Development and Validation of a Risk Prediction Model for Postoperative Pulmonary Infection in Renal Transplant Patients.

BACKGROUND

The incidence of postoperative pulmonary infection following renal transplantation is high; however, there is a paucity of studies focused on developing risk prediction models in this patient population.

OBJECTIVE

To construct a risk prediction model for postoperative pulmonary infection in renal transplant patients and verify the predictive efficacy of the model.

METHODS

This was a retrospective case-control study involving 765 patients who underwent renal transplantation in the Renal Transplantation Department of a tertiary hospital in Shandong Province in China between January 2022 and September 2024. According to a 7:3 ratio, 535 patients who received renal transplants from January 2022 to December 2023 were assigned to the model group, while 230 patients who underwent renal transplantation from January 2024 to September 2024 served as the validation group. Logistic regression analysis was used to explore the influencing factors of postoperative pulmonary infection in renal transplant recipients, and a nomogram-based risk prediction model was constructed and validated.

RESULTS

This study found that the incidence of postoperative pulmonary infection in renal transplant recipients was 24.31%. Logistic regression analysis identified gender (male), history of multiple transplants, comorbid hypertension history, comorbid heart disease history , and postoperative rejection as significant influencing factors. The Hosmer-Lemeshow goodness-of-fit test indicated adequate model calibration (χ²= 9.550, P = 0.145). The receiver operating characteristic (ROC) curve analysis yielded an area under the curve (AUC) of 0.740 [95% confidence interval (CI): 0.692-0.788], with a sensitivity of 88.00%, specificity of 49.20%, accuracy of 78.70%, positive predictive value (PPV) of 84.60%, and negative predictive value (NPV) of 56.20%. The calibration curve demonstrated close alignment between predicted and actual probabilities, with a low mean absolute error (MAE = 0.038), indicating robust calibration performance. In the validation cohort, the AUC was 0.705 [95% CI: 0.631-0.799], with an optimal probability threshold of 0.350, sensitivity of 91.90%, specificity of 36.20%, accuracy of 77.80%, PPV of 81.00%, and NPV of 60.00%.

CONCLUSION

The incidence of postoperative pulmonary infections in renal transplant patients is relatively high and is influenced by various factors (included male, history of multiple transplants, comorbid hypertension history, comorbid heart disease history, and postoperative rejection). The proposed risk prediction model exhibits favorable predictive performance and clinical utility, aiding clinicians in early identification of high-risk patients and guiding targeted preventive interventions.