A Mathematical Model for Interpretable Clinical Decision Support with Applications in Gynecology

摘要: Background: Over time, methods for the development of clinical decision support (CDS) systems have evolved from interpretable and easy-to-use scoring to very complex non-interpretable mathematical models. In order accomplish effective support, CDS should provide information on how model arrives at a certain decision. To address issue incompatibility between performance, interpretability applicability systems, this paper proposes an innovative structure, automatically leading easily applicable The resulting models can be used guide clinicians when deciding upon appropriate treatment, estimating patient-specific risks improve communication with patients. Methods Findings: We propose interval coded (ICS) system, which imposes that effect each variable estimated risk is constant within consecutive intervals. number position intervals are obtained by solving optimization problem, additionally performs selection. visualised means appealing tables color bars. ICS software packages, in smartphone applications, or paper, particularly useful bedside medicine home-monitoring. approach illustrated two gynecological problems: diagnosis malignancy ovarian tumors using dataset containing 3,511 patients, prediction first trimester viability pregnancies 1,435 women. Comparison performance range proposed literature illustrates ability combine optimal simple systems. Conclusions: patient-clinician will additional insights importance influence available variables. Future challenges include extensions methodology towards automated detection interaction effects, multi-class prognosis high-dimensional data.