(Stroke. 1995;26:1150-1152.)
© 1995 American Heart Association, Inc.
Articles |
The End of the Stroke Belt?
It May Be Too Early to Declare Victory!
From the Departments of Public Health Sciences (G.H.) and Neurology (G.H., V.J.H.), Bowman Gray School of Medicine of Wake Forest University, Winston-Salem, NC.
Correspondence to George Howard, DrPH, Department of Public Health Sciences, Bowman Gray School of Medicine of Wake Forest University, Medical Center Blvd, Winston-Salem, NC. E-mail howard@phs.bgsm.wfu.edu.
Key Words: epidemiology • geography • cerebrovascular disorders • mortality
 |
Introduction |
|---|
 Top Introduction A Review of the...The Case for the...References |
|---|
Although there is little debate about whether there has been excess stroke mortality in the southeastern United States,1 2 3 4 5 6 7 8 9 10 there is a growing healthy scientific debate about whether this "Stroke Belt" is dissipating or persisting. With both scientific involvement as cerebrovascular epidemiologists and personal involvement as long-term residents, we greet news of the potential dissipation of the Stroke Belt with great enthusiasm. However, describing the Stroke Belt reminds one of the proverb of blind men probing different parts of the elephant and trying to describe the total animal. Which part of the beast one gropes changes the perception of the animal. Similarly, different methodological approaches to describing the Stroke Belt and temporal changes in this region greatly influence interpretation. Not that different approaches are "wrong"; rather, differences must be appreciated to understand the nature of the beast. As such, the strengths and weaknesses of different methodological approaches should be reviewed.
|
A Review of the Case for the Dissipation of the Stroke Belt |
|---|
|
TopIntroductionA Review of the...The Case for the...References |
|---|
In this issue of Stroke Lanska and Peterson1 report an analysis of state mortality rates and conclude that "the long-standing excess of stroke mortality in the southeastern United States, and particularly the dense concentration of excess stroke mortality along the Atlantic coastal plain, is beginning to dissipate." The authors also note that the rate of decline in stroke mortality rates has been highest in the South and that the magnitude of the difference between high- and low-rate areas has become much smaller. That there is a "shifting" of high stroke rates from the Piedmont region of the Southeast to the Mississippi River valley is also supported by recent publications from Casper and coworkers.2 These recent reports seem to support the position that the Stroke Belt in the southeastern United States is dissipating; however, the interpretation of these results should be carefully reviewed.
A fundamental decision in the analysis of these data is the choice of the unit of analysis. Some investigators report mortality rates on a state basis,1 3 4 5 6 7 whereas others have reported data by groups of counties that relate to each other economically (State Economic Areas).2 8 In this issue of Stroke we report mortality rates in an arbitrary grouping of counties on the coastal plain of North Carolina, South Carolina, and Georgia compared with the remainder of the United States.9 Importantly, all these approaches represent arbitrary representations of the mortality data, with there being no reason to assume that stroke mortality rates will follow arbitrary political boundaries of either states, counties, or arbitrary regions defined by geology.
The major advantage of analyses based on state data is adequate sample size to provide stable estimates of stroke mortality rates. However, analyses based on State Economic Areas show substantial heterogeneity within many states.2 8 Furthermore, systematic differences have been documented in North Carolina, with stroke rates increasing from west to east (from the mountains to the coast).10 If there is substantial heterogeneity within a state, reporting on a state basis provides nonrepresentative estimates for the high and low regions within the state.
This potential heterogeneity of stroke mortality rates further complicates the analysis of temporal changes, particularly if there have been substantial population shifts between regions within the state with different mortality. For example, consider reporting temporal changes for the state of North Carolina. North Carolina has three broad (and substantially different) geographic regions: the coastal plain, the Piedmont, and the mountains. As noted, the stroke mortality rates are higher in the coastal plain than in either the mountains or the Piedmont.10 The five largest cities in North Carolina (Charlotte, Greensboro, Raleigh, Durham, and Winston-Salem) are all in the Piedmont region. In 1960 22% of the North Carolina population was in the counties containing these cities (Mecklenburg, Guilford, Wake, Durham, and Forsyth counties, respectively).11 In the same year an arbitrarily selected group of counties on the southern coastal plain of North Carolina chosen to be from the region with a higher stroke risk2 8 10 had a population representing 16% of the state population (the counties of Beaufort, Bladen, Brunswick, Carteret, Columbus, Craven, Duplin, Greene, Jones, Lenoir, New Hanover, Pamlico, Pender, Robeson, Sampson, Wayne, and Wilson).11 Between 1960 and 1990 there was substantial growth in urban areas in North Carolina, so that these five large urban areas increased to 27% of the state total, while the Stroke Belt region decreased to 14% of the state total.12 This shift of the proportions of the population from the high stroke mortality region to the lower stroke mortality region produces statewide estimates of temporal changes in stroke rates that are confounded by the changes in population size, with increasing weight given to the lower stroke mortality regions in later years.
