducing behaviors and technologies. Certainly the pattern between both female groups indicates that differential mortality remains even when rates are low and relatively stable. What accounts for this persistent differential forms the bulk of health disparities literature today. The above discussion and description of the patterns of crude and age- adjusted mortality reveals that a geographic disparity exists between the 41 county ( and 29 county) region of NC and the remaining counties of the state, with the east experiencing significantly higher rates than RNC. Within ENC, age- adjusted general mortality rates have been declining over the past three decades for the major demographic groups discussed in this chapter. For females of both racial groups the decline is relatively minimal, but for males the decline has been more dramatic, with nonwhite males having the sharpest decrease. Nonwhite males, although experiencing a larger decrease in general mortality rates have begun their downward trek at a much higher beginning rate so that the relative rate disparities between them and the other demographic groups will remain high for the foreseeable future. As previously mentioned, density measures tend to mask other types of information that can be derived from mortality records. The next section focuses on the concept of mortality burden and its measurement. Understanding the impact of premature mortality on county populations can assist in discriminating where disparities of mortality burden are occurring. 9 Mortality Burden Mortality burden can be viewed at different scales of impact. Within a family there is the obvious psychological, social, and economic impact of a member’s death. The decedent’s stage in the life cycle, occupation, resources, and position in society also has relevance in broader local and community scales of social relationships. Implicit in any decedent’s age at death is the tangible and intangible cost, benefit, and potential contribution of that individual’s life to both family and friends, and to the larger extended communities to which he or she belonged. Collectively these mortality experiences can be summarized into one point value: crude mortality rate. This density measure indicates the direct arithmetic impact or burden actual deaths can have on a population. However, a population with an older age structure will naturally have more individuals at risk of dying as they enter the latter stages of their life cycle during a given time interval and so that population may appear to be experiencing a higher burden of mortality. Another way to look at mortality burden is to look at how much potential life is lost, which is a comparison of an observed age- at- death against some expected or standard age at death. Instead of one point value, two point values are used, with greater differences between corresponding to increases in mortality burden. Age- at- death can be used to measure the amount of life lost prematurely from a standard number of years of life that an individual can be expected to live in the population of interest. The typical standard age used in current research is 75 years, which is close to 77.5 years, the life expectancy at birth ( e0) for the US in 2003 ( Arias, 2006) and nearly identical to the mean age of death in North Carolina. The number of deaths and their ages of occurrence before the age of 75 can be accumulated, age- adjusted, and normalized by the underlying population. Greater differences mean greater years of life lost, when calculated in this manner, and indicates a greater level of mortality burden being experienced prematurely. The meaning of a premature mortality rate or years of life lost rate as described above is qualitatively different than for the more commonly used density measures. To illustrate, the age- adjusted mortality rate in North Carolina for female breast cancer was 25.6 per 100,000 and for prostate cancer in males it was 29.1 per 100,000 in 2004. A comparison of these two rates would lead one to the conclusion that prostate was a slightly bigger killer of men than breast cancer is in women. However, when the premature mortality rates∗ for these two causes of death are compared, the number of years of life lost before age 75 is 33.9 years per 10,000 for female breast cancer and 6.4 years for prostat e ∗ Currently premature mortality is typically measured by the number of years of life lost ( YLL) before age 75 per 10,000 people. Each death is aggregated into an age category and the total number of deaths in that category is multiplied by the difference between the age category mean age at death and age 75. The resulting age category YLLs are then summed, divided by the population, and then multiplied by 10,000 to make interpretation easier. The YLL- 75 ( premature mortality) measure can either be crude or age- adjusted. 10 cancer. These values indicate that males tend to die at much later ages from prostate cancer and not prematurely relative to the age of 75. Females tend to die from breast cancer at earlier ages, suffering a greater mortality burden than their male counterparts for a sex- specific disease, with perhaps a greater impact on families and communities. The Spatial Distribution of Premature Mortality from All Causes The national age- adjusted premature mortality rate for the year 2002 is 751 years of life lost per 10,000 people. The lowest state premature mortality rate in this same year is found in Vermont at 568, while the worst state rate belongs to Mississippi at 1088. If the District of Columbia is added as a state it would fall behind Mississippi ranking a distant 51st with a rate of 1323. Within the state rankings, North Carolina is 39th with a rate of 833, and with the exception of Florida and Virginia, has the lowest premature mortality of the remaining southern states. If the 41 county region of ENC is entered into the state rankings, it would rank 47th at 959, with Arkansas, Alabama, Louisiana, Mississippi, and the District of Columbia trailing in the lowest ranks.† The 29- county region of ENC would rank 48th at 975, ousting Tennessee, which moves up to 47th. The Piedmont region compares more favorably as a state with a premature mortality rate of 774 placing it 29th among the states. The Western NC region has a more intermediate premature mortality rate of 805 and ranks 34th. Figure 2.9 is a map of the United States that shows the age- adjusted premature mortality rates for the states with North Carolina’s three regions mapped as ” states”. From this national context we now move to a more specific in- depth discussion of how premature mortality varies by sub- region and county within North Carolina. Figure 2.10 portrays both crude ( fig. 10a) and age- adjusted ( fig 2.10b) premature mortality rates measured as years of life lost before the age of 75 years ( YLL- 75). The maps in this figure describe the distribution of mortality burden for counties. Unlike the maps in figure 2.5, age- adjusting the rates ( i. e., the expected number of deaths) has very little effect on the map pattern of premature mortality. The ENC 41- county region stands out distinctly relative to the other regions of the state with its large number of high premature mortality counties. Table 2.2 bears this out with the age- adjusted rate for premature mortality 22% higher than RNC, and when compared to PNC and WNC, the region is 23% and 17% higher, respectively. Finally, the age- adjusted premature mortality rate for ENC is 27% higher than the rate for the nation, which for the year 2002, is 751.0. When premature mortality is compared on a national and regional level, the counties of North Carolina and ENC do not fare well. Only 14 NC counties in the state have premature mortality rates less than the US 2002 rate, with New † ENC 41 and 29 county regional rates, as well as other NC regional rates, are calculated using the National Center for Health Statistics’ Compressed Mortality File series data for the year 2002. 11 Hanover, at 705.6 years of life lost per 10,000, being the only county in the east to do so. Regionally, 36 of the 41 counties in ENC ( 84%) have rates above the North Carolina rate, while 27 of the 59 counties of the remaining NC counties ( 46%) have rates greater than the state. In terms of population exposed to risk of dying prematurely at a rate higher than the state, the difference between the two regions becomes even more dramatic. In ENC, 84% of the region’s population who are under the age of 75 years live in those counties that have higher rates than the state, while 27% of RNC’s population under 75 live in counties with a higher rate than the state. Moving to the individual county comparisons, Wake County experiences the least years of life lost in the state for the 2000 to 2004 period with a rate of 564.8 years per 10,000, which is 32% lower than the state rate. Robeson County has the least favorable rate for this study period at 1,234.7 years of life lost, 119% greater than the rate for Wake County. When the age- adjusted map for premature mortality for all causes is decomposed into maps focusing on the four demographic groups, differences in their contributions to the overall rates emerge ( see figure 2.11). The greatest contribution to the overall rate is made by nonwhite males ( fig. 2.11b). Like age-adjusted mortality rates, high county rates for this group are a ubiquitous feature throughout North Carolina, with the exception of a few counties in the western part of the state. ( For county locations, see the map in appendix A.) Duplin County, in southern ENC had the highest premature mortality rate in the state at 2,133.4 years of life lost per 10,000 ( see table 2.2). To contrast, white females ( fig. 2.11b) have ubiquitously low county rates with the highest state- wide county rate found in an ENC county, Northampton, at 803.4. Regionally, the lowest rate for white females is found in New Hanover County at 412.4, slightly more than half of the Northampton rate. Overlaying these two contrasting map patterns, are the rate distributions of white males ( fig. 2.11a) and nonwhite females ( fig. 2.11d). Both of these map patterns are more variegated than the previous two. The mapped distribution for white males, though heterogeneous, is weighted more by the higher rate categories concentrated primarily in ENC, but also found distributed throughout the peripheral non- metropolitan counties of the Piedmont, and the western counties. The highest rate for white males, 1,563.0 years of life lost, is found in Robeson County located in southern ENC on the South Carolina border. Like white males, the distribution of high rate categories for nonwhite female culminates in the east, while high rate counties are found scattered to the west of the region. For this group, the highest rate-- 1,517.3 years of life lost-- is found in Perquimans County. While the highest rates for each of the four demographic groups are found in ENC, the lowest rates for any of these groups are located outside of ENC. For white males and females ( fig. 2.11a— b), the lowest premature mortality rates are found in Wake County at rates of 578.1 and 352.9, respectively. The lowest meaningful rates ( i. e., rates calculated from deaths numbering 20 and more) for their nonwhite counterparts are found in McDowell County with nonwhite males at 977.8 years and nonwhite females at 494.0 years in Wilkes County. Both of these counties are found in the western portion of the state. To conclude, there is a discernable geographic difference in 12 mortality burden between ENC and RNC that is driven by the mortality experience of white males and nonwhite females The Temporal Distribution of Premature Mortality from All Causes Figure 2.12 is a comparison of premature mortality trends among regions from the years 1979 to 2004 ( 2002 for the US). Premature mortality for all four regions is declining at approximately the same rates. The relationships among the trends, in terms of their relative ranking in years of life lost rates, remains constant throughout the time series study period. ENC consistently experiences the highest rates of age- adjusted premature mortality but the trend line indicates an approximate 27% decrease from the beginning of the study period in 1979 to 2004, slightly less than the other regions. All regions show a similar pattern of decline, including the gentle oscillations of observed values about their respective trend lines. For the first five or six years, the decline in trends is steeper than any other interval in the series. Thereafter, the observed regional rates decline less steeply and fluctuate very little from their respective trend lines. This suggests that there may be emerging countervailing trends in premature mortality from specific causes, which either balance each other out or have become more stable over time. When ENC’s observed premature general mortality rate series is decomposed into four separate premature mortality series corresponding to each of the four demographic sub- groups, several distinct patterns emerge ( figure 2.13). The greatest decline in premature mortality is experienced by white males at 34%. Although nonwhite males have the largest negative coefficient (- 30.35), indicating the steepest rate of decline, they begin the series with an expected or modeled premature mortality rate ( the intercept) at a level some 72% greater than their white counterparts. The pattern of decline for this group is very similar to the one observed for regions and it may be that the nonwhite male experience is what is driving the patterns seen in the previous figure. The trend line for white males is decreasing more than twice as fast as the trend for nonwhite females and overtakes the latter sometime around the year 2002. The observed rate patterns suggest a convergence— a convergence that has been evident for the 10 years prior to 2002. For the last two or three years of the series the rates appear to be diverging but it is probably not indicative of a reversal in trends. The last demographic group, white females, shows the least amount of decrease ( 17%) over the 26- year period and like the age- adjusted mortality rates for this group they appear to approaching a rate asymptote. If present trends continue for the four demographic groups, the next convergence of premature mortality rates will occur between white males and females around the year 2030. As with age-adjusted mortality, decomposing the general premature mortality rate by demographic groups reveals differences and potential disparities among them. The shift in the county distributions from crude premature mortality rates to age-adjusted premature mortality rates is minimal when compared to west- to- east 13 shift in distributions of the density measures. One reason for this difference in pattern shifting is that ages at death close to 75 years have a small negative impact on the premature mortality outcome measure and a zero impact when deaths occur after that age. Larger numbers of deaths occurring at ages several years prior to 75 indicate a population experiencing a greater share of mortality burden as an outcome. For example, the accumulation of years of life lost due to high infant mortality rates, and earlier ages at death from cardiovascular diseases and cancer can reflect inherent problems with access to appropriate healthcare. ( Density measures essentially treat all deaths as equal in impact and cannot be used to measure the depth of mortality burden.) Regionally, this suggests that although the western region of the state possesses populations with higher relative proportions of elderly, their respective mortality burdens are not greater than expected. This contrasts to what the measure portrays for the eastern 41 counties of the state— a region that not only has a high proportion of elderly population with its attendant mortality but it is also a region that has a disproportionate number of its population dying prematurely. From Empiricism to Explanation: General Mortality Disparities The numerical evidence tells us that mortality is not experienced equally between ENC and the 59 remaining counties of North Carolina ( RNC). From 2000 to 2004, 110,390 deaths occurred in ENC and 249,278 deaths occurred in RNC. The latter region’s population is larger with a 5- year population- at- risk of 29,349,691 individuals compared to ENC’s 12,192,418. Proportionally, the expected number of deaths for ENC numbers would be 103,555. Subtracting the proportionalized mortality from the observed value of 110,390 yields an excess of 6,835 deaths ( 6.6% more) carried by ENC-- a crude measure of a geographic disparity for general mortality between the two regions. However, this does not account for the probable regional differences in age structure. ( Recall that age is the greatest risk factor for any individual dying during a specified time interval.) If age structure is controlled for the two regions, the difference in the number of deaths between the two regions grows to 12,924 ( 12.2% more), nearly twice the observed value and further exacerbating the apparent geographic disparity between the two regions. ENC experiences a greater burden of mortality-- almost 2,600 more deaths per year than would be expected given its population size and its age- structure. Characteristics other than population size and age may affect the observed and adjusted mortality disparity between the two regions. We can hypothesize ( or speculate) that there may be other factors or covariates at work with mortality rates that are also geographically distributed. For illustrative purposes, explanatory variables might include underlying racial and ethnic diversity, poverty, and rurality. The rationale or assumption for the choices of these variables is that income distribution ( related to racial/ ethnic diversity) and measurable financial and physical ( distance) access to health care have some discernable effect or relationship to mortality. However, one can further 14 speculate that these covariates are associated with many other measurable variables such as educational attainment, occupation and associated social relations ( including peer pressure), risky or health promoting behaviors, the value and awareness of health as a personal and social good, and so on. The first three covariates introduced can be thought of as surrogate measures— they are meant to capture and simplify a complex series of relationships among a spectrum of factors that are operating at different scales. Surrogate measures are used to assist the students of public health and mortality in focusing on those relationships with the most explanatory power and in the construction of the most parsimonious ecological model of mortality. Racial/ ethnic diversity, poverty, and rurality can be measured like age- adjusted mortality at the county and county- based regional level. For example, ENC’s county populations are more racially and ethnically diverse when compared to the counties of the rest of the state ( RNC). According to the US Census atlas, Mapping Census 2000: the Geography of U. S. Diversity ( see page 22 in Brewer & Suchan, 2001), 26 of ENC’s 41 counties have diversity index values at or above the US value of 0.49, with a regional index of 0.52. ( The diversity index is a measure of the probability that any two random people chosen from a county’s population will be of a different race.) Only 13 of RNC’s 59 counties are more diverse than the US, with a regional index of 0.39. From the US Census year 2000 ( 1999) data, a little more than 16.0% of ENC’s population is below the poverty line for that census year, which is almost 50% greater than the 10.7% reported for RNC’s population. Rurality is another attribute that distinguishes ENC from RNC. Slightly less than 49% of ENC’s population is classified as rural by the US Census Bureau, which contrasts to slightly more than 36% of RNC’s population being rural. The next step is to determine what influence or how well these proposed variables explain the county distribution of general mortality. To assess the relationships and associations between any two of these variables, we employ a methodology similar to that used in studying the temporal trends of general mortality. The following discussion will describe the linear relationships between the dependent ( age- adjusted mortality from all deaths) variable and each of the independent variables: the diversity index, poverty, and the proportion of rural population.. The interrelationships among the independent variables will also be examined. Exploring the strengths and weaknesses of association among variables is fundamental to hypotheses testing and the development of explanatory models. The correlation coefficient between mortality and the diversity index is 0.61. The adjusted R2 value is 0.365, which translates into more than 36% of the variation in mortality is explained by the variation found in the diversity index alone. The correlation coefficient between mortality and poverty is 0.63 and has an R2 of 0.385. More than 38% of the variation in mortality is explained by poverty alone— about 2% more than the diversity index. The least amount of explanation ( 0.0%) can be attributed to the measure of rurality. The correlation coefficient is 15 only 0.046, which produces the negligible R2 of 0.002. These simple analyses show that ethnic/ racial diversity and poverty have a substantial and direct effect on mortality. The next step would be to determine if there was any direct relationship between diversity and poverty and whether at some indirect level, rurality having some effect. A relationship among these variables would indicate that their effects on mortality were not independent. To get a handle on the amount of interaction between diversity and poverty ( collinearity) we can apply the same method used in the preceding example Lower R2 values will suggest smaller amounts of collinearity, less association, and more independence among the independent variables. For rurality and poverty the R is 0.344 and the R2 is 0.118, which means there is a small level of rurality and poverty associated with each other at the county level. Next, the diversity index and poverty measure yield an R of 0.531, with an R2 of 0.27, which means that racial/ ethnic diversity is more related to poverty than the degree of county rurality. How related is a county’s racial and ethnic diversity to its level of rurality? The R for this comparison is 0.186 with an R2 value of just 0.025. Recall that poverty, in this simple example, offers the greatest explanation of mortality. We now know that while rurality has some effect on poverty, diversity has an even greater effect on this variable. In more elaborate models of explanation, the rurality measure ( as devised here) would not contribute much to explanation and could probably be excluded. The foregoing discussion is meant as a simple example of how empirical descriptions of mortality can provide a basis for research questions and the building statistically oriented explanatory models. However, numerical and graphical descriptions of mortality can also stimulate further research or thinking in non- statistical ways. For example, thoughtfully publicized rate increases in mortality due to automobile accidents or diabetes will raise the awareness of policy makers and citizenry and help promote interventions, funding, and other ameliorative measures. Empirical description and explanatory models each have their own place and can be useful adjuncts to each other in the presentation and understanding of public health and demographic problems. Conclusion Geographically, different ways of measuring and describing general mortality demonstrates that the eastern 41 counties of North Carolina experience both higher comparative levels of death from all causes and a disproportionate share of mortality burden in regional and national contexts. When general mortality rates for ENC 41 are decomposed into four major demographic groups, rate differentials ( or disparities) emerge. The distributions of age- adjusted general mortality rates also have unique characteristics for each of the race- sex sub-populations. Time series depictions ( 1979 to 2004) for both regions and race-sex sub- populations also show that there has been progress, but relatively large gaps or “ disparities” continue to exist. For sub- populations, males of both racial 16 groups have greater relative declines in their rates compared to their female counterparts. All measures, spatially and temporally, indicate that although absolute differences in general mortality has been declining among regions and sub- populations, relative disparities will continue for some time to come. A description and examination of general mortality, which reveals the great disparities observed in our region of interest, naturally leads to further questions about how and why such disparities exist. With this in mind, we enter into the realm of explanation and can begin to consider the relationships and associations of covariates and mortality. Explanatory models are valuable aids for determining where changes can be effected and where healthcare resources can best be allocated. General mortality encompasses a myriad of causes of death, all which have been classified and coded. In this regional atlas of mortality, the subsequent chapters will address the ten leading causes of death as shown in figure 2.1. These ten leading causes of death account for more than 80% of deaths occurring in ENC 41 during the years 2000 through 2004. It is our hope that a consideration of each of these will lead to an increased understanding in the exceptional character of the region’s mortality experience. References American Heart Association. ( 2005). Heart disease and stroke statistics-- 2005 update. Dallas, Texas: American Heart Association. Anderson, R. N., & Rosenberg, H. M. ( 1998). Age standardization of death rates: Implementation of the year 2000 standard. National Vital Statistics Reports, 47( 3) Arias, E. ( 2006). United states life tables, 2003. National Vital Statistics Reports, 54( 14) Brewer, C. A., & Suchan, T. A. ( 2001). Mapping census 2000: The geography of U. S. diversity. Washington, D. C.: U. S. Government Printing Office. 