Risk Factors Analysis And Prediction Model Construction Of PCP In Non-HIV Infected Population

Background With the wide application of early antiretroviral therapy in HIV-positive infected patients,the risk of Pneumocystis pneumonia(PCP)is greatly reduced in this group.Moreover,the number of non-HIV patients,such as using immunosuppressants,malignant tumors,undergoing organ transplantation,infected with PCP appeared rising in recent years,who usually have deteriorated progress rapidly and higher mortality than HIV positive PCP patients.The aim of this study was to analyze the risk factors of PCP infection reported in current publications,screen out the independent risk factors and construct a prediction model,and draw a nomogram as a visual risk prediction tool that could help clinicians identify PCP early and treat appropriately,which would improve the prognosis of PCP.Methods The prospective study were conducted in 9 hospitals of China from September 2018 to January 2021.According to the inclusion and exclusion criteria,a total of 191 non-HIV patients were included.According to PCP diagnostic criteria,they were divided into PCP group(27 cases)and non-PCP group(164 cases).Univariate regression analysis was used to compare the clinical manifestations,laboratory examination and imaging findings between the two groups,and investigate the correlation.Multivariate logistic regression was used to analyze the independent risk factors of PCP,and establish a diagnostic prediction model of PCP.Results According to univariate regression analysis,age,heart rate,oxygen concentration,PCO2,serum LDH,APTT,dyspnea,long-term use of steroids for more than 4 weeks,daily dose of more than 0.5 mg/kg,immunosuppressive agents in the recent 90 days,oxygen inhalation,moderate COPD or above,organ bone marrow transplantation,connective tissue disease,chronic renal failure with need of dialysis,nodule shadow,exudate shadow,admission department,hospitalization department,mechanical ventilation and death rate were significantly different between PCP group and Non-PCP group.The multivariate regression analysis showed the independent risk factors of PCP included three factors: serum LDH level(p=0.000,OR:1.005,95%CI:1.002-1.008),CT showed exudation shadow(p=0.006,OR:56.035,95%CI: 5.001-1940.865),and mechanical ventilation(p=0.012,OR:13.101,95%CI: 1.901-117.711).Logistic regression model was established,with an equation of log(P)=0.004718 × LDH + 4.025968 × whether exudation shadow + 2.572701 ×whether mechanical ventilation.Based on the multivariate regression result,a nomogram model was established to predict the risk of PCP infection with the above three independent risk factors.The model was visual to facilitate clinical operation and calculation,and the area under curve(AUC)of ROC is 0.9695(95% CI: 0.9426-0.9964).when the prediction model was verified by internal data,the specificity,sensitivity,positive predictive value and negative predictive value was 70.4%,98.2%,95.3% and 13.6%,respectively.The total correct rate of this prediction model was 94.2%,which indicated the predictive risk was close to the actual risk.Conclusion The serum LDH level,radiography with exudation shadow and mechanical ventilation are the independent risk factors of PCP.The diagnostic prediction nomogram with these above three factors could better predict the risk of PCP.Because the included variables are simple and easy to obtain clinically,the prediction model has probable high feasibility and good clinical application prospect.However,the prediction model still needs to be verified by multi-center and large sample external data.