There are other substantial differences between these regions in North Carolina. The urban Piedmont region has a population that is more mobile than the more rural coastal plain region. For example, in 1990 62% of the black and white residents of the five major urban counties were North Carolina natives compared with 75% in the 17 (arbitrarily chosen) eastern counties.13 If the stroke risk in the Stroke Belt increases with exposure to the underlying (and unknown) risk factors, coastal plain counties would appear to have higher risk, even given a constant increase in risk per year of exposure. That is, even if lifetime residents of North Carolina all have the same (elevated) stroke risk, the event rate would be lower in the Piedmont because of the greater dilution from nonnatives. These differential in-migration rates may contribute to the observed regional differences in stroke mortality rates, which would be obscured by analysis on a state basis. Importantly, analyses of temporal changes are further confounded by a trend for increasing in-migration rates, leading to a trend for an increase in the number of residents not having long-term exposure to risk factors underlying the Stroke Belt. This trend would be reflected by estimating a decrease in stroke mortality rates over time that is confounded with an increasing in-migration rate.
For both reasons, the rate of decrease in the stroke rate in North Carolina would be overestimated and perhaps lead to a false conclusion that the Stroke Belt is disappearing. Importantly, the same factors are likely operating in other southern states. In South Carolina and Georgia the highest stroke risk is in the coastal plain and southern regions of Georgia.2 8 However, most of the urban regions (more rapid growth and higher in-migration rates) in South Carolina are western regions (Greenville/Spartanburg, the Rock Hill area south of Charlotte, and Columbia to the west but Charleston in the east). Similarly, in Georgia the Atlanta region is (by far) the largest urban region, but it is north and west of the high-stroke area. Therefore, analyses on a state basis may falsely overestimate the rate of the decreased risk in the southeastern states.
Lanska and Peterson1 also modeled temporal changes using a model with a linear term for time. As such, absolute rather than relative changes in stroke rates are reflected in the analysis. While the relationship between stroke rates and year was "approximately linear," this model assumes a constant annual change in stroke rates. Given the dramatic national decline in stroke mortality,14 it may be more reasonable to expect the differences in stroke mortality between any groups to change proportionally, thereby potentially reflecting a constant relative risk of stroke between groups. For example, suppose there are two hypothetical regions (A and B). In region A stroke rate declined from 100 deaths per year to 50 deaths per year over time (a 50% decrease). Suppose the second region began at 200 deaths per year. Then what would an "equal" decrease be? An absolute equal decrease would be to 150 deaths, but this would represent only a 25% decrease. Conversely, a decline to 100 deaths per year would be an identical 50% decrease as in region A but would be larger on an absolute basis than in region A. If the proportional decline in stroke rates is the appropriate model but a linear model is used, the estimates at the "ends" of the data (first and last years) will be overestimated and underestimated, respectively, and the extent of this underestimation will be proportional to the percent decline. Therefore, if the South is decreasing at a faster percentage, the linear model will underestimate the stroke rate more than for other regions with slower percent decreases. Furthermore, if the proportional model is appropriate, it would be expected that "differences between high-rate areas and low-rate areas have lessened dramatically," as observed in the above example.
|
The Case for the Persistence of the Stroke Belt |
|---|
|
TopIntroductionA Review of the...The Case for the...References |
|---|
In this issue of Stroke we report that the relative risk of stroke death for residents of the southeastern coastal plain has remained remarkably constant between 1968 and 1991 for black and white men and women.9 The persistence of this elevated stroke risk in the coastal plain region of the Southeast is also supported by the analyses based on State Economic Areas, where the region is consistently in the highest rate group for men and women and for blacks and whites.2 8 Hence, the persistence of the Stroke Belt in the coastal plain region of the southeastern states appears clear.
Equally clear is the increased stroke mortality in the Mississippi/Ohio river basin.2 8 This increased stroke mortality underlies the reported westward shifting of the Stroke Belt and its northern shifting into Indiana.2 8 14 There is also marginal decline of stroke rates in the Piedmont region of the Southeast. Importantly, it is unknown whether the decline in risk in this region applies to long-term residents of the Piedmont region or is an artifact of increased in-migration rates. Migration data, on an individual basis, will be required to address this important concern. However, we urge caution in declaring the decline in relative stroke risk in the Piedmont region a medical success until the extent of the effect attributable to in-migration can be more completely understood.
