Diagnostic accuracy of a predictive scoring tool for patients who are eligible for home discharge from a palliative care unit

Staying at a preferred place, principally at home, is of great value for dying patients, and palliative care units (PCUs) have an important role in providing adequate support so that patients can be discharged and go home. We attempted to create and validate a scoring tool to predict whether a cancer patient admitted to a PCU will be discharged home.All 369 cancer patients admitted to the PCU of a 533-bed general hospital in Japan from October 2016 to October 2019 were enrolled. As outcomes, we recorded whether patients were discharged to home, died in hospitals, or were discharged to other hospitals. Attending physicians recorded 22 potential scale items at admission, including (I) demographic variables, (II) patient general conditions, (III) vital signs, (IV) medications, and (V) patient symptoms. Training-testing procedure to develop a screening score was performed.Among 369 cancer patients admitted to the PCU, we excluded 10 cases for whom a death location could not be identified. Among the remaining 359 patients, 180 were analyzed in the development phase and 179 in the validation phase. Multivariate logistic regression analysis identified five items as independent factors associated with discharge to home, and a prediction equation was created using the regression coefficients: sex (female, 4 points), calorie intake (520 kcal or more, 19 points), availability of daytime caregivers (11 points), family's preferred place of care (home, 139 points), and symptoms that resulted in hospitalization (not fatigue, 7 points). Using a cutoff point of 155, the area under the curve (AUC) value was 0.949 with 95% confidence intervals of 0.918 to 0.981. In the validation sample, the sensitivity, specificity, negative predictive value (NPV), positive predictive value (PPV), and error rate were 75.3%, 86.3%, 82.2%, 80.6%, and 18.4%, respectively.Whether a patient admitted to a PCU can discharge to home could be predicted using the simple clinical tool. Further validation and outcome studies are warranted.