17 Buescher, P. A. ( 1998). Age- adjusted death rates ( 13th ed.). Raleigh, North Carolina: North Carolina Center for Health Statistics. U. S. Department of Health and Human Services, Centers for Disease Control and Prevention & National Center for Health Statistics. ( 2006). International classification of diseases, tenth revision ( ICD- 10). Retrieved 10/ 20, 2006, from http:// www. cdc. gov/ nchs/ about/ major/ dvs/ icd10des. htm World Health Organization. ( 2006). International statistical classification of diseases and disorders and related problems 10th revision for 2006. Retrieved 10/ 20, 2006, from http:// www. who. int/ classifications/ apps/ icd/ icd10online/ World Health Organization. ( 2004). International statistical classification of diseases and related health problems ( 10th revision, 2nd ed.). Geneva: World Health Organization. Total Cardiovascular Disease 37.0% Malignant Neoplasms 22.3% COPD/ CLRD1 4.9% Diabetes Mellitus 3.5% UMVI2 2.8% AOUIAD3 2.5% Pneu/ Infl4 2.3% NNN5 2.0% Alzheimer’s 1.7% Septicemia 1.7% All Other 19.3% 1Chronic Obstructive Pulmonary Diseases and Allied Conditions/ Chronic Lower Respiratory Disease 2Unintentional Motor Vehicle Injuries 3All Other Unintentional Injuries and Adverse Effects 4Pneumonia and Influenza 5Nephritis, Nephrotic Syndrome, and Nephrosis Figure 2.1: General Mortality in Eastern North Carolina 2000 to 2004 Percent Contributions from the Top Ten Leading Causes of Death to the 5- year Total Number of Deaths: 110,390 ECU, Center for Health Services Research and Development, 2007 0.0 to 856.8 856.8 to 974.2 974.2 to 1064.9 1064.9 to 1188.7 1188.6 to 2018.3 Figure 2.2 US Crude General Mortality Rates1 2001 to 2003 Per 100,000 Population ENC 41 Counties ECU, Center for Health Services Research and Development, 2007 1Data from Compressed Mortality Files 1999 to 2003 0.03 to 0.15 0.15 to 0.18 0.18 to 0.20 0.20 to 0.23 0.23 to 0.42 Figure 2.3 US County Population Proportions 60 Years and Older1 2000 County Proportion GTE 60 Years ENC 41 Counties ECU, Center for Health Services Research and Development, 2007 1Data from US Census 2000 0.0 to 843.0 843.0 to 896.8 896.8 to 944.8 944.8 to 1004.3 1004.3 to 2018.3 Per 100,000 Population Figure 2.4 US Age- Adjusted General Mortality Rates1 2001 to 2003 ENC 41 Counties ECU, Center for Health Services Research and Development, 2007 1Data from Compressed Mortality Files 1999 to 2003 and the 2000 Standard Million Population for the US Per 100,000 Population a. Crude b. Age- Adjusted1 Data Source: Odum Institute, UNC— Chapel Hill Per 100,000 Population 503.6 to 856.8 856.8 to 974.2 974.2 to 1064.9 1064.9 to 1188.6 1188.6 to 1480.9 752.3 to 843.1 843.1 to 896.8 896.8 to 944.8 944.8 to 1004.3 1004.3 to 1133.3 ECU, Center for Health Services Research and Development, 2007 1 Five- Year Average, Age- Adjusted Rates Standardized to US 2000 SM Mortality Rates from All Causes of Death: North Carolina and Eastern North Carolina Total Population, 2000- 2004 Figure 2.5 ECU, Center for Health Services Research and Development, 2007 Age- Adjusted1 Mortality Rates from All Causes of Death: North Carolina and Eastern North Carolina Race- Sex Specific, 2000- 2004 a. White Males b. White Females c. Non- White d. Non- White Per 100,000 Population 1 Five- Year Average, Age- Adjusted Rates Standardized to US 2000 SM 2 in Mitchell County, there were no non- white female deaths Data Source: NC State Center for Health Statistics Figure 2.6 0.02 to 843.1 843.1 to 896.8 896.8 to 944.8 944.8 to 1004.3 1004.3 to 2107.8 ( NWM) Males Females ECU, Center for Health Services Research and Development, 2007 1 Age- Adjusted Rates Standardized to US 2000 SM Figure 2.7 North Carolina: Comparisons among Regions2, 1979 to 2004 Trend Descriptions Age- Adjusted1 Mortality Rate Trends from All Causes of Death ENC 41 16% decrease R2 = 0.92 Y = - 7.04x + 1146 RNC 59 16% decrease R2 = 0.92 Y = - 6.34x + 1023 NC 16% decrease R2 = 0.93 Y = - 6.57x + 1059 US 17% decrease R2 = 0.96 Y = - 7.75x + 1034 800 850 900 950 1000 1050 1100 1150 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 00 01 02 03 04 05 06 07 08 09 10 Age- adjusted mortality rate per 100,000 population ENC 41 RNC 59 NC US Years 2 NC, ENC 41, and RNC 59 1979- 2004 mortality data from NC SCHS & US 1979- 2002 mortality data from NCHS’s Compressed Mortality File 600 700 800 900 1000 1100 1200 1300 1400 1500 1600 1700 1800 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 00 01 02 03 04 05 06 07 08 09 10 Age- adjusted mortality rate per 100,000 population NWM WM NWF WF ECU, Center for Health Services Research and Development, 2007 1 Age- Adjusted Rates Standardized to US 2000 SM Trend Descriptions WM 28% decrease R2 = 0.97 y = - 16.35x + 1497 WF 7% decrease R2 = 0.45 y = - 2.24x + 813 NWM 16% decrease R2 = 0.53 y = - 10.80x + 1751 NWF ------ R2 = 0.09 Y = - 1.16x + 943 Years Figure 2.8 Eastern North Carolina: Comparisons among Race- Sex Groups2, 1979 to 2004 Age- Adjusted1 Mortality Rate Trends from All Causes of Death 2 ENC 41mortality data from NC SCHS 567.6 to 630.9 630.9 to 707.3 707.3 to 788.1 788.1 to 911.7 911.7 to 1088.0 Age- Adjusted1 Years of Potential Life Lost before Age 752 Per 10,000 Population Natural Breaks Regional Variation of Years of Potential Life Lost in North Carolina ECU, Center for Health Services Research and Development, 2007 Figure 2.9 Premature Mortality in the United States 2002 with Selected Rankings Not Shown: AK: 36th DC: 52nd HI: 5th VT: 1st NH: 2nd MN: 3rd IA: 4th NC: 37th AR: 48th AL: 49th LA: 50th MS: 51st ENC 41: 47th PNC: 29th WNC: 34th VA: 22nd SC: 45th US ( 751.0) 2 ENC 41, PNC, WNC, and US 1979- 2002 mortality data from NCHS’s Compressed Mortality File 1 Age- Adjusted Rates Standardized to US 2000 SM DC ( 1323.0) Years of Life Lost Per 10,000 Population a. Crude b. Age- Adjusted1 Data Source: Odum Institute, UNC— Chapel Hill 541.0 to 806.7 806.7 to 878.2 878.2 to 958.4 958.4 to 1073.2 1073.2 to 1273.6 Years of Life Lost Per 10,000 Population 564.8 to 775.2 775.2 to 835.2 835.2 to 924.5 924.5 to 1036.6 1036.6 to 1234.7 ECU, Center for Health Services Research and Development, 2007 1 Five- Year Average, Age- Adjusted Rates Standardized to US 2000 SM Premature Mortality Rates from All Causes of Death: North Carolina and Eastern North Carolina Total Population, 2000- 2004 Figure 2.10 ECU, Center for Health Services Research and Development, 2007 Age- Adjusted1 Premature Mortality Rates from All Causes of Death: North Carolina and Eastern North Carolina Race- Sex Specific, 2000- 2004 Years of Life Lost Per 10,000 Population 1 Five- Year Average, Age- Adjusted Rates Standardized to US 2000 SM 2 in Mitchell County, there were no non- white female deaths Data Source: NC State Center for Health Statistics Figure 2.11 0.02 to 775.2 775.2 to 835.2 835.2 to 924.5 924.5 to 1036.6 1036.6 to 2174.3 ( NWM) a. White Males b. White Females c. Non- White d. Non- White Males Females ECU, Center for Health Services Research and Development, 2007 600 700 800 900 1000 1100 1200 1300 1400 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 00 01 02 03 04 05 06 07 08 09 10 Age- adjusted years of life lost per 10,000 population < 75 years of age ENC 41 RNC 59 NC US Trend Descriptions ENC 41 27% decrease R2 = 0.94 Y = - 12.97x + 1265 RNC 59 30% decrease R2 = 0.94 Y = - 12.43x + 1083 NC 29% decrease R2 = 0.94 Y = - 12.72x + 1139 US 30% decrease R2 = 0.96 Y = - 13.01x + 1053 Years 1 Age- Adjusted Rates Standardized to US 2000 SM Figure 2.12 North Carolina: Comparisons among Regions2, 1979 to 2004 Age- Adjusted1 Mortality Rate Trends from All Causes of Death 2 NC, ENC 41, and RNC 59 1979- 2004 mortality data from NC SCHS & US 1979- 2002 mortality data from NCHS’s Compressed Mortality File ECU, Center for Health Services Research and Development, 2007 400 600 800 1000 1200 1400 1600 1800 2000 2200 2400 2600 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 00 01 02 03 04 05 06 07 08 09 10 Age- adjusted years of life lost per 10,000 population < 75 years of age NWM WM NWF WF WM 34% decrease R2 = 0.91 y = - 18.79x + 1424 WF 17% decrease R2 = 0.75 y = - 4.30x + 670 NWM 32% decrease R2 = 0.81 y = - 30.35x + 2445 NWF 18% decrease R2 = 0.65 Y = - 8.13x + 1168 Trend Descriptions Years 1 Age- Adjusted Rates Standardized to US 2000 SM Figure 2.13 Eastern North Carolina: Comparisons among Race- Sex Groups2, 1979 to 2004 Age- Adjusted1 Mortality Rate Trends from All Causes of Death 2 ENC 41mortality data from NC SCHS ECU, Center for Health Services Research and Development, 2007 Table 2.1 Mortality from All Causes: Eastern North Carolina, 2000- 2004 County Deaths Crude Adjusted Deaths Rate Deaths Rate Deaths Rate Deaths Rate Beaufort 2,692 1185.2 988.2 925 1159.1 975 770.0 382 1418.9 410 929.5 Bertie 1,292 1313.3 1111.2 249 1170.4 300 907.4 373 1541.3 370 899.4 Bladen 1,983 1216.3 1101.7 586 1238.8 614 875.0 390 1551.1 393 971.0 Brunswick 3,806 962.7 842.8 1,741 927.3 1,533 717.3 278 1289.0 254 789.0 Camden 323 856.8 844.1 131 966.5 126 763.4 39 1122.7 27 589.5 Carteret 3,452 1142.1 932.7 1,667 1108.4 1,575 785.2 102 1251.7 108 812.7 Chowan 925 1275.0 949.1 296 1058.6 308 719.4 169 1620.3 152 785.2 Columbus 3,169 1157.3 1069.9 1,063 1316.3 1,066 824.8 531 1488.5 509 927.7 Craven 4,275 931.7 938.0 1,534 975.0 1,578 769.0 567 1448.4 596 974.8 Cumberland 10,093 663.7 1004.3 3,140 1175.3 3,080 832.5 1,965 1274.8 1,908 879.0 Currituck 833 835.9 911.8 367 964.8 365 804.8 48 1453.3 53 985.3 Dare 1,317 822.6 859.7 683 964.3 587 733.9 23 1141.7 24 1045.9 Duplin 2,525 999.3 983.9 807 1095.1 844 768.3 440 1534.7 434 888.7 Edgecombe 3,045 1107.6 1125.6 707 1336.7 778 881.2 772 1563.1 788 924.7 Gates 614 1149.1 1061.1 185 1251.1 190 906.9 121 1331.2 118 890.2 Greene 897 918.4 941.9 290 1256.9 277 722.9 169 1334.1 161 737.8 Halifax 3,310 1167.0 1023.9 785 1184.1 891 769.2 830 1413.4 804 868.8 Harnett 3,807 787.8 944.8 1,448 1136.9 1,477 760.6 459 1375.2 423 815.3 Hertford 1,499 1330.6 1124.5 317 1404.6 336 828.0 432 1621.4 414 903.1 Hoke 1,256 691.0 1015.5 330 1144.7 282 765.5 323 1323.8 321 922.1 Hyde 340 1197.6 905.1 109 1171.5 123 771.9 58 1233.8 50 679.4 Johnston 4,985 752.5 917.7 2,117 1128.0 2,014 739.2 462 1344.1 392 759.1 Jones 587 1138.6 980.2 189 1198.3 171 738.8 98 1260.5 129 928.2 Lenoir 3,546 1201.9 1078.8 1,058 1275.8 1,084 834.1 702 1637.1 702 910.5 Martin 1,605 1275.2 1091.0 444 1388.0 498 881.0 299 1324.0 364 953.6 Nash 4,449 998.1 1003.6 1,414 1122.0 1,591 814.7 721 1516.9 723 900.4 New Hanover 7,084 850.9 832.7 2,732 917.5 2,907 672.9 646 1310.6 799 954.8 Northampton 1,432 1309.9 1007.0 297 1040.5 317 759.2 448 1653.7 370 779.8 Onslow 3,879 518.4 956.7 1,648 1171.1 1,442 789.0 381 1193.9 408 852.2 Pamlico 723 1124.7 815.1 263 931.9 280 678.8 89 1210.4 91 758.0 Pasquotank 1,820 1020.5 925.0 538 1072.7 615 735.0 322 1308.6 345 843.3 Pender 1,877 873.8 830.1 730 945.2 632 687.2 251 1214.7 264 753.9 Perquimans 743 1285.3 929.0 264 998.7 250 731.6 112 1543.4 117 893.9 Pitt 5,269 768.2 955.8 1,510 1058.1 1,700 740.4 1,005 1458.0 1,054 924.6 Robeson 5,904 943.9 1133.3 1,292 1326.2 1,258 876.7 1,674 1464.5 1,680 977.7 Sampson 3,158 1028.3 1013.0 1,073 1253.1 1,006 764.4 549 1410.8 530 895.4 Scotland 1,800 1000.8 1063.6 478 1237.1 551 859.8 384 1622.9 387 903.3 Tyrrell 221 1064.9 864.3 72 1016.5 72 742.0 45 1184.7 32 667.6 Washington 829 1226.9 1027.8 217 1074.5 249 794.5 174 1364.6 189 1000.6 Wayne 5,307 936.0 1046.6 1,686 1199.2 1,742 850.2 898 1394.9 981 974.6 Wilson 3,719 992.3 976.7 1,120 1114.1 1,239 769.2 674 1394.8 686 879.9 ENC 29 61,468 967.7 983.2 19,772 1109.8 20,503 783.8 10,493 1439.7 10,700 892.4 ENC 41 110,390 905.4 972.2 36,502 1100.2 36,923 773.5 18,405 1406.0 18,560 891.6 RNC 59 249,278 849.3 866.1 99,131 1011.2 105,502 703.2 22,467 1284.6 22,178 831.6 PNC 190,449 796.7 871.7 71,626 1009.0 77,309 706.3 20,889 1287.9 20,625 833.9 WNC 58,829 1080.4 854.1 27,505 1032.0 28,193 700.1 1,578 1252.7 1,553 810.3 NC 359,668 865.8 896.5 135,633 1034.0 142,425 720.4 40,872 1336.2 40,738 857.3 US, 2002 2,443,030 847.2 845.5 1,024,966 993.1 1,077,337 701.5 174,016 1118.2 166,711 773.1 White Males White Females 5- Year Race- Sex Specific Age- Adjusted Death Rates Rates 5- Year Totals Non- White Males Non- White Females Center for Health Services Research and Development East Carolina University Source NC Data: Odum Institute-- UNC, Chapel Hill US Data: NCHS ECU, Center for Health Services Research and Development, 2007 Table 2.2 Premature Mortality from All Causes: Years of Life Lost before Age 75 in Eastern North Carolina, 2000- 2004 County Deaths Crude Adjusted Deaths Rate Deaths Rate Deaths Rate Deaths Rate Beaufort 1,265 1119.7 1048.2 526 1181.0 294 595.3 260 1966.2 185 1102.0 Bertie 612 1219.9 1172.9 127 1142.3 94 639.8 242 1931.5 149 821.9 Bladen 996 1205.6 1143.7 334 1239.2 207 642.4 269 2008.4 186 996.9 Brunswick 2,002 930.0 859.5 1,052 1082.3 638 548.0 183 1580.4 129 780.8 Camden 163 801.8 755.6 75 826.5 46 529.4 26 1758.3 16 541.0 Carteret 1,594 1003.2 931.1 880 1184.7 596 673.6 68 1264.7 50 778.6 Chowan 387 977.5 924.5 137 906.1 89 527.9 102 1903.1 59 760.0 Columbus 1,655 1219.4 1166.5 648 1360.1 390 681.4 371 1997.8 246 1020.6 Craven 1,992 843.9 833.5 818 893.9 535 510.8 352 1457.3 287 961.5 Cumberland 5,971 881.3 935.5 2,005 989.4 1,383 627.3 1,471 1380.1 1,112 857.2 Currituck 425 843.6 816.5 217 924.8 165 660.7 28 1397.2 15 654.7 Dare 677 889.2 858.1 409 1098.2 236 588.6 19 1679.8 13 702.8 Duplin 1,236 1024.0 1004.9 469 1043.4 278 576.4 301 2133.6 188 912.7 Edgecombe 1,583 1165.4 1141.8 414 1148.4 263 628.2 521 1866.6 385 893.0 Gates 257 958.4 936.9 91 1000.1 59 745.4 62 1433.7 45 662.7 Greene 458 1021.5 1006.2 170 1131.5 96 603.0 112 1331.4 80 1081.8 Halifax 1,618 1181.7 1155.6 401 1226.5 278 601.8 574 1759.3 365 980.8 Harnett 1,978 844.2 878.7 874 1025.7 560 581.9 331 1464.1 213 821.1 Hertford 691 1241.5 1203.7 153 1350.5 92 535.7 267 1865.1 179 932.6 Hoke 774 1004.0 1069.0 229 1174.3 122 668.9 242 1487.8 181 929.0 Hyde 145 931.0 884.6 57 1015.5 38 614.4 27 1109.8 23 949.6 Johnston 2,620 830.4 833.7 1,298 1003.7 781 525.0 337 1609.1 204 813.7 Jones 276 1012.2 964.2 114 1269.2 53 494.5 58 1330.3 51 977.3 Lenoir 1,779 1273.6 1225.9 595 1307.7 385 758.5 477 2047.6 322 1149.2 Martin 775 1203.4 1143.9 239 1243.3 156 650.7 204 1777.9 176 1120.2 Nash 2,113 970.7 949.9 774 1028.4 507 570.9 484 1572.3 348 961.7 New Hanover 3,169 733.7 705.6 1,423 765.3 947 412.4 432 1614.4 367 938.4 Northampton 679 1266.4 1209.6 153 1138.9 107 803.4 259 1851.0 160 948.7 Onslow 2,315 719.4 816.9 1,091 898.0 693 629.1 283 1156.1 248 801.8 Pamlico 321 887.0 826.6 145 924.9 86 503.4 47 1155.2 43 1165.9 Pasquotank 784 828.0 821.2 251 822.3 184 508.3 189 1289.6 160 862.8 Pender 968 901.3 856.0 429 991.9 259 567.6 161 1557.2 119 709.1 Perquimans 333 1110.1 1055.5 135 1052.2 86 783.7 69 1530.9 43 1517.3 Pitt 2,671 857.1 913.4 844 843.9 586 541.2 711 1692.7 530 1008.8 Robeson 3,290 1219.0 1234.7 820 1563.0 444 728.3 1,184 1649.3 842 962.8 Sampson 1,553 1096.3 1081.3 616 1243.2 342 665.5 347 1725.6 248 1017.3 Scotland 944 1087.9 1077.7 281 1119.3 207 687.2 257 1715.1 199 964.3 Tyrrell 101 1041.3 1007.1 36 909.8 26 765.7 25 1564.1 14 1100.5 Washington 376 1079.7 1036.6 114 1137.4 67 480.6 113 1592.5 82 1072.1 Wayne 2,752 1010.0 1000.0 1,001 1023.5 617 628.7 636 1628.7 498 1061.7 Wilson 1,776 1008.9 983.6 608 1019.7 394 549.3 446 1716.2 328 932.0 ENC 29 30,154 976.4 964.5 11,044 1002.0 7,106 598.1 6,962 1648.4 5,042 965.0 ENC 41 56,074 957.8 951.6 21,053 1014.9 13,386 584.2 12,547 1608.2 9,088 936.7 RNC 59 113,504 794.0 780.8 52,209 885.4 34,378 513.2 15,669 1420.1 11,248 827.1 PNC 88,764 777.5 773.9 38,159 848.3 25,450 499.6 14,602 1420.4 10,553 825.5 WNC 24,740 868.2 814.9 14,050 1026.1 8,928 565.6 1,067 1419.0 695 863.0 NC 169,578 842.2 830.6 73,262 918.9 47,764 531.1 28,216 1495.5 20,336 870.7 US, 2002 1,054,300 755.3 751.0 509,168 878.6 337,327 511.3 120,404 1276.0 87,401 761.1 White Males White Females 5- Year Race- Sex Specific Age- Adjusted Death Rates Rates 5- Year Totals Non- White Males Non- White Females Center for Health Services Research and Development East Carolina University Source NC Data: Odum Institute-- UNC, Chapel Hill US Data: NCHS Pitt Wake Hyde Duplin Bladen Bertie Pender Wilkes Moore Onslow Union Surry Ashe Beaufort Craven Halifax Robeson Nash Sampson Iredell Columbus Swain Carteret Burke Brunswick Johnston Anson Guilford Randolph Harnett Wayne Jones Chatham Macon Rowan Hoke Martin Tyrrell Dare Lee Stokes Stanly Lenoir Franklin Buncombe Warren Granville Davidson Jackson Haywood Gates Person Caldwell Wilson Forsyth Polk Caswell Cumberland Orange Pamlico Rutherford Madison Yadkin Gaston Clay Cherokee Richmond Cleveland Catawba Davie Rockingham McDowell Hertford Alamance Vance Avery Yancey Mecklenburg Northampton Edgecombe Montgomery Durham Graham Scotland Greene Watauga Henderson Washington Transylvania Mitchell Alleghany Currituck Camden Chowan Perquimans Pasquotank New Hanover Lincoln Cabarrus Alexander Western ( WNC) Piedmont ( PNC) Remaining 59- County Region ( RNC 59) Eastern North Carolina 29- County Sub- region ( ENC 29) Eastern North Carolina 12- County Sub- region Eastern North Carolina 41- County Region ( ENC 41) North Carolina County and Regional Locations Center for Health Services Research and Development East Carolina University Greenville, NC ECU, Center for Health Services Research and Development, 2007 Appendix A ECU, Center for Health Services Research and Development, 2007 CARDIOVASCULAR DISEASE MORTALITY The biggest cause of death in both the United States and North Carolina continues to be from diseases of the circulatory system, commonly referred to collectively as cardiovascular disease. Cardiovascular disease ( CVD) includes high blood pressure ( hypertension), coronary heart disease, congestive heart failure, atherosclerosis, and stroke, conditions which often occur in combination. An estimate for the year 2004 indicates that 79 million adult Americans, about 1 of every 3, have one or more types of CVD and mortality from CVD comprises a little more than 36% of the 2.4 million deaths that occurred in the United States ( Writing Group Members et al., 2006). In 2004, CVD in North Carolina accounts for almost 34% of the 72,000 resident deaths that year and in Eastern North Carolina more than 35% of its 22,000 deaths have been attributable to CVD. The impact and burden of CVD is so great that if all its forms were to be eliminated, life expectancy in the United States would rise by almost 7 years. For Americans born today, there is nearly a 50- 50 chance that their eventual death will be due to CVD ( Anderson, 1999). In the present chapter, CVD mortality includes deaths due to heart disease ( HD), coronary heart disease ( CHD), and stroke, in addition to several other less prominent causes of the circulatory system. 1 The largest CVD mortality component is heart disease, which includes rheumatic heart disease, irregular heart rhythms, and diseases of the linings, valves, and vessels of the heart. The latter- most group generally pertains to blockages and constriction of the vessels that supply the heart and can lead to diseases like infarction and ischemia. Mortality from this group is a significant part of HD mortality and is considered separately as CHD. Stroke mortality is a distinct category within CVD that includes intracranial blockages ( resulting in infarctions) and hemorrhages, and other cerebrovascular diseases. Figure 3.1 summarizes the relationships of the TCVD mortality categories for the 41 counties of ENC during the period 2000 to 2004. For this 5- year period, heart disease and stroke comprise nearly 92% of all mortality attributed to TCVD, while CHD alone contributes slightly more than half of all CVD deaths. The less prominent CVD mortality category ( All Other) is not considered in this chapter. A complete listing of ICD10 codes organized by the categories used here can be found in the appendix for this section. 