Abstract
Objectives. We conducted an ecologic analysis to determine whether housing characteristics mediate the associations between concentration of poverty and pediatric injury and between concentration of racial minorities and pediatric injury and whether the association between housing conditions and pediatric injury is independent of other risks.
Methods. We created a hierarchical data set by linking individual-level data for pediatric injury with census data. Effect sizes were estimated with a Poisson model.
Results. After adjustment for owner occupancy and the percentage of housing built before 1950, the association between concentration of poverty and pediatric injury was attenuated. For concentration of racial minorities, only percentage of owner occupancy had some mediating effect. In hierarchical models, housing characteristics remained independent and significant predictors of pediatric injury.
Conclusions. The association between community characteristics and pediatric injury is partially mediated by housing conditions. Risk of pediatric injury associated with housing conditions is independent of other risks.
It is commonly, but not universally, reported that children residing in areas with concentrated poverty or a high concentration of minorities suffer unintentional injury at higher rates than do other children.1–7 However, concentrated poverty and concentration of minorities cannot be considered causal; they are correlates of more proximal determinants of injury. Because young children often suffer injuries at home,8 at least some of these more proximal determinants are likely to be related to housing conditions. To the extent that housing conditions reflect conditions of the larger residential area,9,10 they can be viewed as community-level determinants of injury, but more proximal community-level determinants than social conditions such as poverty. Moreover, because housing conditions are relatively more proximal to injury than social conditions, it is likely that the association between social conditions and injury is, at least partially, mediated by housing conditions. Other factors—particularly those at the individual level, such as parental supervision—are also among the important correlates of pediatric injury. However, the focus of this study was on housing conditions.
Age of housing is a likely determinant of injury.8 Older houses are less likely to be in compliance with building or sanitary codes and may have substandard electrical and heating systems, narrow stairwells, or other safety hazards.11,12 Another likely housing-related determinant of injury is whether a house is owner occupied or rented. Inadequate or deferred maintenance can be a common problem in low-income rental properties,13,14 and high tenant turnover can increase the number of people exposed to these hazards over time.
To date, no studies have examined the association between housing factors and injury in which social conditions (e.g., concentrated poverty) and individual-level determinants (e.g., age, gender) have been simultaneously examined. We present the first multilevel, population-based study of pediatric injury. We also respond to calls for the study of nonfatal injuries,15–17 the most severe of which require hospitalization. Injuries requiring hospitalization are often associated with high treatment and rehabilitation costs18 and appear to have patterns and risks that are distinct from fatal injuries or injuries that do not require hospitalization.16,19 As a consequence, the study of nonfatal, hospitalized injuries can provide information useful for their prevention. Falls and burns are among the most prevalent causes of pediatric injury.20 In our sample, they accounted for 59% of all hospitalized pediatric injuries; after adjustment of hospital charges to true costs, the estimated median cost was $4670 for burns and $2760 for falls.21 As a consequence, these types of injury were used to test the following hypotheses: community-level owner occupancy and age of housing (measured at the zip code level) mediate the association between concentration of poverty and concentration of minorities and risk for pediatric injury, and the association between pediatric injury and community-level housing conditions is independent of individual- and other community-level determinants of injuries requiring hospitalization.
METHODS
Sources of Data
Data from 2 different sources were used in this study. Information for all hospital discharges in the state of Illinois for 1990 through 2000 was abstracted from administrative hospital discharge data compiled as part of an Illinois state mandate22 (as described elsewhere23), and the 1990 US census collected housing information by zip code, including number of owner-occupied units, number of residents living below the federal poverty limit, number of housing units built before 1950, and number of residents, by race. We linked these 2 data sets by zip code and created a hierarchical data structure with both individual- and zip code–level data.
