Revisiting the decision rule of cost–effectiveness analysis under certainty and uncertainty

The classical decision rule of cost–effectiveness analysis uses a threshold cost–effectiveness ratio as a cut-off point for resources allocation. One assumption of this decision rule is complete divisibility of health care programs. In this article, we argue that health care programs cannot be completely divisible since individuals are not divisible. Consequently, instead of a linear programming approach, an integer programming approach to budget allocation is suggested. The integer programming framework can be extended to include uncertainty in the analysis. An objective function (expected aggregate effects) is maximised subject to the constraint that the probability of exceeding the budget is limited to an arbitrary level (e.g., 0.05). In case the budget is exceeded, the objective function is penalised in order to account for the opportunity costs of the additional resource requirements.