A binary decision tree approach for pharmaceutical project portfolio management

Abstract

In pharmaceutical R&D, decision makers face a large portion of uncertainty in portfolio decisions. Existing optimization tools are mostly based on expected values, ignoring uncertainty, and most often deliver a single, “best” selection of projects. A decision tree approach can aid decision makers in the evaluation of various project selection alternatives. The decision tree can be used to efficiently evaluate the feasibility of the possible selection sets and their respective fulfillment of the objective according to qualitative, deterministic and/or stochastic criteria. One or more of the feasible sets with the best value can thereafter be selected for further analysis. The basis for this capability is a novel optimization model that can manage the uncertainty in the data and a solver that delivers information on all feasible selection sets. A fully deterministic approach cannot include information about uncertainty and hence may lead to non-optimal decisions. An innovative tool has been developed, based on a stochastic optimization model and using a novel binary decision tree solver has been developed. The results show that the value of the portfolio can be significantly increased by presenting multiple options for portfolio decisions including the effects of uncertainty.