Diagnostic codes were based on the International Classification of Diseases, 9th Revision,24 in which E-codes reflect external causes of injury, poisoning, or other adverse events. An observation was defined as a fall if 1 of the principal or secondary diagnosis codes contained an E-code between 880 and 888. An observation was defined as a burn if it had an E-code between 890 and 899, 924.0 and 924.9, or 925.0. E-codes 925.1, 925.2, 925.8, and 925.9 were defined as nondomestic burns and were excluded. Sixty-nine observations had 2 types of injury E-codes, in which case the first recorded E-code was used. Only children 6 years old and younger were included in the analysis. A total of 241 (2%) observations were excluded because they could not be linked to the 1990 US census. Of these, 96 observations were in zip codes created after 1990, and 29 observations were in zip codes that were either exclusively for post office boxes or for commemorative postal issues. The zip codes of 116 (1%) observations could not be identified in the Postal Bulletin archives, and these codes were also removed. Of the 1240 zip codes listed for the state of Illinois in 1990, 12 had no households or no children 6 years old or younger. Thus, 11 735 observations in 1228 zip codes remained for analysis.
Statistical Analyses
Data for this study involved counts of discharged injuries nested within a zip code. Because counts of injuries at the zip code level are bounded at zero and approximate a binomial distribution with a large number of trials and a small probability of success, the data assumed the shape of a Poisson distribution. Therefore, Poisson regression was used to model the rate of injury as a function of individual and community variables. An assumption of the Poisson distribution is that each individual observation is an independent event. In this study, some observations may have represented the same person or perhaps children from the same family, thus violating this assumption. We examined the extent of this violation, and our analysis indicated that multiple discharges for the same child were rare. Random samples of 1000 falls and 1000 burns were matched by birth date (month, day, and year), gender, and insurance type. Only 2 falls and 10 burns matched with 1 other observation on all 3 characteristics. When we used only birth date and gender, 3 falls and 13 burns had multiple observations. Thus, only about 1% of the falls and burns might have been repeat discharges of the same person. Moreover, the low probability of falls or burns in the population further limited the likelihood that the assumption of independence was violated. For these reasons, we assumed that the events were independent. Also, whereas violation of independence may have led to overestimation of injury rates, lack of independence will not bias rate ratios so long as the extent of overestimation is comparable across subgroups. Finally, offsets for rate calculations were based on the 1990 census population of children by age and gender.
The hierarchical structure of the data led to correlation between counts that were nested within zip codes. Thus, estimates of rates and rate ratios of injuries and their associated 95% confidence intervals were calculated using Generalized Estimating Equations, a method that allows for specification of the Poisson distribution and accommodates correlated data.25 An exchangeable correlation structure was assumed, and standard errors for the parameters were based on the empirical estimator of the covariance matrix of the estimated coefficients. Rate ratios and their 95% confidence intervals were calculated in SAS with the GENMOD26 procedure.
Chi-squared goodness-of-fit tests27 indicated some lack of fit, so the standardized Pearson residuals were examined, and 3 outlying zip codes that were exerting undue influence on the coefficients were removed. Next, residual plots versus the predicted values and the covariates were examined. No systematic biases were detected that would indicate nonlinearity.
The distributions of concentrated poverty and concentration of minorities were also divided into decile indicator variables, and saturated models were fit. Because plots of the point estimates and confidence intervals evinced a slight curvature, both of these variables were divided into tertiles. This procedure was informed by data from the 1990 US Census. In 1990, approximately 12% of the population was African American28; this proportion was unchanged in 2000.29 Between 1990 and 2000, the percentage of children living in poverty declined from approximately 20% to 16%.30 Thus, the 3 categories for each of the census indicators corresponded roughly to below average, average, and above average for concentrated poverty and percentage African American in the United States.
Multivariate analyses followed the study aims. First, hierarchical models were fit to examine whether owner occupancy and age of housing mediated the association between concentration of poverty or concentration of minorities and pediatric injury. For example, the mediating effect of owner occupancy was examined by entering this variable in a model that, in addition to the individual-level variables, included either zip code–level concentrated poverty or concentrated minority. A significant change in the coefficient for concentrated poverty or concentrated minority indicated that its association with injury was mediated by owner occupancy (see D’agostino31 for the significance test). Multivariate hierarchical models also were developed both with and without interactions.
