OBJECTIVES: It is recognised that unpaid caregiving can affect health-related quality of life (HRQoL). It is also understood that using EQ-5D-5L instead of EQ-5D-3L to measure patient HRQoL shifts mean utility scores up the scale towards full health and compresses them into a smaller range.

We compared the average treatment effect of different caregiving intensities on EQ-5D-3L and EQ-5D-5L, to understand whether the effect of caregiving on HRQoL differs between the two measures.

METHODS: We used causal inference methods in our cross-sectional observational UK dataset of 3,536 carers and 39,901 non-carers. We conditioned on potential confounders: sex, age, family status, education, and diagnosis of a health condition. We used inverse probability weighting with a multinomial logit model to predict the probability of belonging to the weekly hours of caregiving intensity classes. We predicted EQ-5D-3L and -5L index scores using regression and the probability of reporting full health using a probit model. We checked for balance and overlap.

RESULTS: People providing upwards of 5 hours/week unpaid care had statistically significantly worse EQ-5D-3L and 5L index scores than non-carers. The average treatment effect was smaller for all intensities using 5L than 3L: a maximum of -0.157 for 50+ hours for 3L compared to -0.109 for 5L. Carers were less likely to report full health for 3L and 5L and the effect size was bigger for 3L: up to -0.165 compared a maximum of -0.133 for 5L.

CONCLUSIONS: Unpaid carers have worse HRQoL than non-carers and the utility decrement increases as intensity of care increases. Using EQ-5D-5L instead of EQ-5D-3L reduces the utility decrement associated with caregiving and reduces the decrease in the proportion of respondents reporting full health. Using EQ-5D-5L instead of EQ-5D-3L will reduce the cost-effectiveness of interventions that decrease the caregiving burden, all other things being equal.