Does Inclusion of Interactions Result in Higher Precision of Estimated Health State Values?

Most preference-based instruments producing overall values for health states are devised on the simplifying assumption that the overall effect of distinct health-related quality of life domains (attributes) of the instrument equals the sum of the attributes. Nevertheless, health attributes are often inter-related and depend on each other.

To investigate whether inclusion of second-order interactions in the three-level EuroQol five-dimensional questionnaire (EQ-5D-3L) value function would result in better fit and lead to different health state values than a model with main effects only.

Using an efficient design, 400 pairs of EQ-5D-3L health states were generated in a pairwise choice format. We analyzed responses of 4000 people from the general population using a conditional logit model, and we tested goodness of fit using pseudo R², Akaike information criterion, differences in log-likelihood, and likelihood ratio. We compared accuracies of models' predictions based on root mean square error and mean absolute error.

The interaction-effects model showed systematically lower values than the main-effects model. Inclusion of interactions resulted only in a slightly better model fit. Interactions comprising mobility and self-care were the most salient.

For the EQ-5D-3L, a value function based on interactions produces systematically lower values than a main-effects model, meaning that the effect of two or more health problems combined is stronger than the sum of the individual main effects.