Individual health, neighborhood characteristics, and allocation of primary health care resources

Abstract

Aims: To examine whether neighborhood education and neighborhood income predict incidence rates of coronary heart disease (CHD), beyond individual characteristics (study 1); to examine whether neighborhood deprivation, measured with Care Need Index (CNI), predicts CHD incidence rates, beyond individual characteristics (study 2); to examine whether low scores in a social participation index predict CHD incidence rates, after adjustment for individual characteristics (study 3); and to examine the relationship between CNI and poor self-reported health at neighborhood level, to examine whether the transformed CNI can be used for a total allocation of primary health care resources and to compare the transformed CNI with the official Stockholm model (study 4).

Methods: In study 1 25,319 individuals between 1986 and 1993 from the Swedish Annual Level of Living Survey (SALLS) were followed until December 31, 1997, for CHD incidence events. Neighborhood level characteristics were defined by the use of neighborhood education and neighborhood income. Individual level characteristics were defined as age, sex, and education or income. Multilevel Cox proportional hazard models were used to analyze the data. In study 2 the whole Swedish population, aged 40 64, was followed from December 31, 1995, to December 31, 1999, for CHD incidence events. Multilevel logistic regression was used in the analysis with individual level characteristics (age, income) at the first level and neighborhood deprivation, measured by CNI, at the second level. In study 3 6,861 individuals from SALLS, interviewed in 1990/91 were followed until December 31, 2000, for CHD incidence events. Individual characteristics were age, sex, education, housing tenure, smoking habits, and a social participation index. Cox regression was used in the statistical analysis. In study 4 the population in Stockholm County was divided into deciles by CNI, according to the level of neighborhood deprivation. CNI ratios were calculated for each decile by dividing the CNI means in each decile 2 10 by the CNI mean in decile 1. A sample from SALLS was used to estimate Odds Ratios (OR) for poor self-reported health in the deciles. The ORs were then compared with the CNI ratios. Hierarchical logistic regression was used in the statistical analysis.

Results: In study 1 each neighborhood measure predicted CHD incidence rates after adjustment for individual characteristics (hazard ratios 1.32 and 1.25). CHD events would hypothetically be reduced by 25 26 percent for women and 10 15 percent for men if everyone had the same CHD risk as those living in the most affluent neighborhoods. In study 2 the risk of developing CHD was 87 percent higher for women and 42 percent higher for men in the most deprived neighborhoods than in the most affluent neighborhoods, after accounting for individual characteristics. In both study 1 and study 2 the variance at neighborhood level was small but significant, indicating that there was a neighborhood effect on CHD beyond the individual effect. In study 3 there was a gradient between the social participation index and CHD. After adjustment for individual characteristics, the risk of CHD for persons with low social participation remained high (hazard ratio = 1.69). In study 4 the CNI was transformed into a positive scale. CNI ratios corresponded to the ORs of poor self-reported health in the deciles. The transformed CNI showed a high degree of agreement with the official model.

Conclusions: CHD prevention needs to combine both individual- and neighborhood-level approaches, in order to reduce socioeconomic disparities in CHD. Although the neighborhood effect was small, it is of importance since the outcome, CHD, is highly prevalent among the entire population. The CNI model, which is an exclusively need-based tool, constitutes an attractive approach for the total allocation of primary health care resources.