Eliciting regret improves decision making at the end of life

Introduction: Management choices at the end of life are high-stake decisions fraught with emotions, chief among is regret. Our objective in this paper is to test the utility of a regret-based model to facilitate referral to hospice care while helping patients clarify their preferences on how they wish to spend the remaining days of their lives. Methods: A prospective cohort study that enrolled consecutive adult patients (n = 178) aware of the terminal nature of their disease. The patients were at the point in care where they had to decide between continuing potentially 'curative/life-prolonging' treatment (Rx) versus hospice care. Preferences were elicited using a Dual Visual Analog Scale regarding the level of regret of omission versus commission (RgO/RgC) towards hospice care and Rx. Each patient's RgO/RgC was contrasted against the predictive probability of death to suggest a management plan, which was then compared with the patient's actual choice. The probability of death was estimated using validated Palliative Performance Scale predictive model. Results: Eighty-five percent (151/178) of patients agreed with the model's recommendations (p < 0.000001). Model predicted the actual choices for 72% (128/178) of patients (p < 0.00001). Logistic regression analysis showed that people who were initially inclined to be referred to hospice and were predicted to choose hospice over disease-directed treatment by the regret model have close to 98% probability of choosing hospice care at the end of their lives. No other factors (age, gender, race, educational status and pain level) affected their choice. Conclusions: Using regret to elicit choices in the end-of-life setting is both descriptively and prescriptively valid. People with terminal disease who are initially inclined to choose hospice and do not regret such a choice will select hospice care with high level of certainty. (C) 2016 The Author(s). Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license.