1 ICD9 Codes 390- 459; ICD10 Codes I00- I99 Cardiovascular Disease Mortality 1 ECU, Center for Health Services Research and Development, 2007 CVD mortality and its three major component diseases discussed in this chapter can be accessed below. CARDIOVASCULAR DISEASE MORTALITY Spatial Distribution of Cardiovascular Disease Mortality Temporal Distribution of Cardiovascular Disease Mortality HEART DISEASE MORTALITY Spatial Distribution of Heart Disease Mortality Temporal Distribution of Heart Disease Mortality CORONARY HEART DISEASE MORTALITY Progress towards Coronary Heart Disease Mortality Reduction Spatial Distribution of Coronary Heart Disease Mortality Temporal Distribution of Coronary Heart Disease Mortality STROKE MORTALITY Progress towards Stroke Mortality Reduction Spatial Distribution of Stroke Mortality Temporal Distribution of Stroke Mortality SUMMARY References As can be seen from the chart Six Leading Causes of Mortality in the US 1900 to 2001 ( figure 1.2), heart disease has emerged as the nation’s leading cause of death in the 1920s and continues to be the leading cause into the early 21st century. The chart also shows how the decline of infectious and communicable diseases in the first several decades of the twentieth century paved the way for this emergence. If both stroke mortality and HD mortality rates depicted in figure 1.2 were combined, then the combined rate would account for the largest share of general mortality since the turn of the 20th century ( with the exception of the influenza pandemic of 1918). The diminishing effect of infectious and communicable diseases on the mortality experience of the first half of the 20th century in the United States has given way to the rising prominence of death from heart disease in the latter half. The Epidemiologic Transition ( Omran, 1977) discussed in chapter one ( introduction) describes the secular decline of infectious/ communicable diseases and the concomitant rise of chronic disease mortality and its demographic consequences. The increase seen in HD mortality is more than likely the result of the rise in the proportion of people surviving the onslaughts of communicable diseases. Communicable diseases have their impact on both ends of the age Cardiovascular Disease Mortality 2 ECU, Center for Health Services Research and Development, 2007 spectrum. Over time, survivors of childhood diseases swell older age groups which have increasing susceptibility to HD and other cardiovascular problems. This pattern is repeated wherever infectious/ communicable diseases are brought under control with various public health measures and interventions. However, the demographic responses and outcomes can vary geographically and culturally. It is interesting to note that the states with the lowest rates, Minnesota, Alaska, and New Mexico are quite different in regard to their demographic attributes; investigation of the role of culture is suggested. The US Department of Health and Human Service’s document, Healthy People 2010 ( U. S. Department of Health and Human Services, 2000) provides target rates for the two major mortality categories of CVD: coronary heart disease, and stroke. Objective maps are included in this chapter for these two causes of death. Time series charts ( 1979 to 2004) are also included for each CVD mortality category ( including total CVD). For the coronary heart disease and stroke mortality time series charts, the HP 2010 targets are indicated. Spatial Distribution of Cardiovascular Disease Mortality The 2002 age- adjusted mortality rate for CVD ( ICD- 10: I00- I99) for the United States is 319 deaths per 100,000 population but there is remarkable geographic variation across the nation. State rankings2 ( including the District of Columbia) place Minnesota, Alaska, and New Mexico first, second, and third with the lowest respective age- adjusted rates per 100,000 of 237.7, 242.1, and 255.9 per 100,000, respectively. The highest rates are found for Tennessee, Oklahoma, and Mississippi, ( rates of 380.8, 398.8, and 420.7, which placed 50th, 51st, and 52nd respectively). The rate for North Carolina in 2002 was 327.0, ranking it 33rd in the nation. The 2002 average age- adjusted rate for the 41- county region within Eastern North Carolina ( ENC) is 366.6. If this region were treated as a state, it would rank 45th. For the 5- year period 2000- 2004, seven counties in ENC ranked worse than the state of Mississippi in 2002.3 The maps at the top of figure 3.2 shows the spatial distribution of CVD crude mortality rates for the 100 counties of North Carolina and the 41- county ENC region. CVD crude mortality has its greatest impact in the northeastern part of the state in those counties that comprise the 29- county hospital service area and sub- region. ( For county locations and names, see appendix A.) From Table 3.1, three counties-- Chowan, Perquimans, and Washington— have 5- year ( 2000- 2004) crude rates above 500 per 100,000. This translates to an average of 5 CVD deaths per 1,000 people per year living in those counties. Many counties 2 These rankings are based on calculations made at East Carolina University’s Center for Health Services Research and Development. The data for combined state and regional comparisons are from the National Center for Health Statistics Compressed Mortality Files ( 1999- 2002). 3 Calculations for county comparisons use primary data from North Carolina’s State Center for Health Statistics via University of North Carolina— Chapel Hill’s Odum Institute. Cardiovascular Disease Mortality 3 ECU, Center for Health Services Research and Development, 2007 with relatively high observed crude rates also have relatively small numbers of people and may be proportionally older, which naturally leads to their increased susceptibility to more chronic conditions like CVD. Crude mortality rates are a kind of density measure— the number of deaths normalized ( or divided by) the population of interest and do not account for age structure. Their depiction on maps is for the purpose of focusing the reader to areas where the mortality burden is greatest ( see chapter 1 for more discussion). Maps of crude rates are useful in the development of policy, intervention measures, and determining the allocation of health care resources. The age- adjusted mortality rate maps found at the bottom of figure 3.2 permit comparisons among counties and population groups which may have different age structures ( see chapter 1). The state map shows a sharper distinction in the disparity of county age- adjusted rates between the state’s eastern 41 counties and the remaining counties to the west. As regions, 41- county ENC’s age-adjusted rate of 367.3 is 19% greater that the 59- county region of NC at 308.0 deaths per 100,000 ( see table 3.1). In 2002, the age- adjusted rate for the US was 319.0, less than 2% of the 2000- 2004 rate for the state and less than 13% of ENC’s rate. From another perspective, if ENC 41 had the same mortality rate as RNC 59 during the years 2000 to 2004, 6,590 lives would have been spared from death due to CVD. Figure 3.3 shows age- adjusted mortality by race and sex using the same rate classification cut points found in the age- adjusted map in figure 3.2. These maps provide a visual sense of group contributions to the overall CVD mortality rate and distribution. For white males, the heaviest concentration of high rate counties is found in the east, while some metropolitan counties to the west and a chain of mountain counties tend towards lower rates. Within ENC, the county with the highest rate for white males is Hertford at 576.3 and 131 observed deaths ( table 3.1). High rates are ubiquitous throughout the state for non- white males with the highest found in the ENC county of Currituck at 644.3 and 21 deaths. The highest rates for white females are found scattered throughout ENC with Washington County having the highest rate in this region at 374.6 ( 124 deaths). Currituck County also had the highest rate for non- white females at 528.2 ( 29 deaths). Statewide, ENC is home to the largest concentrations of high rate counties for these four demographic groups. For males of both races there appears to be little difference between ENC and the rest of the state. ENC becomes distinct as a high rate region because of the influence of regional white and non- white female rates. Temporal Distribution of Cardiovascular Disease Mortality The decline in CVD is hinted at in figure 1.2 using the large proportional effects ( 72.1%) of HD mortality as a surrogate. This figure depicts the secular trend in heart disease ( HD) mortality reaching its peak in the 1960s and soon after, crude stroke mortality rates begin to decline. ( Together, these two diseases currently Cardiovascular Disease Mortality 4 ECU, Center for Health Services Research and Development, 2007 comprise more than 90% [ see figure 3.1] of CVD mortality and so gives a good approximation of the patterns of burden and progress made with respect to this disease.) Figure 3.4 is a closer, comparative look at how ENC has been faring over time with respect to CVD mortality over the last two decades of the 20th century and the early years of the 21st. It charts the continuing decline in age-adjusted CVD mortality rates for ENC, the remaining 59 counties of North Carolina ( RNC), North Carolina, and the United States, from 1979 to 2004 ( US: 1979 to 2002). Within the 26- year period, ENC’s annual rates are the highest, followed by the state, the nation, and the remaining 59- county region, each showing very similar patterns of decline. ( The state values are a weighted average between ENC and RNC and will always have intermediate values.) The negative coefficients found in the equations of the lines, listed in the chart ( figure 3.4), show that ENC’s rate of decline is slightly greater than RNC’s rate with the relative gap between the regions’ fitted rates growing from 9% in 1979 to 13% in 2004. This represents a relative 44% increase in regional disparity for CVD mortality. In absolute terms, these same line equations show that the expected or fitted rate differences in age- adjusted death rates declined from 51 deaths per 100,000 in 1979 to 41 deaths per 100,000, which translates into a 24% decrease in regional disparity. Figure 3.5 depicts the 26- year trend of CVD mortality among the four major demographic groups in ENC. It is immediately apparent that the age- adjusted rates are declining for all groups. ENC white males show the greatest decreasing trend-- a decrease of 52%, which on average saves 16.7 lives per annum. This compares favorably to the 42% decline for white females; a saving of 8 lives per year. With R2 values around 0.90 one can make projection into the not- too- distant future with a fair amount of confidence. If the same trends continue, the age- adjusted CVD rates for white males and white females will converge around the year 2015 with an age- adjusted rate of approximately 184 per 100,000. The age- adjusted rates for both non- white men and non- white women are also converging but with their age- adjusted rate trends not projected to converge until sometime around the year 2030, when both non- white sexes attain the rate of approximately 188. In this scenario, it takes non- whites almost 15 years longer to achieve a projected rate similar to that of whites. Recall that the calculations are based on simplifying assumptions concerning the behavior of rates over time and any projections will have an increasing range of error as they move more distant in time from the last observed rate year. However such exercises can be viewed as another way of describing disparities and the amount of relative effort that would be required to achieve parity measured over time. Although mortality due to CVD is declining, its greatest impact is on the county populations of ENC. White males appear to do better in the large metropolitan counties of the Piedmont. However, these lower rates are comparable to the highest rates found in white female population. The highest rates for this latter group are concentrated in the counties of ENC. High rates of mortality for non-white males are nearly ubiquitous within the state, with low rates interspersed in Cardiovascular Disease Mortality 5 ECU, Center for Health Services Research and Development, 2007 the mountain counties. ( Low rates here are probably due to the small numbers of non- whites in this region.) For non- white females, high rates are concentrated in ENC, as well as the south- central portion of the state. Trend analysis covering the period 1979 to 2004 show a dramatic 45% decrease in regional rates for CVD mortality ( figure 3.4). The decrease in the age-adjusted rate for ENC roughly parallels the declining rates for the other regions, but there is a relative increase in regional disparity during this time— an artifact that results from using decreasing bases. When the CVD time series trend line for ENC is broken down into four race- sex trend lines, two patterns emerge: divergence in mortality rates between the two racial groups and convergence between the sexes for each racial group. HEART DISEASE MORTALITY Proportionally, heart disease ( HD) comprises more than 70% of all TCVD deaths for the period 2000 to 2004 ( see figure 3.1). The spatial and temporal patterns of HD mortality, therefore, should correlate strongly to those patterns observed for CVD. Any observable differences in these patterns will probably be due to the effects of stroke mortality, the next largest category outside of HD accounting for almost 20% of all CVD mortality. The ICD- 10 definitions for HD can be found at the end of this section in appendix B. Spatial Distribution of Heart Disease Mortality A comparison of the crude and age- adjusted maps for HD ( figure 3.6) and CVD ( figure 3.2) mortality does show strong similarities in patterns of mortality. ( Note that the cut- points of HD mortality rate categories in the legends for both crude and age- adjusted maps are approximately 70% of the ranges observed for CVD mortality.) The crude map of HD mortality shows concentrations of higher rates in the extreme northeastern and western portions of the state, with smaller concentrations in the southeast and south. Age- adjustment produces a larger concentration of high rates in ENC, de- emphasizing HD mortality rates in the western region of the state. Comparisons of regional age- adjusted HD mortality rates illustrate the continuing presence of geographic disparities. From table 3.2, ENC’s 2000- 2004 age-adjusted rate ( 263.5) is 13% higher than the US rate ( 240.8) and 19% greater than the rate for RNC ( 221.9). The coastal counties of Dare and Pamlico possess the lowest rates at 187.9 ( 286 deaths) and 190.4 ( 174 deaths), respectively. ( For county locations and names, see appendix A) These counties compare favorably to RNC’s rate for the same period. Moving inland, the highest age- adjusted HD rates are found in two county clusters. The first cluster is found in the southern part of the 41- county ENC region. Here, the counties of Bladen ( 319.2), Columbus ( 347.5), Robeson ( 315.6), and Scotland ( 310.4) experience Cardiovascular Disease Mortality 6 ECU, Center for Health Services Research and Development, 2007 12.7% of ENC’s mortality attributable to HD while 10.1% of the region’s aggregated estimated population from 2000 to 2004 lives in those counties. The proportional disparity grows when we move to the next high rate cluster of counties found in the northern part of the region. The high rates for Beaufort ( 309.1), Edgecombe ( 305.5), Martin ( 311.0), and Washington ( 314.8) counties comprise 8.2% of the region’s HD deaths, but comprise only 5.7% of the region’s population. Given their respective populations sizes, these two county clusters have a disproportionate share of ENC’s HD mortality. 4 Figure 3.7 depicts the spatial distribution of age- adjusted mortality rates for HD ( 2000- 2004) broken down into four race- sex groups. The observed spatial patterns closely resemble those for CVD ( figure 3.3) and indicate similar regional effects among the four groups: higher rates for females of both racial groups are again more concentrated in the eastern portion of the state, while high white male rates are found throughout the state with the exception of the Piedmont’s metropolitan counties, and non- white male rates are ubiquitously high with the exception of several counties in the west. From table 3.2, the highest regional age- adjusted county rate for white males is Columbus at 424.1 with 335 dying from HD over five years. For the same period, Washington County is the deadliest for white females who experience 96 HD deaths and an age- adjusted rate of 294.0 per 100,000. Non- white males experience their highest rate of age-adjusted HD mortality in Currituck County at 480.7 per 100,000 but this is the result of only 16 individuals dying during that period— Perquimans County has the next highest rate at 441.1 and a more statistically stable death count of 32. In Columbus County, 189 non- white females died from HD producing the highest age- adjusted county rate of 339.1 during the years 2000 to 2004. The total age-adjusted HD mortality rate Columbus County is weighted largely by deaths contributed from females of both racial groups, although white males also make a significant contribution. The high CVD rate experienced by non- white males in Edgecombe County appears to be heavily influenced by the HD component for this race- sex group. Within ENC, the lowest statistically reliable age- adjusted rate for any race- sex group is that found for white females in Greene County at 165.0. Temporal Distribution of Heart Disease Mortality Figure 3.8 shows trend lines for age- adjusted HD mortality among the four regions for the period 1979 to 2004. The slope of the lines all follow the same pattern of decline observed in figure 3.4 for CVD. Closer observation shows, however, that with the exception of the ENC trend line, the relative positions of the other three regions have shifted slightly. For CVD ( figure 3.4), North Carolina has been consistently above the US rate, but for HD the state emerges 4 Because age- adjusted rates can be used for making comparisons, they can be helpful in targeting areas where problems might exist. In this case, two county clusters have been identified and their count data are used to create proportions, which can be used to calculate the relative amount of mortality burden. Cardiovascular Disease Mortality 7 ECU, Center for Health Services Research and Development, 2007 with rates slightly less than the nation. ( This is probably due to the impact of stroke mortality in ENC, which tends to be higher and has a significant additive effect to the state rate for CVD.) Rates for RNC have been consistently below the declining trend for the US, whereas for CVD the trend lines closely matched one another. The impact of HD mortality on RNC’s population is less than it is for the nation as a whole. ENC’s age- adjusted mortality rates for HD are clearly higher throughout the 26- year time series with a slightly greater rate of decrease among all the regions. The pattern of HD mortality decline witnessed here is a good example of the secular trend in HD mortality burden observed during the 20th century ( see figure 1.2). Both observed and modeled trend lines for race- sex groups ( figure 3.9) show patterns of decline similar to CVD ( figure 3.5). What emerges in the pr
Object Description
Description
Title | Eastern North Carolina health care atlas a resource for healthier communities. |
Other Title | ENC health care atlas; Health care atlas; Resource for healthier communities; Eastern North Carolina atlas of mortality |
Date | 2007 |
Description | 2006 |
Digital Characteristics-A | 12.5 KB; 147 p. |
Digital Format | application/pdf |
Pres Local File Path-M | \Preservation_content\StatePubs\pubs_borndigital\images_master\ |
Full Text | Introduction to the 2006 Eastern North Carolina Atlas of Mortality A tombstone in an eastern North Carolina church cemetery is inscribed: In Memory of James Bonner Foreman Who was born The 1st of December 1785 And died The 22nd of December 1807 Aged 22 Years and 21 Days Come view my Tomb as you pass by, As you are now so once was I; As I am now so must you be, Therefore prepare to follow me. Death is a personal event that we will all eventually experience. It is also something fundamentally empirical, recordable, and therefore measurable. The tradition and culture of recordkeeping varies throughout the world and in the west some countries have been compiling data on peoples’ lives for centuries either for ecclesiastical or secular purposes. One extremely important secular purpose is the amassing of individual records over time and place into part of North Carolina’s vital statistics collection. Eventually, every North Carolina resident shows up in the vital statistics registry “ book” as a single data record, an abstraction, of a life once lived. Unlike Mr. Foreman’s epigraph two centuries ago, more data and information pertaining to the circumstances of the mortal event are recorded. In addition to date of birth and death ( i. e., age at death), these include the decedent’s location at death, cause of death, race, sex, and residence. The data recording the circumstances surrounding people’s deaths can be formed into a picture about the conditions of living in their period of time and their society when aggregated at various scales and dimension. The atlas format is an appropriate means of display and description of vital events such as mortality. The present chapter is an introduction to the approach and concepts used in the current edition of the Eastern North Carolina Atlas of Mortality. Specifically addressed topics can be found using the following linked headings. Overview of the Atlas Portraying Geographic Data Data Sources Mapping with GIS Software Maps in the Atlas Time Series Charts in the Atlas Overview of the Atlas The 2007 edition of the Eastern North Carolina Atlas of Mortality is a narrated collection of such statistical pictures that describe the spatial and temporal facets— the descriptive geography-- of death in the eastern- most 41 counties of North Carolina ( ENC). Over the last three decades, this region has seen thousands of individuals dying in excess of what would be expected or experienced in other parts of the country. The underlying motivation for this work is to bring this ongoing tragedy to light and to show health professionals and policy makers where and on what problems need their attention. The information presented in this atlas will allow the reader to form a coherent image in his or her mind of the history and future of mortality in Eastern North Carolina. It is hoped that these statistical images will lead to not only an increased awareness of the conditions of life-- and death-- in ENC but that it will also stimulate thinking about hypotheses, research questions, policy, and strategies for making life better in our region. In this work, the geographical distributions of mortality from leading causes are aggregated and portrayed for the years 2000 to 2004 ( 5 years) and chronicled over a 26- year time series beginning in the year 1979 an ending in 2004. From 2004, rate projections ( linear best fit lines) are included. Figure 1.1 portrays the 100 counties of North Carolina and delineates the major regions used in this Atlas. The regional focus is the eastern- most 41 counties whose western boundary is approximated by I- 95 and extends to the coastline. ENC 41 also corresponds to the physiographic province of the Coastal Plain. The 41- county region is further divided into two sub- regions: ENC 29, comprised of the northeastern- most 29 counties of ENC 41, and a remaining southern 12- county region. ENC 29 corresponds spatially to the county service area of University Health Systems of Eastern Carolina. ENC 41 possesses North Carolina’s greatest levels of poverty and ethnic diversity, while population and economic growth lags behind the remaining western 59 counties. To contrast and compare mortality rates with the rest of the state, the remaining 59 counties are grouped into two regions corresponding to the Piedmont ( PNC) and the western mountain region ( WNC). Over the last 30 to 40 years, PNC and WNC have experienced rather different population and economic trajectories than the east and this is reflected in their more favorable mortality outcomes. The Atlas traces the spatial and temporal domains of ENC’s mortality