That the clustering of high stroke rates is on the coastal plain further complicates the understanding of the underlying factors for the Stroke Belt. As we discussed in the article in this issue,9 none of the 81 counties selected in the National Health and Nutrition Examination Survey (NHANES)15 are in the coastal plain region. The four counties in North Carolina (Forsyth, Guilford, Mecklenburg, and Granville) and the single South Carolina county (Chester) are in the western-central area, with the Stroke Belt lying to their east. The four counties selected in Georgia (Cobb, Clayton, Spalding, and Twiggs) are in the north-central part of the state, with the Stroke Belt to their south. Hence, we would suggest that while there are data describing risk factors in Stroke Belt states, there are not national data from the high-stroke regions within the states.
The increased relative stroke risk in the Mississippi/Ohio river region should be a matter of national concern. However, the risk in the coastal plain of the Southeast appears persistent, and we remain cautious in declaring that risk is declining in the Piedmont region. As such, we encourage continued study of the persistent Stroke Belt in the southeastern United States.
|
References |
|---|
|
TopIntroductionA Review of the...The Case for the...References |
|---|
1. Lanska DJ, Peterson PM. Geographic variation in the decline of stroke mortality in the United States. Stroke. 1995;26:1159-1165.
2.
Casper ML, Wing S, Anda R, Knowles M, Pollard RA.
The shifting Stroke Belt: changes in geographic pattern of
stroke mortality in the United States. Stroke. 1995;26:755-760.
3. Borhani NO. Changes and geographic distribution of mortality from cerebrovascular disease. Am J Public Health. 1965;55:673-681.
4.
Lanska DJ. Geographic distribution of stroke
mortality in the United States: 1939-1941 to 1979-1981.
Neurology. 1993;43:1839-1851.
5. Feinleib M, Ingster L, Rosenberg H, Maurer J, Singh G, Kochanek K. Time trends, cohort effects, and geographic patterns in stroke mortality: United States. Ann Epidemiol. 1993;3:458-465. [Medline] [Order article via Infotrieve]
6.
Lanska DJ, Kryscio R. Geographic distribution
of hospitalization rates, case fatality, and mortality from stroke in
the United States. Neurology. 1994;44:1541-1550.
7. Lanska DJ, Peterson PM. Effects of interstate migration on geographic distribution of stroke mortality in the United States. Stroke. 1995;26:544-561.
8.
Wing S, Casper M, Davis WB, Pellom A, Riggan W,
Tyroler HA. Stroke mortality maps: United States whites aged
35–74 years, 1962–1982. Stroke. 1988;19:1507-1513.
9.
Howard G, Evans GW, Pearce K, Howard VJ, Bell RA,
Mayer EJ, Burke GL. Is the Stroke Belt disappearing? An
analysis of racial, temporal, and age effects.
Stroke. 1995;26:1153-1158.
10.
Heyman A, Tyroler HA, Cassel JC, O'Fallon WM,
Davis L, Muchbaier L. Geographic differences in mortality from
stroke in North Carolina, I: analysis of death
certificates. Stroke. 1976;7:41-45.
11. Bureau of the Census. The 1960 Census of the Population: Characteristics of the Population: North Carolina. Washngton, DC: US Dept of Commerce, Bureau of the Census; 1960.
12. Bureau of the Census. The 1990 Census of the Population: Characteristics of the Population: North Carolina. Washington, DC: US Dept of Commerce, Bureau of the Census; 1990.
13. Bureau of the Census. The 1990 Census of the Population: Social and Economic Characteristics: North Carolina, Sections 1 and 2. US Dept of Commerce, Bureau of the Census; 1993.
14. Thom T. The Stroke Belt: mortality since 1920. Presented at the 35th Annual Conference on Cardiovascular Disease Epidemiology and Prevention; March 8-11, 1995; San Antonio, Tex. Circulation. In press. Abstract.
15. Plan and Operation of the Third National Health and Nutrition Examination Survey, 1988-94. Hyattsville, Md: US Dept of Health and Human Services, Public Health Service, Centers for Disease Control and Prevention, National Center for Health Statistics; 1994. Series 1, Programs and Collections Procedures, No. 32.
This article has been cited by other articles:
 |
|
 |
|
  D. T. Lackland, D. L. Bachman, T. D. Carter, D. L. Barker, S. Timms, and H. Kohli The Geographic Variation in Stroke Incidence in Two Areas of the Southeastern Stroke Belt : The Anderson and Pee Dee Stroke Study Stroke, October 1, 1998; 29(10): 2061 - 2068. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
L. W. Pickle, M. Mungiole, and R. F. Gillum Geographic Variation in Stroke Mortality in Blacks and Whites in the United States Stroke, August 1, 1997; 28(8): 1639 - 1647. [Abstract] [Full Text]
|
|
|
|
|
|
D. J. Lanska and P. M. Peterson Comparison of Additive and Multiplicative Models of Regional Variation in the Decline of Stroke Mortality in the United States Stroke, June 1, 1996; 27(6): 1055 - 1059. [Abstract] [Full Text]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||