| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
RESEARCH AND PRACTICE |
At the time of the study, Adam Karpati and Sandro Galea were with the Department of Population and International Health, Harvard School of Public Health, Boston, Mass. Adam Karpati is also with the Epidemiology Program Office, Centers for Disease Control and Prevention, Atlanta, Ga. Sandro Galea is also with the Center for Urban Epidemiologic Studies, New York Academy of Medicine, New York. Tamara Awerbuch and Richard Levins are with the Department of Population and International Health, Harvard School of Public Health.
Correspondence: Requests for reprints should be sent to Adam Karpati, MD, MPH, Bureau of Community HealthWorks, New York City Department of Health, 40 Worth St, Room 1607, New York, NY 10013 (e-mail: aek3{at}cdc.gov).
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMETHODSRESULTSDISCUSSIONReferences |
|---|
Objectives. We examined variability in disease rates to gain understanding of the complex interactions between contextual socioeconomic factors and health.
Methods. We compared mortality rates between New York and California counties in the lowest and highest quartiles of socioeconomic status (SES), assessed rate variability between counties for various outcomes, and examined correlations between outcomes’ sensitivity to SES and their variability.
Results. Outcomes with mortality rates that differed most by county SES were among those whose variability across counties was high (e.g., AIDS, homicide, cirrhosis). Lower-SES counties manifested greater variability among outcome measures.
Conclusions. Differences in health outcome variability reflect differences in SES impact on health. Health variability at the ecological level might reflect the impact of stressors on vulnerable populations. (Am J Public Health. 2002;92:1768–1772)
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONReferences |
|---|
Analyses of community factors attempt to elucidate how context affects the health of individuals.4 Although multilevel analysis allows statistical determination of the relative effect of individual and community factors,5 the manner in which these measures exert their effects on public health is likely to be more complex than is suggested by generalized multilevel linear models.6,7 A more accurate understanding of the interplay between individuals and their environments requires construction of models that take into account our knowledge of interactions on various levels, contextual and otherwise, and the fact that system components are interconnected and likely display feedback loops.8–10
One approach to understanding complex systems is to examine variability among their components. Variability refers to the extent to which a characteristic of a complex system (e.g., heart rate or stock prices) changes over time or space. Variability in a complex system might reflect the effect of external influences ("stressors") through their interaction with the system’s homeostatic mechanisms.11
Most evaluations of variability in complex physiological systems have been done in the context of individual clinical characteristics. For example, a decrease in heart rate variability has been shown to predict mortality after myocardial infarction.12 For public health surveillance or for epidemiological analysis, variability in population health or its determinants may be a more informative characteristic than the absolute level of particular components. Similarly, for policy or program evaluation, variability might be a useful measure of the relative effects of different interventions.
We studied the relation between contextual effects and population health outcomes by examining mortality rates associated with several conditions in counties in New York and California. We hypothesized that, first, certain diseases or health outcomes (e.g., traumatic events or communicable diseases) are more sensitive to population socioeconomic factors than are others, reflecting the degree to which those outcomes are avoidable or preventable. Second, the rates of the outcomes that are most sensitive to socioeconomic factors also vary the most among counties, reflecting the wide distribution of responses to the stressors to which populations are exposed.
|
METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONReferences |
|---|
Data
We used New York State Department of Health and California Department of Health Services data to obtain age-adjusted mortality rates in each county for the various outcomes.13,14 Table 1
presents the mortality rates for the outcomes studied. We selected outcomes on the basis of data availability, range of clinical conditions, and consistency with previously published studies. Rates for New York were from 1997; rates for California were either from 1997 alone or were an average of rates from 1995 to 1997.
|
Analysis
We analyzed counties in California and New York separately. We stratified counties into quartiles by each of the SES measures, calculated the average rate of each health outcome for the bottom and top quartiles, and obtained the rate ratio for a given health outcome by comparing counties in the lowest and highest socioeconomic quartiles.
We also calculated the variability of each health outcome across all counties in each state. Following Levins and Lopez, we used the range of values divided by the mean value as the measure of variability; this measure provides a useful estimate for qualitative analyses.11 The larger the range divided by the mean value, the higher the variability of a particular health outcome across counties. No statistical inferences were based on this measure of variability. We calculated Pearson correlation coefficients between rate ratios and variability measures and examined variability in outcomes across counties, stratified by SES.
We also calculated smoothed countyspecific rates, in which the observed rate in a county was "stabilized" by replacing it with the weighted average of the county rate and all adjacent county rates; weights were proportionate to population size.18 We repeated all of the analyses described here on the smoothed rate estimates. Finally, we compared outcome rankings and correlations derived using the range-divided-by-mean measure with rankings obtained using 2 other variability measures: interquartile range divided by mean, and the coefficient of variation (SD / mean x 100%).