experience with the use of maps, tables, and time series charts. These three components of the Atlas are built on measures that summarize the population’s mortality experience. Summary measures like mortality rates are calculated from several of the descriptive elements of the individual death record. The resulting rate calculations are then tabulated by county, region, and time period. In contrast to the simple table, maps are a 2- dimensional spatial ordering of mortality rates that describe a place’s mortality experience and burden. Time series charts portray the temporal order of mortality rates for regions, counties, and their constituent population groups. These charts show general parallel, convergent, or divergent 2 trends among regions and population groups. Relative and absolute mortality rate comparisons can be made from the maps, tables, and charts to determine progress toward the elimination of rate disparities and mortality burden over space and time. Portraying Geographic Data Maps are the most important feature of a geographical atlas. Along with other graphical means of communication, a wide range of topical literature has evolved that discuss the nature of maps and the geographic information and meaning that they portray from a variety of technical and philosophical both within and without the discipline of geography. A good discussion of the foregoing, which also includes Information Theory, can be found in Poore and Chrisman’s Order from Noise: Toward a Social Theory of Geographic Information ( Poore & Chrisman, 2006). The more salient and general points concerning maps and time series data found in this work are discussed below. For a more technical treatment of charts, with a strong emphasis on the proper construction of graphics that convey meaningful information from quantitative data, the reader is directed to the works of Tufte ( Tufte, 1995; Tufte, 1997; Tufte, 2001; Tufte, 2006). Pragmatically, different aspects of various techniques and perspectives necessarily come together in the development of any atlas and how they come together may distinguish one atlas’s approach from another. In this Atlas, our approach is one of description and chronicling in such a way that the reader can make meaningful geographical comparisons of the regional mortality experience. One functional definition of geography considers both space and time as referential systems. Borrowing terminology from Werlen ( Werlen, 1993), a space can be defined as a three dimensional container. This type of space orders events ( an occurrence or areas with given attributes like mortality rates) by measuring their positional relationships ( the x and y axes) and their sizes or magnitudes ( the z axis). Another dimension can be added that orders those events temporally and therefore, sequentially. The 2- dimensional or 3- dimensional static map can be stacked or sequenced along a temporal axis to form a time series of maps. As long ago as 1964, Berry ( Berry, 1964) described and operationalized a very similar concept as the geographical data matrix, where the matrix is the container of geographically referenced data— attributes/ characteristics ( or mortality rates) that are linked to places or areas. With some modifications, this prosaic and functional conceptualization describes how spatially referenced data are managed in modern Geographic Information Systems ( GISs). With a GIS, these data can be stacked or sequenced in temporal order very quickly to create a moving picture of a geographic process. Because of space constraints, only the most current 5- year maps of mortality rates are provided in this Atlas, but they are accompanied by charts that show temporal trends among regions and population groups. 3 Geographical referencing and the binding together of attribute data over points in time or sequence of time periods are a means to the comparative study of trends in mortality processes. In both spatial and temporal referential systems, there is a well- known tendency for objects within the system that are nearer to one another to be more alike than those more distant or, as stated in Waldo Tobler’s first law of geography, “… everything is related to everything else, but near things are more related than distant things.” ( Tobler, 1970) This notion of propinquity and similarity is important for understanding relationships among demographic, social, biological, and physical attributes of places. For example, a group of neighboring counties such as those found in Eastern North Carolina will tend to have similar age, race, and sex structures because they have had similar economic and demographic histories or, more generally, have experienced similar social relations and processes as well as live within similar spatial structures ( Gregory & Urry, 1985). Since age is the greatest risk factor for mortality we would also expect a group of neighboring counties that share a similar age structure to have similar mortality rates. In varying degrees, these same counties may also have similarities in other known risk factors such as certain occupations, race, housing, and poverty. Within the spatial analytical line of inquiry, this well known propensity in geography is extremely useful for constructing hypotheses, modeling, and theory testing. Maps can be thought of as models of real- world patterns and processes at a given point in time. They reduce reality to a set of graphical and geometric objects that have an a priori common meaning, which is necessary for interpretation and communication. This reality is not produced, reproduced, or experienced in exactly the same way by any two persons or reflected in individual death records but collectively similarities and patterns can emerge and be traced for population aggregates. A map as a representation allows a way for the user to apprehend a myriad of facts about places and order them both spatially and temporally into one coherent mental picture. Once geographic data have been integrated into a suitable level of coherency, assessment and analyses can begin with a certain set of well- grounded assumptions. These assumptions might include Tobler’s first law of geography ( the closer, the more similar) or considered in conjunction with certain risk factors such as age or diabetes with certain mortality outcomes. However, it should always be borne in the mind of the map user or analyst that these newly acquired understandings and cognitive models are ultimately based on a reduced reality— that is, in the time- worn phrase: the map is not the territory. Finally, maps can be used either as arguments to make a case for further study into the etiology of the causes of mortality and morbidity or they can be used as propositions ( or hypotheses) addressing potential causes of observed mortality and morbidity patterns ( Koch, 2005). To illustrate, given the range of social and structural inequalities that exist among certain demographic groups in the US and particularly in the South, the Atlas provides evidence for the argument that differences in the underlying social fabric will manifest themselves in the 4 observed patterns of mortality for Whites and Non- whites in eastern North Carolina and for all Eastern North Carolinians versus the rest of the state. The case can be made by employing maps, tables, and charts that permit comparisons among the race- sex groups at county, regional, and national scales. Maps of related demographic and socio- economic variables are either included or referenced in the Atlas as propositions about relationships underlying the observed mortality patterns. As a tool for integrating disparate data, either as argument or proposition, the Atlas can assist in developing research questions for topics on health disparities, health resources, and economic development. Representational data used in the construction of maps are of two distinct classes. The first data class is made up of a limited set of geometrical objects that are used to represent a large range of real- world features on a map. The most basic of these data is the geometric point that is located on a geometric plane. The point can represent an event, institution, or place, for example. On this same plane an additional point will define a line and a series of lines can represent features such as road networks, stream systems, or social relationships and connections. Three or more points will define a polygon and can represent real- world entities such as counties or urban areas. In some maps polyhedra or solids defined by four or more polygons can be constructed to represent specific types of features. These geometrical representations ( or features) have some measurable quality or attribute assigned to them, which provides the basis for making comparisons and discerning patterns. Points, lines, and polygons can be assigned an attribute, quality, or quantity that describes map features. This second class of data can be partitioned into three categories: nominal, ordinal, and interval/ ratio ( Earickson & Harlin, 1994). Nominal data refers to the binary presence/ absence of a quality or one or more types of a given feature, such as vegetation cover or soil. Ordinal data are ranked in ascending or descending order and can be used to describe a hierarchical system of, for example, health states or levels of care quality measured as poor, fair, good, or excellent. Finally, interval/ ratio scale ( or metric scale) data measure quantities like mortality rates, dentist to population ratios, or disease prevalence. For interval data the difference between any pair of values is always the same no matter where they are located along the metric scale. There is a small but important distinction when considering either interval or ratio data. Interval data can include values that are less than an arbitrarily defined zero, such as temperature or elevation. However, unlike elevation, one cannot speak of a temperature being twice as cold or hot as another. These data are strictly interval in nature. Ratio data are interval data that can be compared meaningfully. For example, one could make the statement that the mortality rate for female breast cancer in county A is 33% greater than the rate in county B. Interval data can be evaluated as “ twice as much,” “ half as great,” or as some percent or proportion of one value in relation to another. 5 Data Sources The predominant types of data employed in this Atlas are polygons bound or joined to interval/ ratio data attributes. Polygons are used to represent counties, which are the basic units of analysis and are the building blocks for larger multi-county regions. County- level polygon data ( i. e., boundary files) are obtainable from the geography page of the US Census website. These data are available in several formats and are ready for use with most GIS packages. Because boundary files have unique county identifiers, they are also ready to “ join” or link to attribute data. A wide variety of county- level attribute data are employed in this work. Demographic and socio- economic data can be obtained from the American FactFinder section of the US Census website and the NC State Data Center. In the Atlas, mortality rates by leading causes of death are calculated from two sources. The North Carolina source is located at the University of North Carolina’s Odum Institute, which provides the most up to date vital statistics for the state. Mortality data for the nation and other areas of the county are calculated from data found in the Compressed Mortality File ( CMF) series produced by the National Center for Health Statistics. These data tend to be 3 to 4 years behind the latest year for North Carolina. Mapping with GIS Software Today, nearly all data required for GIS and mapping exist in a digital form. Many printed tabular data sources, collected in more remote periods of time, have been archived either on paper or microforms. These data sources can be scanned or imaged into formats suitable for optical character recognition ( OCR) programs or other software tools that will transform the printed character or numeral into a digital rendition. Once obtained, the data need to be stored in some type of database. Storage can be in a large relational enterprise level database such as MS- SQL ® or Oracle ® with member tables distributed according to function anywhere on the globe or data storage can simply be in a spreadsheet “ database” residing on a desktop PC. In Microsoft’s Excel ® , one or more data ranges ( i. e., columns × rows) described in a worksheet can behave as individual database tables within a workbook. These data ranges and tables loosely correspond to Berry’s geographic data matrices. ( Berry, 1964) Using a small set of basic database functions in Excel, it is possible to link and match records ( table rows) in a way similar to what is done in a true relational database. In order to match records, there must be a field serving as an index. An index field contains rows of unique identifiers and is common to all tables that will be linked or joined. In this Atlas, we use either the unique county name within the state or the Federal Information Processing Standard ( FIPS) code that uniquely identifies any county among the more than 3,000 counties in the US. These same identifiers are used to match attribute data to county polygons prior to mapping in a GIS. 6 Map- making today is largely done using GIS software that integrates a wide variety of disparate data sources and data types. The construction of maps is actually one of many functions a modern GIS can perform. Other functions include spatial querying, spatial analyses, modeling, as well as layering and combining spatial objects and their attribute data to develop new data. For the purposes of descriptive spatial epidemiology and ultimately the comparisons that will be made, the Atlas here employs the primary and more basic functions of a GIS which manage geographically referenced data and quickly generate map layers with accompanying cartographic elements. Cartographic elements include the legend or map key derived from data and feature classification and symbology. Data in an atlas of mortality are typically rates and percentages ( interval/ ratio data). A GIS is able to partition and classify a data distribution with a choice of automated default methods ( e. g. quantile, equal interval, natural breaks, or statistical) or the user can classify the data manually. The choice of method is based on the purpose of the map ( e. g., statistical description, proposition, or argument) and the intended audience of map readers ( Wilson & Buescher, 2002). The GIS also provides color palettes for selecting a hue for each theme. A hue can be further divided into a series of graded shades with hue saturation corresponding logically to category ranges. Analysis proceeds by examining the resulting patterns of categorized rates represented as shades: do counties with more saturated shades tend to cluster together? Or are they more dispersed, demonstrating no real comprehensible pattern? Such basic analyses can yield ideas for the development of hypotheses or intervention strategies if something is known about the processes that created them. Different ideas about presentation of map data and experimentation with categories can proceed quickly with a GIS. What took several days to produce by hand as recently as twenty years ago today only takes several minutes. The maps in this Atlas were created in ESRI’s ArcGIS 9.1 and 9.2. Maps in the Atlas The Atlas is organized in a way that invites the assessment of patterns in both the spatial and temporal domains. Maps show the distribution of categorized county rates of mortality for the years 2000 to 2004. Mortality rates are, in effect, measures of density. They measure the density of events ( deaths by selected causes) in relation to the population producing those events. Both crude and age- adjusted rates are employed for making regional comparisons in those maps depicting total deaths by cause, while only age- adjusted rates are used for making county and regional comparison by race- sex groups. Crude rates are constructed by dividing the number of events ( or case mortality by cause) in a county by that county’s total population, and then multiplying the result by 100,000, which has the effect of reducing in a certain time period the number of decimal places and thereby making the rate more easily understood. A crude rate is the actual rate and is useful for measuring the burden of disease mortality 7 in an area and time period. However, making comparisons among counties with crude rates is problematic because the differences in their respective age-structures can confound interpretation. For example, knowing that increased age is the greatest risk factor for dying in a given time period, a county with a larger proportion of elderly ( e. g., retirees) will naturally produce a greater crude rate than a county where there are larger proportions of college- age students or individuals stationed on military bases. To make meaningful comparisons, a county’s age structure ( the numbers of people in each previously defined age group) must be adjusted. Essentially this adjustment is a re- weighting of a county’s population that produces an expected, as opposed to actual or observed, number of deaths for that population. The weights are based on an external or synthetic population structure known as the standard million population. Age- specific death rates based on the weights are calculated for each group in the age structure and then summed to produce an age- adjusted rate ( Buescher, 1998). An age- adjusted county rate is the rate a county would have if it had the same age structure as the external or standard million population and renders this county’s rate comparable to any other county using the same standard million population. It should be emphasized that age-adjusted rates used in making comparisons are not the actual observed rates but are the rates that would be expected if each county and region had the same age structure. The external population used in this work is the US Standard Million for the year 2000. Knowing which standard million population is used is extremely important when comparing rates calculated from mortality data from different states and time periods, otherwise the rates are simply not comparable. Time Series Charts in the Atlas Time series graphs for the years 1979 to 2004 provide a synoptic view of mortality trends for regions and race- sex groups. Age- adjusted rates are used to make comparisons among the 41 counties of ENC, the remaining 59 counties of the state ( RNC), North Carolina, and the US ( 1979 to 2002). Time series plots for four ENC race- sex groups ( male and female Whites and Nonwhites) are provided on an additional graph. Best- fit lines are incorporated into the time series plots for both regional and population charts so that the user can assess differences and trends. How well the trend and projection line fits the data is described by the coefficient of determination, R2. ( R2 is a statistic with values 0.0 to 1.0; the closer to 1.0, the better the fit.) For some leading causes of death there are Healthy People 2010 goals, which are age- adjusted target rates for the year 2010 ( U. S. Department of Health and Human Services, 2000). Where applicable, target values are included in the chart and can be used in conjunction with the projected trend lines. This permits the user to make comparisons among regions and population groups in terms of the amount of progress that is being made against a nationally recognized standard. 8 Over the course of many years, mortality rates will ebb and flow with small annual perturbations deviating from the general trend. A larger view over many decades may show gradually decreasing ( the ebb) or increasing ( the flow) trends for chronic diseases and intermittent spiking for epidemics during that period of time when communicable and infectious diseases were predominant causes of death ( see figure 1.2). Long term directional changes and pattern shifts in mortality rates are known as secular trends.∗ These trends are both responding and contributing to the underlying long term shifts in demographic, socio-economic, and environmental processes. One of the best examples is the nearly complete decline of mortality due to infectious diseases in the early part of the twentieth century. Infectious diseases tend to carry off larger proportions of susceptible young as well as those in the older age groups. Socio- economic and environmental processes such as improved access to better food and nutrition, improved sanitation, and generally better living conditions resulted in fewer deaths of the young as a result of contagion. In turn, a gradual shift in demographics occurred: more children survived into adulthood and into later life. This demographic shift— the result of more individuals now surviving into the older age groups-- is a major influence on the rise of the crude mortality rate from cardiovascular disease ( with the exception of stroke) in the early- mid twentieth century. These kinds of changes are described in Omran’s work on the epidemiologic transition ( Omran, 2005). The long term mortality trends resulting from different causes of death may not all be the same. Generally, mortality rates over the long term trace curvilinear patterns. As these patterns are examined more closely, parts of the curve begin to take on a more linear form. To simplify and give a general snapshot of recent trends, the mortality time series depicted in this Atlas models the data linearly. The benefit to this is that it provides easily understandable summary measures of mortality events occurring over three decades. However, the reader is cautioned to examine the general pattern of the entire series, giving more weight to events that have occurred later in the series than earlier. The maps, tables, and charts found in the Eastern North Carolina Atlas of Mortality form an armamentarium for understanding, integration, and synthesis of the region’s mortality burden and experience. Singly, an individual’s death, like the one found in an obscure corner of a church cemetery may appear to be a random event. However, when lone events like these are amassed into numerators and then rates, meaningful pictures about the conditions of life in a place can be created. In the end, it should always be kept in mind when gazing upon the abstract representation of the mortality map that ultimately it was the lives rather than the deaths of people that generated the observed patterns. ∗ The term secular as used here refers to a characteristic pattern for a given age or time period in population history. For example, until the First World War in the United States infectious and communicable diseases had a much more prominent role in observed mortality patterns than they do today. The last several decades of the twentieth century has seen a gradual decline for certain chronic diseases like those of the heart and some cancers. 