Combined Use of Clinician Prediction of Survival and the Palliative Prognostic Index for Short‐Term Mortality Screening in Advanced Cancer Patients in an Acute Care Hospital

Background In acute care oncology, timely recognition of short‐term mortality risk is essential for initiating care planning, discharge coordination, and transition to palliative care. Clinician prediction of survival (CPS) is widely used in practice, but may fail to identify some patients at risk of death within weeks. This study evaluated the short‐term mortality screening performance of CPS, the palliative prognostic index (PPI), and their combination among patients with advanced cancer transitioning to best supportive care (BSC), with particular emphasis on whether PPI could identify additional patients at risk of short‐term mortality who were not identified by CPS alone. Methods This single‐center retrospective observational study screened adult patients with cancer referred to the palliative care team between April 2020 and November 2025. Patients who had transitioned to BSC were included in the primary analysis if a quantifiable CPS and a confirmed date of death were available. CPS was converted into survival days using predefined rules, and PPI was retrospectively assessed at the same index date. Diagnostic performance for death within 21 and 42 days was evaluated for CPS, PPI, and an “either‐positive” combined rule. Receiver operating characteristic analyses and Kaplan–Meier analyses according to four CPS/PPI classification groups were also performed. Results Of 750 patients referred to the palliative care team, 620 had transitioned to BSC and 357 were included in the primary analysis. Death within 21 and 42 days occurred in 158 (44.3%) and 237 (66.4%) patients, respectively. CPS and PPI yielded similar area under the curve (AUC) estimates for 21‐day mortality (AUC, 0.732 vs. 0.742; p = 0.760) and 42‐day mortality (0.727 vs. 0.731; p = 0.923). For 21‐day mortality, sensitivity/specificity were 37.3%/89.4% for CPS, 58.9%/72.4% for PPI, and 72.2%/66.3% for the combined rule. For 42‐day mortality, the corresponding values were 48.9%/78.3%, 73.4%/58.3%, and 82.3%/47.5%, respectively. Among patients who died within 21 and 42 days but were classified as negative by CPS, PPI additionally identified 55 of 99 (55.6%) and 79 of 121 (65.3%) patients, respectively. Among CPS‐negative patients, Kaplan–Meier analyses showed significantly shorter survival in PPI‐positive than in PPI‐negative patients for both the 3‐ and 6‐week classifications (log rank p < 0.001 for both). However, 25.0% and 42.4% of patients classified as negative by both measures died within 21 and 42 days, respectively. Conclusion Adding PPI to CPS identified additional short‐term deaths missed by CPS alone, although this gain in sensitivity was accompanied by reduced specificity and only limited improvement in overall accuracy. Given the substantial residual mortality among patients classified as negative by both measures, the combined assessment should not be used to exclude short‐term mortality. In acute care oncology, PPI may serve as a practical adjunct to CPS by prompting reassessment of short‐term prognosis and consideration of the urgency of care planning, including discharge coordination and discussions regarding goals of care.