Mathias Meijer

Through the 1990s there has been increasing scientific interest in how contexts, especially neighborhoods, influence individual health. Research suggests that peoples’ health is shaped not only by individual-level factors such as biology, demography and socioeconomic status (SES), but also by the neighborhoods in which they live. Individual health behavior is affected through interaction with fellow inhabitants and by the physical characteristics of the neighborhood. Both the social and the physical determinants are influenced by the ethnic and socioeconomic composition of inhabitants. This thesis consists of four papers analyzing the association between neighborhoods and health. The aims were to

• to conduct a systematic review of multilevel studies controlling for individual SES to evaluate if neighborhoods affect mortality and cancer-specific incidence and to conduct a meta-analysis investigating the association between area-level socioeconomic status (ALSES) and all-cause mortality.

• investigate how best to measure ALSES with a single indicator and to conduct multilevel modeling investigating how ALSES and population density affect individual all-cause mortality.

• conduct multilevel modeling evaluating the effects of population density and ALSES on breast, lung and prostate cancer incidence.

• construct an empirically based Danish deprivation index capable of explaining variation in health on a small-scale geographic level.

A systematic review was conducted to investigate if neighborhoods affect mortality and cancer-specific incidence (paper 1). By searching five databases a total of 40 studies were found eligible for the systematic review while 18 studies qualified for the meta-analysis. No clear associations were found for income inequality or social cohesion. Studies including more than one area level suggested that all levels contribute to variation in mortality. It was also found that studies including lag time between neighborhood influences and health outcomes found greater effects. In the meta-analysis all-cause mortality was found to be significantly higher among inhabitants living in areas with low SES (OR=1.05, 95% CI=1.04-1.06) compared to those living in affluent areas. Associations were stronger for men, younger age groups and in studies analyzing geographical units with fewer inhabitants. No effects were found for the type of welfare state regime in which the studies were conducted or for the number of covariates controlled for.

Two multilevel analyses were conducted to investigate area effects on all-cause mortality (paper 2) and cancer-specific incidence (paper 3) in Denmark. All individuals with residence in Denmark in 2004 between 30-81 years (all-cause mortality), 30-83 years (breast cancer incidence) and 50-83 years (prostate and lung cancer incidence) were followed through 2006 (all cause) and 2008 (breast, prostate and lung cancer incidence). Frailty modeling was conducted and age, sex, marital status, education, disposable income and occupational SES were adjusted for on the individual level. On the area-level the effect of population density and ALSES was examined.

Results in paper 2 showed that living in areas with the lowest population density was associated with reduced mortality among individuals between 30 and 49 years, HR:0.85 (95% CI=0.76-0.95), compared to those living in areas with the highest population density. The effects were HR:0.81 (95% CI=0.76-0.86) and HR:0.86 (95% CI=0.83-0.89) for individuals aged 50-64 years and 65-81 years respectively. In addition, living in the most deprived areas was associated with excess mortality in the two older age groups, HR:1.05 (95% CI=1.01-1.09) and HR:1.05 (95% CI=1.02-1.07) respectively. No association was found between all-cause mortality and ALSES for individuals aged 30-49 years.

In paper 3 a reduced risk of breast cancer in areas with lower population density was found HR:0.93 (0.86-0.99) compared to areas with higher population density. There was no effect of ALSES. For prostate cancer higher risk was found among inhabitants in affluent areas HR:1.14 (95% CI=1.08-1.21) compared to those living in deprived areas. No effect was found for population density. Lung cancer risk was lower in the least densely populated areas HR:0.80 (95% CI=0.74-0.85) and in affluent areas HR:0.88 (95% CI=0.84-0.92) while being controlled for each other and for individual-level characteristics.

In paper 4 an area-based deprivation index was constructed for Denmark. Individual-level data of all Danes (N=5,391,995) were aggregated to 11 indicators describing parishes (N=2,113) by characteristics in income, employment, education, health, housing, demography and crime. A principal component analysis was conducted to determine the relative weights of the variables and to reduce them to a set of components. The index was validated using standardized mortality ratios (SMRs) in all parishes in 2005. To compare the strength of the index with the Townsend index and to evaluate to what extent the index could be used as a proxy for individual-level SES factors, index scores were applied to 2.7 million individuals in Denmark in a shared frailty model evaluating the risk of death between 2004 and 2006 and compared to a similar model containing individual education, income and occupation-based SES. The index measures material deprivation on one dimension and socioeconomic deprivation on another dimension. There were clear gradients in SMRs in both dimensions: SMR for men in materially deprived parishes were 1.06 (1.05-1.07) and 0.79 (0.76-0.82) in affluent areas. In parishes deprived on the socioeconomic dimension SMR for men was 1.14 (1.12-1.16) and 0.88 (0.87-0.89) in the more affluent areas. The same pattern existed among women although the effects were smaller. The index was better at explaining variation in all-cause mortality compared to the Townsend index (76% vs. 69.6% of frailty variation) and was able to account for as much variation as individual SES factors could (76% of frailty variation). The index can be used to identify Danish parishes by their material and socioeconomic status and as a predictor of mortality on the area-level. The index provides policy makers with a tool to allocate health related resources and can assist in planning area-specific health interventions. It can also be used in epidemiological studies investigating or adjusting for area-effects on individual health outcomes.

Timeframe: 01 July 2008 - 01 July 2011
Supervisors: Christiane Stock, SDU; Kim Bloomfield, Aarhus University
External funding: Helsefonden and Danish Research Council

Download (pdf) full PhD dissertation without articles