RESULTS
In the state of Illinois, from January 1, 1990, through September 30, 2000, there were 11 735 hospital discharges of children with nonfatal injuries coded as occurring at home. The annual incidence rates for the 2 most prevalent types of injury, falls (43.5%) and burns (15.2%), were 3.93/10 000 population aged 6 years or younger (95% confidence interval [CI] = 3.83, 4.04) and 1.37/10 000 population aged 6 years or younger (95% CI = 1.31, 1.44), respectively. Fifty-eight percent of burns occurred among children aged 1 through 2 years, whereas falls were fairly evenly distributed across age groups. The median length of hospital stay for burns and falls was 5 and 2 days, respectively.
Bivariate analyses of individual-level variables indicated that compared with children aged 5 through 6 years, infants (< 1 year) were more likely to suffer an injury resulting from a fall, and toddlers (aged 1 through 2 years) were most likely to suffer a burn. Males were at a higher risk for both types of injury. Bivariate analyses of the zip code–level variables for tenancy and age of housing demonstrated the change in risk for falling or being burned following a 10% increase in the continuous explanatory variable. For every 10% increase in the proportion of owneroccupied units, risk for falling decreased by 16% and risk for being burned decreased by 27%. A 10% increase in the proportion of housing built before 1950 was associated with a 17% increase in risk for falling and a 34% increase in risk for being burned. Children residing in zip codes with the highest concentrations of poverty, compared with residents of zip codes with the lowest concentrations of poverty, were significantly more likely to sustain a fall or burn. Similarly, children residing in zip codes with the highest concentrations of minorities were significantly more likely to sustain a fall or burn than residents of zip codes with the lowest concentrations of minorities.
Next we examined the mediating effect of housing conditions on risk for injury associated with concentrated poverty and a concentration of minorities (Figures 1a ▶–2b ▶). The association between concentrated poverty and risk for falling or being burned was mediated by both owner occupancy and age of housing; the mediation was considerably larger and significant for the top tertile. When both housing conditions were included in the model for falls, the risk ratio for areas with a medium concentration of poverty became insignificant, and the risk ratio for areas with a high concentration of poverty remained only marginally significant. For burns, the risk ratios for middle and high levels of poverty were reduced but remained significant when both housing conditions were included in the model.
FIGURE 1—
Old housing and owner occupancy as mediators of risk of (a) falls and (b) burns associated with poverty.
aThe estimate is significantly different from that of the crude model.
FIGURE 2—
Old housing and owner occupancy as mediators of risk of (a) falls and (b) burns associated with minority concentration.
aThe estimate is significantly different from that of the crude model.
For concentrated minority populations, only owner occupancy appeared to be a mediator. Again, the mediation was larger and significant for the top tertile. Inclusion of age of housing actually increased the risk ratio for the middle tertile and only slightly reduced the risk ratio for the top tertile. The full model for falls (i.e., including both housing conditions) rendered significant the associations for areas with both medium and high minority concentration. The full model for burns evinced a significant association for areas with a high minority concentration, but the risk ratio for areas with a medium concentration became insignificant.
In the final hierarchical models (Table 1 ▶), both owner occupancy and age of housing remained significant predictors of both types of injury. We also found that owner occupancy modified risk for injury associated with concentration of poverty and minority concentration. In particular, owner occupancy was more protective in areas with higher concentrations of poverty and was less protective in areas with higher minority concentration and old housing.
