Long-Term Survival After Acute Myocardial Infarction Is Lower in More Deprived Neighborhoods
http://www.100md.com
《循环学杂志》
the Departments of Environmental Health (C.T., J.S., S.M., H.S.) and Epidemiology (C.T., J.S., M.M.)
Harvard School of Public Health, Boston, Mass; Beth Israel Deaconess Medical Center, Boston, Mass (M.M.)
University of Massachusetts Medical School, Worcester (R.G.).
Abstract
Background— As part of the Worcester Heart Attack Study, a community-wide study examining changes over time in the incidence and long-term case-fatality rates of greater Worcester, Mass, residents hospitalized with confirmed acute myocardial infarction (AMI), we investigated the hypothesis that census tract–level socioeconomic position is an important predictor of survival after hospital discharge for AMI, after adjusting for demographic and clinical characteristics.
Methods and Results— Data were available for 3423 confirmed cases of AMI among metropolitan Worcester residents during the 4 study years of 1995, 1997, 1999, and 2001 who were followed up through the end of 2002. The mean age among patients was 69 years, and 58% were men. Using a multilevel Cox proportional hazards regression model, we estimated a 30% higher death rate after AMI for patients living in census tracts with the most residents living below the poverty line compared with patients living in the wealthiest census tracts (relative risk=1.30; 95% CI, 1.08 to 1.56). Similarly, patients living in census tracts with the highest proportion of residents with less than a high school education experienced a 47% higher death rate than patients living in census tracts with the lowest proportion of residents with less than a high school education (relative risk=1.47; 95% CI, 1.15 to 1.88).
Conclusions— Within a medium-sized urban area, there are important variations in survival after hospital discharge for AMI that are associated with socioeconomic position. These associations persist after adjustment for demographic and clinical characteristics. Reasons for these differences warrant further investigation.
Key Words: epidemiology ; follow-up studies ; myocardial infarction ; survival
Introduction
Mortality from coronary heart disease has been associated with socioeconomic position (SEP) in a number of different countries,1–3 despite substantial differences in the degree of social inequality and access to healthcare from country to country. In comparison, far less is known about the effect of SEP on prognosis after acute myocardial infarction (AMI). Associations between individual SEP and long-term prognosis after AMI have been demonstrated in Finland and the United Kingdom4,5 and in relation to functional status (NYHA class) in the United States.6 Associations between area-based or community-level measures of socioeconomic characteristics and cardiovascular health have been found that are independent of individual-level socioeconomic factors.1,7,8 These findings suggest that community-level socioeconomic variables capture information above and beyond those at the individual level and do not serve simply as proxies for individual-level SEP. Alter and colleagues9 observed an association between area-based income and 1-year mortality after AMI in Ontario residents. In the United States, area-based measures of SEP at the census tract or block group level allow for investigation of the effects of community-level SEP at substantially finer resolution. Despite these observations, few studies have investigated the association between SEP and prognosis after AMI in the United States.10 We investigated socioeconomic differences in long-term survival after hospital discharge for AMI in a population-based sample of patients with independently confirmed AMI. Specifically, we investigated the hypothesis that census tract–level SEP is an important predictor of survival after adjustment for demographic and clinical characteristics that may affect post-AMI survival.
See p 3020
Methods
The Worcester Heart Attack Study is an ongoing community-wide investigation examining changes over time in the incidence, hospital, and long-term case-fatality rates of greater Worcester, Mass, residents hospitalized with confirmed AMI.11–13 The present investigation was limited to residents aged 25 years of the Worcester Metropolitan Statistical Area (MSA), a Census Bureau–defined urban area representing a core urban area and adjacent communities with a high degree of economic and social integration.14 Cases of possible AMI were admitted to any of the 11 acute care general hospitals in the Worcester MSA during 1995, 1997, 1999, and 2001. The medical records of all patients with a discharge diagnosis of AMI (ICD-9 code 410) were reviewed and independently validated according to preestablished diagnostic criteria described previously.11–13 These criteria included a suggestive clinical history, increased cardiac enzyme levels above each hospital’s normal range, and serial ECG findings indicative of AMI. At least 2 of the 3 criteria were necessary for study inclusion. Information was abstracted from hospital medical records with regard to demographic and clinical characteristics, medical history, survival status during acute hospitalization, AMI order (initial versus prior), and AMI type (Q wave versus non–Q wave). Patients’ residential addresses at the time of AMI were collected from medical records and submitted to a commercial firm for geocoding at the census tract level. Patients residing in Worcester County for whom the census tract was uncertain after geocoding were excluded from the analysis (n=111), as were patients who died during the index hospitalization. Survival status after hospital discharge was ascertained by reviewing records for additional hospitalizations and a statewide and national search of death certificates for Greater Worcester residents. Some form of additional follow-up was obtained for >99% of patients discharged from Greater Worcester hospitals after AMI. The study was approved by the Committee for the Protection of Human Subjects at the University of Massachusetts Medical School and the Human Subjects Committee at Harvard School of Public Health.
We obtained 2000 Census data from the US Bureau of the Census Summary File 3. Census tracts have an average population of 4000 persons and are defined by the Census Bureau as small statistical subdivisions of counties with generally stable boundaries, designed to have relatively homogeneous demographic and economic characteristics.15 Area-based measures of SEP were derived from 2000 Census data and included the following16: median income (median household income in 1999); poverty (percentage of persons below the federally defined poverty line, a threshold that depends on family size and age of children); education (percentage of persons aged 25 years whose highest degree was less than a high school diploma or equivalency); and crowding (percentage of households with >1 person per room). A composite of the 4 derived measures of SEP was also considered. We chose area-based measures of SEP that were most likely to capture deprivation of material resources and that have been shown previously to identify differences in health outcomes at the census tract level.2
Data Analysis
Categorical SEP variables were created by dividing census tracts into categories according to median income, poverty, education, and degree of crowding. Cut points were based on percentile distributions (quintiles) or a priori considerations of meaningful differences in SEP that would facilitate comparison with other studies.2 A composite measure of SEP was created by summing the Z scores from the distributions of the 4 SEP variables.17 Posthospital discharge survival curves were calculated according to quintile or category of each SEP measure as unadjusted Kaplan-Meier curves.
