022 Health in small places: deriving and validating estimates of common mental disorders at the lower super output areas using microsimulation procedures.

Background There are few data sets that provide detailed population health data at a geographic scale below Government Office Region, and few estimates at a finer scale are validated against real-world data. The focus of the paper is on testing the validity of a methodological approach to create small area population health profiles that may be used in academic and policy research to explore the spatial patterning of health, and repeated as new data become available. The objectives of the study are to 1) simulate estimates of common mental disorders (CMD) in adults at a small area-level and 2) to validate estimates against small area health and socio-economic measures. Methods A deterministic reweighting methodology assigns probabilities of respondents from the 2004–6 annual Health Survey for England (HSE) to live in small areas (Lower Super Output Areas, or LSOAs) based on matching sociodemographic attributes available in both the HSE and the 2001 Population Census. These attributes are chosen because they are strong predictors of CMD (measured by GHQ-12). Gender, social class, economic activity and marital status were used to create estimates of people reporting CMD for each LSOA. These estimates were correlated to LSOA indicators composing the “health domain” of the Index of Multiple Deprivation 2007 (IMD2007) and to other socio-economic information. LSOA estimates were then aggregated at the Local Authority (LA) level and proportions of people reporting CMD were computed; these were then compared to observed prevalence of CMD at the LA level (based on 30 304 HSE respondents nested in 352 LA). Results LSOA CMD estimates were correlated at 0.68 (p<0.001) with adults suffering from mood or anxiety disorders and at 0.83 (p<0.001) with comparative illness and disability ratio. Significant positive correlations between CMD estimates and overall, and domain specific, scores of the IMD2007 were observed. In 90.6% of LA, discrepancies between microsimulated and observed prevalence of CMD were less than 10%. LA where discrepancies were greater than 10% were mostly characterised by small HSE sample size, which may explain why estimates were more inaccurate in these localities. Conclusion The findings indicate that spatial microsimulation might be an appropriate methodological approach for replicating social and demographic patterns of mental health in order to create a small-scale spatial data set. The validation of simulated area-based estimates of mental health presents a viable and cost-effective alternative to local level surveys. [ABSTRACT FROM PUBLISHER]