9 The next chapter addresses general mortality. In this chapter, the leading causes of death for ENC 41 are delineated for the 5- year period, 2000 to 2004. Discussion of the spatial and temporal distributions of mortality from all causes ( i. e., general mortality) follows, including a more in- depth treatment of rates and measures in light of the observed data. Subsequent chapters address the 10 leading causes of death for the region and will generally follow the pattern of discussion found in the chapter on general mortality. References Berry, B. J. L. ( 1964). Approaches to regional analysis: A synthesis. Annals of the Association of American Geographers, 54( 1), 2- 11. Buescher, P. A. ( 1998). Age- adjusted death rates. Raleigh, North Carolina: North Carolina Center for Health Statistics. Earickson, R., & Harlin, J. M. ( 1994). Geographic measurement and quantitative analysis. New York : Macmillan ; Toronto; New York: Maxwell Macmillan Canada; Maxwell Macmillan International. Gregory, D., & Urry, J. ( 1985). Social relations and spatial structures. New York: St. Martin's Press. Koch, T. ( 2005). Cartographies of disease : Maps, mapping, and medicine ( 1st ed.). Redlands, Calif.: ESRI Press. OMRAN, A. R. ( 2005). The epidemiologic transition: A theory of the epidemiology of population change. Milbank Quarterly, 83( 4), 731- 757. Poore, B. S., & Chrisman, N. R. ( 2006). Order from noise: Toward a social theory of geographic information. Annals of the Association of American Geographers, 96( 3), 508- 523. Tobler, W. R. ( 1970). A computer movie simulating urban growth in the detroit region. Economic Geography, 46( Supplement: Proceedings. International Geographical Union. Commission on Quantitative Methods), 234- 240. Tufte, E. R. ( 2006). Beautiful evidence. Cheshire, Conn.: Graphics Press. Tufte, E. R. ( 2001). The visual display of quantitative information ( 2nd ed.). Cheshire, Conn.: Graphics Press. 10 11 Tufte, E. R. ( 1997). Visual explanations : Images and quantities, evidence and narrative. Cheshire, Conn.: Graphics Press. Tufte, E. R. ( 1995). Envisioning information ( 5th printing, August 1995 ed.). Cheshire, Conn.: Graphics Press. Werlen, B. ( 1993). Society action and space : An alternative human geography [ Gesellschaft, Handlung und Raum.] . London ; New York: Routledge. Wilson, J. L., & Buescher, P. A. ( 2002). Mapping mortality and morbidity rates. Raleigh, North Carolina: North Carolina Center for Health Statistics. Pitt Wake Hyde Duplin Bladen Bertie Pender Wilkes Moore Onslow Union Surry Ashe Beaufort Craven Halifax Robeson Nash Sampson Iredell Columbus Swain Carteret Burke Brunswick Johnston Anson Guilford Randolph Harnett Wayne Jones Chatham Macon Rowan Hoke Martin Tyrrell Dare Lee Stokes Stanly Lenoir Franklin Buncombe Warren Granville Davidson Jackson Haywood Gates Person Caldwell Wilson Forsyth Polk Caswell Cumberland Orange Pamlico Rutherford Madison Yadkin Gaston Clay Cherokee Richmond Cleveland Catawba Davie Rockingham McDowell Hertford Alamance Vance Avery Yancey Mecklenburg Northampton Edgecombe Montgomery Durham Graham Scotland Greene Watauga Henderson Washington Transylvania Mitchell Alleghany Currituck Camden Chowan Perquimans Pasquotank New Hanover Lincoln Cabarrus Alexander Western ( WNC) Piedmont ( PNC) Remaining 59- County Region ( RNC 59) Eastern North Carolina 29- County Sub- region ( ENC 29) Eastern North Carolina 12- County Sub- region Eastern North Carolina 41- County Region ( ENC 41) North Carolina County and Regional Locations Center for Health Services Research and Development East Carolina University Greenville, NC ECU, Center for Health Services Research and Development, 2007 Figure 1.1 Six Leading Causes of Mortality in the US 1900 to 2001 0 100 200 300 400 500 600 700 1900 1910 1920 1930 1940 1950 1960 1970 1980 1990 2000 Year Heart Disease Cancer ( All Types) Pneumonia & Influenza Tuberculosis ( All Forms) Diarrhea & Enteritis Three Infectious/ Communicable and Three Chronic Diseases Deaths per 100,000 Population* Sources: Leading Causes of Death, 1900- 1998 http:// www. cdc. gov/ datawh/ statab/ unpubd/ mortabs/ hist- tab. htm ( Last accessed Dec. 29, 2005) Data for 1999- 2001 from NCHS’s Compressed Mortality Files Stroke * Rates are not age- adjusted Center for Health Services Research and Development East Carolina University Greenville, NC ECU, Center for Health Services Research and Development, 2007 Figure 1.2 General Mortality in Eastern North Carolina 2000 to 2004 The chapter on general mortality is divided into several topics related to mortality from all causes for eastern North Carolina. They can be accessed directly with the following links. Introduction The Spatial Distribution of Crude and Age- Adjusted General Mortality Rates The Temporal Distribution of Age- Adjusted General Mortality Rates Mortality Burden The Spatial Distribution of Premature Mortality from All Causes The Temporal Distribution of Premature Mortality from All Causes From Empiricism to Explanation: General Mortality Disparities Conclusion Introduction General mortality includes all causes of death over a specified time interval. Causes of death are further defined and classified into internationally recognized series of grouped codes, such as the International Statistical Classification of Diseases and Disorders and Related Problems, 10th Revision or ICD- 10 ( World Health Organization, 2004). ( For the most recent revision of codes, see the electronic version at the World Health Organization’s website ( World Health Organization, 2006).) Periodically, revisions are made to incorporate changes in medical knowledge and to incorporate and facilitate improved coding rules ( see U. S. Department of Health and Human Services, Centers for Disease Control and Prevention, & National Center for Health Statistics, 2006). Standardized coding, in conjunction with using standard populations for age- adjustment, permits comparability of rates among different time periods and geographical units. Once the cause of death has been coded, each record is accumulated into a time and place- specific total number of deaths. The accumulated totals are then used to determine the relative ranking and importance of leading causes of death for a county or a region. Figure 2.1 portrays the resulting 5- year totals for the ten leading causes of death proportionally to the total number of deaths in ENC from 2000 to 2004. In this figure, two general classes of mortality causes dominate the mortality experience of ENC: Total Cardiovascular Disease ( TCVD) and Malignant Neoplasms ( All Cancers). From 2000 to 2004, more than 59% or 65,442 deaths have occurred due to these two disease categories. The remaining eight leading causes of death account for just 21.4% or 23,605 deaths during this same period. The number one leading cause of death in ENC for the study period is TCVD, which accounts for 37.0% ( 40,820) of the region’s 110,390 deaths. ( The TCVD category is based on the definitions proposed by the American Heart Association ( American Heart Association, 2005) and includes mortality due to stroke.) Death from malignant neoplasms is the second of the ten leading causes of death and 2 accounts for 22.3% ( 24,622) of all regional mortality. A distant third leading cause is attributed to Chronic Obstructive Pulmonary Disease and Chronic Lower Respiratory Disease COPD/ CLRD with 4.9% ( 5,384) of all deaths from this cause. Mortality from Diabetes Mellitus follows with 3.5% ( 3,904) of all ENC deaths. In fifth place, death from Unintentional Motor Vehicle Injuries ( UMVI), accounts for 2.8% ( 3,047) of the region’s deaths. Septicemia is the tenth ranking cause of death claiming 1.7% of all deaths. The ten leading causes of death are followed by a single category, All Other, which accounts for 19.3% ( 21,343) of General Mortality. Within this final category, 1,378 people have committed suicide, 1,193 people have died from chronic liver disease and cirrhosis, 1,104 people have been murdered, and 776 people have died from AIDS due to HIV ( Human Immuno- deficiency Virus). Regionally, deaths from specific causes in the All Other category make up very small percentages within general mortality. Nevertheless, when counties are examined separately, the seemingly insignificant causes of death at the regional scale can be important causes of death at the more local county level scale. It is therefore important to monitor at the “ basement” level so that emerging mortality trends at regional and local scales can be detected. The present chapter is organized around three general topics. The first two topics describe patterns of mortality from all causes, but using two different approaches in its portrayal. The first approach examines the spatial and temporal patterns of two density measures: crude and age- adjusted general mortality rates. These two measures describe mortality quantities in relation to population sizes and their distributions in space and time. However, density measures do not provide information about what part of a population is being affected. Mortality, whether from specific or general causes, can affect populations in a differential manner across spatial and temporal dimensions. Measuring the cumulative differences of age- at- death of individuals that occur before an accepted standard age- at- death ( say, 75 years) produces information about the level of premature mortality. Larger amounts of years of potential life lost in a population signify greater levels of mortality burden being shouldered by that population. The second topic covered in this chapter addresses the distributions of premature mortality in eastern North Carolina and the state. Finally, we move from empirical descriptions to a brief discussion of how patterns of general mortality can be explained by their relationships to other factors. The Spatial Distribution of Crude and Age- Adjusted General Mortality Rates A map of crude mortality rates will draw the map- reader’s attention to those areas that are experiencing the highest numbers of deaths relative to their local populations. The crude mortality rate measures the density of resident deaths occurring in an area in relation to the population of that area. It is a summary measure representing the proportion of a population that has died over a given interval of time. Because this proportion is frequently a very small value, it is multiplied by a larger number ( of persons) like 1,000, 10,000, 100,000, or even 3 1,000,000 for extremely rare causes of death. ( This atlas will employ the multiple of 100,000 persons when discussing and comparing density measures.) Because age is the greatest risk factor for dying, the map of crude mortality rates is also, to some degree, a map of the underlying spatial distribution of population age structures. Controlling for the effects of age variation will permit the map reader to make comparisons of mortality rates among different areas on the map. This is accomplished through the technique of age- adjustment, which adjusts the observed number of deaths to an expected number of deaths if the population under study had the same age structure as some external reference population ( Buescher, 1998). In this atlas, the US Standard Million for the year 2000 is employed ( Anderson & Rosenberg, 1998). It is extremely important that the standard population used in each case is the same when comparing age-adjusted maps from one period of time to another, or when comparing maps of age- adjusted rates from other states. Different model or standard populations will generate different age- adjusted rates even when the actual or observed distribution of deaths across the population age distribution remains constant. Figure 2.2 shows the mapped distribution of crude mortality rates from all causes for counties in the contiguous US from 2001 to 2003. The category classification is based on the extension of the classification scheme used in the North Carolina mortality maps discussed later in this chapter.∗ Higher rates of general mortality in this map are concentrated in the central part of the nation, which includes the Great Plains and Midwest, the South, and the outlying high rate counties in the Far West. There is also a significant cluster of counties centered in mountainous West Virginia and eastern Kentucky. Recall that age is the greatest risk for dying. Figure 2.3 is a map that shows the distribution of county level proportions of people 60 years of age and older for the 2000 US Census year. The distribution of the proportions of elderly is similar to the distribution of the higher crude mortality rates seen in the previous figure. Statistically, the relationship, measured as a correlation, results in an r- value of 0.81 and an R2 of 0.66, which means that 66% of the variation in county crude rates of mortality is explained by just the proportions of individuals greater than 60 years of age. Although the ∗ The North Carolina rates ( crude and age- adjusted) are based on more current numbers from the State Center for Health Statistics and State Data Center and use a five- year period such as 2000 to 2004. Because numbers for the entire US are usually available three years behind the state’s and that there are a significant number of US counties that experience small numbers of mortality events, US county rates ( crude and age- adjusted) in this work are based on three- year aggregations ( 2001 and 2003) from the National Center for Health Statistics’ Compressed Mortality File ( CMF) 1999 to 2003. The center point year or fulcrum year for the US county rate maps is 2002. That same year is the fulcrum for the period 2000 to 2004, which is the period used in the state and regional discussions throughout the Atlas. However, in the state and regional comparisons, the US value for one year is 2002, because their numbers are sufficiently large not to warrant aggregation. It should also be noted that the rates generated for North Carolina counties in the three- year US map will be slightly different than those generated for the five- year NC maps seen elsewhere in the Atlas. This is the result of using different numbers of data points ( three and five) and slight differences found in the denominators ( i. e., county populations) between the US ( CMF 1999- 2003) and NC ( state demographic estimates) data sources. 4 crude mortality rate map depicts where mortality is occurring in relation to population age structure, this map cannot be used to make meaningful comparisons among individual counties because their respective age structures are different. Figure 2.4 shows the effect that age- adjustment has on the county mortality pattern using an external standard population ( US 2000 Standard Million). The high rate counties shift and concentrate their spatial distribution to the Ozarks, Lower Mississippi Valley, the southern Coastal Plain, and the south-central Appalachian region of West Virginia and eastern Kentucky. A few outlying high rate counties are found scattered throughout the west, which generally correspond to Indian reservations. The inset map in figure 2.4 shows that ENC is a northern extension of the high rates of age- adjusted mortality found in the southern Coastal Plain. To contrast, the national age- adjusted map also shows that most of the remaining counties on North Carolina ( RNC) are part of the southern extension of the much more favorable mortality conditions of the Northeast. Maps showing the spatial distribution of crude and age- adjusted mortality rates from all causes in both NC and ENC for the years 2000 to 2004 are found in figure 2.5. Individual county and regional mortality rates are listed in table 2.1 and their locations can be found using the map in appendix A of this chapter. The state map for crude rates shows that the greatest mortality burden is experienced at both ends of the state. Western North Carolina ( WNC) has the highest crude general mortality rate of 1,080 deaths per 100,000 people. The next highest crude mortality rate is found in the northeast 29- county region of North Carolina ( ENC 29) with a rate of 967.7-- 7% higher than the 41- county ENC regional rate of 905.4. The highest county rates in ENC cluster together along a northwest- southeast axis. In this cluster, Hertford County experiences the highest rate of 1,330.6 and the second highest rate is found in its southern neighboring county, Bertie, at 1,313.3. Onslow County’s rate is the lowest in the region ( 518.4), experiencing mortality at just 39% of Hertford’s level. The greatest local impact of mortality from all causes is felt in the northern county cluster of ENC which also possesses an older aging- in- place population. This contrasts starkly to Onslow County where a significant portion of its population is made up of young, transient, military service- age people. When the county rates are age- adjusted ( figure 2.5b), the high mortality categories become more concentrated in the east and less so in the state’s mountainous west-- where the relative older ages of those county populations played a role in that region’s observed higher crude rates. The effect of age adjustment on relatively youthful county populations with low crude mortality rates can be quite dramatic. For example, when the crude rates for Cumberland and Onslow-- two counties with large military aged populations-- are age- adjusted, there is an apparent jump in mortality rates of 51% and 84%, respectively. At the regional level, age- adjustment widens the disparity for general mortality between ENC and RNC to more than 12% from the crude rate difference of 7%. Within the ENC region, counties with the highest age- adjusted rates form two centers, 5 one in the northeastern 29- county sub- region of ENC and the other in the remaining southern 12- county sub- region. In the 29- county sub- region, Edgecombe County possesses the highest rate ( 1125.6) among the eight county cluster found there. Other counties in this cluster include Halifax ( 1023.9), Northampton ( 1007.0), Hertford ( 1124.5), Gates ( 1061.1), Bertie ( 1111.2), Martin ( 1091.0), and Washington ( 1027.8).∗ The highest age- adjusted general mortality rate is found in the southern sub- region. Robeson County with 1133.3 age-adjusted deaths per 100,000 has the highest general mortality rate in the state and forms the core of the southern- center- high- rate- county- cluster. This county cluster includes Scotland ( 1063.6), Hoke ( 1015.5), Bladen ( 1101.7), and Columbus ( 1069.9) counties. A north- northeast linear series of adjacent high rate counties ( Sampson, Wayne, and Lenoir) continues from the northeastern border of Bladen County. Immediately adjacent to the east of the southern cluster is a three county cluster of the lowest age- adjusted general mortality rates found in the entire 41 county region. The counties in this cluster include New Hanover ( 832.7), Brunswick ( 842.8), and Pender ( 830.1). All three of these counties have rates that are more favorable than the RNC 59 county rate of 866.1 age- adjusted deaths per 100,000. The examination and comparison of crude and age- adjusted general mortality in North Carolina yields two conclusions. First, the higher crude rates found in the east like in the western counties, can partly be explained as a function of greater proportions of elderly. Second, when general mortality rates are adjusted for age, 16 of the 20 highest rate counties are found in ENC 41, which clearly demarcates this region as one experiencing a greater mortality burden in both an absolute and relative sense. Many of the counties in this discussion will be seen again in later sections of the Atlas when the geographic patterns of mortality from specific causes are explored. Figure 2.6 shows the contributions that race- sex specific age- adjusted general mortality make to the overall pattern of general mortality in North Carolina seen in the age- adjusted map ( fig. 2.5b). Applying the same age- adjusted rate category classification found in figure 2.5b to the rate distributions of each of the four demographic groups in figure 2.6 produces four distinct map patterns. Males of both races have higher rates ( i. e. they occupy the highest rate category: 1,004.3 to 2,107.8) of general mortality throughout the state. White males ( fig. 2.6c) have a state rate of 1,034.0 and a regional ( ENC 41) rate of 1,100.2 compared to nonwhite males whose rates are 1,336.2 and 1,406.0, respectively. Counties with larger proportions of retirement age populations, found within each of the state’s three physiographic provinces, as well as the larger metropolitan counties of the Piedmont have lower rates of death from all causes for white males. For nonwhite males, 95 of the state’s 100 counties are in the highest rate category. The remaining five counties are found in the westernmost portion of the state, and their lower rates for nonwhite males are probably the result of fewer people being in this race- sex group in the western region. The age- adjusted death rates ∗ In the three year average ( 2001 to 2003) for the 3,100 plus counties of the US, Martin County raked 15th highest in the nation at 1313 age- adjusted deaths per 100,000. 