The data used in this analysis have a 2-level structure in which individuals are nested within census tracts; therefore, we used a multilevel analysis. Within a given neighborhood, individuals are more alike with respect to unmeasured variables then they are between neighborhoods. Consequently, individuals within a neighborhood provide less independent information than if they were randomly distributed geographically, allowing for potential correlation in the errors of individuals in the same census tract, a violation of a basic assumption of regression.18 A random intercept for each census tract was modeled to address this autocorrelation. A multilevel Cox proportional hazards regression model was used to estimate the association between survival after hospital discharge for AMI and census tract–level SEP. The first level of the model operated at the individual level, explaining differences in survival with the use of individual-level demographic and clinical predictors. The second level of the model attempts to explain variation in the census tract–specific intercepts with the use of census tract–level measures of SEP. In the first stage of the model, the individual level regression was assumed to have the following form: log(Hazard Ratioij) = Xi · ; + uj, where Hazard Ratioij is the relative risk of death for the ith subject living in the jth census tract, and uj is a random census tract intercept that captures mean differences in risk within a census tract that are not explained by the vector of individual-level covariates Xi. Our individual-level covariates included the following potential confounders: patient age; sex; white race; medical history of hypertension, diabetes, angina, stroke, congestive heart failure; clinical complications of atrial fibrillation, congestive heart failure, and cardiogenic shock that occurred during the index hospitalization; and AMI order and type.19 A dummy variable for discharge hospital was also included as an individual-level covariate. In addition, we adjusted for an age-by-sex interaction because younger women who survived hospitalization for AMI were found to have higher long-term mortality than men according to previous data from the Worcester Heart Attack Study.20 Potential modification of the association between census tract–level SEP and long-term prognosis by age and sex was also investigated by including interactions between continuous or binary measures of SEP and age and sex.
The second, census tract–level, model had 111 degrees of freedom versus 3423 for the individual level. At the census tract level, we assumed uj = 0 + 1 · SEPj + ej, where SEPj is the SEP measure in the jth census tract; thus, for categorical SEP values there was 1 regression coefficient for each level. The regression models were fit with the use of Splus statistical software, with a gamma frailty term used for the census tract–level random intercepts to allow for the possibility of a skewed, nonnormal distribution of the random intercepts.
Results
There are 111 census tracts in the Worcester MSA. Within the MSA, the average household income was $47 949, 1 in 10 residents was living below the poverty line, and 1 in 6 residents had less than a high school education. Approximately 1 in 40 residents lived in households with >1 occupant per room (Table 1).
Subjects living in the most deprived census tracts had the lowest survival rate 1 year after hospital discharge for AMI and throughout the majority of subsequent follow-up (Figure). Although no single measure of SEP was clearly better at stratifying patients by risk, the unadjusted survival curves showed a relatively consistent pattern across the 4 measures of SEP.
Unadjusted Kaplan-Meier survival curves by measure of SEP for full duration of follow-up. Median household income is in thousands of dollars.
After adjustment for the covariates listed in the footnote of Table 3, the Cox proportional hazards model showed that each of the indicators of low SEP were significantly associated with lower survival after AMI. Race (white versus nonwhite) did not confound the association between census tract–level SEP and long-term prognosis, likely because of the small number of nonwhite patients in the study population. Therefore, we did not adjust for race in the final multivariable models. When we used the percentage of census tract residents living below the poverty line as our measure of SEP, the association with survival was mostly concentrated in subjects living in the most deprived census tracts. In the 20% of patients living in census tracts with the highest percent of residents living below the poverty line, the death rate after discharge for AMI was 30% higher (95% CI, 8% to 56%) than that of patients living in the wealthiest census tracts. A clearer exposure-response relationship was observed for the other measures of SEP. The 20% of patients living in census tracts with the lowest median household income experienced a 38% higher (95% CI, 14% to 67%) death rate than those living in census tracts with the highest median household income. Among patients living in census tracts with the highest percentage of residents with less than a high school education, the death rate was 47% higher (95% CI, 15% to 88%) than those living in census tracts with the lowest proportion of residents with less than a high school education. The composite measure of SEP yielded estimates of the association between SEP and prognosis after hospital discharge for AMI similar to the other measures of SEP. Patients living in census tracts with the highest score for the composite SEP measure had a 38% higher (95% CI, 14% to 67%) death rate than those living in the least deprived census tracts. Crowded living conditions had the strongest association with survival after hospital discharge for AMI; patients living in census tracts with the most crowding had an 82% higher (95% CI, 39% to 238%) death rate than patients living in census tracts with the lowest proportion of crowded households. We also included a quadratic age term in our models and found that the parameter estimates for all SEP measures changed only minimally. No significant modification of the association between SEP and post-AMI prognosis by age or sex was observed. However, there was a suggestion that census tract–level median household income was more protective for women than for men. A violation of the proportional hazards assumption was observed only after >6.8 years of follow-up and was limited to the poverty SEP measure.
As a sensitivity analysis, the Cox regression models were repeated with the use of SEP measures at the block group, an area presumably more homogeneous with respect to SEP than a census tract. The point estimates were essentially identical to those calculated with the use of census tract–level measures of SEP. To explore the possibility that the observed associations were due to differences in access to follow-up care due to lack of health insurance, we restricted the analysis to patients aged 65 years who would be eligible for Medicare. Each measure of SEP remained significantly associated with survival after hospital discharge among patients living in the most deprived census tracts, and the overall exposure-response pattern was very similar to that observed in the full study population.
Discussion
In this community-based study, we found important differences in survival after hospital discharge for AMI according to neighborhood-level SEP. Patients living in the most materially deprived neighborhoods had significantly higher all-cause mortality after hospital discharge, after adjustment for a variety of characteristics that may affect long-term prognosis after AMI. The unadjusted Kaplan-Meier curves showed a consistent pattern across the 4 different measures of SEP in which patients living in the most deprived census tracts had the poorest survival after 1 year as well as over a more extended period of follow-up. In the regression analysis, living in a census tract with a high level of crowding had the largest association with long-term survival. However, because our measure of crowded conditions was based on a priori cut points, only a small number of patients lived in census tracts with the most crowded conditions (n=122). Our overall ability to identify the SEP measure that best captured the association with long-term survival was limited. Neighborhood-level SEP is difficult to measure; therefore, we used 4 different derived SEP variables that potentially capture different facets of SEP as well as a composite measure of SEP. No single measure appeared to risk stratify patients better than the others. Relative to other known predictors of survival after hospital discharge for AMI, such as congestive heart failure, left ventricular hypertrophy, and a lower left ventricular ejection fraction, the observed associations with measures of census tract level SEP were modest.19,21
