Generating small-area prevalence of psychological distress and alcohol consumption: validation of a spatial microsimulation method

Abstract

Public mental health surveillance data are rarely available at a fine geographic scale. This study applies a spatial microsimulation procedure to generate small-area (lower super outputs areas [LSOA]) estimates of psychological distress and alcohol consumption. The validity of LSOA estimates and their associations with proximal and broader socioeconomic conditions are examined. A deterministic reweighting methodology assigns prevalence estimates for psychological distress and heavy alcohol consumption through a process of matching individuals from a large, population-representative dataset (Health Survey for England) to known LSOA populations (from the 2001 population Census). 'goodness-of-fit' of LSOA estimates is assessed by their comparison to observed prevalence of these health indicators at higher levels of aggregation (local authority districts [LAD]). Population prevalence estimates are correlated to the mental health needs index (MINI) and other health indicators; ordered logistic regression is applied to investigate their associations with proximal and broader socioeconomic conditions. Performance of microsimulation models is high with no more than 10% errors in at least 90% of LAD for psychological distress and moderate and heavy alcohol consumption. The MINI is strongly correlated with psychological distress (r = 0.910; p value < 0.001) and moderately with heavy drinking (r = 0.389; p value < 0.001). Psychological distress and heavy alcohol consumption are differently associated with socioeconomic and rurality indicators at the LSOA level. Associations further vary at the LAD level and regional variations are apparent. Spatial microsimulation may be an appropriate methodological approach for replicating social and demographic health patterns at the local level.