6 for females of either race are significantly less than their male counterparts. White females ( fig. 2.6b) have a state rate of 720.4 and a regional ( ENC 41) rate of 773.5 compared to nonwhite females ( fig. 2.6d) whose rates are 857.3 and 891.6, respectively. White females have rates in the lowest map category throughout the state. Eight out of eleven of this group’s highest rate counties are found in ENC. Nonwhite females have the most complex spatial distribution of mortality. A wide range of rates are observed throughout the state with the largest concentration of high rate counties found in ENC. Another large concentration of higher rate counties can be found along a north- south axis in the central Piedmont. In some counties, these high rates may be attributed to the smaller representation of this demographic group and thereby the potential effects of random variation of rates due to small numbers. Overall, there is very little geographic effect on nonwhite males with respect to the age- adjusted general mortality map patterns. White males, and females of both racial groups appear to shape or delimit the regional distribution of mortality from all causes, while the relatively greater proportion of nonwhite males in ENC further accentuates the high general mortality rates found in that region. When general mortality rates for North Carolina are age- adjusted for the years 2000 to 2004, 35 of ENC’s 41 counties ( 85%) emerge with rates above the state rate of 896.5, while 26 of RNC’s 59 counties ( 44%) do so. Partitioning the general mortality map for the total population into four separate maps based on race and sex reveals how the distribution of rates for the total population is weighted and shaped by its constituent sub- populations. Later chapters of the atlas will show the impacts of specific leading causes of death on these sub-populations and their subsequent contribution to the observed spatial patterns of general mortality. The age- adjusted general mortality map of NC and ENC represents the integration of the patterns produced by component leading causes of death. It is also the culmination of many different mortality processes that have been operating at their own characteristic scales, tempos, and modes. The next section discusses how some of these processes have affected the observed pattern of mortality in ENC over time. The Temporal Distribution of Age- Adjusted General Mortality Rates The following two figures ( 2.7 and 2.8) show how mortality has evolved over the 26- year time period from 1979 to 2004. The last five data points ( the years 2000 through 2004) in the ENC 41, RNC 59, and NC time series illustrate the amount of variation in annual rates that are subsumed into the single age- adjusted five-year ( 2000- 2004) rate seen in the preceding table and maps. A trend line, shown by dashes in the figure, is fitted for each of the time series and extended to the year 2010. The trend line is calculated based on information from the entire series of data points ( i. e., annual rates). Additional information about the trend line is also provided below the figure. This information includes the percent change in rates from the initial year to the latest year in the time series. The R2 value is a measure of how well the fitted trend line corresponds to the observed 7 series. The equation of the line, also shown, generates the trend line that allows the investigator to calculate an expected value for a given point in time. Time series trend lines can diverge, converge, or run parallel to one another. To make analysis easier, linearity of the observed data is assumed for the 26- year period in these time series graphs. However, broader temporal scales of observation show that mortality from any number of causes is generally non- linear ( see figure 1.2). With the simplifying assumption of linearity, it is possible to calculate an approximate time when two series will have the same rate ( convergence) or when two series began to separate ( divergence) from each other by setting the two equations of the line equal to one another. However this should be done only when R2 values are high ( i. e., approaching 1.00) and when making projections into the near future or more recent past. Making projections too far into the future, or past, over- extends the more limiting and linear perspective of recent mortality trends, resulting in the danger of making spurious conclusions about long- term and, most, likely non- linear processes. For example, using the equations- of- the- line in the trends description section found in figure 2.7, the age- adjusted general mortality rate for ENC 41 and RNC 59 will not be equal or converge until the year 2154! Clearly, the use of linear trend lines should only be used short term prognostication. Their utility lies in permitting the researcher to make summary assessments and examine potentially meaningful trends, emerging differences or improvements in rate disparities. Figure 2.7 illustrates four solid trends in regional declines of general mortality. The goodness- of- fit lines are all above 0.90, indicating that from 1979 to 2004 there are very tight fits to the modeled trend line and that predictions for the next several years could be reasonably and confidently made. Over this 26- year study period, age- adjusted mortality rates have declined by 16 and 17 percent for all four regions. The greatest decreasing coefficient belongs to the US (- 7.75) and the least to RNC 59 (- 6.34). This translates into an average growing disparity of age- adjusted general mortality rates of about 1.4 age- adjusted deaths per 100,000 per year over the course of the last 26 years. Although all trends are certainly favorable in absolute terms, the ENC 41 trend line stands out well above the others with the line equations demonstrating persistent relative disparity in mortality rate trends between this region and RNC 59. A closer look at the mortality experience of ENC 41 reveals substantial differences by race and gender. Figure 2.8 shows relatively flat trends ( from negligible to 7% decrease) for females, with only a slight growth in disparity by race ( see trend descriptions) over the 26- year period. White males have had the greatest amount of rate decline with a 28% decrease from 1979 to 2004. The trend for white males is very consistent over time and can probably be used reliably as an indicator of mortality scenarios in the near future. Nonwhite males follow with a more modest 16% decrease and a less confident trend line than their white counterparts. Although the trend lines for males from either racial 8 group are decreasing, the relative rate disparity between them, as measured by the equations- of- the- line, increases from 17% in 1979 to 37% in 2004. Since 1979, age- adjusted general mortality has been improving for all males in the region, while rates have remained relatively flat or changing little for regional females. The female pattern suggests that mortality rates may reach an asymptotic level for a period of time. One reason for this flattening out might be that all benefits from current health technologies, innovations, knowledge concerning care and behaviors have been nearly realized for that group over the last two to three decades. There may also be a certain amount of intra- regional “ balancing out” or counteracting of high and low rates among counties in different parts of ENC 41. The trend lines for males are converging on the trend lines for females— with white males approximating the mortality rates for nonwhite females some time around the year 2014 or 2015. It will be interesting to see if white males, and probably much later for nonwhite males, begin to approach a similar mortality asymptote as has been the case for females. It is likely that the reasons for the relatively low rates for females have yet to be completely realized for males, but the rates show that they are still in the process of responding to or adopting mortality reducing behaviors and technologies. Certainly the pattern between both female groups indicates that differential mortality remains even when rates are low and relatively stable. What accounts for this persistent differential forms the bulk of health disparities literature today. The above discussion and description of the patterns of crude and age- adjusted mortality reveals that a geographic disparity exists between the 41 county ( and 29 county) region of NC and the remaining counties of the state, with the east experiencing significantly higher rates than RNC. Within ENC, age- adjusted general mortality rates have been declining over the past three decades for the major demographic groups discussed in this chapter. For females of both racial groups the decline is relatively minimal, but for males the decline has been more dramatic, with nonwhite males having the sharpest decrease. Nonwhite males, although experiencing a larger decrease in general mortality rates have begun their downward trek at a much higher beginning rate so that the relative rate disparities between them and the other demographic groups will remain high for the foreseeable future. As previously mentioned, density measures tend to mask other types of information that can be derived from mortality records. The next section focuses on the concept of mortality burden and its measurement. Understanding the impact of premature mortality on county populations can assist in discriminating where disparities of mortality burden are occurring. 9 Mortality Burden Mortality burden can be viewed at different scales of impact. Within a family there is the obvious psychological, social, and economic impact of a member’s death. The decedent’s stage in the life cycle, occupation, resources, and position in society also has relevance in broader local and community scales of social relationships. Implicit in any decedent’s age at death is the tangible and intangible cost, benefit, and potential contribution of that individual’s life to both family and friends, and to the larger extended communities to which he or she belonged. Collectively these mortality experiences can be summarized into one point value: crude mortality rate. This density measure indicates the direct arithmetic impact or burden actual deaths can have on a population. However, a population with an older age structure will naturally have more individuals at risk of dying as they enter the latter stages of their life cycle during a given time interval and so that population may appear to be experiencing a higher burden of mortality. Another way to look at mortality burden is to look at how much potential life is lost, which is a comparison of an observed age- at- death against some expected or standard age at death. Instead of one point value, two point values are used, with greater differences between corresponding to increases in mortality burden. Age- at- death can be used to measure the amount of life lost prematurely from a standard number of years of life that an individual can be expected to live in the population of interest. The typical standard age used in current research is 75 years, which is close to 77.5 years, the life expectancy at birth ( e0) for the US in 2003 ( Arias, 2006) and nearly identical to the mean age of death in North Carolina. The number of deaths and their ages of occurrence before the age of 75 can be accumulated, age- adjusted, and normalized by the underlying population. Greater differences mean greater years of life lost, when calculated in this manner, and indicates a greater level of mortality burden being experienced prematurely. The meaning of a premature mortality rate or years of life lost rate as described above is qualitatively different than for the more commonly used density measures. To illustrate, the age- adjusted mortality rate in North Carolina for female breast cancer was 25.6 per 100,000 and for prostate cancer in males it was 29.1 per 100,000 in 2004. A comparison of these two rates would lead one to the conclusion that prostate was a slightly bigger killer of men than breast cancer is in women. However, when the premature mortality rates∗ for these two causes of death are compared, the number of years of life lost before age 75 is 33.9 years per 10,000 for female breast cancer and 6.4 years for prostat e ∗ Currently premature mortality is typically measured by the number of years of life lost ( YLL) before age 75 per 10,000 people. Each death is aggregated into an age category and the total number of deaths in that category is multiplied by the difference between the age category mean age at death and age 75. The resulting age category YLLs are then summed, divided by the population, and then multiplied by 10,000 to make interpretation easier. The YLL- 75 ( premature mortality) measure can either be crude or age- adjusted. 10 cancer. These values indicate that males tend to die at much later ages from prostate cancer and not prematurely relative to the age of 75. Females tend to die from breast cancer at earlier ages, suffering a greater mortality burden than their male counterparts for a sex- specific disease, with perhaps a greater impact on families and communities. The Spatial Distribution of Premature Mortality from All Causes The national age- adjusted premature mortality rate for the year 2002 is 751 years of life lost per 10,000 people. The lowest state premature mortality rate in this same year is found in Vermont at 568, while the worst state rate belongs to Mississippi at 1088. If the District of Columbia is added as a state it would fall behind Mississippi ranking a distant 51st with a rate of 1323. Within the state rankings, North Carolina is 39th with a rate of 833, and with the exception of Florida and Virginia, has the lowest premature mortality of the remaining southern states. If the 41 county region of ENC is entered into the state rankings, it would rank 47th at 959, with Arkansas, Alabama, Louisiana, Mississippi, and the District of Columbia trailing in the lowest ranks.† The 29- county region of ENC would rank 48th at 975, ousting Tennessee, which moves up to 47th. The Piedmont region compares more favorably as a state with a premature mortality rate of 774 placing it 29th among the states. The Western NC region has a more intermediate premature mortality rate of 805 and ranks 34th. Figure 2.9 is a map of the United States that shows the age- adjusted premature mortality rates for the states with North Carolina’s three regions mapped as ” states”. From this national context we now move to a more specific in- depth discussion of how premature mortality varies by sub- region and county within North Carolina. Figure 2.10 portrays both crude ( fig. 10a) and age- adjusted ( fig 2.10b) premature mortality rates measured as years of life lost before the age of 75 years ( YLL- 75). The maps in this figure describe the distribution of mortality burden for counties. Unlike the maps in figure 2.5, age- adjusting the rates ( i. e., the expected number of deaths) has very little effect on the map pattern of premature mortality. The ENC 41- county region stands out distinctly relative to the other regions of the state with its large number of high premature mortality counties. Table 2.2 bears this out with the age- adjusted rate for premature mortality 22% higher than RNC, and when compared to PNC and WNC, the region is 23% and 17% higher, respectively. Finally, the age- adjusted premature mortality rate for ENC is 27% higher than the rate for the nation, which for the year 2002, is 751.0. When premature mortality is compared on a national and regional level, the counties of North Carolina and ENC do not fare well. Only 14 NC counties in the state have premature mortality rates less than the US 2002 rate, with New † ENC 41 and 29 county regional rates, as well as other NC regional rates, are calculated using the National Center for Health Statistics’ Compressed Mortality File series data for the year 2002. 11 Hanover, at 705.6 years of life lost per 10,000, being the only county in the east to do so. Regionally, 36 of the 41 counties in ENC ( 84%) have rates above the North Carolina rate, while 27 of the 59 counties of the remaining NC counties ( 46%) have rates greater than the state. In terms of population exposed to risk of dying prematurely at a rate higher than the state, the difference between the two regions becomes even more dramatic. In ENC, 84% of the region’s population who are under the age of 75 years live in those counties that have higher rates than the state, while 27% of RNC’s population under 75 live in counties with a higher rate than the state. Moving to the individual county comparisons, Wake County experiences the least years of life lost in the state for the 2000 to 2004 period with a rate of 564.8 years per 10,000, which is 32% lower than the state rate. Robeson County has the least favorable rate for this study period at 1,234.7 years of life lost, 119% greater than the rate for Wake County. When the age- adjusted map for premature mortality for all causes is decomposed into maps focusing on the four demographic groups, differences in their contributions to the overall rates emerge ( see figure 2.11). The greatest contribution to the overall rate is made by nonwhite males ( fig. 2.11b). Like age-adjusted mortality rates, high county rates for this group are a ubiquitous feature throughout North Carolina, with the exception of a few counties in the western part of the state. ( For county locations, see the map in appendix A.) Duplin County, in southern ENC had the highest premature mortality rate in the state at 2,133.4 years of life lost per 10,000 ( see table 2.2). To contrast, white females ( fig. 2.11b) have ubiquitously low county rates with the highest state- wide county rate found in an ENC county, Northampton, at 803.4. Regionally, the lowest rate for white females is found in New Hanover County at 412.4, slightly more than half of the Northampton rate. Overlaying these two contrasting map patterns, are the rate distributions of white males ( fig. 2.11a) and nonwhite females ( fig. 2.11d). Both of these map patterns are more variegated than the previous two. The mapped distribution for white males, though heterogeneous, is weighted more by the higher rate categories concentrated primarily in ENC, but also found distributed throughout the peripheral non- metropolitan counties of the Piedmont, and the western counties. The highest rate for white males, 1,563.0 years of life lost, is found in Robeson County located in southern ENC on the South Carolina border. Like white males, the distribution of high rate categories for nonwhite female culminates in the east, while high rate counties are found scattered to the west of the region. For this group, the highest rate-- 1,517.3 years of life lost-- is found in Perquimans County. While the highest rates for each of the four demographic groups are found in ENC, the lowest rates for any of these groups are located outside of ENC. For white males and females ( fig. 2.11a— b), the lowest premature mortality rates are found in Wake County at rates of 578.1 and 352.9, respectively. The lowest meaningful rates ( i. e., rates calculated from deaths numbering 20 and more) for their nonwhite counterparts are found in McDowell County with nonwhite males at 977.8 years and nonwhite females at 494.0 years in Wilkes County. Both of these counties are found in the western portion of the state. To conclude, there is a discernable geographic difference in 12 mortality burden between ENC and RNC that is driven by the mortality experience of white males and nonwhite females The Temporal Distribution of Premature Mortality from All Causes Figure 2.12 is a comparison of premature mortality trends among regions from the years 1979 to 2004 ( 2002 for the US). Premature mortality for all four regions is declining at approximately the same rates. The relationships among the trends, in terms of their relative ranking in years of life lost rates, remains constant throughout the time series study period. ENC consistently experiences the highest rates of age- adjusted premature mortality but the trend line indicates an approximate 27% decrease from the beginning of the study period in 1979 to 2004, slightly less than the other regions. All regions show a similar pattern of decline, including the gentle oscillations of observed values about their respective trend lines. For the first five or six years, the decline in trends is steeper than any other interval in the series. Thereafter, the observed regional rates decline less steeply and fluctuate very little from their respective trend lines. This suggests that there may be emerging countervailing trends in premature mortality from specific causes, which either balance each other out or have become more stable over time. When ENC’s observed premature general mortality rate series is decomposed into four separate premature mortality series corresponding to each of the four demographic sub- groups, several distinct patterns emerge ( figure 2.13). The greatest decline in premature mortality is experienced by white males at 34%. Although nonwhite males have the largest negative coefficient (- 30.35), indicating the steepest rate of decline, they begin the series with an expected or modeled premature mortality rate ( the intercept) at a level some 72% greater than their white counterparts. The pattern of decline for this group is very similar to the one observed for regions and it may be that the nonwhite male experience is what is driving the patterns seen in the previous figure. The trend line for white males is decreasing more than twice as fast as the trend for nonwhite females and overtakes the latter sometime around the year 2002. The observed rate patterns suggest a convergence— a convergence that has been evident for the 10 years prior to 2002. For the last two or three years of the series the rates appear to be diverging but it is probably not indicative of a reversal in trends. The last demographic group, white females, shows the least amount of decrease ( 17%) over the 26- year period and like the age- adjusted mortality rates for this group they appear to approaching a rate asymptote. If present trends continue for the four demographic groups, the next convergence of premature mortality rates will occur between white males and females around the year 2030. As with age-adjusted mortality, decomposing the general premature mortality rate by demographic groups reveals differences and potential disparities among them. The shift in the county distributions from crude premature mortality rates to age-adjusted premature mortality rates is minimal when compared to west- to- east 13 shift in distributions of the density measures. One reason for this difference in pattern shifting is that ages at death close to 75 years have a small negative impact on the premature mortality outcome measure and a zero impact when deaths occur after that age. Larger numbers of deaths occurring at ages several years prior to 75 indicate a population experiencing a greater share of mortality burden as an outcome. For example, the accumulation of years of life lost due to high infant mortality rates, and earlier ages at death from cardiovascular diseases and cancer can reflect inherent problems with access to appropriate healthcare. ( Density measures essentially treat all deaths as equal in impact and cannot be used to measure the depth of mortality burden.) Regionally, this suggests that although the western region of the state possesses populations with higher relative proportions of elderly, their respective mortality burdens are not greater than expected. This contrasts to what the measure portrays for the eastern 41 counties of the state— a region that not only has a high proportion of elderly population with its attendant mortality but it is also a region that has a disproportionate number of its population dying prematurely. From Empiricism to Explanation: General Mortality Disparities The numerical evidence tells us that mortality is not experienced equally between ENC and the 59 remaining counties of North Carolina ( RNC). From 2000 to 2004, 110,390 deaths occurred in ENC and 249,278 deaths occurred in RNC. The latter region’s population is larger with a 5- year population- at- risk of 29,349,691 individuals compared to ENC’s 12,192,418. Proportionally, the expected number of deaths for ENC numbers would be 103,555. Subtracting the proportionalized mortality from the observed value of 110,390 yields an excess of 6,835 deaths ( 6.6% more) carried by ENC-- a crude measure of a geographic disparity for general mortality between the two regions. However, this does not account for the probable regional differences in age structure. ( Recall that age is the greatest risk factor for any individual dying during a specified time interval.) If age structure is controlled for the two regions, the difference in the number of deaths between the two regions grows to 12,924 ( 12.2% more), nearly twice the observed value and further exacerbating the apparent geographic disparity between the two regions. ENC experiences a greater burden of mortality-- almost 2,600 more deaths per year than would be expected given its population size and its age- structure. Characteristics other than population size and age may affect the observed and adjusted mortality disparity between the two regions. We can hypothesize ( or speculate) that there may be other factors or covariates at work with mortality rates that are also geographically distributed. For illustrative purposes, explanatory variables might include underlying racial and ethnic diversity, poverty, and rurality. The rationale or assumption for the choices of these variables is that income distribution ( related to racial/ ethnic diversity) and measurable financial and physical ( distance) access to health care have some discernable effect or relationship to mortality. However, one can further 14 speculate that these covariates are associated with many other measurable variables such as educational attainment, occupation and associated social relations ( including peer pressure), risky or health promoting behaviors, the value and awareness of health as a personal and social good, and so on. The first three covariates introduced can be thought of as surrogate measures— they are meant to capture and simplify a complex series of relationships among a spectrum of factors that are operating at different scales. Surrogate