TABLE 1—
Hierarchical Fall and Burn Models
| Fall Rate Ratio | (95% CIa) | Burn Rate Ratio | (95% CIa) | |
| Individual | ||||
| Age, y | ||||
| < 1 | 2.13 | (1.94, 2.34) | 8.16 | (6.77, 9.83) |
| 1–2 | 1.11 | (1.01, 1.23) | 8.28 | (6.92, 9.92) |
| 3–4 | 0.97 | (0.88, 1.07) | 2.02 | (1.64, 2.49) |
| 5–6 (reference group) | 1.00 | . . . | 1.00 | . . . |
| Gender | ||||
| Male | 1.45 | (1.35, 1.55) | 1.32 | (1.18, 1.48) |
| Female (reference group) | 1.00 | . . . | 1.00 | . . . |
| Zip Code | ||||
| Percentage owner-occupied housingb | 0.94 | (0.90, 0.99) | 0.92 | (0.84, 1.00) |
| Percentage housing built before 1950b | 1.10 | (1.06, 1.15) | 1.11 | (1.04, 1.18) |
| Concentrated poverty | ||||
| High | 1.05 | (0.86, 1.27) | 2.10 | (1.56, 2.83) |
| Middle | 1.02 | (0.83, 1.24) | 1.79 | (1.36, 2.36) |
| Low (reference group) | 1.00 | . . . | 1.00 | . . . |
| African American population | ||||
| High | 1.92 | (1.55, 2.36) | 2.64 | (1.84, 3.79) |
| Middle | 1.43 | (1.17, 1.74) | 1.24 | (0.88, 1.74) |
| Low (reference group) | 1.00 | . . . | 1.00 | . . . |
aWald confidence interval.
bRate ratios per 10% increase in the value of the independent variable.
DISCUSSION
Our findings indicate that the community-level concentration of owner-occupied housing and age of housing are significantly associated with rates of nonfatal hospitalized pediatric injury. This is in line with earlier findings that a variety of negative health outcomes, including fatalities caused by house fires, all-cause mortality, and elevated lead blood levels, are significantly more prevalent among renters than among homeowners.1,2,32,33 This study further illuminates the association between housing conditions and health in 2 important ways. First, our results indicate that housing conditions mediate the association between community characteristics, such as concentrated poverty, and pediatric injury. Second, the results of our hierarchical analyses demonstrate that the association between housing conditions and pediatric injury is independent of both individual- and other community-level determinants of injury.
However, the association between these community characteristics and injury does remain significant after control for housing conditions. That the association between concentrated poverty and injury is only partially mediated by housing conditions may be in part the result of the fact that during most of the period covered by this study, some of the zip codes with concentrated poverty also had a concentration of federally owned or subsidized housing, which must meet minimum maintenance standards. Partial mediation of the association between concentration of minorities and injury is consistent with the findings of previous studies of racial disparities in health. These studies found that a significant proportion of health-related racial disparities remain unexplained even after control for social factors, probably because of the myriad of pathways by which race relations can influence health and because of the difficulties in capturing all of these influences in 1 study.34–39
The implications of these findings should be considered in light of the study’s limitations and strengths. A common shortcoming of ecologic studies such as this is the use of broadly defined and heterogeneous geographic areas (zip codes in our case) as the unit of analysis.40,41 This misspecification can dilute the effect of interest and result in underestimation of the true effects. This misspecification also can compromise the accuracy by which community-level variables allow measurement of the underlying construct of interest. Given the heterogeneity of zip codes, underestimation of the relative risks and a degree of residual confounding are both possibilities in this study.
The main threat to validity of any ecological study is ecological fallacy, and we addressed this issue both methodologically and logically. In addition to controlling for key confounders, we limited our inferences at the ecological level (i.e., inferences regard group rates and not individuals’ risk).42,43 As outlined below, the validity of our conclusions are bolstered by their logical plausibility, an important criterion of validity for any study.44 Arguably the most prominent source of ecological fallacy that remains unaccounted for in this study is parental child supervision and its correlates such as race/ethnicity and income. Clearly the link between poor supervision of children and injury (e.g., placing high chairs close to windows without screens) can masquerade as an effect of housing conditions. (It can be logically argued that given the same degree of parental supervision, children residing in well-maintained homes would be less likely to suffer injuries.) It is also reasonable to question whether housing conditions are themselves proxies for poverty and compromised ability to supervise children or are proxies for racial/ethnic norms about parenting.
In regard to poverty, although several developmental models evince a negative correlation between income and the quality of parenting,45,46 material deprivation is not uniformly linked with poor child health outcomes or parenting practices. For example, lowincome Hispanic women can have better pregnancy outcomes than wealthier residents of the same area, a paradox that may be attributable to social cohesion and the presence of extended family members among this population.47–49 We suggest that social cohesion and the presence of extended family members also can promote better child supervision.