Although we were able to adjust for many comorbidities related to coronary heart disease, the observed association between census tract–level SEP and post-AMI survival may reflect the higher prevalence of other diseases among persons living in low-SEP neighborhoods. Ample evidence exists to suggest that individuals with lower SEP and those living in deprived neighborhoods live shorter lives and are generally of poorer health.22–27 Poorer overall health may be compounded by psychosocial factors including occupational stress, social isolation, and depression.28 Differences in access to effective cardiac medications and interventional procedures, cardiovascular risk factors, and compliance with prescribed medications may also contribute to the observed associations. We selected area-based measures of SEP that were most likely to reflect lack of material resources and characteristics of the physical environment of the neighborhood. Although the precise pathways by which neighborhood-level SEP affects survival after hospital discharge for AMI have not been identified, this association may be due in part to exposure to traffic-related pollutants, housing conditions, or other aspects of the physical environment.29
Our findings are consistent with the results of other studies investigating the relationship between socioeconomic factors and post-AMI survival in other populations. In a national sample of Medicare beneficiaries hospitalized for AMI between 1994 and 1996, those living in ZIP codes with the lowest median household income had a 5% higher 1-year mortality rate after hospital discharge than those living in the highest-income ZIP codes.10 A community-based study among residents of Malm;, Sweden, used a composite score to measure the socioeconomic characteristics of a residential area. The study investigators found that residents of areas with high socioeconomic scores had markedly lower mortality (40%) than those living in areas with the lowest score.30 Similarly, a population-based study of patients with AMI who were admitted to hospitals in Ontario, Canada, showed that socioeconomic position was significantly associated with mortality during the first year after AMI.9 A $10 000 increase in area-level median income was associated with a 10% reduction in the risk of dying during the first year after AMI. Notably, income measures were based on Forward Sortation Areas, which have an average of 7000 households and can include up to 50 000 households31 and therefore are likely to result in a large degree of misclassification of income at the neighborhood level. The present analysis takes advantage of finer-resolution SEP data available through the US Census by using census tract–level SEP data. Moreover, census tract–level SEP has been shown to perform more consistently than ZIP code–level measures with respect to detecting socioeconomic gradients in mortality.2
Beyond serving as proxies for individual-level SEP, area-based measures in their own right reflect the socioeconomic context of the neighborhood or community. These neighborhood-level effects may represent differences in the availability and cost of healthful foods, access to healthcare clinics, differences in traffic burden, and availability of public space, as well as differences in the neighborhood social environment.1,7,32 Ideally, an investigation of the effect of SEP on prognosis after AMI would separate the neighborhood- and individual-level effects and identify for which individuals context matters. We adjusted for many important individual-level predictors of survival after AMI such as age, sex, comorbidities, and prior AMI; however, because we lacked individual-level data on socioeconomic factors, we were not able to effectively distinguish between the influence of SEP at the individual level and that which may occur at the neighborhood level. Unmeasured factors influencing where an individual chooses to live, apart from socioeconomics, may also have led to confounding of the association between neighborhood-level SEP and prognosis.33,34 In observational studies such as ours, unmeasured or residual confounding due to systematic differences in individual characteristics across census tracts limits our ability to derive causal inferences and therefore recommend specific interventions. Despite our inability to isolate the neighborhood-level influence of SEP, neighborhood-level SEP effects on prognosis after AMI are almost certain to be at work and to be of substantive interest. Cardiovascular epidemiology has traditionally focused on individual-level biological and behavioral risk factors for coronary disease, often without consideration of the environments or social contexts in which these risk factors may have developed or may have been exacerbated.29 Consequently, prevention and treatment strategies have been targeted primarily at the individual level. However, the effectiveness of individually focused prevention and risk factor modification programs may be strongly influenced by the context in which individuals experience their daily lives, specifically as it applies to the initiation of behavior change and adherence. Thus, it has been argued that changing behavior in isolation of an individual’s context may be less effective than modifying the environments to promote healthy behaviors.35 Further research is needed to identify the relative effect of individual- versus neighborhood-level deprivation on cardiovascular outcomes and the importance of both levels in treatment and prevention strategies.
Additional limitations of our analysis included the lack of data to adjust for important behavioral factors, such as cigarette smoking and alcohol use, as well as psychosocial factors that may affect prognosis after AMI.28,36,37 We also lacked data on the use of medications during follow-up and additional secondary prevention measures. The surveillance approach used only captures data on greater Worcester residents who sought care for AMI at hospitals in the Greater Worcester area and slightly beyond. Therefore, we do not have data on patients with silent or unrecognized AMI, which can account for more than a quarter of all AMIs,38 or on out-of-hospital sudden cardiac deaths. It is unclear whether the care-seeking patterns for these patients would differ from the patients who were hospitalized for AMI and how this pattern would depend on area-level SEP. Additionally, we could not account for the duration of time patients spent in their census tract before the onset of AMI or if they moved to another tract after their AMI. Assuming that residents were affected by the SEP of their census tract of residence at the time the event occurred is likely to lead to some misclassification of SEP and consequent underestimation of the association. We focused on the association between SEP and survival after hospital discharge for AMI. If subjects living in more deprived neighborhoods were more likely to have a fatal AMI before or during hospitalization, this would also lead to an underestimation of the association of SEP with survival after hospital discharge for AMI. Finally, because this was an observational study, there are likely to be a number of unmeasured confounders of the association between census tract–level SEP and prognosis after hospital discharge for AMI. We therefore are unable to make inferences about the reasons for the observed socioeconomic differences in post-AMI survival. The primary strength of our study is the complete enrollment of a community-wide population of patients who were hospitalized with independently confirmed AMI and the use of relatively recent data that reflect current medical practices.
In conclusion, we observed that long-term prognosis after hospital discharge for AMI is associated with census tract–level SEP, particularly for those living in the most impoverished neighborhoods. The association persisted after adjustment for many important demographic, medical history, and clinical characteristics. Notwithstanding the limitations of this analysis, there are significant differences in post-AMI survival according to census tract–level SEP, suggesting that simple area-based measures of SEP may be useful in targeting individuals at highest risk of developing adverse long-term outcomes after AMI. Further research is needed to characterize the relative effect of individual- versus neighborhood-level SEP on post-AMI survival and to elucidate the precise pathways by which neighborhood-level SEP affects cardiovascular morbidity and mortality.
Acknowledgments
This study was supported by grants RO1 ES011636 and ES0002, provided by the National Institute of Environmental Health Sciences, and RO1 HL35434 and T32 HL07427, provided by the National Heart, Lung, and Blood Institute. We thank the administration, medical records, and cardiology departments of participating Worcester area hospitals for their cooperation.
References
Smith GD, Hart C, Watt G, Hole D, Hawthorne V. Individual social class, area-based deprivation, cardiovascular disease risk factors, and mortality: the Renfrew and Paisley study. J Epidemiol Community Health. 1998; 52: 399–405.
Krieger N, Chen JT, Waterman PD, Soobader M-J, Subramanian SV, Carson R. Geocoding and monitoring of US socioeconomic inequalities in mortality and cancer incidence: does the choice of area-based measure and geographic level matter; Am J Epidemiol. 2002; 156: 471–482.
Logue EE, Jarjoura D. Modeling heart disease mortality with census tract rates and social class mixtures. Soc Sci Med. 1990; 31: 545–550.
Salomaa V, Miettinen H, Niemela M, Ketonen M, Mahonon M, Immonen-Raiha P, Lehto S, Vuorenmaa T, Koskinen S, Palomaki P, Mustaneimi H, Kaarsalo E, Arstila M, Torppa J, Kuulasmaa K, Puska P, Pyorala K, Tuomilehto J. Relation of socioeconomic position to the case fatality, prognosis and treatment of myocardial infarction events: the FINMONICA MI Register Study. J Epidemiol Community Health. 2001; 55: 475–482.
