Objective It is important for health policy and expenditure projections to fully understand the relationship between age, death and expenditure for health care (HC). Evidence shows that older age groups incur less acute HC costs than previously anticipated and that time to death (TTD) may be a much stronger indicator for HC expenditure than age alone. Our objective was to undertake research into the relationship between TTD, age and HC expenditure in Scotland.

Methods We use a longitudinal data set (the Renfrew/Paisley study, one of the Midspan studies) covering a period of 35 years, and including baseline survey data linked to subsequent hospital admissions (Scottish Morbidity Records) and death records, to estimate the independent effect that age and TTD have on expenditure for acute inpatient care. We include variables previously omitted in other studies such as deprivation category. This is of special interest in Scotland, with high levels of deprivation in some areas and a poor record of premature deaths. We estimate HC costs in the last 60 quarters before death. We use speciality and hospital specific cost data from the Scottish Cost Book. We estimate HC expenditure conditional on having had a hospitalisation within a particular quarter before death. Due to the skewed nature of our cost variable we employ a generalised linear regression model with a Gamma distribution and a log link and use clustering on patient identifier to control for unobserved within-patient effects.

Results Regression results for age and its quadratic form show that on average older individuals incur higher costs (p<0.001) but at a decreasing rate, which means that for the very old costs are declining. After including a measure of remaining TTD we find that age at death has a non significant association with costs but TTD itself is a significant predictor for costs (p<0.001). We find deprivation category to be a main influencing factor for hospital costs.

Conclusion Our study confirms findings from other national research, and in addition shows interesting results for the effect that socio-economic status has on costs. This may be due to the restricted geographical area we are studying (Renfrew/Paisley), and leaves scope for further analysis using a sample representative for the whole of Scotland.