Development and assessment of a mortality risk prediction nomogram model for pneumocystis disease in ICU within 28 days.

To develop and assess a nomogram predictive model for evaluating the 28-day mortality risk in patients diagnosed with Pneumocystis who have been admitted to the intensive care unit (ICU). From 2008 to 2022, clinical data on patients with Pneumocystis were collected using the American Critical Care Medical Information Database IV (MIMIC-IV). Initially, 63 significant predictive indicators were included, with ICU admission as the time node and all-cause mortality within 28 days as the outcome. Using complete data modeling, the variable selection approach combines two methods: Lasso regression (glmnet package) and collinearity screening (car package). Use the bootstrap method 1000 times for internal validation. Calculate the AUC, mean sensitivity, and specificity of 1000 resamplings, as well as the 95% confidence interval, and then plot the ROC, calibration, and DCA curves. The patients were split into two groups based on their 28-day survival status: 83 cases (67.48%) in the survival group and 40 instances (32.52%) in the death group. Five variables-the history of malignant tumors, Lods scores, Oasis scores, and complications of shock and severe renal injury-were eventually included in the entire sample after screening. According to Receiver Operating Characteristic (ROC) analysis, the model's sensitivity was 0.600 (95% CI 0.448-0.752), specificity was 0.904 (95% CI 0.840-0.967), and AUC was 0.814 (95% CI 0.732-0.897). The DCA curve indicates that the model application has high accuracy, which leads to a net benefit for population prediction, and the model calibration curve demonstrates good calibration accuracy. The bootstrap model's 1000 internal validation results demonstrate that the model's calibration performance is good and that its accuracy is highest when the probability of patient outcome events falls between 35 and 60%, which yields the highest net benefit for population prediction. This study developed a nomogram utilizing MIMIC-IV clinical big data to predict the 28-day mortality risk in patients with Pneumocystis disease. Incorporating five critical factors, the nomogram offers a user-friendly, visual method for calculating personalized risk scores based on patient-specific information, including medical history, laboratory results, and clinical scores. Demonstrating robust discrimination and calibration, this tool provides clinicians with a valuable resource for assessing prognosis and making evidence-based treatment decisions.