Cross-classifying by state of birth and state of residence, we find that excess risk of stroke mortality associated with living in the Southern SB is apparent even among people who were born in the SB but no longer lived there in adulthood. Individuals who were born outside of the SB but lived there in adulthood appear to have modestly elevated odds of stroke mortality compared to individuals who did not reside in the SB at birth or in adulthood. The greatest elevation in odds of stroke mortality was evident in those exposed to the SB at birth and in adulthood. This pattern was evident in all 3 years examined—1980, 1990, and 2000—despite substantial secular declines in absolute stroke mortality rates. SB exposure was associated with excess mortality among men and women; black and white subjects; and those younger or older than 60. In the year 2000, education appeared to contribute but not fully explain the effect of SB birth. These results suggest that SB birth and adulthood both contribute to excess stroke risk for black and white subjects, but for white subjects the contribution of SB birth predominates.
The current analyses are limited by the geographic and temporal detail available in the census and death certificates. There is substantial heterogeneity in stroke risk within states.25
SB exposure of recent migrants may be misclassified. In the 2000 census, approximately 93% of SB residents reported living in the SB 5 years previously (calculations from Integrated Public Use Microdata Series 5% sample).20
Similarly, many individuals do not reside long in their birth state; thus state of birth may not accurately represent childhood exposure. However, data from the 1940 Census 1% microsample indicates that the majority of people born in the SB likely resided in the SB at least through childhood. For example, 88% of 15-year-olds lived in their state of birth.20
Misclassification may also result from incorrect recall of birth state. For the education-adjusted analyses, we used only 3 levels of education to facilitate data linkage, but this rough measure fails to capture the full continuous range of education or multiple dimensions of education quality that probably historically differ in SB states.26,27
Cause of death data on mortality records are imperfect and classification rules change over time.28
The declines in stroke mortality rates evident here may be misleading if the cause of death coding patterns have evolved to give greater emphasis to underlying conditions such as atherosclerosis or diabetes rather than stroke. It is difficult to rule out this possibility with the current data. However, the pattern of reporting changes would have to be highly complex in order for this potential bias to account for our primary finding that childhood and adult residence appear to independently contribute to stroke risk. Although cause of death classification norms may differ regionally, we expect this would be patterned by place of death, not place of birth. Any possible bias from regional differences in cause of death recording should have similar effects for the group who were born and died in the SB as for the group who were born outside the SB but died in the SB. We also have no information on major stroke risk factors, stroke subtypes, or nonfatal stroke, precluding consideration of a number of interesting etiologic questions.29,30
Analyses of ecologic data routinely face 2 important methodologic challenges: avoiding ecologic fallacies and disentangling contextual from compositional effects. The ecologic fallacy arises when one infers individual level relationships from ecologic relationships; however, in the current study the primary independent variables are ecologic exposures. This obviates the risk of ecologic fallacy because there are no conclusions regarding individual level variables. The concern regarding distinguishing contextual from compositional effects is key, and our analyses cannot address this. People born in SB states may be more vulnerable to stroke because of individual genetic or behavioral risk factors (compositional effects) or because of environmental toxins or toxic social dynamics (contextual effects). Our results cannot pinpoint a specific explanation, but they rule out explanations that would not affect early life (e.g., acute medical care). There are both contextual and compositional explanations that might operate from childhood forward, although we note that compositional explanations relating to race are not adequate. Selective migration also complicates interpretation of these results.31
In the 2000 census, SB in-migrant black subjects had higher socioeconomic status in terms of education, income, and occupational characteristics than all other black subjects. Similarly, SB in-migrant white subjects were of higher socioeconomic status than other white subjects (details available from the authors, estimates based on reference20
). The excess stroke risk among the in-migrants compared to the non-migrants may therefore be underestimated in our analyses. The effect estimates for the SB born (, model 1) are the least vulnerable to bias arising from selective migration.
Our findings confirm prior evidence on the persistence of the US SB and extend these results by demonstrating that the timing of SB exposure may influence the risk conferred.1,32
Although most prior SB work focuses on state of residence at stroke onset to define SB exposure,33,34
several studies indicate that place of birth and childhood residence may influence stroke risk. The importance of place of birth has been shown in the United Kingdom35
and, more directly related to the SB, in a study of South Carolina residents4
and one national US study.16
Our findings differ from these studies in suggesting that although either exposure at birth or in adulthood elevates risk of fatal stroke, the greatest risk is among individuals who resided in the SB at both time points. Our results also extend prior evidence demonstrating a geographic patterning of circulatory disease mortality among black subjects stratified by migration status17
by focusing specifically on the SB states, showing temporal stability of the patterns through 2000, and presenting models both with and without stratification by place of death. The only other national analyses incorporating place of birth demonstrated the importance of nativity but were limited in that neither focused on the SB region.17,18
Prior research has considered whether racial disparities in stroke arise in part because black subjects live in higher risk areas than white subjects.36
Our results suggest that the most important geographic characteristic is place of residence in childhood. In 1980 and 1990, the OR associated with adult SB residence for black subjects was greater than the OR associated with adult SB residence among white subjects, but this appeared to be largely because nearly all black adult SB residents were born in the SB, while a substantial fraction of white adult SB residents were born outside the SB. The ORs associated with SB birth is similar for black and white subjects.
These results suggest independent risk associated with childhood or adult SB residence. Place of residence may affect stroke through access to medical care, physical risks associated with environmental conditions, social norms affecting behaviors, socioeconomic conditions created by local macroeconomic factors, or psychosocial pathways stemming from features of social organization in communities.37
The finding that the harm of SB residence accumulates over life implicates health-promoting resources such as preventive medical care, social conditions, or environmental exposures. Acute medical treatments provided after stroke onset, which would more likely be predicted by place of residence at stroke onset, rather than place of birth or childhood residence, are unlikely explanations. Behavioral norms regarding diet, physical activity, and smoking are profoundly influenced by childhood social conditions,38
and many of these behaviors appear to be largely determined before adulthood. Such risk factors may affect prevalence of silent infarcts,39
increasing vulnerability to clinical strokes among Southerners much later in life.
These results do not support genetic explanations because the phenomenon appears to affect both black and white subjects. Given the large fraction of northern African Americans whose families immigrated from the South during the 20th century,40
a genotype shared by southern black and white subjects but not northern black subjects seems unlikely.
Our understanding and ability to eliminate the US SB may be advanced by focusing on the timing of exposure. Future research should address whether there is truly a “dose” phenomenon, in which the harm accumulates with increasing exposure, or if the timing of exposure at specific critical periods is of greater importance. Our results suggest that conceptualizing SB exposure based on place of current residence is inadequate.