Greenwood D, Packham C, Muir K, Madeley R. How do economic status and social support influence survival after initial recovery from acute myocardial infarction; Soc Sci Med. 1995; 40: 639–647.
Ickovics JR, Viscoli CM, Horwitz RI. Functional recovery after myocardial infarction in men: the independent effects of social class. Ann Intern Med. 1997; 127: 518–525.
Diez-Roux A, Merkin SS, Arnett D, Chambless L, Massing M, Nieto FJ, Sorlie P, Szklo M, Tyroler HA, Watson RL. Neighborhood of residence and incidence of coronary heart disease. N Engl J Med. 2001; 345: 99–106.
Leclere FB, Rogers RG, Peters K. Neighborhood social context and racial differences in women’s heart disease mortality. J Health Soc Behav. 1998; 39: 91–107.
Alter DA, Naylor CD, Austin P, Tu JV. Effects of socioeconomic status on access to invasive cardiac procedures and on mortality after acute myocardial infarction. N Engl J Med. 1999; 341: 1359–1367.
Rao SV, Schulman KA, Curtis LH, Gersh BJ, Jollis JG. Socioeconomic status and outcome following acute myocardial infarction in elderly patients. Arch Intern Med. 2004; 164: 1128–33.
Goldberg RJ, Gore JM, Alpert J, Dalen J. Recent changes in attack and survival rates of acute myocardial infarction (1975 through 1981): the Worcester Heart Attack Study. JAMA. 1986; 255: 2774–2779.
Goldberg RJ, Yarzebski J, Lessard D, Gore JM. A two-decades (1975 to 1995) long experience in the incidence, in-hospital and long-term case-fatality rates of acute myocardial infarction: a community-wide perspective. J Am Coll Cardiol. 1999; 33: 1533–1539.
Goldberg RJ, Gore JM, Alpert J, Dalen J. Incidence and case fatality rates of acute myocardial infarction (1975–1984): the Worcester Heart Attack Study. Am Heart J. 1988; 115: 761–767.
Federal Registrar. Part IX Office of Management and Budget. Standards for defining metropolitan and micropolitan statistical areas. 2000:Vol 65(249). Available at: http://www.census.gov/population/www/estimates/00-32997.txt. Accessed June 16, 2003.
Bureau of the Census [homepage on Internet]. Introduction to Census 2000 Data Products. US Dept of Commerce; 2000. Available at: http://www.census.gov/prod/2001pubs/mso-01icdp.pdf. Accessed June 16, 2003.
Bureau of the Census [homepage on Internet]. 2000 Census of Population and Housing Summary File3. Appendix B. Definitions of Subject Characteristics. US Dept of Commerce; 2000:B-1–B-77. Available at: http://www.census.gov/prod/cen2000/doc/sf3.pdf. Accessed May 23, 2003.
Gilthorpe MS. The importance of normalisation in the construction of deprivation indices. J Epidemiol Community Health. 1995; 49 (suppl 2): S45–S50.
Duncan C, Jones K, Moon G. Context, composition and heterogeneity: using multilevel models in health research. Soc Sci Med. 1998; 46: 97–117.
Kaplan RC, Heckbert SR, Furberg CD, Psaty BM. Predictors of subsequent coronary events, stroke, and death among survivors of first hospitalized myocardial infarction. J Clin Epidemiol. 2002; 55: 654–664.
Vaccarino V, Krumholz H, Yarzebski J, Gore JM, Goldberg RJ. Sex differences in 2-year mortality after hospital discharge for myocardial infarction. Ann Intern Med. 2001; 134: 239–241.
Schwammenthal E, Adler Y, Amichai K, Sagie A, Behar S, Hod H, Feinberg MS. Prognostic value of global myocardial performance indices in acute myocardial infarction: comparison to measures of systolic and diastolic left ventricular function. Chest. 2003; 124: 1645–1651.
Schrijvers CTM, Mackenbach JP, Lutz J-M, Quinn MJ, Coleman MP. Deprivation, stage at diagnosis and cancer survival. Int J Cancer. 1995; 63: 324–329.
Kaplan GA, Keil JE. Socioeconomic factors and cardiovascular disease: a review of the literature. Circulation. 1993; 88: 1973–1998.
Pollock AM, Vickers N. Breast, lung and colorectal cancer incidence and survival in South Thames Region, 1987–1992: the effect of social deprivation. J Public Health Med. 1997; 19: 288–294.
Diez-Roux A, Nieto FJ, Tyroler HA, Crum LD, Szklo M. Social inequalities and atherosclerosis: the Atherosclerosis Risk in Communities Study. Am J Epidemiol. 1995; 141: 960–972.
Townsend P, Phillimore P, Beattie A. Health and Deprivation: Inequality and the North. London, UK: Croom Helm; 1988.
Haan MN, Kaplan G, Syme SL. Socioeconomic status and health: old observations and new thoughts. In: Bunker JP, Gomby DS, Kehrer BH, eds. Pathways to Health: The Role of Social Factors. Menlo Park, Calif: Henry J. Kaiser Family Foundation; 1989.
Hemingway H, Marmot M. Psychosocial factors in the aetiology and prognosis of coronary heart disease: systematic review of prospective cohort studies. Br Med J. 1999; 318: 1460–1467.
Diez-Roux A. Residential environments and cardiovascular risk. J Urban Health. 2003; 80: 569–589.
Tyden P, Hansen O, Engstrom G, Hedblad B, Janzon L. Myocardial infarction in an urban population: worse long term prognosis for patients from less affluent residential areas. J Epidemiol Community Health. 2002; 56: 785–790.
Statistics Canada [homepage on Internet]. 2001 Census Dictionary. 2002. Available at: http://www.statcan.ca/english/census2001/dict/geo035.htm. Accessed August 30, 2003.
Krieger N, Williams DR, Moss NE. Measuring social class in US public health research: concepts, methodologies, and guidelines. Annu Rev Publ Health. 1997; 18: 341–378.
Oakes JM. The (mis)estimation of neighborhood effects: casual inference for a practicable social epidemiology. Soc Sci Med. 2004; 58: 1929–1952.
Greenland S. Ecologic versus individual-level sources of bias in ecologic estimates of contextual health effects. Int J Epidemiol. 2001; 30: 1343–1350.
Syme SL. Strategies for health promotion. Prev Med. 1986; 15: 492–507.
Welin C, Lappas G, Wilhelmsen L. Independent importance of psychosocial factors for prognosis after myocardial infarction. J Intern Med. 2000; 247: 629–639.
Frasure-Smith N, Lesperance F. Depression and other psychological risks following myocardial infarction. Arch Gen Psychiatry. 2003; 60: 627–636.
