Predicting in-hospital death in pneumonic COPD exacerbation via BAP-65, CURB-65 and machine learning

Abstract

IntroductionThere is no established clinical prediction model for in-hospital death among patients with pneumonic chronic obstructive pulmonary disease (COPD) exacerbation. We aimed to externally validate BAP-65 and CURB-65 and to develop a new model based on the eXtreme Gradient Boosting (XGBoost) algorithm.MethodsThis multicentre cohort study included patients aged ≥40 years with pneumonic COPD exacerbation. The input data were age, sex, activities of daily living, mental status, systolic and diastolic blood pressure, respiratory rate, heart rate, peripheral blood eosinophil count, and blood urea nitrogen. The primary outcome was in-hospital death. BAP-65 and CURB-65 underwent external validation using the area under the receiver operating characteristic curve (AUROC) in the whole dataset. We used XGBoost to develop a new prediction model. We compared the AUROCs of XGBoost with that of BAP-65 and CURB-65 in the test dataset using bootstrap sampling.ResultsWe included 1190 patients with pneumonic COPD exacerbation. The in-hospital mortality was 7% (88/1190). In the external validation of BAP-65 and CURB-65, the AUROCs (95% confidence interval [CI]) of BAP-65 and CURB-65 were 0.69 (0.66–0.72, and 0.69 (0.66–0.72), respectively. XGBoost showed an AUROC of 0.71 (0.62–0.81) in the test dataset. There was no significant difference in the AUROCs of XGBoost versus BAP-65 (absolute difference, 0.054; 95% CI, −0.057–0.16) or versus CURB-65 (absolute difference, 0.0021; 95% CI, −0.091–0.088).ConclusionBAP-65, CURB-65, and XGBoost showed low predictive performance for in-hospital death in pneumonic COPD exacerbation. Further large-scale studies including more variables are warranted.

A pooled analysis of the risk prediction models for mortality in acute exacerbation of chronic obstructive pulmonary disease

“…A total of 4376 pieces of literature were obtained through database searching, and 37 pieces of literature 11–47 were finally included after the screening. The process and results are shown in Figure 1.…”

“…A list of excluded studies and reasons for exclusion are provided in Table S4. This study included 26 studies 11–36 that developed models with or without validation and 11 studies 37–47 that only validated models. Fifteen studies 30,32,34,36–47 were finally included for quantitative statistical analysis.…”

“…This study included 26 studies 11–36 that developed models with or without validation and 11 studies 37–47 that only validated models. Fifteen studies 30,32,34,36–47 were finally included for quantitative statistical analysis.…”

“…We found 26 studies 11–36 describing the development of risk prediction models for mortality in patients with AECOPD. In‐hospital death ( n = 15) was the most common predictor of mortality.…”

“…We evaluated 37 studies 11–47 for the risk of bias using the PROBAST checklist, and all studies were at a high risk of bias, as shown in Figure 3. The main sources of risk were failure to correctly assess predictive model performance ( n = 29), insufficient sample size ( n = 27), selection of predictors using univariate analysis (n = 17), inappropriate data sources ( n = 17), lack of internal validation ( n = 8), continuous predictors handled inappropriately ( n = 7) and missing data not handled appropriately ( n = 2).…”