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONReferences |
|---|
shows the mean, range, and range divided by mean across counties in both states. In New York, the largest variability in outcomes was in AIDS mortality (range / mean = 10.9), followed by homicide (6.3), all-cause mortality among persons aged 10–24 years (2.6), and mortality from cirrhosis (2.0). The smallest variability was observed in all-cause mortality across all ages (0.3) and among persons aged more than 75 years (0.4), as well as in mortality from neoplastic disease (0.5) and mortality from cardiovascular disease (0.8).
In California, variability was highest for AIDS (range / mean = 7.3), followed by homicide (2.6), mortality from cirrhosis (2.4), and mortality among persons aged 15–24 years (2.1). The lowest variability was in rates for all-cause mortality across all ages and for persons aged more than 75 years as well as mortality from neoplastic disease (0.5 for each) and mortality from cardiovascular (0.8) or cerebrovascular (0.9) disease. The ordering of diseases by their intercounty variability was similar between the 2 states.
Rate ratios comparing mean disease-specific mortality rates between counties in the lower and upper quartiles of various socioeconomic markers are shown in Table 2. Rate ratios greater than 1.0 imply that counties with lower SES have higher disease-specific mortality than do those with higher SES.
|
In California, ratios ranged from 0.40 to 2.61. The mean ratio across economic indicators for neoplastic disease rates was 1.05 (0.90 for female breast cancer and 1.12 for lung cancer). The highest mean ratios were for mortality rates from motor vehicle accidents (2.03) and for homicide rates (1.85). The mean ratio for cirrhosis was 1.27. All-cause mortality for persons aged more than 75 years, suicide, pneumonia and influenza mortality, and female breast cancer mortality each had rate ratios of less than 1.10 for all economic markers. In addition, ratios for AIDS mortality ranged from 0.40 to 0.88. Mortality from all causes in persons aged 10–24 years had a mean rate ratio of 1.31, whereas for persons aged more than 75 years the mean ratio was 0.93. The mean ratios for all outcomes across economic measures ranged from 1.13 (high school graduation rate) to 1.27 (percentage of persons aged < 18 years in poverty).
In both states, the variability (range / mean) of health outcomes across counties was strongly correlated with the mean ratio of rates between counties in the lowest and highest quartiles of economic status (measured by percentage of children < 18 years living in poverty). In New York, the Pearson correlation coefficient was 0.97; in California, it was 0.70 (0.01 when AIDS was included in the calculation).
Figure 1 shows the relation between the homicide rates in New York counties and their socioeconomic status. Homicide rates for each county were plotted against tertiles of poverty (percentage of persons aged < 18 years living in poverty). Counties of lower economic status displayed greater variability in their rates of homicide than do counties with high economic status. The general trend of the relation was linear, with a positive slope; however, among counties with the lowest economic status, there were both low and high rates.
|
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONReferences |
|---|
All-cause mortality and certain neoplastic disease mortality rates did not differ greatly between poorer and wealthier counties. The underlying mechanisms for these health outcomes might account for the findings. Diseases with an incidence or course that is influenced by behavioral or environmental factors would be expected to exhibit sensitivity to SES, whereas diseases with genetic or other nonmodifiable causes would not. Our analysis is consistent with earlier findings and builds on previous small-area analyses in Kansas, Saskatchewan, and Cuba.11
A general process underlying these observations has been articulated by Link and Phelan, who postulated that access to protections and avoidance of harms underlie health outcomes that are sensitive to SES.19 For example, in California, lung cancer mortality (largely a consequence of smoking) had a higher mean ratio than did female breast cancer mortality. In New York, cirrhosis mortality (largely a consequence of alcohol misuse) had a higher rate ratio than did neoplasms. A comparison of age-specific mortality further supports this hypothesis. In New York and California, county rates between economic strata exhibit low ratios for mortality among older persons and high ratios for youth mortality. Youth mortality has more potentially modifiable behavioral and social causes at its root than does mortality among older persons.
The economic sensitivity of AIDS was reversed in New York and California. In New York State, counties in the lowest economic quartiles had an average of 2.96 times the AIDS rates of counties in the highest quartiles, and AIDS is particularly prevalent in poor communities of New York City. By contrast, in California, the populations with the lowest SES are both urban and rural, and AIDS incidence is more widely distributed in populations of varying SES.
Our principal measures of interest were variabilities in outcomes rather than absolute rates. Variability in mortality rates across counties was highest for the outcomes with larger SES rate ratios. AIDS mortality and homicide were the outcomes with the largest variability in the 2 states. Rates for all-cause mortality across ages and for mortality in older persons, as well as rates for neoplastic disease mortality, exhibited small variability and were generally not sensitive to socioeconomic conditions.
