Patterns of long-term conditions in older age and subsequent mortality: a national study of inequalities in health

Abstract

Background: Multiple long-term health conditions in older people are associated with increased mortality. The study aims to identify patterns of long-term health in a national ageing population using a census-based self-reported indicator of long-term health conditions. We assessed associations with subsequent mortality and socio-economic and demographic risk factors.

Methods: Using linked administrative data from the Northern Ireland Mortality Study, we assessed the presence of latent classes of morbidity in self-reported data on 11 long-term health conditions in a population aged 65 or more (N = 244 349). These classes were associated with demographic and socio-economic predictors using multi-nomial logistic regression. In a 3.75-year follow-up, all-cause and cause-specific mortality were regressed on morbidity patterns.

Results: Four latent classes of long-term ill-health conditions were derived, and labelled: 'low impairment'; 'pain/mobility'; 'cognitive/mental'; 'sensory impairment'. Groupings reflecting higher levels of long-term ill-health were associated with class-specific increases in all-cause and cause-specific mortality. Strongest effects were found for the 'cognitive/mental' group, which predicted all-cause mortality [hazard ratio (HR) = 2.96: 95% confidence interval (CI) = 2.83, 3.10) as well as some cause-specific mortality (i.e. dementia-related death: HR = 10.78: 95% CI = 9.39, 12.15). Class membership was predicted by a range of socio-demographic factors. Lower socio-economic status was associated with poorer health.

Conclusion: Results indicate that long-term ill-health clusters in specific patterns, which are both predicted by socio-demographic factors and are themselves predictive of mortality in the elderly. The syndromic nature of long-term ill-health and functioning in ageing populations has implications for healthcare planning and public health policy in older populations.