Statistical Models for Patient-Centered Outcomes in Clinical Studies

Days alive and out of hospital is recommended as a patient-centered outcome in perioperative clinical studies. It is defined as the number of days, out of the first M postoperative days, that the patient has been discharged from hospital, or zero if the patient dies within M days of surgery. This composite measure presents statistical challenges in its unusual distributional shape, and its inability to distinguish between the qualitatively different outcomes of death, and a hospital stay longer than M days. We propose a mixed binary-continuous model that overcomes these difficulties, and illustrate its use on a clinical trial of a drug administered in cardiac surgery.