In regard to racial/ethnic differences in parenting, the literature generally indicates that correlates of poor parenting skills and supervision are not differentially distributed by race/ethnicity. Although certain parenting practices are more prevalent among various racial/ethnic groups, these differences are not necessarily correlated with poor developmental outcomes.50 Moreover, the influence of race/ethnicity is often confounded by socioeconomic status and recency of immigration.51,52 However, it appears that the effect of socioeconomic status is stronger and absorbs that of race/ethnicity.53,54
Another persuasive piece of evidence for the independent contribution of housing conditions to risk for pediatric injury is the significant and sustained reduction in pediatric falls and burns that is attributed to the installation of window guards and sprinklers.55,56 These passive measures address building defects and not parenting skills. In these cases, even when a high chair is placed next to a window, the window guard will prevent a fall regardless of the parents’ income or skills. Thus, we suggest that the reported mediation by housing conditions is not simply a reflection of ecological fallacy. Future studies could be more persuasive by adjusting for more individual-level correlates of injury.
Finally, because our results are from racially segregated geographic areas, the generalizability of our findings may be limited to such areas. However, racial segregation is unfortunately a defining characteristic of many US cities.35,39 Subsequently, our findings are likely to be generalizable to a broad crosssection of residential areas in the United States.
A strength of this study is our use of both individual- and community-level data in a hierarchical design. To our knowledge, this is the first study of injury that has used such a design. Another strength is our analysis of incident cases of acute health outcomes. This feature discounts 2 competing explanations that often limit inference regarding the effect of place on health. First is the argument that the health outcomes attributed to a place may be simply long-term consequences of individual-level characteristics.57 Although this argument can be persuasive in the case of prevalent chronic illnesses, the argument that acute events such as injury among children age 6 years or younger are long-term consequences of individual-level characteristics is far less compelling. The second competing explanation, “social drift,”57 suggests that those in poor health, through loss of economic status, “drift” to poorer residential areas. In this study, by examining incident cases among young children, we have rendered social drift an unlikely explanation for the disparate concentration of injury in poor and wealthy areas.
The association between housing conditions and pediatric injury has both immediate and long-term implications. The immediate implications regard the efficacy of intervention and prevention programs.58 These programs can increase their efficiency and efficacy by considering both housing and socioeconomic characteristics of the community. Although not the focus of this study, we did find evidence of some interesting interactions. For example, we found that owner occupancy is less protective in areas with a high proportion of old housing. We also found owner occupancy to be more protective in areas with a higher concentration of poverty. This indicates that where a high proportion of homes were built before 1950, interventions should target old housing, whereas in high-poverty areas, the first priority should be non–owner-occupied homes.
The long-term implications are predicated on replication of these findings in longitudinal studies that consider more individual-level variables, including parenting practices, conducted with smaller units of analysis. If such studies, as we suspect, further support our conclusions, then the long-term implications pertain to remediation of social disparities in health through remediation of differences in housing. Housing is an indicator (at times even a leading indicator) of conditions prevalent in the larger residential area. For example, age of housing reflects neighborhood economic conditions and can foretell future economic patterns. In 1990, housing units in high-poverty US census tracts, compared with low-poverty census tracts, were on average 11 years older and had correspondingly lower market value.9 The units’ low market values are a disincentive for maintaining them and a harbinger of worse physical conditions and further reductions in value. Concentration of units with low market value can initiate a cycle of private and institutional divestments, such as “redlining” by financial institutions, which can lead to the deterioration of neighborhoods’ social, financial, and physical resources.9,10 In turn, poor housing conditions have important and wide-ranging health implications.9,33,59–61
Vigorous application of already existing programs62 to improve housing conditions may have benefits beyond the immediate residences and their occupants. This study indicates that social disparities in health may be addressed, at least partially, through remediation of social disparities in housing—a remediation that also can benefit future generations.63,64
Acknowledgments
Supported in part by grants from National Center for Healthy Housing (M. J. Brown and E. D. Shenassa) and the Harvard Injury Control Research Center (E. D. Shenassa).
We thank Drs Kimberly Lochner and Stephen Gilman for their insightful comments on an earlier version of the manuscript, and Meredith Eastman for her invaluable assistance with data management.