measures are used to assist the students of public health and mortality in focusing on those relationships with the most explanatory power and in the construction of the most parsimonious ecological model of mortality. Racial/ ethnic diversity, poverty, and rurality can be measured like age- adjusted mortality at the county and county- based regional level. For example, ENC’s county populations are more racially and ethnically diverse when compared to the counties of the rest of the state ( RNC). According to the US Census atlas, Mapping Census 2000: the Geography of U. S. Diversity ( see page 22 in Brewer & Suchan, 2001), 26 of ENC’s 41 counties have diversity index values at or above the US value of 0.49, with a regional index of 0.52. ( The diversity index is a measure of the probability that any two random people chosen from a county’s population will be of a different race.) Only 13 of RNC’s 59 counties are more diverse than the US, with a regional index of 0.39. From the US Census year 2000 ( 1999) data, a little more than 16.0% of ENC’s population is below the poverty line for that census year, which is almost 50% greater than the 10.7% reported for RNC’s population. Rurality is another attribute that distinguishes ENC from RNC. Slightly less than 49% of ENC’s population is classified as rural by the US Census Bureau, which contrasts to slightly more than 36% of RNC’s population being rural. The next step is to determine what influence or how well these proposed variables explain the county distribution of general mortality. To assess the relationships and associations between any two of these variables, we employ a methodology similar to that used in studying the temporal trends of general mortality. The following discussion will describe the linear relationships between the dependent ( age- adjusted mortality from all deaths) variable and each of the independent variables: the diversity index, poverty, and the proportion of rural population.. The interrelationships among the independent variables will also be examined. Exploring the strengths and weaknesses of association among variables is fundamental to hypotheses testing and the development of explanatory models. The correlation coefficient between mortality and the diversity index is 0.61. The adjusted R2 value is 0.365, which translates into more than 36% of the variation in mortality is explained by the variation found in the diversity index alone. The correlation coefficient between mortality and poverty is 0.63 and has an R2 of 0.385. More than 38% of the variation in mortality is explained by poverty alone— about 2% more than the diversity index. The least amount of explanation ( 0.0%) can be attributed to the measure of rurality. The correlation coefficient is 15 only 0.046, which produces the negligible R2 of 0.002. These simple analyses show that ethnic/ racial diversity and poverty have a substantial and direct effect on mortality. The next step would be to determine if there was any direct relationship between diversity and poverty and whether at some indirect level, rurality having some effect. A relationship among these variables would indicate that their effects on mortality were not independent. To get a handle on the amount of interaction between diversity and poverty ( collinearity) we can apply the same method used in the preceding example Lower R2 values will suggest smaller amounts of collinearity, less association, and more independence among the independent variables. For rurality and poverty the R is 0.344 and the R2 is 0.118, which means there is a small level of rurality and poverty associated with each other at the county level. Next, the diversity index and poverty measure yield an R of 0.531, with an R2 of 0.27, which means that racial/ ethnic diversity is more related to poverty than the degree of county rurality. How related is a county’s racial and ethnic diversity to its level of rurality? The R for this comparison is 0.186 with an R2 value of just 0.025. Recall that poverty, in this simple example, offers the greatest explanation of mortality. We now know that while rurality has some effect on poverty, diversity has an even greater effect on this variable. In more elaborate models of explanation, the rurality measure ( as devised here) would not contribute much to explanation and could probably be excluded. The foregoing discussion is meant as a simple example of how empirical descriptions of mortality can provide a basis for research questions and the building statistically oriented explanatory models. However, numerical and graphical descriptions of mortality can also stimulate further research or thinking in non- statistical ways. For example, thoughtfully publicized rate increases in mortality due to automobile accidents or diabetes will raise the awareness of policy makers and citizenry and help promote interventions, funding, and other ameliorative measures. Empirical description and explanatory models each have their own place and can be useful adjuncts to each other in the presentation and understanding of public health and demographic problems. Conclusion Geographically, different ways of measuring and describing general mortality demonstrates that the eastern 41 counties of North Carolina experience both higher comparative levels of death from all causes and a disproportionate share of mortality burden in regional and national contexts. When general mortality rates for ENC 41 are decomposed into four major demographic groups, rate differentials ( or disparities) emerge. The distributions of age- adjusted general mortality rates also have unique characteristics for each of the race- sex sub-populations. Time series depictions ( 1979 to 2004) for both regions and race-sex sub- populations also show that there has been progress, but relatively large gaps or “ disparities” continue to exist. For sub- populations, males of both racial 16 groups have greater relative declines in their rates compared to their female counterparts. All measures, spatially and temporally, indicate that although absolute differences in general mortality has been declining among regions and sub- populations, relative disparities will continue for some time to come. A description and examination of general mortality, which reveals the great disparities observed in our region of interest, naturally leads to further questions about how and why such disparities exist. With this in mind, we enter into the realm of explanation and can begin to consider the relationships and associations of covariates and mortality. Explanatory models are valuable aids for determining where changes can be effected and where healthcare resources can best be allocated. General mortality encompasses a myriad of causes of death, all which have been classified and coded. In this regional atlas of mortality, the subsequent chapters will address the ten leading causes of death as shown in figure 2.1. These ten leading causes of death account for more than 80% of deaths occurring in ENC 41 during the years 2000 through 2004. It is our hope that a consideration of each of these will lead to an increased understanding in the exceptional character of the region’s mortality experience. References American Heart Association. ( 2005). Heart disease and stroke statistics-- 2005 update. Dallas, Texas: American Heart Association. Anderson, R. N., & Rosenberg, H. M. ( 1998). Age standardization of death rates: Implementation of the year 2000 standard. National Vital Statistics Reports, 47( 3) Arias, E. ( 2006). United states life tables, 2003. National Vital Statistics Reports, 54( 14) Brewer, C. A., & Suchan, T. A. ( 2001). Mapping census 2000: The geography of U. S. diversity. Washington, D. C.: U. S. Government Printing Office. 17 Buescher, P. A. ( 1998). Age- adjusted death rates ( 13th ed.). Raleigh, North Carolina: North Carolina Center for Health Statistics. U. S. Department of Health and Human Services, Centers for Disease Control and Prevention & National Center for Health Statistics. ( 2006). International classification of diseases, tenth revision ( ICD- 10). Retrieved 10/ 20, 2006, from http:// www. cdc. gov/ nchs/ about/ major/ dvs/ icd10des. htm World Health Organization. ( 2006). International statistical classification of diseases and disorders and related problems 10th revision for 2006. Retrieved 10/ 20, 2006, from http:// www. who. int/ classifications/ apps/ icd/ icd10online/ World Health Organization. ( 2004). International statistical classification of diseases and related health problems ( 10th revision, 2nd ed.). Geneva: World Health Organization. Total Cardiovascular Disease 37.0% Malignant Neoplasms 22.3% COPD/ CLRD1 4.9% Diabetes Mellitus 3.5% UMVI2 2.8% AOUIAD3 2.5% Pneu/ Infl4 2.3% NNN5 2.0% Alzheimer’s 1.7% Septicemia 1.7% All Other 19.3% 1Chronic Obstructive Pulmonary Diseases and Allied Conditions/ Chronic Lower Respiratory Disease 2Unintentional Motor Vehicle Injuries 3All Other Unintentional Injuries and Adverse Effects 4Pneumonia and Influenza 5Nephritis, Nephrotic Syndrome, and Nephrosis Figure 2.1: General Mortality in Eastern North Carolina 2000 to 2004 Percent Contributions from the Top Ten Leading Causes of Death to the 5- year Total Number of Deaths: 110,390 ECU, Center for Health Services Research and Development, 2007 0.0 to 856.8 856.8 to 974.2 974.2 to 1064.9 1064.9 to 1188.7 1188.6 to 2018.3 Figure 2.2 US Crude General Mortality Rates1 2001 to 2003 Per 100,000 Population ENC 41 Counties ECU, Center for Health Services Research and Development, 2007 1Data from Compressed Mortality Files 1999 to 2003 0.03 to 0.15 0.15 to 0.18 0.18 to 0.20 0.20 to 0.23 0.23 to 0.42 Figure 2.3 US County Population Proportions 60 Years and Older1 2000 County Proportion GTE 60 Years ENC 41 Counties ECU, Center for Health Services Research and Development, 2007 1Data from US Census 2000 0.0 to 843.0 843.0 to 896.8 896.8 to 944.8 944.8 to 1004.3 1004.3 to 2018.3 Per 100,000 Population Figure 2.4 US Age- Adjusted General Mortality Rates1 2001 to 2003 ENC 41 Counties ECU, Center for Health Services Research and Development, 2007 1Data from Compressed Mortality Files 1999 to 2003 and the 2000 Standard Million Population for the US Per 100,000 Population a. Crude b. Age- Adjusted1 Data Source: Odum Institute, UNC— Chapel Hill Per 100,000 Population 503.6 to 856.8 856.8 to 974.2 974.2 to 1064.9 1064.9 to 1188.6 1188.6 to 1480.9 752.3 to 843.1 843.1 to 896.8 896.8 to 944.8 944.8 to 1004.3 1004.3 to 1133.3 ECU, Center for Health Services Research and Development, 2007 1 Five- Year Average, Age- Adjusted Rates Standardized to US 2000 SM Mortality Rates from All Causes of Death: North Carolina and Eastern North Carolina Total Population, 2000- 2004 Figure 2.5 ECU, Center for Health Services Research and Development, 2007 Age- Adjusted1 Mortality Rates from All Causes of Death: North Carolina and Eastern North Carolina Race- Sex Specific, 2000- 2004 a. White Males b. White Females c. Non- White d. Non- White Per 100,000 Population 1 Five- Year Average, Age- Adjusted Rates Standardized to US 2000 SM 2 in Mitchell County, there were no non- white female deaths Data Source: NC State Center for Health Statistics Figure 2.6 0.02 to 843.1 843.1 to 896.8 896.8 to 944.8 944.8 to 1004.3 1004.3 to 2107.8 ( NWM) Males Females ECU, Center for Health Services Research and Development, 2007 1 Age- Adjusted Rates Standardized to US 2000 SM Figure 2.7 North Carolina: Comparisons among Regions2, 1979 to 2004 Trend Descriptions Age- Adjusted1 Mortality Rate Trends from All Causes of Death ENC 41 16% decrease R2 = 0.92 Y = - 7.04x + 1146 RNC 59 16% decrease R2 = 0.92 Y = - 6.34x + 1023 NC 16% decrease R2 = 0.93 Y = - 6.57x + 1059 US 17% decrease R2 = 0.96 Y = - 7.75x + 1034 800 850 900 950 1000 1050 1100 1150 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 00 01 02 03 04 05 06 07 08 09 10 Age- adjusted mortality rate per 100,000 population ENC 41 RNC 59 NC US Years 2 NC, ENC 41, and RNC 59 1979- 2004 mortality data from NC SCHS & US 1979- 2002 mortality data from NCHS’s Compressed Mortality File 600 700 800 900 1000 1100 1200 1300 1400 1500 1600 1700 1800 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 00 01 02 03 04 05 06 07 08 09 10 Age- adjusted mortality rate per 100,000 population NWM WM NWF WF ECU, Center for Health Services Research and Development, 2007 1 Age- Adjusted Rates Standardized to US 2000 SM Trend Descriptions WM 28% decrease R2 = 0.97 y = - 16.35x + 1497 WF 7% decrease R2 = 0.45 y = - 2.24x + 813 NWM 16% decrease R2 = 0.53 y = - 10.80x + 1751 NWF ------ R2 = 0.09 Y = - 1.16x + 943 Years Figure 2.8 Eastern North Carolina: Comparisons among Race- Sex Groups2, 1979 to 2004 Age- Adjusted1 Mortality Rate Trends from All Causes of Death 2 ENC 41mortality data from NC SCHS 567.6 to 630.9 630.9 to 707.3 707.3 to 788.1 788.1 to 911.7 911.7 to 1088.0 Age- Adjusted1 Years of Potential Life Lost before Age 752 Per 10,000 Population Natural Breaks Regional Variation of Years of Potential Life Lost in North Carolina ECU, Center for Health Services Research and Development, 2007 Figure 2.9 Premature Mortality in the United States 2002 with Selected Rankings Not Shown: AK: 36th DC: 52nd HI: 5th VT: 1st NH: 2nd MN: 3rd IA: 4th NC: 37th AR: 48th AL: 49th LA: 50th MS: 51st ENC 41: 47th PNC: 29th WNC: 34th VA: 22nd SC: 45th US ( 751.0) 2 ENC 41, PNC, WNC, and US 1979- 2002 mortality data from NCHS’s Compressed Mortality File 1 Age- Adjusted Rates Standardized to US 2000 SM DC ( 1323.0) Years of Life Lost Per 10,000 Population a. Crude b. Age- Adjusted1 Data Source: Odum Institute, UNC— Chapel Hill 541.0 to 806.7 806.7 to 878.2 878.2 to 958.4 958.4 to 1073.2 1073.2 to 1273.6 Years of Life Lost Per 10,000 Population 564.8 to 775.2 775.2 to 835.2 835.2 to 924.5 924.5 to 1036.6 1036.6 to 1234.7 ECU, Center for Health Services Research and Development, 2007 1 Five- Year Average, Age- Adjusted Rates Standardized to US 2000 SM Premature Mortality Rates from All Causes of Death: North Carolina and Eastern North Carolina Total Population, 2000- 2004 Figure 2.10 ECU, Center for Health Services Research and Development, 2007 Age- Adjusted1 Premature Mortality Rates from All Causes of Death: North Carolina and Eastern North Carolina Race- Sex Specific, 2000- 2004 Years of Life Lost Per 10,000 Population 1 Five- Year Average, Age- Adjusted Rates Standardized to US 2000 SM 2 in Mitchell County, there were no non- white female deaths Data Source: NC State Center for Health Statistics Figure 2.11 0.02 to 775.2 775.2 to 835.2 835.2 to 924.5 924.5 to 1036.6 1036.6 to 2174.3 ( NWM) a. White Males b. White Females c. Non- White d. Non- White Males Females ECU, Center for Health Services Research and Development, 2007 600 700 800 900 1000 1100 1200 1300 1400 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 00 01 02 03 04 05 06 07 08 09 10 Age- adjusted years of life lost per 10,000 population < 75 years of age ENC 41 RNC 59 NC US Trend Descriptions ENC 41 27% decrease R2 = 0.94 Y = - 12.97x + 1265 RNC 59 30% decrease R2 = 0.94 Y = - 12.43x + 1083 NC 29% decrease R2 = 0.94 Y = - 12.72x + 1139 US 30% decrease R2 = 0.96 Y = - 13.01x + 1053 Years 1 Age- Adjusted Rates Standardized to US 2000 SM Figure 2.12 North Carolina: Comparisons among Regions2, 1979 to 2004 Age- Adjusted1 Mortality Rate Trends from All Causes of Death 2 NC, ENC 41, and RNC 59 1979- 2004 mortality data from NC SCHS & US 1979- 2002 mortality data from NCHS’s Compressed Mortality File ECU, Center for Health Services Research and Development, 2007 400 600 800 1000 1200 1400 1600 1800 2000 2200 2400 2600 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 00 01 02 03 04 05 06 07 08 09 10 Age- adjusted years of life lost per 10,000 population < 75 years of age NWM WM NWF WF WM 34% decrease R2 = 0.91 y = - 18.79x + 1424 WF 17% decrease R2 = 0.75 y = - 4.30x + 670 NWM 32% decrease R2 = 0.81 y = - 30.35x + 2445 NWF 18% decrease R2 = 0.65 Y = - 8.13x + 1168 Trend Descriptions Years 1 Age- Adjusted Rates Standardized to US 2000 SM Figure 2.13 Eastern North Carolina: Comparisons among Race- Sex Groups2, 1979 to 2004 Age- Adjusted1 Mortality Rate Trends from All Causes of Death 2 ENC 41mortality data from NC SCHS ECU, Center for Health Services Research and Development, 2007 Table 2.1 Mortality from All Causes: Eastern North Carolina, 2000- 2004 County Deaths Crude Adjusted Deaths Rate Deaths Rate Deaths Rate Deaths Rate Beaufort 2,692 1185.2 988.2 925 1159.1 975 770.0 382 1418.9 410 929.5 Bertie 1,292 1313.3 1111.2 249 1170.4 300 907.4 373 1541.3 370 899.4 Bladen 1,983 1216.3 1101.7 586 1238.8 614 875.0 390 1551.1 393 971.0 Brunswick 3,806 962.7 842.8 1,741 927.3 1,533 717.3 278 1289.0 254 789.0 Camden 323 856.8 844.1 131 966.5 126 763.4 39 1122.7 27 589.5 Carteret 3,452 1142.1 932.7 1,667 1108.4 1,575 785.2 102 1251.7 108 812.7 Chowan 925 1275.0 949.1 296 1058.6 308 719.4 169 1620.3 152 785.2 Columbus 3,169 1157.3 1069.9 1,063 1316.3 1,066 824.8 531 1488.5 509 927.7 Craven 4,275 931.7 938.0 1,534 975.0 1,578 769.0 567 1448.4 596 974.8 Cumberland 10,093 663.7 1004.3 3,140 1175.3 3,080 832.5 1,965 1274.8 1,908 879.0 Currituck 833 835.9 911.8 367 964.8 365 804.8 48 1453.3 53 985.3 Dare 1,317 822.6 859.7 683 964.3 587 733.9 23 1141.7 24 1045.9 Duplin 2,525 999.3 983.9 807 1095.1 844 768.3 440 1534.7 434 888.7 Edgecombe 3,045 1107.6 1125.6 707 1336.7 778 881.2 772 1563.1 788 924.7 Gates 614 1149.1 1061.1 185 1251.1 190 906.9 121 1331.2 118 890.2 Greene 897 918.4 941.9 290 1256.9 277 722.9 169 1334.1 161 737.8 Halifax 3,310 1167.0 1023.9 785 1184.1 891 769.2 830 1413.4 804 868.8 Harnett 3,807 787.8 944.8 1,448 1136.9 1,477 760.6 459 1375.2 423 815.3 Hertford 1,499 1330.6 1124.5 317 1404.6 336 828.0 432 1621.4 414 903.1 Hoke 1,256 691.0 1015.5 330 1144.7 282 765.5 323 1323.8 321 922.1 Hyde 340 1197.6 905.1 109 1171.5 123 771.9 58 1233.8 50 679.4 Johnston 4,985 752.5 917.7 2,117 1128.0 2,014 739.2 462 1344.1 392 759.1 Jones 587 1138.6 980.2 189 1198.3 171 738.8 98 1260.5 129 928.2 Lenoir 3,546 1201.9 1078.8 1,058 1275.8 1,084 834.1 702 1637.1 702 910.5 Martin 1,605 1275.2 1091.0 444 1388.0 498 881.0 299 1324.0 364 953.6 Nash 4,449 998.1 1003.6 1,414 1122.0 1,591 814.7 721 1516.9 723 900.4 New Hanover 7,084 850.9 832.7 2,732 917.5 2,907 672.9 646 1310.6 799 954.8 Northampton 1,432 1309.9 1007.0 297 1040.5 317 759.2 448 1653.7 370 779.8 Onslow 3,879 518.4 956.7 1,648 1171.1 1,442 789.0 381 1193.9 408 852.2 Pamlico 723 1124.7 815.1 263 931.9 280 678.8 89 1210.4 91 758.0 Pasquotank 1,820 1020.5 925.0 538 1072.7 615 735.0 322 1308.6 345 843.3 Pender 1,877 873.8 830.1 730 945.2 632 687.2 251 1214.7 264 753.9 Perquimans 743 1285.3 929.0 264 998.7 250 731.6 112 1543.4 117 893.9 Pitt 5,269 768.2 955.8 1,510 1058.1 1,700 740.4 1,005 1458.0 1,054 924.6 Robeson 5,904 943.9 1133.3 1,292 1326.2 1,258 876.7 1,674 1464.5 1,680 977.7 Sampson 3,158 1028.3 1013.0 1,073 1253.1 1,006 764.4 549 1410.8 530 895.4 Scotland 1,800 1000.8 1063.6 478 1237.1 551 859.8 384 1622.9 387 903.3 Tyrrell 221 1064.9 864.3 72 1016.5 72 742.0 45 1184.7 32 667.6 Washington 829 1226.9 1027.8 217 1074.5 249 794.5 174 1364.6 189 1000.6 Wayne 5,307 936.0 1046.6 1,686 1199.2 1,742 850.2 898 1394.9 981 974.6 Wilson 3,719 992.3 976.7 1,120 1114.1 1,239 769.2 674 1394.8 686 879.9 ENC 29 61,468 967.7 983.2 19,772 1109.8 20,503 783.8 10,493 1439.7 10,700 892.4 ENC 41 110,390 905.4 972.2 36,502 1100.2 36,923 773.5 18,405 1406.0 18,560 891.6 RNC 59 249,278 849.3 866.1 99,131 1011.2 105,502 703.2 22,467 1284.6 22,178 831.6 PNC 190,449 796.7 871.7 71,626 1009.0 77,309 706.3 20,889 1287.9 20,625 833.9 WNC 58,829 1080.4 854.1 27,505 1032.0 28,193 700.1 1,578 1252.7 1,553 810.3 NC 359,668 865.8 896.5 135,633 1034.0 142,425 720.4 40,872 1336.2 40,738 857.3 US, 2002 2,443,030 847.2 845.5 1,024,966 993.1 1,077,337 701.5 174,016 1118.2 166,711 773.1 White Males White Females 5- Year Race- Sex Specific Age- Adjusted Death Rates Rates 5- Year Totals Non- White Males Non- White Females Center for Health Services Research and Development East Carolina University Source NC Data: Odum Institute-- UNC, Chapel Hill US Data: NCHS ECU, Center for Health Services Research and Development, 2007 Table 2.2 Premature Mortality from All Causes: Years of Life Lost before Age 75 in Eastern North Carolina, 2000- 2004 County Deaths Crude Adjusted Deaths Rate Deaths Rate Deaths Rate Deaths Rate Beaufort 1,265 1119.7 1048.2 526 1181.0 294 595.3 260 1966.2 185 1102.0 Bertie 612 1219.9 1172.9 127 1142.3 94 639.8 242 1931.5 149 821.9 Bladen 996 1205.6 1143.7 334 1239.2 207 642.4 269 2008.4 186 996.9 Brunswick 2,002 930.0 859.5 1,052 1082.3 638 548.0 183 1580.4 129 780.8 Camden 163 801.8 755.6 75 826.5 46 529.4 26 1758.3 16 541.0 Carteret 1,594 1003.2 931.1 880 1184.7 596 673.6 68 1264.7 50 778.6 Chowan 387 977.5 924.5 137 906.1 89 527.9 102 1903.1 59 760.0 Columbus 1,655 1219.4 1166.5 648 1360.1 390 681.4 371 1997.8 246 1020.6 Craven 1,992 843.9 833.5 818 893.9 535 510.8 352 1457.3 287 961.5 Cumberland 5,971 881.3 935.5 2,005 989.4 1,383 627.3 1,471 1380.1 1,112 857.2 Currituck 425 843.6 816.5 217 924.8 165 660.7 28 1397.2 15 654.7 Dare 677 889.2 858.1 409 1098.2 236 588.6 19 1679.8 13 702.8 Duplin 1,236 1024.0 1004.9 469 1043.4 278 576.4 301 2133.6 188 912.7 Edgecombe 1,583 1165.4 1141.8 414 1148.4 263 628.2 521 1866.6 385 893.0 Gates 257 958.4 936.9 91 1000.1 59 745.4 62 1433.7 45 662.7 Greene 458 1021.5 1006.2 170 1131.5 96 603.0 112 1331.4 80 1081.8 Halifax 1,618 1181.7 1155.6 401 1226.5 278 601.8 574 1759.3 365 980.8 Harnett 1,978 844.2 878.7 874 1025.7 560 581.9 331 1464.1 213 821.1 Hertford 691 1241.5 1203.7 153 1350.5 92 535.7 267 1865.1 179 932.6 Hoke 774 1004.0 1069.0 229 1174.3 122 668.9 242 1487.8 181 929.0 Hyde 145 931.0 884.6 57 1015.5 38 614.4 27 1109.8 23 949.6 Johnston 2,620 830.4 833.7 1,298 1003.7 781 525.0 337 1609.1 204 813.7 Jones 276 1012.2 964.2 114 1269.2 53 494.5 58 1330.3 51 977.3 Lenoir 1,779 1273.6 1225.9 595 1307.7 385 758.5 477 2047.6 322 1149.2 Martin 775 1203.4 1143.9 239 1243.3 156 650.7 204 1777.9 176 1120.2 Nash 2,113 970.7 949.9 774 1028.4 507 570.9 484 1572.3 348 961.7 New Hanover 3,169 733.7 705.6 1,423 765.3 947 412.4 432 1614.4 367 938.4 Northampton 679 1266.4 1209.6 153 1138.9 107 803.4 259 1851.0 160 948.7 Onslow 2,315 719.4 816.9 1,091 898.0 693 629.1 283 1156.1 248 801.8 Pamlico 321 887.0 826.6 145 924.9 86 503.4 47 1155.2 43 1165.9 Pasquotank 784 828.0 821.2 251 822.3 184 508.3 189 1289.6 160 862.8 Pender 968 901.3 856.0 429 991.9 259 567.6 161 1557.2 119 709.1 Perquimans 333 1110.1 1055.5 135 1052.2 86 783.7 69 1530.9 43 1517.3 Pitt 2,671 857.1 913.4 844 843.9 586 541.2 711 1692.7 530 1008.8 Robeson 3,290 1219.0 1234.7 820 1563.0 444 728.3 1,184 1649.3 842 962.8 Sampson 1,553 1096.3 1081.3 616 1243.2 342 665.5 347 1725.6 248 1017.3 Scotland 944 1087.9 1077.7 281 1119.3 207 687.2 257 1715.1 199 964.3 Tyrrell 101 1041.3 1007.1 36 909.8 26 765.7 25 1564.1 14 1100.5 Washington 376 1079.7 1036.6 114 1137.4 67 480.6 113 1592.5 82 1072.1 Wayne 2,752 1010.0 1000.0 1,001 1023.5 617 628.7 636 1628.7 498 1061.7 Wilson 1,776 1008.9 983.6 608 1019.7 394 549.3 446 1716.2 328 932.0 ENC 29 30,154 976.4 964.5 11,044 1002.0 7,106 598.1 6,962 1648.4 5,042 965.0 ENC 41 56,074 957.8 951.6 21,053 1014.9 13,386 584.2 12,547 1608.2 9,088 936.7 RNC 59 113,504 794.0 780.8 52,209 885.4 34,378 513.2 15,669 1420.1 11,248 827.1 PNC 88,764 777.5 773.9 38,159 848.3 25,450 499.6 14,602 1420.4 10,553 825.5 WNC 24,740 868.2 814.9 14,050 1026.1 8,928 565.6 1,067 1419.0 695 863.0 NC 169,578 842.2 830.6 73,262 918.9 47,764 531.1 28,216 1495.5 20,336 870.7 US, 2002 1,054,300 755.3 751.0 509,168 878.6 337,327 511.3 120,404 1276.0 87,401 761.1 White Males White Females 5- Year Race- Sex Specific Age- Adjusted Death Rates Rates 5- Year Totals Non- White Males Non- White Females Center for Health Services Research and Development East Carolina University Source NC Data: Odum Institute-- UNC, Chapel Hill US Data: NCHS Pitt Wake Hyde Duplin Bladen Bertie Pender Wilkes Moore Onslow Union Surry Ashe Beaufort Craven Halifax Robeson Nash Sampson Iredell Columbus Swain Carteret Burke Brunswick Johnston Anson Guilford Randolph Harnett Wayne Jones Chatham Macon Rowan Hoke Martin Tyrrell Dare Lee Stokes Stanly Lenoir Franklin Buncombe Warren Granville Davidson Jackson Haywood Gates Person Caldwell Wilson Forsyth Polk Caswell Cumberland Orange Pamlico Rutherford Madison Yadkin Gaston Clay Cherokee Richmond Cleveland Catawba Davie Rockingham McDowell Hertford Alamance Vance Avery Yancey Mecklenburg Northampton Edgecombe Montgomery Durham Graham Scotland Greene Watauga Henderson Washington Transylvania Mitchell Alleghany Currituck Camden Chowan Perquimans Pasquotank New Hanover Lincoln Cabarrus Alexander Western ( WNC) Piedmont ( PNC) Remaining 59- County Region ( RNC 59) Eastern North Carolina 29- County Sub- region ( ENC 29) Eastern North Carolina 12- County Sub- region Eastern North Carolina 41- County Region ( ENC 41) North Carolina County and Regional Locations Center for Health Services Research and Development East Carolina University Greenville, NC ECU, Center for Health Services Research and Development, 2007 Appendix A ECU, Center for Health Services Research and Development, 2007 CARDIOVASCULAR DISEASE MORTALITY The biggest cause of death in both the United States and North Carolina continues to be from diseases of the circulatory system, commonly referred to collectively as cardiovascular disease. Cardiovascular disease ( CVD) includes high blood pressure ( hypertension), coronary heart disease, congestive heart failure, atherosclerosis, and stroke, conditions which often occur in combination. An estimate for the year 2004 indicates that 79 million adult Americans, about 1 of every 3, have one or more types of CVD and mortality from CVD comprises a little more than 36% of the 2.4 million deaths that occurred in the United States ( Writing Group Members et al., 2006). In 2004, CVD in North Carolina accounts for almost 34% of the 72,000 resident deaths that year and in Eastern North Carolina more than 35% of its 22,000 deaths have been attributable to CVD. The impact and burden of CVD is so great that if all its forms were to be eliminated, life expectancy in the United States would rise by almost 7 years. For Americans born today, there is nearly a 50- 50 chance that their eventual death will be due to CVD ( Anderson, 1999). In the present chapter, CVD mortality includes deaths due to heart disease ( HD), coronary heart disease ( CHD), and stroke, in addition to several other less prominent causes of the circulatory system. 