A User-Customizable Hybrid Framework for Targeted Medical Decision-Making

Abstract

Targeted medical decision-making is a current strategy for addressing the heterogeneity in the patient population, especially when patients' preferences are included in the decision-making process. In this paper, we propose a user-customizable hybrid framework that can be adjusted at the patient group level to target a medical decision process. Our framework provides a flexible design, capable of balancing the gain from the reduction of provider time against the cost of prediction inaccuracy resulting from group customization. The framework combines a descriptive process, used to group the patients based on preference-based subjective features, with a predictive process, which uses objective features to match a new patient with a group. We illustrate our approach by applying it to the colorectal cancer-screening problem. The provider chooses what level of trade-off is appropriate, as a function of the acceptable error level. The group customization process allows decision makers to better allocate scarce resources, by potentially shortening the time-consuming process of modelling patients' preferences using individualised stability analysis. The proposed framework might be applied, with minor changes, to various medical decisions, or even to broader provider-user scenarios, in which targeted decision-making that includes user preferences is advantageous.