Kannel WB, Abbott RD. Incidence and prognosis of unrecognized myocardial infarction: an update on the Framingham Study. N Engl J Med. 1984; 311: 1144–1147.(Cathryn Tonne, MPH; Joel )
Harvard School of Public Health, Boston, Mass; Beth Israel Deaconess Medical Center, Boston, Mass (M.M.)
University of Massachusetts Medical School, Worcester (R.G.).
Abstract
Background— As part of the Worcester Heart Attack Study, a community-wide study examining changes over time in the incidence and long-term case-fatality rates of greater Worcester, Mass, residents hospitalized with confirmed acute myocardial infarction (AMI), we investigated the hypothesis that census tract–level socioeconomic position is an important predictor of survival after hospital discharge for AMI, after adjusting for demographic and clinical characteristics.
Methods and Results— Data were available for 3423 confirmed cases of AMI among metropolitan Worcester residents during the 4 study years of 1995, 1997, 1999, and 2001 who were followed up through the end of 2002. The mean age among patients was 69 years, and 58% were men. Using a multilevel Cox proportional hazards regression model, we estimated a 30% higher death rate after AMI for patients living in census tracts with the most residents living below the poverty line compared with patients living in the wealthiest census tracts (relative risk=1.30; 95% CI, 1.08 to 1.56). Similarly, patients living in census tracts with the highest proportion of residents with less than a high school education experienced a 47% higher death rate than patients living in census tracts with the lowest proportion of residents with less than a high school education (relative risk=1.47; 95% CI, 1.15 to 1.88).
Conclusions— Within a medium-sized urban area, there are important variations in survival after hospital discharge for AMI that are associated with socioeconomic position. These associations persist after adjustment for demographic and clinical characteristics. Reasons for these differences warrant further investigation.
Key Words: epidemiology ; follow-up studies ; myocardial infarction ; survival
Introduction
Mortality from coronary heart disease has been associated with socioeconomic position (SEP) in a number of different countries,1–3 despite substantial differences in the degree of social inequality and access to healthcare from country to country. In comparison, far less is known about the effect of SEP on prognosis after acute myocardial infarction (AMI). Associations between individual SEP and long-term prognosis after AMI have been demonstrated in Finland and the United Kingdom4,5 and in relation to functional status (NYHA class) in the United States.6 Associations between area-based or community-level measures of socioeconomic characteristics and cardiovascular health have been found that are independent of individual-level socioeconomic factors.1,7,8 These findings suggest that community-level socioeconomic variables capture information above and beyond those at the individual level and do not serve simply as proxies for individual-level SEP. Alter and colleagues9 observed an association between area-based income and 1-year mortality after AMI in Ontario residents. In the United States, area-based measures of SEP at the census tract or block group level allow for investigation of the effects of community-level SEP at substantially finer resolution. Despite these observations, few studies have investigated the association between SEP and prognosis after AMI in the United States.10 We investigated socioeconomic differences in long-term survival after hospital discharge for AMI in a population-based sample of patients with independently confirmed AMI. Specifically, we investigated the hypothesis that census tract–level SEP is an important predictor of survival after adjustment for demographic and clinical characteristics that may affect post-AMI survival.
See p 3020
Methods
The Worcester Heart Attack Study is an ongoing community-wide investigation examining changes over time in the incidence, hospital, and long-term case-fatality rates of greater Worcester, Mass, residents hospitalized with confirmed AMI.11–13 The present investigation was limited to residents aged 25 years of the Worcester Metropolitan Statistical Area (MSA), a Census Bureau–defined urban area representing a core urban area and adjacent communities with a high degree of economic and social integration.14 Cases of possible AMI were admitted to any of the 11 acute care general hospitals in the Worcester MSA during 1995, 1997, 1999, and 2001. The medical records of all patients with a discharge diagnosis of AMI (ICD-9 code 410) were reviewed and independently validated according to preestablished diagnostic criteria described previously.11–13 These criteria included a suggestive clinical history, increased cardiac enzyme levels above each hospital’s normal range, and serial ECG findings indicative of AMI. At least 2 of the 3 criteria were necessary for study inclusion. Information was abstracted from hospital medical records with regard to demographic and clinical characteristics, medical history, survival status during acute hospitalization, AMI order (initial versus prior), and AMI type (Q wave versus non–Q wave). Patients’ residential addresses at the time of AMI were collected from medical records and submitted to a commercial firm for geocoding at the census tract level. Patients residing in Worcester County for whom the census tract was uncertain after geocoding were excluded from the analysis (n=111), as were patients who died during the index hospitalization. Survival status after hospital discharge was ascertained by reviewing records for additional hospitalizations and a statewide and national search of death certificates for Greater Worcester residents. Some form of additional follow-up was obtained for >99% of patients discharged from Greater Worcester hospitals after AMI. The study was approved by the Committee for the Protection of Human Subjects at the University of Massachusetts Medical School and the Human Subjects Committee at Harvard School of Public Health.
We obtained 2000 Census data from the US Bureau of the Census Summary File 3. Census tracts have an average population of 4000 persons and are defined by the Census Bureau as small statistical subdivisions of counties with generally stable boundaries, designed to have relatively homogeneous demographic and economic characteristics.15 Area-based measures of SEP were derived from 2000 Census data and included the following16: median income (median household income in 1999); poverty (percentage of persons below the federally defined poverty line, a threshold that depends on family size and age of children); education (percentage of persons aged 25 years whose highest degree was less than a high school diploma or equivalency); and crowding (percentage of households with >1 person per room). A composite of the 4 derived measures of SEP was also considered. We chose area-based measures of SEP that were most likely to capture deprivation of material resources and that have been shown previously to identify differences in health outcomes at the census tract level.2
Data Analysis
Categorical SEP variables were created by dividing census tracts into categories according to median income, poverty, education, and degree of crowding. Cut points were based on percentile distributions (quintiles) or a priori considerations of meaningful differences in SEP that would facilitate comparison with other studies.2 A composite measure of SEP was created by summing the Z scores from the distributions of the 4 SEP variables.17 Posthospital discharge survival curves were calculated according to quintile or category of each SEP measure as unadjusted Kaplan-Meier curves.