1 The largest CVD mortality component is heart disease, which includes rheumatic heart disease, irregular heart rhythms, and diseases of the linings, valves, and vessels of the heart. The latter- most group generally pertains to blockages and constriction of the vessels that supply the heart and can lead to diseases like infarction and ischemia. Mortality from this group is a significant part of HD mortality and is considered separately as CHD. Stroke mortality is a distinct category within CVD that includes intracranial blockages ( resulting in infarctions) and hemorrhages, and other cerebrovascular diseases. Figure 3.1 summarizes the relationships of the TCVD mortality categories for the 41 counties of ENC during the period 2000 to 2004. For this 5- year period, heart disease and stroke comprise nearly 92% of all mortality attributed to TCVD, while CHD alone contributes slightly more than half of all CVD deaths. The less prominent CVD mortality category ( All Other) is not considered in this chapter. A complete listing of ICD10 codes organized by the categories used here can be found in the appendix for this section. 1 ICD9 Codes 390- 459; ICD10 Codes I00- I99 Cardiovascular Disease Mortality 1 ECU, Center for Health Services Research and Development, 2007 CVD mortality and its three major component diseases discussed in this chapter can be accessed below. CARDIOVASCULAR DISEASE MORTALITY Spatial Distribution of Cardiovascular Disease Mortality Temporal Distribution of Cardiovascular Disease Mortality HEART DISEASE MORTALITY Spatial Distribution of Heart Disease Mortality Temporal Distribution of Heart Disease Mortality CORONARY HEART DISEASE MORTALITY Progress towards Coronary Heart Disease Mortality Reduction Spatial Distribution of Coronary Heart Disease Mortality Temporal Distribution of Coronary Heart Disease Mortality STROKE MORTALITY Progress towards Stroke Mortality Reduction Spatial Distribution of Stroke Mortality Temporal Distribution of Stroke Mortality SUMMARY References As can be seen from the chart Six Leading Causes of Mortality in the US 1900 to 2001 ( figure 1.2), heart disease has emerged as the nation’s leading cause of death in the 1920s and continues to be the leading cause into the early 21st century. The chart also shows how the decline of infectious and communicable diseases in the first several decades of the twentieth century paved the way for this emergence. If both stroke mortality and HD mortality rates depicted in figure 1.2 were combined, then the combined rate would account for the largest share of general mortality since the turn of the 20th century ( with the exception of the influenza pandemic of 1918). The diminishing effect of infectious and communicable diseases on the mortality experience of the first half of the 20th century in the United States has given way to the rising prominence of death from heart disease in the latter half. The Epidemiologic Transition ( Omran, 1977) discussed in chapter one ( introduction) describes the secular decline of infectious/ communicable diseases and the concomitant rise of chronic disease mortality and its demographic consequences. The increase seen in HD mortality is more than likely the result of the rise in the proportion of people surviving the onslaughts of communicable diseases. Communicable diseases have their impact on both ends of the age Cardiovascular Disease Mortality 2 ECU, Center for Health Services Research and Development, 2007 spectrum. Over time, survivors of childhood diseases swell older age groups which have increasing susceptibility to HD and other cardiovascular problems. This pattern is repeated wherever infectious/ communicable diseases are brought under control with various public health measures and interventions. However, the demographic responses and outcomes can vary geographically and culturally. It is interesting to note that the states with the lowest rates, Minnesota, Alaska, and New Mexico are quite different in regard to their demographic attributes; investigation of the role of culture is suggested. The US Department of Health and Human Service’s document, Healthy People 2010 ( U. S. Department of Health and Human Services, 2000) provides target rates for the two major mortality categories of CVD: coronary heart disease, and stroke. Objective maps are included in this chapter for these two causes of death. Time series charts ( 1979 to 2004) are also included for each CVD mortality category ( including total CVD). For the coronary heart disease and stroke mortality time series charts, the HP 2010 targets are indicated. Spatial Distribution of Cardiovascular Disease Mortality The 2002 age- adjusted mortality rate for CVD ( ICD- 10: I00- I99) for the United States is 319 deaths per 100,000 population but there is remarkable geographic variation across the nation. State rankings2 ( including the District of Columbia) place Minnesota, Alaska, and New Mexico first, second, and third with the lowest respective age- adjusted rates per 100,000 of 237.7, 242.1, and 255.9 per 100,000, respectively. The highest rates are found for Tennessee, Oklahoma, and Mississippi, ( rates of 380.8, 398.8, and 420.7, which placed 50th, 51st, and 52nd respectively). The rate for North Carolina in 2002 was 327.0, ranking it 33rd in the nation. The 2002 average age- adjusted rate for the 41- county region within Eastern North Carolina ( ENC) is 366.6. If this region were treated as a state, it would rank 45th. For the 5- year period 2000- 2004, seven counties in ENC ranked worse than the state of Mississippi in 2002.3 The maps at the top of figure 3.2 shows the spatial distribution of CVD crude mortality rates for the 100 counties of North Carolina and the 41- county ENC region. CVD crude mortality has its greatest impact in the northeastern part of the state in those counties that comprise the 29- county hospital service area and sub- region. ( For county locations and names, see appendix A.) From Table 3.1, three counties-- Chowan, Perquimans, and Washington— have 5- year ( 2000- 2004) crude rates above 500 per 100,000. This translates to an average of 5 CVD deaths per 1,000 people per year living in those counties. Many counties 2 These rankings are based on calculations made at East Carolina University’s Center for Health Services Research and Development. The data for combined state and regional comparisons are from the National Center for Health Statistics Compressed Mortality Files ( 1999- 2002). 3 Calculations for county comparisons use primary data from North Carolina’s State Center for Health Statistics via University of North Carolina— Chapel Hill’s Odum Institute. Cardiovascular Disease Mortality 3 ECU, Center for Health Services Research and Development, 2007 with relatively high observed crude rates also have relatively small numbers of people and may be proportionally older, which naturally leads to their increased susceptibility to more chronic conditions like CVD. Crude mortality rates are a kind of density measure— the number of deaths normalized ( or divided by) the population of interest and do not account for age structure. Their depiction on maps is for the purpose of focusing the reader to areas where the mortality burden is greatest ( see chapter 1 for more discussion). Maps of crude rates are useful in the development of policy, intervention measures, and determining the allocation of health care resources. The age- adjusted mortality rate maps found at the bottom of figure 3.2 permit comparisons among counties and population groups which may have different age structures ( see chapter 1). The state map shows a sharper distinction in the disparity of county age- adjusted rates between the state’s eastern 41 counties and the remaining counties to the west. As regions, 41- county ENC’s age-adjusted rate of 367.3 is 19% greater that the 59- county region of NC at 308.0 deaths per 100,000 ( see table 3.1). In 2002, the age- adjusted rate for the US was 319.0, less than 2% of the 2000- 2004 rate for the state and less than 13% of ENC’s rate. From another perspective, if ENC 41 had the same mortality rate as RNC 59 during the years 2000 to 2004, 6,590 lives would have been spared from death due to CVD. Figure 3.3 shows age- adjusted mortality by race and sex using the same rate classification cut points found in the age- adjusted map in figure 3.2. These maps provide a visual sense of group contributions to the overall CVD mortality rate and distribution. For white males, the heaviest concentration of high rate counties is found in the east, while some metropolitan counties to the west and a chain of mountain counties tend towards lower rates. Within ENC, the county with the highest rate for white males is Hertford at 576.3 and 131 observed deaths ( table 3.1). High rates are ubiquitous throughout the state for non- white males with the highest found in the ENC county of Currituck at 644.3 and 21 deaths. The highest rates for white females are found scattered throughout ENC with Washington County having the highest rate in this region at 374.6 ( 124 deaths). Currituck County also had the highest rate for non- white females at 528.2 ( 29 deaths). Statewide, ENC is home to the largest concentrations of high rate counties for these four demographic groups. For males of both races there appears to be little difference between ENC and the rest of the state. ENC becomes distinct as a high rate region because of the influence of regional white and non- white female rates. Temporal Distribution of Cardiovascular Disease Mortality The decline in CVD is hinted at in figure 1.2 using the large proportional effects ( 72.1%) of HD mortality as a surrogate. This figure depicts the secular trend in heart disease ( HD) mortality reaching its peak in the 1960s and soon after, crude stroke mortality rates begin to decline. ( Together, these two diseases currently Cardiovascular Disease Mortality 4 ECU, Center for Health Services Research and Development, 2007 comprise more than 90% [ see figure 3.1] of CVD mortality and so gives a good approximation of the patterns of burden and progress made with respect to this disease.) Figure 3.4 is a closer, comparative look at how ENC has been faring over time with respect to CVD mortality over the last two decades of the 20th century and the early years of the 21st. It charts the continuing decline in age-adjusted CVD mortality rates for ENC, the remaining 59 counties of North Carolina ( RNC), North Carolina, and the United States, from 1979 to 2004 ( US: 1979 to 2002). Within the 26- year period, ENC’s annual rates are the highest, followed by the state, the nation, and the remaining 59- county region, each showing very similar patterns of decline. ( The state values are a weighted average between ENC and RNC and will always have intermediate values.) The negative coefficients found in the equations of the lines, listed in the chart ( figure 3.4), show that ENC’s rate of decline is slightly greater than RNC’s rate with the relative gap between the regions’ fitted rates growing from 9% in 1979 to 13% in 2004. This represents a relative 44% increase in regional disparity for CVD mortality. In absolute terms, these same line equations show that the expected or fitted rate differences in age- adjusted death rates declined from 51 deaths per 100,000 in 1979 to 41 deaths per 100,000, which translates into a 24% decrease in regional disparity. Figure 3.5 depicts the 26- year trend of CVD mortality among the four major demographic groups in ENC. It is immediately apparent that the age- adjusted rates are declining for all groups. ENC white males show the greatest decreasing trend-- a decrease of 52%, which on average saves 16.7 lives per annum. This compares favorably to the 42% decline for white females; a saving of 8 lives per year. With R2 values around 0.90 one can make projection into the not- too- distant future with a fair amount of confidence. If the same trends continue, the age- adjusted CVD rates for white males and white females will converge around the year 2015 with an age- adjusted rate of approximately 184 per 100,000. The age- adjusted rates for both non- white men and non- white women are also converging but with their age- adjusted rate trends not projected to converge until sometime around the year 2030, when both non- white sexes attain the rate of approximately 188. In this scenario, it takes non- whites almost 15 years longer to achieve a projected rate similar to that of whites. Recall that the calculations are based on simplifying assumptions concerning the behavior of rates over time and any projections will have an increasing range of error as they move more distant in time from the last observed rate year. However such exercises can be viewed as another way of describing disparities and the amount of relative effort that would be required to achieve parity measured over time. Although mortality due to CVD is declining, its greatest impact is on the county populations of ENC. White males appear to do better in the large metropolitan counties of the Piedmont. However, these lower rates are comparable to the highest rates found in white female population. The highest rates for this latter group are concentrated in the counties of ENC. High rates of mortality for non-white males are nearly ubiquitous within the state, with low rates interspersed in Cardiovascular Disease Mortality 5 ECU, Center for Health Services Research and Development, 2007 the mountain counties. ( Low rates here are probably due to the small numbers of non- whites in this region.) For non- white females, high rates are concentrated in ENC, as well as the south- central portion of the state. Trend analysis covering the period 1979 to 2004 show a dramatic 45% decrease in regional rates for CVD mortality ( figure 3.4). The decrease in the age-adjusted rate for ENC roughly parallels the declining rates for the other regions, but there is a relative increase in regional disparity during this time— an artifact that results from using decreasing bases. When the CVD time series trend line for ENC is broken down into four race- sex trend lines, two patterns emerge: divergence in mortality rates between the two racial groups and convergence between the sexes for each racial group. HEART DISEASE MORTALITY Proportionally, heart disease ( HD) comprises more than 70% of all TCVD deaths for the period 2000 to 2004 ( see figure 3.1). The spatial and temporal patterns of HD mortality, therefore, should correlate strongly to those patterns observed for CVD. Any observable differences in these patterns will probably be due to the effects of stroke mortality, the next largest category outside of HD accounting for almost 20% of all CVD mortality. The ICD- 10 definitions for HD can be found at the end of this section in appendix B. Spatial Distribution of Heart Disease Mortality A comparison of the crude and age- adjusted maps for HD ( figure 3.6) and CVD ( figure 3.2) mortality does show strong similarities in patterns of mortality. ( Note that the cut- points of HD mortality rate categories in the legends for both crude and age- adjusted maps are approximately 70% of the ranges observed for CVD mortality.) The crude map of HD mortality shows concentrations of higher rates in the extreme northeastern and western portions of the state, with smaller concentrations in the southeast and south. Age- adjustment produces a larger concentration of high rates in ENC, de- emphasizing HD mortality rates in the western region of the state. Comparisons of regional age- adjusted HD mortality rates illustrate the continuing presence of geographic disparities. From table 3.2, ENC’s 2000- 2004 age-adjusted rate ( 263.5) is 13% higher than the US rate ( 240.8) and 19% greater than the rate for RNC ( 221.9). The coastal counties of Dare and Pamlico possess the lowest rates at 187.9 ( 286 deaths) and 190.4 ( 174 deaths), respectively. ( For county locations and names, see appendix A) These counties compare favorably to RNC’s rate for the same period. Moving inland, the highest age- adjusted HD rates are found in two county clusters. The first cluster is found in the southern part of the 41- county ENC region. Here, the counties of Bladen ( 319.2), Columbus ( 347.5), Robeson ( 315.6), and Scotland ( 310.4) experience Cardiovascular Disease Mortality 6 ECU, Center for Health Services Research and Development, 2007 12.7% of ENC’s mortality attributable to HD while 10.1% of the region’s aggregated estimated population from 2000 to 2004 lives in those counties. The proportional disparity grows when we move to the next high rate cluster of counties found in the northern part of the region. The high rates for Beaufort ( 309.1), Edgecombe ( 305.5), Martin ( 311.0), and Washington ( 314.8) counties comprise 8.2% of the region’s HD deaths, but comprise only 5.7% of the region’s population. Given their respective populations sizes, these two county clusters have a disproportionate share of ENC’s HD mortality. 4 Figure 3.7 depicts the spatial distribution of age- adjusted mortality rates for HD ( 2000- 2004) broken down into four race- sex groups. The observed spatial patterns closely resemble those for CVD ( figure 3.3) and indicate similar regional effects among the four groups: higher rates for females of both racial groups are again more concentrated in the eastern portion of the state, while high white male rates are found throughout the state with the exception of the Piedmont’s metropolitan counties, and non- white male rates are ubiquitously high with the exception of several counties in the west. From table 3.2, the highest regional age- adjusted county rate for white males is Columbus at 424.1 with 335 dying from HD over five years. For the same period, Washington County is the deadliest for white females who experience 96 HD deaths and an age- adjusted rate of 294.0 per 100,000. Non- white males experience their highest rate of age-adjusted HD mortality in Currituck County at 480.7 per 100,000 but this is the result of only 16 individuals dying during that period— Perquimans County has the next highest rate at 441.1 and a more statistically stable death count of 32. In Columbus County, 189 non- white females died from HD producing the highest age- adjusted county rate of 339.1 during the years 2000 to 2004. The total age-adjusted HD mortality rate Columbus County is weighted largely by deaths contributed from females of both racial groups, although white males also make a significant contribution. The high CVD rate experienced by non- white males in Edgecombe County appears to be heavily influenced by the HD component for this race- sex group. Within ENC, the lowest statistically reliable age- adjusted rate for any race- sex group is that found for white females in Greene County at 165.0. Temporal Distribution of Heart Disease Mortality Figure 3.8 shows trend lines for age- adjusted HD mortality among the four regions for the period 1979 to 2004. The slope of the lines all follow the same pattern of decline observed in figure 3.4 for CVD. Closer observation shows, however, that with the exception of the ENC trend line, the relative positions of the other three regions have shifted slightly. For CVD ( figure 3.4), North Carolina has been consistently above the US rate, but for HD the state emerges 4 Because age- adjusted rates can be used for making comparisons, they can be helpful in targeting areas where problems might exist. In this case, two county clusters have been identified and their count data are used to create proportions, which can be used to calculate the relative amount of mortality burden. Cardiovascular Disease Mortality 7 ECU, Center for Health Services Research and Development, 2007 with rates slightly less than the nation. ( This is probably due to the impact of stroke mortality in ENC, which tends to be higher and has a significant additive effect to the state rate for CVD.) Rates for RNC have been consistently below the declining trend for the US, whereas for CVD the trend lines closely matched one another. The impact of HD mortality on RNC’s population is less than it is for the nation as a whole. ENC’s age- adjusted mortality rates for HD are clearly higher throughout the 26- year time series with a slightly greater rate of decrease among all the regions. The pattern of HD mortality decline witnessed here is a good example of the secular trend in HD mortality burden observed during the 20th century ( see figure 1.2). Both observed and modeled trend lines for race- sex groups ( figure 3.9) show patterns of decline similar to CVD ( figure 3.5). What emerges in the pr |
OCLC number | 300039044 |