Table 1

Summary of published studies of area level socioeconomic factors and health outcomes

Reference Outcome Sample Neighbourhood definition (average or stated size) Neighbourhood SES measures Individual SES measures Cofactors Type of model Significant area level effects after controlling for individual SES and cofactors
1 Mortality studies
 Andersonet al, 199755 All cause mortality233 600 black or white adults over age 25 from US National Longitudinal Mortality StudyCensus tract (4000)Median family incomeAnnual family incomeSex, age, raceCox proportional hazards model using robust covariance matrixAmong whites: RR of lowv high income = 1.26 for men, 1.16 for women
Among blacks: RR = 1.49 for men, no effect in black women or in either ethnic group over age 65
Davey Smith et al, 199748 All cause mortality5766 men aged 35–64 recruited from 27 workplaces in West ScotlandPostcode sector (4660)Carstair's index of deprivationCumulative lifetime social class, based on current, early adulthood and father's occupationAgeCox proportional hazardsNone, with cumulative lifetime socioeconomic status in the model
Davey Smithet al, 199856 1-150 All cause mortality14 952 adult residents of West Scotland, aged 45–64Postcode sector (4660)Carstair's index of deprivationOccupational classSex, ageCox proportional hazardsFor most deprived v least deprived areas: in men RR = 1.34 for all cause mortality, 1.26 for CVD mortality; in women RR = 1.29 for all cause mortality, 1.33 for CVD mortality
Haan et al, 198757 All cause mortality1811 residents of Alameda County, California, over age 35Federal poverty area (151 000)Residence within poverty areav notIncome, education, employment, health insuranceSex, age, race, healthLogistic regressionRRs from 1.46–1.53 for residence in poverty area, adjusting for age, sex, race, health, and single measures of individual SES
Kaplan, 199658 As for: Haan, 1987As for: Haan, 1987Census tract (4000)4 scales of neighbourhood type, based on factor analysisIncomeSex, ageCox proportional hazardsRelative hazards from 1.3–2.0 for highv low quartile in all 4 scales
LeClereet al, 199759 All cause mortality346 917 white, black or Hispanic men and women over age 18 from US National Health Interview SurveyCensus tract (4000)Proportion black and Hispanic, median income, education, female headed households, poor female headed householdsEducation, income-to-needs ratioSex, age, race, marital statusCox proportional hazardsMultiple models, eg: for men, RR for high proportion African-American tracts v low proportion = 1.13 to 1.18. For women, RR for high proportion female headed households v low proportion = 1.16, for high v low education = 0.84–0.86
LeClere et al, 199860 Heart disease mortality199 221 white, black or Hispanic women over age 18 from US National Health Interview SurveyCensus tract (4000)Female headed household (female headship), black race, public assistance, poverty, poor female headed households, unemployment, median incomePer capita household income, education, employment statusAge, race, marital status, BMI, pre-existing heart conditions or hypertension, limiting activityCox proportional hazardsMultiple models, eg: in women under 65, effects of all other tract level variables become non-significant when tract level % of female headship included in models. RR from 1.47–1.85 for highest quartile of female headship
Effect of % female headship weaker in older women
Sloggett and Joshi, 199461 All cause mortality before age 70Population-based, longitudinal sample of 292 860 adult residents of England and Wales, ages 16–65Electoral ward (5500)Index based on unemployment, access to car, rental housing, social classOccupational social class, housing tenure, access to carSex, age, region, marital status, presence of spouse, time periodLogistic regressionNo significant area level effects in men, RR = 1.02 in women
Sloggett and Joshi, 199862 1-150 All cause mortality before age 70Population-based, longitudinal sample of over 300 000 residents of England and Wales, ages 10–64Electoral ward (5500)Index based on unemployment, access to car, rental housing, social classOccupational social class, unemployment, housewife status, housing tenure, access to carSex, age, region, time periodLogistic regressionComparing most deprived to least deprived areas: for men RR = 1.02, for women, RR = 1.04
Waitzman and Smith, 199863 All cause mortality, and cardiovascular, cancer, external, and other cause mortality10 161 black or white adults from US National Health and Nutrition Examination Study IFederal poverty area (not given)Residence within poverty area v notEducation, incomeSex, age, race, marital status, alcohol use, BMI, smoking, exercise, health status, hypertension, cholesterolCox proportional hazardsNo significant effect in adults aged 55–74. In adults, 25–54 years, RR = 1.78 for all cause mortality, 1.9 for cardiovascular mortality, 1.95 for cancer mortality
2 Morbidity studies
(a) Infant and child health (See also RR for low birth weight in Sloggett and Joshi, 1998)
 Morgan and Chinn, 198364 Asthma in last year, bronchitis in last year, respiratory illness, hospital stay in last year, physician consult in last 2 weeks4902 children aged 5–11 years resident in England and Scotland from National Study of Health and GrowthEnumeration district (500)ACORN group: classification of area based on factor analysis of 6 SES variablesFather or male guardian's occupational social classSex, ageLogistic regressionEffect of ACORN on bronchitis but not other outcomes.