Although we hypothesized that variability reflects system-specific conditions (i.e., the balance among vulnerability, stressors, and protectors), variability may also be the product of random events, especially when the outcome of interest is rare or the population within which it occurs is small. Moreover, there may be confounding of the SES–sensitivity/variability relationship if rare events are also more sensitive to SES. To address these possibilities, we repeated all analyses using smoothed county-specific rates, which reduce intercounty population variability, as well as robust measures of variability, which have low sensitivity to outliers. The observed rate variability was attenuated under these circumstances, but trends in health outcome variability and associations with county-level SES were preserved.
In their examination of ecological factors contributing to adverse health effects, Levins and Lopez suggested that the relation between economic deprivation and variability in health status might be mediated by the vulnerability of populations.11 They cited an observation by a geneticist, I. I. Schmalhausen, that "a system at the boundary of its tolerance along any dimension of its existence is more vulnerable to small differences in circumstance along any dimension."20 Populations enduring social or economic deprivation will be more vulnerable to potential stressors than will populations of higher status. Thus, acute outbreaks of infectious diseases, environmental risks, or transient gaps in public health services will likely affect an economically deprived population to a greater degree than they would a less marginalized population.
We note, however, that these external stressors are not uniformly distributed across all disadvantaged communities, and therein might lie the source of the observed variability. Although vulnerability might result from chronic economic deprivation, the range of adverse health outcomes will depend on the degree to which each community experiences stressors and the distribution within communities of counteracting protective factors.
Variability in biological systems is increasingly seen as a marker for stresses to systems in homeostasis. Applying this insight to communities and health, we postulate that economic deprivation produces vulnerability to stressors whose nonuniform distribution across populations manifests as variability in health outcomes. One possible implication of this model is that interventions to improve public health might exert the greatest effect not by targeting particular stressors, but rather by focusing on improving general social and economic well-being, thus reducing populations’ overall vulnerability.
|
Acknowledgments |
|---|
|
Footnotes |
|---|
Accepted for publication January 10, 2002.
|
References |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONReferences |
|---|
2. O’Campo P, Xue X, Wang MC, Caughy M. Neighborhood risk factors for low birthweight in Baltimore: a multilevel analysis. Am J Public Health. 1997;87:1113–1118.
3. Yen IH, Kaplan GA. Neighborhood social environment and risk of death: multilevel evidence from the Alameda County Study. Am J Epidemiol. 1999;149:898–907.
4. Susser M. The logic in ecological, I: the logic of analysis. Am J Public Health. 1994;84:825–829.
5. Diez-Roux AV. Bringing context back into epidemiology: variables and fallacies in multilevel analysis. Am J Public Health. 1998;88: 216–222.
6. Philippe P, Mansi O. Nonlinearity in the epidemiology of complex health and disease processes. Theor Med Bioeth. 1998;19:591–607.[Medline]
7. Goldenfeld N, Kadanoff LP. Simple lessons from complexity. Science. 1999;284:87–89.
8. Sandberg S, Awerbuch TE, Gonin R. Simplicity vs complexity in deterministic models: an application to AIDS data. J Biol Systems. 1994;4:61–81.
9. Koopman JS, Longini IM Jr. The ecological effects of individual exposures and nonlinear disease dynamics in populations. Am J Public Health. 1994;84:836–842.
10. Seely AJ, Christou NV. Multiple organ dysfunction syndrome: exploring the paradigm of complex nonlinear systems. Crit Care Med. 2000;28:2193–2200.[Medline]
11. Levins R, Lopez C. Toward an ecosocial view of health. Int J Health Serv. 1999;29:261–293.[Medline]
12. Kleiger RE, Miller JP, Bigger JT Jr, Moss AJ. Decreased heart rate variability and its association with increased mortality after acute myocardial infarction. Am J Cardiol. 1987;59:256–262.[Medline]
13. New York State Department of Health. Available at: http://www.health.state.ny.us/nysdoh/research/research.htm. Accessed April 2, 2002.
14. California Department of Health Services. Available at: http://www.dhs.ca.gov/hisp. Accessed April 2, 2002.
15. U.S. Bureau of the Census. Available at http://www.census.gov/hhes/www/saipe/estimatetoc.html. Accessed April 2, 2002.
16. Dalaker J. Poverty in the United States: 1998. Washington, DC: US Census Bureau, Current Population Reports; 1999. Series P60-207.
17. Cubbin C, Pickle LW, Fingerhut L. Social context and geographic patterns of homicide among US black and white males. Am J Public Health. 2000;90:579–587.
18. Devine O, Parrish RG. Monitoring the health of a population. In: Stroup DF, Teutsch SM, eds. Statistics in Public Health: Quantitative Approaches to Public Health Problems. New York, NY: Oxford University Press; 1998:58–91.
19. Link BG, Phelan JC. Evaluating the fundamental cause explanation for social disparities in health. In: Bird CE, Conrad P, Fremont AM, eds. Handbook of Medical Sociology. 5th ed. Upper Saddle River, NJ: Prentice Hall; 2000:33–46.
20. Schmalhausen II. Factors of Evolution. Philadelphia, Pa: Blakiston Company; 1949.
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