Human Participant Protection This study was exempt from human subject review because the secondary data used for the analysis did not contain personal data.
Contributors E. D. Shenassa contributed to the conceptual development of the study, the data analysis plan, and the writing of the article. A. Stubbendick contributed to the development of the data analysis plan, the analysis of data, and the Methods and Results sections of the article. M. J. Brown contributed to the conceptual development of the study, the data analysis plan, and the writing of the article.
Peer Reviewed
References
- 1.Baker S, O’Neill B, Ginsburg M, Guohua L. The Injury Fact Book. 2nd ed. New York, NY: Oxford University Press; 1991.
- 2.Mallone S, Istre G, Rosenberg M, et al. Surveillance and prevention of residential-fire injuries. N Engl J Med. 1996;335(1):27–31. [DOI] [PubMed] [Google Scholar]
- 3.Istre G, McCoy M, Osborn L, Barnard J, Bolton A. Deaths and injuries from house fires. N Engl J Med. 2001;344:1911–1916. [DOI] [PubMed] [Google Scholar]
- 4.Office of Population Censuses and Surveys. Occupational Mortality: Childhood Supplement. London: HMSO; 1988.
- 5.Hippisley-Cox J, Groom L, Kendrick D, Coupland C, Webber E. Cross sectional survey of socioeconomic variations in severity and mechanism of childhood injuries in Trent 1992–7. BMJ. 2002;324:132–1138. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Mierley M, Baker S. Fatal house fires in an urban population. JAMA. 1983;249(11):1466–1468. [PubMed] [Google Scholar]
- 7.Lyons R, Delahunty A, Heaven M, McCabe M, Howard A, Nash P. Incidence of childhood fractures. BMJ. 2000;320:149. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Macfarland A. Child deaths from accidents: place of accident. Popul Trends. 1979;15:10–15. [Google Scholar]
- 9.Jargowsky P. Poverty and Place: Ghettos, Barrios and the American City. New York, NY: Russell Sage Foundation; 1996.
- 10.Wilson W. When Work Disappears: The World of the New Urban Poor. New York, NY: Vintage Books; 1997.
- 11.Runyan C, Bangdiwala S, Linzer M, Sacks J, Butts J. Risk factors for fatal residential fires. N Engl J Med. 1992;327(12):859–863. [DOI] [PubMed] [Google Scholar]
- 12.Ashley DL, Menkedick JR, Wooton JV. Healthy Home Issue: Injury. External Review Draft, Version 2. Washington, DC: US Department of Housing; 2002.
- 13.Cummins S, Jackson R. The built environment and children’s health. Pediatr Clin North Am. 2001;48(2):1241–1256. [DOI] [PubMed] [Google Scholar]
- 14.Ian R. Deaths from injuries in house fires: Fanning the flames of child health advocacy? BMJ. 1995;311:1381–13828520258 [Google Scholar]
- 15.Grossman D. The history of injury control and the epidemiology of child and adolescent injuries. Future Child. 2000;10(1):23–52. [PubMed] [Google Scholar]
- 16.Runyan C, Gerken E. Epidemiology and prevention of adolescent injury: A review and research agenda. JAMA. 1989;262:2273–2279. [PubMed] [Google Scholar]
- 17.Peel NM, Kassulke DJ, McClure RJ. Population based study of hospitalised fall related injuries in older people. Inj Prev. 2002;8:280–283. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Centers for Disease Control and Prevention. Cost of injury – United States: a report to Congress, 1989. MMWR Morb Mortal Wkly Rep. 1989;38(43):743–746. [PubMed] [Google Scholar]
- 19.Pickett W, Harting L, Brison R. A population-based study of hospitalized injuries in Kingston, Ontario, identified via the Canadian Hospitals injury reporting and prevention program. Chronic Dis Can. 1997;18:61–69. [PubMed] [Google Scholar]
- 20.Rivara F. Pediatric injury control in 1999. Pediatrics. 1999;103(4):883–888. [PubMed] [Google Scholar]
- 21.Haddis AC, Teutsch SM, Schaffer PA, and Dunet DO, eds. Prevention Effectiveness: A Guide to Decision Analysis and Economic Evaluation. London: Oxford University Press; 1996.
