Development and validation of an early warning score for refractory

Background: Methods: A retrospective study was performed using clinical data of children with MPP admitted during the period of January 2023 to December 2024. The patients were categorized into the general MPP (GMPP) group (56 cases) and RMPP group (34 cases) depending on the occurrence of progression to RMPP. The two groups were compared with regard to clinical features, laboratory findings, radiographic appearances, and features of the pathogen. Multivariate logistic regression analysis was performed to identify independent risk factors, and an RMPP early warning scoring system (RMPP-EWS) was developed. The receiver operating characteristic (ROC) curve analysis was used to assess model performance. To mitigate overfitting and quantify the optimism in predictive performance, an internal bootstrap validation with 1,000 resamples was conducted. Furthermore, a sensitivity analysis was performed to assess the impact of excluding fever duration from the prediction model, and a comparative evaluation was carried out between an equal-weighted scoring system and a coefficient-based weighted scoring system. Results: The RMPP group had a significantly increased pre-admission fever duration, C-reactive protein (CRP), lactate dehydrogenase (LDH), D-dimer level, and more atelectasis and A2063G mutation than the GMPP group (P<0.05). Multivariate analysis showed that duration of fever ≥7 days [odds ratio (OR) =6.40, 95% confidence interval (CI): 2.30-17.82], CRP ≥20 mg/L (OR =3.43), LDH ≥400 U/L (OR =4.29), D-dimer ≥500 µg/L (OR =3.07), atelectasis (OR =5.35), and A2063G mutation (OR =3.84) were predictors of RMPP. The RMPP-EWS (0-12 points) was developed based on these indicators. The analysis of the ROC curve showed an area under the curve (AUC) value of 0.892 (95% CI: 0.823-0.961), and the optimal cutoff value was ≥6 points, with a sensitivity of 85.3% and a specificity of 82.1%. The optimism-corrected AUC from bootstrap validation was 0.871 (95% CI: 0.802-0.940), with a shrinkage factor of 0.92. The five-variable model (excluding fever duration) demonstrated an AUC of 0.854 (95% CI: 0.771-0.937), and the coefficient-based weighted score showed an AUC of 0.896 (95% CI: 0.828-0.964), which was not significantly different from the equal-weighted score (DeLong test, P=0.68). The risk stratification included the following: low risk (0-4 points, RMPP incidence 10.7%), moderate risk (5-8 points, 36.8%), and high risk (9-12 points, 70.8%). Conclusions: The RMPP-EWS shows a predictive value and has the capacity to help in the early detection of high-risk RMPP patients to inform individual treatment decisions. This score may assist clinicians in early risk stratification. External validation in a larger multicenter cohort is warranted before clinical implementation.