Development and temporal external validation of a prediction model for unplanned readmission in hospitalized patients with cancer and coronary heart disease
Year: 2026
Type: preprint
Study design: Observational Study
Abstract: Background Unplanned readmission is a substantial burden in patients with cancer, particularly when coronary heart disease (CHD) coexists, yet no risk tool has been developed for this population. We developed and temporally validated a prediction model for unplanned readmission based on routine data available at admission. Methods We conducted a single-center retrospective cohort study of adults with cancer and CHD discharged between January 2022 and June 2024 (438 admissions; 407 unique patients). The outcome was unplanned readmission (33 events, 7.5%). Thirtyfour candidate predictors with < 5% missing data were screened by univariable analysis and least absolute shrinkage and selection operator (LASSO) regression with 10-fold × 5-repeat cross-validation (1-SE rule) and refitted in an unpenalized multivariable logistic model. Discrimination was assessed with the c-index and 1000-bootstrap optimism correction; calibration with the Hosmer-Lemeshow test, calibration slope, and intercept; and clinical utility with decision curve analysis. The model was applied without refitting to a temporally non-overlapping cohort discharged from October 2023 to June 2024 (n = 118; 6 events). Results The final model comprised serum sodium, diastolic blood pressure, hemoglobin, and digestive cancer (overall likelihood-ratio test χ2 = 16.55, P = 0.002). The apparent and optimism-corrected c-indices were 0.709 and 0.682 (95% confidence interval [CI] 0.599–0.766), and the temporal validation c-index was 0.841 (bootstrap 95% CI 0.643–0.985). Calibration was adequate in both cohorts (Hosmer-Lemeshow P = 0.932 and 0.606). At the Youden-optimal threshold (predicted probability 0.089), sensitivity and specificity were 0.61 and 0.75. The model outperformed the Padua score (c-index 0.538) and NRS2002 (0.502). Conclusion A four-variable model using routine admission data provides moderate discrimination with adequate calibration for unplanned readmission in patients with cancer and CHD and remained
Source: Research Square
Institution: Army Medical University
Language: English
Subfield: Oncology
Field: Medicine
Domain: Health Sciences
Open Access status: green
OpenAlex ID: W7225348226