- 22.Illinois Health Financing Reform Act. Public Act 83–1243.
- 23.Shenassa E, Catlin S, Buka S. Lethality of firearms relative to other suicide methods: a population based study. J Epidemiol Community Health. 2003;57:120–124. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24.International Classification of Diseases, 9th Revision. Geneva, Switzerland: World Health Organization; 1980.
- 25.Liang K, Zeger, S. Longitudinal data analysis using generalized linear models. Biometrika. 1986;73:13–22. [Google Scholar]
- 26.SAS/STAT User’s Guide, Version 8. Cary, NC: SAS Institute Inc; 1999.
- 27.Agresti A. An Introduction to Categorical Data Analysis. New York, NY: Wiley; 1996.
- 28.Shea M. Dynamics of Economic Well-Being: Poverty, 1990 to 1992. US Bureau of Census Current Population Reports P70–42. Washington, DC: US Government Printing Office; 1995.
- 29.Census 2000: General Population Characteristics. Washington, DC: US Bureau of the Census; 2002. Available at: http://www.census.gov. Accessed November 21, 2002.
- 30.Dalaker J. Poverty in the United States: 2000. US Census Bureau Current Population Reports, Series P60–214. Washington, DC: Government Printing Office; 2001.
- 31.D’Agostino RB Sr, Grundy S, Sullivan LM, Wilson P. Validation of Framingham coronary heart disease prediction score: results of a multiple ethnic groups investigation. JAMA. 2001;286:180–187. [DOI] [PubMed] [Google Scholar]
- 32.Sargent J, Brown MJ, Freeman J, et al. Childhood lead poisoning in Massachusetts communities: its association with sociodemographic and housing characteristics. Am J Public Health. 1995;85:528–534. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 33.Filakti H, Fox J. Differences in mortality by housing tenure and by car access from the OPCS Longitudinal Study. Popul Trends. 1995;81:27–30. [PubMed] [Google Scholar]
- 34.Otten MW, Teutsch SM, Williamson DF, Marks JM. The effects of known risk factors on the excess mortality of black adults in the United States. JAMA. 1990;263(6):845–850. [PubMed] [Google Scholar]
- 35.Massey D, Denton N. American Apartheid: Segregation and the Making of the Underclass. Cambridge, Mass: Harvard University Press; 1996.
- 36.Jackson S, Anderson R, Johnson N, Sorlie P. The relation of residential segregation to all cause mortality: a study in black and white. Am J Pub Health. 2000(4);90:615–616. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 37.Massey DS, Gross AB, Eggers ML. Segregation, the concentration of poverty and the life chances of individuals. Soc Sci Res. 1991;20:397–420. [Google Scholar]
- 38.Lanphear BP, Weitzman M, Eberly S. Racial differences in urban children’s environmental exposure to lead. Am J Public Health. 1996;86:1460–1463. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 39.Polednak AP. Segregation, Poverty, and Mortality in Urban African Americans. New York, NY: Oxford University Press; 1997.
