The influence of life course socioeconomic position on cognitive function and cognitive decline in older age: the impact of missing data

摘要

Cognitive function has been associated with many physical and mental health conditions, as well as mortality. Cognitive decline is one aspect of ageing that causes anxiety among the general population. Understanding the risk factors which affect cognitive function over the life course is therefore important. One potential risk factor is socioeconomic position (SEP). This thesis investigates the impact of SEP across the life course on crystallized cognitive function and memory decline. The 1946 British birth cohort study and Whitehall II study of British civil servants were used for these analyses. Missing data is a potential source of bias in longitudinal studies, with both SEP and cognitive function predictive of dropout. This thesis therefore considers the impact of methods for dealing with missing data on the findings. A complete case analysis is compared with multiple imputation and Heckman selection models. To compare the suitability of these methods a simulation study was carried out. The Heckman selection method did not perform well in the simulation study. Multiple imputation was the best method of the three considered for data missing not at random. The impact of SEP on cognitive function varied by cohort, as well as SEP and cognitive measures, with father‟s occupational SEP, but not childhood household amenities, associated with crystallized cognitive function in the NSHD after adjustment for later life SEP. Accumulation models were usually supported when considering the life course hypotheses. In some analyses the conclusions varied depending on the missing data methodology utilized. Overall, there was no consistent conclusion as to whether childhood SEP remained a significant predictor of cognitive function in adulthood, but it was not a significant predictor of cognitive decline in Whitehall II after adjustment for later life SEP. Multiple imputation was found to be an appropriate method of dealing with missing data in most situations.