Local Disparities in the Risk of Serious Mental Illness in Three Nations: Israel, New Zealand, and the United States

BACKGROUND AND PURPOSE. Considerable research on mental health disparities has extensively documented dramatic differentials in serious mental illness based on SES, age, ethnicity, and various other predictors. When these are included in multivariate models, the use of statistical controls have revealed complex, and sometimes counter-intuitive interactions, ones that nonetheless enable considerable levels of predictability. These models are not only of theoretical interest, but have numerous practical applications for social workers. Of particular interest are their applications in local needs assessments, since with the development of regression-based small area estimation techniques, it is now possible to use these methods with national surveys, to reliably estimate local rates of mental illness when the statistical power of the national surveys are insufficient to enable direct local estimates. This study’s purpose, therefore, has been to replicate one such approach, originally implemented in the U.S., and compare the results of studies in Israel and New Zealand.

      METHODS. This project employs a secondary analysis of three national household surveys of mental illness: (i) The National Comorbidity Survey Replication, conducted in the U.S. (2001/2002), with 5,593 adults, aged 18+; (ii) The Te Re Hinegaro Mental Health Survey (2003/2004) in New Zealand, included 12,992 adults aged 16+; (ii) and the Israel National Health Survey (2003/2004), with 4,859 adults age 21+. These studies employed complex probability samples, and state-of-the-art interviewing techniques, using the World Health Organization’s Composite International Diagnostic Inventory (WHO-CIDI). The small area estimation approach consisted of “regression synthetic estimation fitted using area-level covariates”, consisting of a 3-stage methodology for the: (i) estimation of a logistic model of socioeconomic and demographic predictors of serious mental illness; (ii) use of the model, with national census data, to estimate local area rates of serious mental illness; (iii) validation of local estimates, using alternative indicators of serious mental illness, e.g. psychiatric hospitalization rates.

      RESULTS. All three nations have similar levels of SMI, ranging from 4.7% to 5.4%. Gender, age, race/ethnicity, marital status, employment, income, education, and occupational status were used to model individual variations in SMI in all three nations, with sensitivity ranging from 16.9% in New Zealand to 26.9% in Israel, and the Area-under-the-curve ranging from .73-.80 between these same nations. Pearson r’s for estimated rates with validators ranged from .51 (p<.001) in New Zealand to .75 p<.001) in the U.S. In Israel, there was strong validation in predominantly Jewish localities (.63; p<.001), but not in Arab localities (.09; n.s.). Disparities in rates between localities were dramatic in U.S. and Israel, but less so in New Zealand.

      CONCLUSION AND IMPLICATIONS. These studies demonstrate pervasive disparities in SMI rates based on place of residence, reflecting the combined impact of socioeconomic and demographic conditions. They also support the the use of models of such disparities to estimating local rates of serious mental illness. Such estimates are an urgent need for public mental health authorities to guide their allocation of mental health resources between local areas, and to plan for particular forms of programming and outreach in these areas.