- 40.Krieger N. Use of census-based aggregate variables to proxy for socioeconomic group: evidence from national samples. Am J Epidemiol. 1999;150(8):892–896. [DOI] [PubMed] [Google Scholar]
- 41.Pickett K, Pearl M. Multilevel analyses of neighborhood socioeconomic context and health outcomes: a critical review. J Epidemiol Community Health. 2001; 55:111–122. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 42.Susser M. The logic in ecological: II. The logic of design. Am J Public Health. 1994;84(5):830–835. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 43.Morgenstern H. Uses of ecologic analysis in epidemiologic research. Am J Public Health. 1982;72:1336–1344. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 44.Cohen B. In defense of ecologic studies for testing a linear no threshold theory. Am J Epidemiol. 1994;139:765–768. [DOI] [PubMed] [Google Scholar]
- 45.Brody GH, Flor DL. Maternal resources, parenting practices, and child competence in rural, single-parent African American families. Child Dev. 1998;69:803–816. [PubMed] [Google Scholar]
- 46.McLoyd VC. The impact of economic hardship on African American families and children: psychological distress, parenting, and socioemotional development. Child Dev. 1990;61:311–346. [DOI] [PubMed] [Google Scholar]
- 47.Balcazar H, Aoyama C, Xi C. Interpretative views on Hispanics’ prenatal problems of low birth weight and prenatal care. Public Health Rep. 1991;106:420–426. [PMC free article] [PubMed] [Google Scholar]
- 48.Gorman B. Racial and ethnic variation in low birthweight in the United States: individual and contextual determinants. Health Place. 1999;5(3):195–207. [DOI] [PubMed] [Google Scholar]
- 49.Palloni A, Morenoff JD. Interpreting the paradoxical in the Hispanic paradox: demographic and epidemiologic approaches. Ann N Y Acad Sci. 2001;954:140–174. [DOI] [PubMed] [Google Scholar]
- 50.Ogbu JU. A cultural ecology of competence among inner-city blacks. In: Spencer ME, Brookins GK, Allen WR, eds. Beginnings: The Social and Affective Development of Black Children. Hillsdale, NJ: Erlbaum; 1985:45–66.
- 51.García Coll C, Lamberty G, Jenkins R, et al. An integrative model for the study of developmental competencies in minority children. Child Dev. 1996;67:1891–1914. [PubMed] [Google Scholar]
- 52.García Coll C, Magnusson K. Cultural influences on child development: are we ready for a paradigm shift? In: Matsen AS, ed. Cultural Processes in Child Development. The Minnesota Symposia on Child Psychology. Vol. 29. Mahwah, NJ: Erlbaum; 1999:1–24.
- 53.García Coll C, Meyer E, Britten S. Ethnic and minority parenting. In: Bornstein MH, ed. Handbook of Parenting. Vol. 2. Biology and Ecology of Parenting. Mahwah, NJ: Erlbaum; 1995:189–210.
- 54.Duncan GJ, Brooks-Gunn J. Consequences of Growing Up Poor. New York, NY: Russell Sage Foundation; 1997.
- 55.Bradley RH, Corwyn RF, McAdoo HP, García Coll C. The home environments of children in the United States, Part I: variations by age, ethnicity, and poverty status. Child Dev. 2001;72(6):1844–1867. [DOI] [PubMed] [Google Scholar]
- 56.Spiegel CN, Lindaman FC. Children can’t fly: a program to prevent childhood morbidity and mortality from window falls. Am J Public Health. 1977;67:1143–1147. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 57.Diez Roux AV. Invited commentary: Places, people, and health. Am J Epidemiol. 2002;155(6):516–519. [DOI] [PubMed] [Google Scholar]
- 58.Shenassa E. Delivering the goods: the relevance of classification accuracy to the design of community intervention and prevention programs. J Community Psychol. 2002;30:197–210. [Google Scholar]
- 59.Evans G, Chan H-Y, Wells N, Saltzman H. Housing quality and mental health. J Consult Clin Psychol. 2000;68(3):526–530. [DOI] [PubMed] [Google Scholar]
- 60.Calsyn R, Winter J. Social support, psychiatric symptoms and housing: a causal analysis. J Community Psychol. 2002;30(3):247–259. [Google Scholar]
- 61.Clark C, Bornschein P, Succop S, Que Hee S, Hammond P, Peace B. Condition and type of housing as an indicator of potential environmental lead exposure and pediatric blood lead level. Environ Res. 1985;38:46–52. [DOI] [PubMed] [Google Scholar]
- 62.Brown MJ, Gardner J, Sargent J, Swartz K, Hu H, Timperi R. The effectiveness of housing policies in reducing children’s lead exposure. Am J Public Health. 2001;91(4):621–624. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 63.Rose G. Sick Individuals and sick populations. Int J Epidemiol. 1985;14:32–38. [DOI] [PubMed] [Google Scholar]
- 64.Shenassa E. Society, physical health and modern epidemiology. Epidemiology. 2001;12:467–470. [DOI] [PubMed] [Google Scholar]