Development and Validation of a Diagnostic Nomogram for Pneumocystis jirovecii Pneumonia in Non-HIV-Infected Pneumonia Patients Undergoing Oral Glucocorticoid Treatment

We developed a nomogram with good diagnostic power for PJP diagnosis in pneumonia patients receiving oral glucocorticoids. This nomogram may help promote timely treatment of PJP and thus reduce the mortality rate in these patients.

This study was a retrospective, cross-sectional research. The development group included 434 patients who were admitted with pneumonia from 6 hospitals. Demographics, symptomatic features, laboratory and computed tomography data were analyzed using the least absolute shrinkage and selection operator (LASSO) to select potential diagnostic indicators. Binary logistic regression was used to develop a diagnostic nomogram. Another 119 patients with pneumonia admitted at Sichuan Provincial People's Hospital was used as the validation group. The diagnostic performance of the nomogram was measured by area under the receiver-operating-characteristics curve (AUC), calibration curves, and the net benefit by decision curve.

Pneumocystis jirovecii pneumonia (PJP) is an opportunistic but potentially fatal infection with increasing prevalence in HIV-free patients. Glucocorticoid therapy is one of the most important risk factors for PJP. The delay in diagnosis contributes to poor outcomes. Hence, the aim of this study was to develop and validate a nomogram for the diagnosis of PJP in patients with non-HIV-infected pneumonia who are undergoing oral glucocorticoid treatment. Patients and

PJP prevalence was 25.3% in the development group. LASSO regression revealed that age, lymphocyte count, fever, dry cough, respiratory failure, ground-glass opacity in lungs, glucocorticoid therapy duration, and immunosuppressive therapy were indicators of PJP. The nomogram showed robust discrimination, with an AUC of 0.82 (95% CI 0.77-0.86) in the development group and an AUC of 0.87 (95% CI 0.80-0.94) in the validation group, both showing acceptable calibration. In the decision curve analysis, our model consistently achieved a greater net benefit across almost all ranges of clinical thresholds.