The data used in this analysis have a 2-level structure in which individuals are nested within census tracts; therefore, we used a multilevel analysis. Within a given neighborhood, individuals are more alike with respect to unmeasured variables then they are between neighborhoods. Consequently, individuals within a neighborhood provide less independent information than if they were randomly distributed geographically, allowing for potential correlation in the errors of individuals in the same census tract, a violation of a basic assumption of regression.18 A random intercept for each census tract was modeled to address this autocorrelation. A multilevel Cox proportional hazards regression model was used to estimate the association between survival after hospital discharge for AMI and census tract–level SEP. The first level of the model operated at the individual level, explaining differences in survival with the use of individual-level demographic and clinical predictors. The second level of the model attempts to explain variation in the census tract–specific intercepts with the use of census tract–level measures of SEP. In the first stage of the model, the individual level regression was assumed to have the following form: log(Hazard Ratioij) = Xi · ; + uj, where Hazard Ratioij is the relative risk of death for the ith subject living in the jth census tract, and uj is a random census tract intercept that captures mean differences in risk within a census tract that are not explained by the vector of individual-level covariates Xi. Our individual-level covariates included the following potential confounders: patient age; sex; white race; medical history of hypertension, diabetes, angina, stroke, congestive heart failure; clinical complications of atrial fibrillation, congestive heart failure, and cardiogenic shock that occurred during the index hospitalization; and AMI order and type.19 A dummy variable for discharge hospital was also included as an individual-level covariate. In addition, we adjusted for an age-by-sex interaction because younger women who survived hospitalization for AMI were found to have higher long-term mortality than men according to previous data from the Worcester Heart Attack Study.20 Potential modification of the association between census tract–level SEP and long-term prognosis by age and sex was also investigated by including interactions between continuous or binary measures of SEP and age and sex.
The second, census tract–level, model had 111 degrees of freedom versus 3423 for the individual level. At the census tract level, we assumed uj = 0 + 1 · SEPj + ej, where SEPj is the SEP measure in the jth census tract; thus, for categorical SEP values there was 1 regression coefficient for each level. The regression models were fit with the use of Splus statistical software, with a gamma frailty term used for the census tract–level random intercepts to allow for the possibility of a skewed, nonnormal distribution of the random intercepts.
Results
There are 111 census tracts in the Worcester MSA. Within the MSA, the average household income was $47 949, 1 in 10 residents was living below the poverty line, and 1 in 6 residents had less than a high school education. Approximately 1 in 40 residents lived in households with >1 occupant per room (Table 1).
Subjects living in the most deprived census tracts had the lowest survival rate 1 year after hospital discharge for AMI and throughout the majority of subsequent follow-up (Figure). Although no single measure of SEP was clearly better at stratifying patients by risk, the unadjusted survival curves showed a relatively consistent pattern across the 4 measures of SEP.
Unadjusted Kaplan-Meier survival curves by measure of SEP for full duration of follow-up. Median household income is in thousands of dollars.
After adjustment for the covariates listed in the footnote of Table 3, the Cox proportional hazards model showed that each of the indicators of low SEP were significantly associated with lower survival after AMI. Race (white versus nonwhite) did not confound the association between census tract–level SEP and long-term prognosis, likely because of the small number of nonwhite patients in the study population. Therefore, we did not adjust for race in the final multivariable models. When we used the percentage of census tract residents living below the poverty line as our measure of SEP, the association with survival was mostly concentrated in subjects living in the most deprived census tracts. In the 20% of patients living in census tracts with the highest percent of residents living below the poverty line, the death rate after discharge for AMI was 30% higher (95% CI, 8% to 56%) than that of patients living in the wealthiest census tracts. A clearer exposure-response relationship was observed for the other measures of SEP. The 20% of patients living in census tracts with the lowest median household income experienced a 38% higher (95% CI, 14% to 67%) death rate than those living in census tracts with the highest median household income. Among patients living in census tracts with the highest percentage of residents with less than a high school education, the death rate was 47% higher (95% CI, 15% to 88%) than those living in census tracts with the lowest proportion of residents with less than a high school education. The composite measure of SEP yielded estimates of the association between SEP and prognosis after hospital discharge for AMI similar to the other measures of SEP. Patients living in census tracts with the highest score for the composite SEP measure had a 38% higher (95% CI, 14% to 67%) death rate than those living in the least deprived census tracts. Crowded living conditions had the strongest association with survival after hospital discharge for AMI; patients living in census tracts with the most crowding had an 82% higher (95% CI, 39% to 238%) death rate than patients living in census tracts with the lowest proportion of crowded households. We also included a quadratic age term in our models and found that the parameter estimates for all SEP measures changed only minimally. No significant modification of the association between SEP and post-AMI prognosis by age or sex was observed. However, there was a suggestion that census tract–level median household income was more protective for women than for men. A violation of the proportional hazards assumption was observed only after >6.8 years of follow-up and was limited to the poverty SEP measure.
As a sensitivity analysis, the Cox regression models were repeated with the use of SEP measures at the block group, an area presumably more homogeneous with respect to SEP than a census tract. The point estimates were essentially identical to those calculated with the use of census tract–level measures of SEP. To explore the possibility that the observed associations were due to differences in access to follow-up care due to lack of health insurance, we restricted the analysis to patients aged 65 years who would be eligible for Medicare. Each measure of SEP remained significantly associated with survival after hospital discharge among patients living in the most deprived census tracts, and the overall exposure-response pattern was very similar to that observed in the full study population.
Discussion
In this community-based study, we found important differences in survival after hospital discharge for AMI according to neighborhood-level SEP. Patients living in the most materially deprived neighborhoods had significantly higher all-cause mortality after hospital discharge, after adjustment for a variety of characteristics that may affect long-term prognosis after AMI. The unadjusted Kaplan-Meier curves showed a consistent pattern across the 4 different measures of SEP in which patients living in the most deprived census tracts had the poorest survival after 1 year as well as over a more extended period of follow-up. In the regression analysis, living in a census tract with a high level of crowding had the largest association with long-term survival. However, because our measure of crowded conditions was based on a priori cut points, only a small number of patients lived in census tracts with the most crowded conditions (n=122). Our overall ability to identify the SEP measure that best captured the association with long-term survival was limited. Neighborhood-level SEP is difficult to measure; therefore, we used 4 different derived SEP variables that potentially capture different facets of SEP as well as a composite measure of SEP. No single measure appeared to risk stratify patients better than the others. Relative to other known predictors of survival after hospital discharge for AMI, such as congestive heart failure, left ventricular hypertrophy, and a lower left ventricular ejection fraction, the observed associations with measures of census tract level SEP were modest.19,21