O'Campo et al, 199747 Low birth weight50 757 births in Baltimore, MarylandCensus tract (4000)Home ownership, number of community groups, unemployment, housing violations, crime rate, wealth, incomeEducation, insurance statusMaternal age, prenatal care, education, health insurance statusHierarchical logistic regressionRR = 1.11 for low per capita income, interaction effects of unemployment and maternal age, crime rate and maternal education, wealth and prenatal care, unemployment and prenatal care, per capita income and health insurance status
Roberts, 199725 Low birth weight112 327 births in the Chicago Statistical Metropolitan AreaCommunity areaSES index, hardship, demography, residential stability, median rent, overcrowding, % black, % HispanicEducationAge, race, marital status, prenatal care, parity, smoking, alcohol useLogistic regressionEconomic hardship and median rent positively associated with low birth weight, area level SES, overcrowding, proportion of young residents and proportion of black residents negatively associated with low birth weight
(b) Chronic disease among adults
 Davey Smith et al, 199856 1-150 Diastolic blood pressure, cholesterol, height, BMI, FEV, bronchitis, angina, ECG ischaemia Cross referenced above Trend test of age adjusted means and proportionsIn women, all outcomes except blood pressure and cholesterol were related to deprivation at p<0.05. Among men, BMI and ECG ischaemia were also not significant.
Diez-Roux, 199769 1-150 Coronary heart disease, serum cholesterol, systolic blood pressure12 601 black or white adults aged 45–64 years from 4 US communities in North Carolina, Minnesota, Mississippi, MarylandCensus block group (1000)Education, median income, occupation, housing valueIncome or education or occupation (each modelled only with area level analogue)Age, field centreHierarchical linear and logistic regressionMultiple models: significant effects of area level SES on blood pressure (mean differences (MD) from 1.8–3.2) and cholesterol (MD from 4.4–5.2). CHD (RR for area level variables from 1.61–2.82) was only significant in women living in white communities but not in predominantly black communities or in men
Humphreys and Carr-Hill, 199165 Self assessed health, long term illness, respiratory function, symptom scores4352 British adults aged over 18 from Health and Lifestyle SurveyElectoral ward (5500)Clusters based on factor analysis of birthplace, unemployment, housing tenure, overcrowding, access to car, social classEducation, overcrowdingSex, age, ethnicity, alcohol use, smoking, exercise, access to gardenHierarchical linear regressionMultiple models: significant negative effects of unemployment, rental housing, no access to car, low social class on self assessed health, symptoms score; effect only of lack of access to car on respiratory function. No effects found for long term illness.
