Development of a mortality prognostic model based on the Pneumonia Severity Index for older adults with pneumonia in integrated medical and older adult care institutions: a retrospective study

Background Population aging has increased the demand for integrated medical and long-term care services, particularly in China. Age-related respiratory impairment and immunosenescence render institutionalized older adults at a high risk for pneumonia-associated with poorer outcomes than community-acquired pneumonia. We aimed to investigate the prognostic value of the Pneumonia Severity Index (PSI) for mortality in older adults with pneumonia in integrated medical and elderly care facilities, and to develop a prognostic prediction model. Methods Medical records of older adult patients with pneumonia admitted to the integrated medical-elderly care ward of the Department of Geriatrics, Hefei First People's Hospital from June 2019 to June 2024 were retrospectively retrieved. A total of 312 patients were enrolled according to predefined inclusion and exclusion criteria. Patients were divided into a survival group ( n = 232) and a non-survival group ( n = 80) based on post-treatment clinical outcomes. Univariate analyses were performed to compare clinical characteristics between the two groups. All clinical variables were screened using Least Absolute Shrinkage and Selection Operator (LASSO) regression. Variables with non-zero coefficients were subsequently incorporated into Firth's bias-corrected logistic regression to construct the prediction model, followed by model validation. Results Univariate analysis revealed significant intergroup differences in age ≥80 years, coronary heart disease, white blood cell count (WBC),neutrophil count, multidrug-resistant organism (MDRO) infection, PSI score, hemoglobin, creatinine, coronavirus disease 2019 (COVID-19)infection, Clinical Frailty Scale (CFS) score >7, Prognostic Nutritional Index (PNI) < 35, and C-reactive protein (CRP) (all P ≤ 0.05). LASSO regression with λ = 0.0678392 identified six non-zero predictor variables: White blood cell count, MDRO infection, PSI score, CFS score >7, PNI <35, and CRP. A nomogram was developed using the coefficients obtained from Firth's bias-corrected logistic regression. The ROC analysis yielded an AUC of 0.889, sensitivity of 80%, specificity of 89%,and optimal cutoff value of 0.719. The model exhibited significantly better discriminative performance than PSI alone. Internal validation with 1000 bootstrap resamples produced a corrected AUC of 0.878, calibration slope of 0.9316, and intercept of −0.0584. Decision curve analysis (DCA) showed substantial net benefit within the threshold probability range of 0–0.6. The calibration curve demonstrated good agreement between predicted and observed mortality risks of the nomogram. Conclusion PSI is an independent risk factor for mortality among older adults with pneumonia in integrated medical and elderly care facilities. The model combining PSI with white blood cell count, MDRO infection, CFS score >7, PNI <35, and CRP can effectively predict mortality risk in this population and has superior clinical utility compared with PSI alone.