Abstract
BACKGROUND: Prognostic factors for ambulatory oncology patients have been described, including Eastern Cooperative Oncology Group (ECOG), tumor stage and malnutrition. However, there is no firm evidence on which variables best predict mortality in hospitalized patients receiving active systemic treatment. Our main goal was to develop a predictive model for 90-day mortality upon admission. METHODS: Between 2020 and 2022, we prospectively collected data from three sites for cancer patients with hospitalizations. Those with metastatic disease receiving systemic therapy in the 6 months before unplanned admission were eligible to this study. The least absolute shrinkage and selection operator (LASSO) method was used to select the most relevant factors to predict 90-day mortality at admission. A multivariable logistic regression was fitted to create the PROgnostic Score for Hospitalized Cancer Patients (PROMISE) score. The score was developed in a single-center training cohort and externally validated. FINDINGS: Of 1658 hospitalized patients, 1009 met eligibility criteria. Baseline demographics, patient and disease characteristics were similar across cohorts. Lung cancer was the most common tumor type in both cohorts. Factors associated with higher 90-day mortality included worse ECOG, stable/progressive disease, low levels of albumin, increased absolute neutrophil count, and high lactate dehydrogenase. The c-index after bootstrap correction was 0.79 (95% CI, 0.75-0.82) and 0.74 (95% CI, 0.68-0.80) in the training and validation cohorts, respectively. A web tool (https://promise.vhio.net/) was developed to facilitate the clinical deployment of the model. INTERPRETATION: The PROMISE tool demonstrated high performance for identifying metastatic cancer patients who are alive 90 days after an unplanned hospitalization. This will facilitate healthcare providers with rational clinical decisions and care planning after discharge. FUNDING: Merck S.L.U., Spain.