Although we were able to adjust for many comorbidities related to coronary heart disease, the observed association between census tract–level SEP and post-AMI survival may reflect the higher prevalence of other diseases among persons living in low-SEP neighborhoods. Ample evidence exists to suggest that individuals with lower SEP and those living in deprived neighborhoods live shorter lives and are generally of poorer health.22–27 Poorer overall health may be compounded by psychosocial factors including occupational stress, social isolation, and depression.28 Differences in access to effective cardiac medications and interventional procedures, cardiovascular risk factors, and compliance with prescribed medications may also contribute to the observed associations. We selected area-based measures of SEP that were most likely to reflect lack of material resources and characteristics of the physical environment of the neighborhood. Although the precise pathways by which neighborhood-level SEP affects survival after hospital discharge for AMI have not been identified, this association may be due in part to exposure to traffic-related pollutants, housing conditions, or other aspects of the physical environment.29
Our findings are consistent with the results of other studies investigating the relationship between socioeconomic factors and post-AMI survival in other populations. In a national sample of Medicare beneficiaries hospitalized for AMI between 1994 and 1996, those living in ZIP codes with the lowest median household income had a 5% higher 1-year mortality rate after hospital discharge than those living in the highest-income ZIP codes.10 A community-based study among residents of Malm;, Sweden, used a composite score to measure the socioeconomic characteristics of a residential area. The study investigators found that residents of areas with high socioeconomic scores had markedly lower mortality (40%) than those living in areas with the lowest score.30 Similarly, a population-based study of patients with AMI who were admitted to hospitals in Ontario, Canada, showed that socioeconomic position was significantly associated with mortality during the first year after AMI.9 A $10 000 increase in area-level median income was associated with a 10% reduction in the risk of dying during the first year after AMI. Notably, income measures were based on Forward Sortation Areas, which have an average of 7000 households and can include up to 50 000 households31 and therefore are likely to result in a large degree of misclassification of income at the neighborhood level. The present analysis takes advantage of finer-resolution SEP data available through the US Census by using census tract–level SEP data. Moreover, census tract–level SEP has been shown to perform more consistently than ZIP code–level measures with respect to detecting socioeconomic gradients in mortality.2
Beyond serving as proxies for individual-level SEP, area-based measures in their own right reflect the socioeconomic context of the neighborhood or community. These neighborhood-level effects may represent differences in the availability and cost of healthful foods, access to healthcare clinics, differences in traffic burden, and availability of public space, as well as differences in the neighborhood social environment.1,7,32 Ideally, an investigation of the effect of SEP on prognosis after AMI would separate the neighborhood- and individual-level effects and identify for which individuals context matters. We adjusted for many important individual-level predictors of survival after AMI such as age, sex, comorbidities, and prior AMI; however, because we lacked individual-level data on socioeconomic factors, we were not able to effectively distinguish between the influence of SEP at the individual level and that which may occur at the neighborhood level. Unmeasured factors influencing where an individual chooses to live, apart from socioeconomics, may also have led to confounding of the association between neighborhood-level SEP and prognosis.33,34 In observational studies such as ours, unmeasured or residual confounding due to systematic differences in individual characteristics across census tracts limits our ability to derive causal inferences and therefore recommend specific interventions. Despite our inability to isolate the neighborhood-level influence of SEP, neighborhood-level SEP effects on prognosis after AMI are almost certain to be at work and to be of substantive interest. Cardiovascular epidemiology has traditionally focused on individual-level biological and behavioral risk factors for coronary disease, often without consideration of the environments or social contexts in which these risk factors may have developed or may have been exacerbated.29 Consequently, prevention and treatment strategies have been targeted primarily at the individual level. However, the effectiveness of individually focused prevention and risk factor modification programs may be strongly influenced by the context in which individuals experience their daily lives, specifically as it applies to the initiation of behavior change and adherence. Thus, it has been argued that changing behavior in isolation of an individual’s context may be less effective than modifying the environments to promote healthy behaviors.35 Further research is needed to identify the relative effect of individual- versus neighborhood-level deprivation on cardiovascular outcomes and the importance of both levels in treatment and prevention strategies.
Additional limitations of our analysis included the lack of data to adjust for important behavioral factors, such as cigarette smoking and alcohol use, as well as psychosocial factors that may affect prognosis after AMI.28,36,37 We also lacked data on the use of medications during follow-up and additional secondary prevention measures. The surveillance approach used only captures data on greater Worcester residents who sought care for AMI at hospitals in the Greater Worcester area and slightly beyond. Therefore, we do not have data on patients with silent or unrecognized AMI, which can account for more than a quarter of all AMIs,38 or on out-of-hospital sudden cardiac deaths. It is unclear whether the care-seeking patterns for these patients would differ from the patients who were hospitalized for AMI and how this pattern would depend on area-level SEP. Additionally, we could not account for the duration of time patients spent in their census tract before the onset of AMI or if they moved to another tract after their AMI. Assuming that residents were affected by the SEP of their census tract of residence at the time the event occurred is likely to lead to some misclassification of SEP and consequent underestimation of the association. We focused on the association between SEP and survival after hospital discharge for AMI. If subjects living in more deprived neighborhoods were more likely to have a fatal AMI before or during hospitalization, this would also lead to an underestimation of the association of SEP with survival after hospital discharge for AMI. Finally, because this was an observational study, there are likely to be a number of unmeasured confounders of the association between census tract–level SEP and prognosis after hospital discharge for AMI. We therefore are unable to make inferences about the reasons for the observed socioeconomic differences in post-AMI survival. The primary strength of our study is the complete enrollment of a community-wide population of patients who were hospitalized with independently confirmed AMI and the use of relatively recent data that reflect current medical practices.
In conclusion, we observed that long-term prognosis after hospital discharge for AMI is associated with census tract–level SEP, particularly for those living in the most impoverished neighborhoods. The association persisted after adjustment for many important demographic, medical history, and clinical characteristics. Notwithstanding the limitations of this analysis, there are significant differences in post-AMI survival according to census tract–level SEP, suggesting that simple area-based measures of SEP may be useful in targeting individuals at highest risk of developing adverse long-term outcomes after AMI. Further research is needed to characterize the relative effect of individual- versus neighborhood-level SEP on post-AMI survival and to elucidate the precise pathways by which neighborhood-level SEP affects cardiovascular morbidity and mortality.
Acknowledgments
This study was supported by grants RO1 ES011636 and ES0002, provided by the National Institute of Environmental Health Sciences, and RO1 HL35434 and T32 HL07427, provided by the National Heart, Lung, and Blood Institute. We thank the administration, medical records, and cardiology departments of participating Worcester area hospitals for their cooperation.
References
Smith GD, Hart C, Watt G, Hole D, Hawthorne V. Individual social class, area-based deprivation, cardiovascular disease risk factors, and mortality: the Renfrew and Paisley study. J Epidemiol Community Health. 1998; 52: 399–405.
Krieger N, Chen JT, Waterman PD, Soobader M-J, Subramanian SV, Carson R. Geocoding and monitoring of US socioeconomic inequalities in mortality and cancer incidence: does the choice of area-based measure and geographic level matter; Am J Epidemiol. 2002; 156: 471–482.
Logue EE, Jarjoura D. Modeling heart disease mortality with census tract rates and social class mixtures. Soc Sci Med. 1990; 31: 545–550.
Salomaa V, Miettinen H, Niemela M, Ketonen M, Mahonon M, Immonen-Raiha P, Lehto S, Vuorenmaa T, Koskinen S, Palomaki P, Mustaneimi H, Kaarsalo E, Arstila M, Torppa J, Kuulasmaa K, Puska P, Pyorala K, Tuomilehto J. Relation of socioeconomic position to the case fatality, prognosis and treatment of myocardial infarction events: the FINMONICA MI Register Study. J Epidemiol Community Health. 2001; 55: 475–482.
