Cost-Effectiveness and Heterogeneity: Using Finite Mixtures of Disease Activity Models to Identify and Analyze Phenotypes

Heterogeneity in patient populations is an important issue in health economic evaluations, as the cost-effectiveness of an intervention can vary between patient subgroups, and an intervention which is not cost-effective in the overall population may be cost-effective in particular subgroups. Identifying such subgroups is of interest in the allocation of healthcare resources. Our aim was to develop a method for cost-effectiveness analysis in heterogeneous chronic diseases, by identifying subgroups (phenotypes) directly relevant to the cost-effectiveness of an intervention, and by enabling cost-effectiveness analyses of the intervention in each of these phenotypes. We identified phenotypes based on healthcare resource utilization, using finite mixtures of underlying disease activity models: first, an explicit disease activity model, and secondly, a model of aggregated disease activity. They differed with regards to time-dependence, level of detail, and what interventions they could evaluate. We used them for cost-effectiveness analyses of two hypothetical interventions. Allowing for different phenotypes improved model fit, and was a key step towards dealing with heterogeneity. The cost-effectiveness of the interventions varied substantially between phenotypes. Using underlying disease activity models for identifying phenotypes as well as cost-effectiveness analysis appears both feasible and useful in that they guide the decision to introduce an intervention.