Jones and Duncan, 19954 Self reported health, symptoms of heart disease, respiratory function9003 British adults, aged over 18 from Health and Lifestyle SurveyElectoral ward (5500) and constituencyUrban environment, index of deprivation based on employment, social class, housing tenure and access to car, average household incomeSocial class, employment status, housing class, incomeSex, age, height, smoking, alcohol use, eating, exercisingHierarchical logistic regressionSignificant effect of ward deprivation on all outcomes. For respiratory function, a significant interaction between smoking and urban environment, for self rated health, an interaction between deprivation index and social class, for heart disease symptoms interaction between individual income and deprivation index
Krieger, 199231 1-150 Hypertension, height14 240 black or white members of Northern California Kaiser Permanente Medical Care ProgramCensus tract (4000) and block group (1000)Proportion working class, impoverished and undereducatedWorking class occupationSex, age, race, BMI (for hypertension only)Multiple linear and logistic regressionResults shown for block group level only: coefficient for non-working class people in non-working class block groupv working class block group = −0.9 for height in centimetres, RR = 1.2 for hypertension
Reijneveld, 199866 Self rated health, physical complaints, longterm disability, BMIRandom sample of 5121 adults over 16 years resident in AmsterdamBorough (not given)Mean income, household income below minimum, unemploymentIncome, education, occupationSex, age, standardised mortality ratio of boroughHierarchical logistic regressionRR = 1.2 for most deprived household incomev least deprived and obese BMI, RR = 1.3 for intermediate mean income v least deprived mean income for longterm disability
Robert, 199867 Number of chronic conditions, functional limitations, self rated health3617 adults, over 25 years from Americans' Changing Lives studyCensus tractHouseholds receiving public assistance, unemployment, family income <$30 000, economic disadvantage indexEducation, family income, family assetsSex, age, raceMultiple linear regressionWhen all individual level SES variables in model: economic disadvantage index is a significant predictor of number of chronic conditions, households receiving public assistance is a significant predictor of self rated health
Shouls et al, 199668 Limiting long term illness10% random subsample of adults over age 16 years, from Sample of Anonymised Records from British censusLocal government districtFactors and clusters of census socioeconomic variablesSocial class, deprivation scoreSex, age, ethnicity, marital statusHierarchical logistic regressionMultiple models: health inequalities most pronounced in affluent areas, deprivation most important area level variable, interaction with sex—stronger area level effects in men
Sloggett and Joshi, 199862 1-150 Long term limiting illness, low birth weight Cross referenced above Gender, age, regionLogistic regressionFor long term limiting illness, RR = 1.08 for women and RR = 1.09 for men. For low birth weight, RR = 1.04. All RRs are <1.1 but are statistically significant, although much smaller than RR for individual level variables.
(c) Mental health
 Reijneveld and Schene, 199870 Mental disordersAs for Reijneveld, 1998 aboveSex, age, standardised mortality ratio of borough, plus family type, chronic physical limitationHierarchical logistic regressionNone, with all individual level SES variables in model
3 Health behaviour studies
 Curryet al, 199324 Attitudes towards smoking, alcohol use and dietary fat intakeRandom survey of 8849 adults, aged over 18 in 15 communities in Western USASelf defined geographical communitiesResidence, education, unemployment, residential stabilityEducation, employment, household income, residential stabilitySex, age, race, marital status, community attitudes towards health behaviours, health behavioursMultiple linear regressionSignificant effects of community attitudes and education on attitude to smoking and of residential stability on attitude to fat consumption
Diez-Roux et al, 199769 1-150 Smoking Cross referenced above In mostly white communities: for men RR for smoking = 1.44 for low education, 1.6 for low household income, 1.68 for low housing value, 1.61 for low occupation; for women, RR for smoking = 1.42 for low occupation. In mostly black communities, significant effects in women for all area level effects but not in men, except for low occupation (RR = 1.54)
Karvonen and Rimpela, 199636 Smoking, alcohol use, physical activity, consumption of full fat milkPopulation-based sample of 9121 Finnish teenagers aged 16 and 18 yearsMunicipalityOccupational structure, self sufficiency of employment, level of servicesOccupation of head of household, type of school, type of familyAge, sexLogistic regressionRR for high level of neighbourhood services and full fat milk use = 0.59, RR for low self sufficiency of employment and full fat milk use = 0.59 in boys, girls in areas of high occupational structure more likely to consume full fat milk, RR = 2.36. In boys, effect of education on alcohol use, modified by level of services, RR for alcohol use = 1.35 in areas with low self sufficiency of employment, RR for alcohol use = 1.47 for girls in highly educated areas
Kleinschmidtet al, 199546 SmokingRandom sample of 7866 adults from England's North West Thames regional health surveyElectoral wardCarstair's Index, based on unemployment, overcrowding, access to car, social classSES measured by a 17-level categorical compositeSex, ageHierarchical logistic regressionRR for smoking in most v least deprived areas = 1.52
Krieger, 199231 1-150 Smoking Cross referenced above RR of smoking = 1.4 for non-working class women in non-working class block groupv working class block group
O'Campoet al, 199537 Partner perpetrated violence in childbearing year157 low income women from Baltimore, MarylandCensus tractIncome, home ownership, unemployment, per capita crime rateEmployment, education, incomeAge, race, marital status, social support, parity, partner drug useLogistic regression and hierarchical logistic regressionRR for violence: unemployment = 3.4, low income = 4.4
Reijneveld, 199866 1-150 Smoking Cross referenced above RR for smoking: 1.19 for low average area level income, 1.23 for high % of household poverty, no effect of neighbourhood unemployment rate
  • 1-150 Study is also cross referenced under another outcome.