Greenwood D, Packham C, Muir K, Madeley R. How do economic status and social support influence survival after initial recovery from acute myocardial infarction; Soc Sci Med. 1995; 40: 639–647.
Ickovics JR, Viscoli CM, Horwitz RI. Functional recovery after myocardial infarction in men: the independent effects of social class. Ann Intern Med. 1997; 127: 518–525.
Diez-Roux A, Merkin SS, Arnett D, Chambless L, Massing M, Nieto FJ, Sorlie P, Szklo M, Tyroler HA, Watson RL. Neighborhood of residence and incidence of coronary heart disease. N Engl J Med. 2001; 345: 99–106.
Leclere FB, Rogers RG, Peters K. Neighborhood social context and racial differences in women’s heart disease mortality. J Health Soc Behav. 1998; 39: 91–107.
Alter DA, Naylor CD, Austin P, Tu JV. Effects of socioeconomic status on access to invasive cardiac procedures and on mortality after acute myocardial infarction. N Engl J Med. 1999; 341: 1359–1367.
Rao SV, Schulman KA, Curtis LH, Gersh BJ, Jollis JG. Socioeconomic status and outcome following acute myocardial infarction in elderly patients. Arch Intern Med. 2004; 164: 1128–33.
Goldberg RJ, Gore JM, Alpert J, Dalen J. Recent changes in attack and survival rates of acute myocardial infarction (1975 through 1981): the Worcester Heart Attack Study. JAMA. 1986; 255: 2774–2779.
Goldberg RJ, Yarzebski J, Lessard D, Gore JM. A two-decades (1975 to 1995) long experience in the incidence, in-hospital and long-term case-fatality rates of acute myocardial infarction: a community-wide perspective. J Am Coll Cardiol. 1999; 33: 1533–1539.
Goldberg RJ, Gore JM, Alpert J, Dalen J. Incidence and case fatality rates of acute myocardial infarction (1975–1984): the Worcester Heart Attack Study. Am Heart J. 1988; 115: 761–767.
Federal Registrar. Part IX Office of Management and Budget. Standards for defining metropolitan and micropolitan statistical areas. 2000:Vol 65(249). Available at: http://www.census.gov/population/www/estimates/00-32997.txt. Accessed June 16, 2003.
Bureau of the Census [homepage on Internet]. Introduction to Census 2000 Data Products. US Dept of Commerce; 2000. Available at: http://www.census.gov/prod/2001pubs/mso-01icdp.pdf. Accessed June 16, 2003.
Bureau of the Census [homepage on Internet]. 2000 Census of Population and Housing Summary File3. Appendix B. Definitions of Subject Characteristics. US Dept of Commerce; 2000:B-1–B-77. Available at: http://www.census.gov/prod/cen2000/doc/sf3.pdf. Accessed May 23, 2003.
Gilthorpe MS. The importance of normalisation in the construction of deprivation indices. J Epidemiol Community Health. 1995; 49 (suppl 2): S45–S50.
Duncan C, Jones K, Moon G. Context, composition and heterogeneity: using multilevel models in health research. Soc Sci Med. 1998; 46: 97–117.
Kaplan RC, Heckbert SR, Furberg CD, Psaty BM. Predictors of subsequent coronary events, stroke, and death among survivors of first hospitalized myocardial infarction. J Clin Epidemiol. 2002; 55: 654–664.
Vaccarino V, Krumholz H, Yarzebski J, Gore JM, Goldberg RJ. Sex differences in 2-year mortality after hospital discharge for myocardial infarction. Ann Intern Med. 2001; 134: 239–241.
Schwammenthal E, Adler Y, Amichai K, Sagie A, Behar S, Hod H, Feinberg MS. Prognostic value of global myocardial performance indices in acute myocardial infarction: comparison to measures of systolic and diastolic left ventricular function. Chest. 2003; 124: 1645–1651.
Schrijvers CTM, Mackenbach JP, Lutz J-M, Quinn MJ, Coleman MP. Deprivation, stage at diagnosis and cancer survival. Int J Cancer. 1995; 63: 324–329.
Kaplan GA, Keil JE. Socioeconomic factors and cardiovascular disease: a review of the literature. Circulation. 1993; 88: 1973–1998.
Pollock AM, Vickers N. Breast, lung and colorectal cancer incidence and survival in South Thames Region, 1987–1992: the effect of social deprivation. J Public Health Med. 1997; 19: 288–294.
Diez-Roux A, Nieto FJ, Tyroler HA, Crum LD, Szklo M. Social inequalities and atherosclerosis: the Atherosclerosis Risk in Communities Study. Am J Epidemiol. 1995; 141: 960–972.
Townsend P, Phillimore P, Beattie A. Health and Deprivation: Inequality and the North. London, UK: Croom Helm; 1988.
Haan MN, Kaplan G, Syme SL. Socioeconomic status and health: old observations and new thoughts. In: Bunker JP, Gomby DS, Kehrer BH, eds. Pathways to Health: The Role of Social Factors. Menlo Park, Calif: Henry J. Kaiser Family Foundation; 1989.
Hemingway H, Marmot M. Psychosocial factors in the aetiology and prognosis of coronary heart disease: systematic review of prospective cohort studies. Br Med J. 1999; 318: 1460–1467.
Diez-Roux A. Residential environments and cardiovascular risk. J Urban Health. 2003; 80: 569–589.
Tyden P, Hansen O, Engstrom G, Hedblad B, Janzon L. Myocardial infarction in an urban population: worse long term prognosis for patients from less affluent residential areas. J Epidemiol Community Health. 2002; 56: 785–790.
Statistics Canada [homepage on Internet]. 2001 Census Dictionary. 2002. Available at: http://www.statcan.ca/english/census2001/dict/geo035.htm. Accessed August 30, 2003.
Krieger N, Williams DR, Moss NE. Measuring social class in US public health research: concepts, methodologies, and guidelines. Annu Rev Publ Health. 1997; 18: 341–378.
Oakes JM. The (mis)estimation of neighborhood effects: casual inference for a practicable social epidemiology. Soc Sci Med. 2004; 58: 1929–1952.
Greenland S. Ecologic versus individual-level sources of bias in ecologic estimates of contextual health effects. Int J Epidemiol. 2001; 30: 1343–1350.
Syme SL. Strategies for health promotion. Prev Med. 1986; 15: 492–507.
Welin C, Lappas G, Wilhelmsen L. Independent importance of psychosocial factors for prognosis after myocardial infarction. J Intern Med. 2000; 247: 629–639.
Frasure-Smith N, Lesperance F. Depression and other psychological risks following myocardial infarction. Arch Gen Psychiatry. 2003; 60: 627–636.
Kannel WB, Abbott RD. Incidence and prognosis of unrecognized myocardial infarction: an update on the Framingham Study. N Engl J Med. 1984; 311: 1144–1147.(Cathryn Tonne, MPH; Joel )