The impact of ignoring population heterogeneity when Markov models are used in cost-effectiveness analysis.

Many factors related to the spread and progression of diseases vary throughout a population. This heterogeneity is frequently ignored in cost-effectiveness analyses by using average or representative values or by considering multiple risk groups. The author explores the impact that such simplifying assumptions may have on the results and interpretation of cost-effectiveness analyses when Markov models are used to calculate the costs and health impact of interventions. A discrete-time Markov model for a disease is defined, and 5 potential interventions are considered. Health benefits, costs, and incremental cost-effectiveness ratios are calculated for each intervention. It is assumed that the population is heterogeneous with respect to the probability of becoming sick. Ignoring this heterogeneity may lead to optimistic or pessimistic estimates of cost-effectiveness ratios, depending on the intervention and, in some cases, the parameter values. Implications are discussed of this finding on the use of league tables and on comparisons of cost-effectiveness ratios versus commonly accepted threshold values.