Understanding policy amenable risk factors: Alcohol consumption and long-term care use among people over 65 years old

This study aims to explore the effect of past alcohol consumption frequency on formal and informal long-term care (LTC) use in old age and explore the different channels through which it may affect LTC use. The existing literature has mainly focused on risk factors associated with a nursing home entry, but this evidence is outdated, not UK-focused, and does not look into other types of care, such as informal care. The results of this study will help in modelling the future demand for various types of care and the corresponding public spending. We use the English Longitudinal Study of Ageing (ELSA) (2002-2017) dataset to conduct longitudinal, individual-level analysis. We explore how the previous frequency of alcohol consumption affects formal and informal care use. We focus on people aged 65 and over with no previous LTC use and run regressions with and without instrumental variables (IV) to estimate how alcohol consumption patterns in the previous wave (2 years before) affect formal and informal care use. For IV regressions, we use the polygenic score for alcohol use, available for a subsample of ELSA respondents, as an instrument while also accounting for sociodemographic characteristics, lifestyle choices, and health conditions. The main IV estimates suggest that frequent alcohol consumption has a weakly significant positive effect on the onset of formal LTC care use compared to none/rare drinking. This relationship diminishes and is not statistically significant when we directly control for health status. We find no statistically significant effect towards informal LTC use. These results contrast with the estimates without IV, which suggest that frequent alcohol consumption is negatively associated with informal care use and no or weakly negative association with formal care use. Our findings suggest that unobserved confounding is important when studying the relationship between alcohol consumption and LTC. We hypothesise that primarily alcohol effects LTC through its adverse effect on health. In addition, unobserved factors like preferences towards seeking care, social behaviour may be related to alcohol consumption and affect access to care. We speculate alcohol may have a damaging effect on personal relationships and could indicate the burden eventually falling on formal care. In as far as the polygenic score IV can account for unobserved preference-behaviour differences, the results (weakly) support the hypothesis that these latter processes are relevant, especially for informal care use.