This paper is a study of variations of serious mental illness (SMI) throughout New Zealand and is part of a multinational test, which includes the U.S. and Israel, of the application of small area methods to estimating local SMI rates.
Study Objective. This study aims to test and apply a model of variations in risk to estimating prevalence of SMI in local service areas throughout New Zealand using SAE methods.
Methods. This study employs data from the Te Rau Hunengaro Mental Health Survey of 12,992 adults, aged 16+, from the household population. The data collected is based on the Composite International Diagnostic Inventory (WHO-CIDI). This particular project uses small area estimation (SAE) methods, specifically: (i) estimation and testing of a multivariate logistic model of risk of serious mental illness; (ii) use of the foregoing equation for computing estimates, using census data, for local areas; (iii) validation of these estimates against other indicators of SMI prevalence.
Results. The model estimated employed age, ethnicity, marital status, employment, and income to successfully predict 92.2% of respondents' SMI statuses, with a sensitivity of 95.9%, specificity of 16.9%, and an AUC index of .73. The resulting estimates for the 21 District Board areas ranged between 4.1% and 5.7% (95% CIs : +/-.3% to +/- 1.1%). The estimates demonstrated a correlation of .51 (p = .028) against rates of psychiatric hospitalization. The most outstanding disparities were found on the level of the 73 territorial authorities, the rates of which ranged from 3.9% to 6.2%.
Conclusion. The use of SAE methods demonstrated the capacity for deriving local prevalence rates of serious mental illness, ones that can be partially validated against available indicators, the results of which are largely comparable to parallel studies conducted in the U.S. and Israel. The results demonstrate not only the feasibility of the use in local needs assessments, but also show that the most outstanding disparities in risk profiles exist at the local level, between smaller towns and neighborhoods. Thus, the results highlight the importance of targeting resources to areas of greatest need, and of individualizing service planning to local communities. In addition, the results provide powerful confirmation the importance of socio-econonomic status, age, ethnicity, employment, and family supports in determining individual risk for serious mental illness. Additional research is needed to refine the methods used, particularly, the methods for determining confidence intervals of the projected estimates.