Prediction of Multiple Organ Failure Complicated by Moderately Severe or Severe Acute Pancreatitis Based on Machine Learning: A Multicenter Cohort Study

Abstract

Background. Multiple organ failure (MOF) may lead to an increased mortality rate of moderately severe (MSAP) or severe acute pancreatitis (SAP). This study is aimed to use machine learning to predict the risk of MOF in the course of disease. Methods. Clinical and laboratory features with significant differences between patients with and without MOF were screened out by univariate analysis. Prediction models were developed for selected features through six machine learning methods. The models were internally validated with a five-fold cross-validation, and a series of optimal feature subsets were generated in corresponding models. A test set was used to evaluate the predictive performance of the six models. Results. 305 (68%) of 455 patients with MSAP or SAP developed MOF. Eighteen features with significant differences between the group with MOF and without it in the training and validation set were used for modeling. Interleukin-6 levels, creatinine levels, and the kinetic time were the three most important features in the optimal feature subsets selected by K-fold cross-validation. The adaptive boosting algorithm (AdaBoost) showed the best predictive performance with the highest AUC value (0.826; 95% confidence interval: 0.740 to 0.888). The sensitivity of AdaBoost (80.49%) and specificity of logistic regression analysis (93.33%) were the best scores among the six models in the test set. Conclusions. A predictive model of MOF complicated by MSAP or SAP was successfully developed based on machine learning. The predictive performance was evaluated by a test set, for which AdaBoost showed a satisfactory predictive performance. The study is registered with the China Clinical Trial Registry (Identifier: ChiCTR1800016079).

Automated Machine Learning for the Early Prediction of the Severity of Acute Pancreatitis in Hospitals

“…In the test dataset, the onset of AP and SAP were also more commonly seen in male than in female patients and the median age ranged from 44 to 47 years. Consistent with what Xu et al (2021) reported, biliary sludge or gallstones (39.49%) was the most frequent etiology of AP in our cohorts, followed by hypertriglyceridemia (17.87%). No statistical differences were observed in sex, age, smoke, history of hypertension, and diabetes in two groups of three datasets (p > 0.05).…”

“…Machine learning (ML) applied in medicine, both supervised and unsupervised, is becoming increasingly popular based on its efficient computing algorithms to learn from massive clinical data (Deo, 2015). Previous studies (Han et al, 2022;Liu et al, 2022;Xu et al, 2021;Yang et al, 2021;Zhang et al, 2021;Zhao et al, 2021;Zhou et al, 2021) have confirmed that ML has great potential in building models for disease diagnosis, prognosis prediction, survival analysis, etc. Traditional ML includes logistic regression (LR), support vector machine (SVM), random forest, etc.…”

Usefulness of Random Forest Algorithm in Predicting Severe Acute Pancreatitis

“…Choi et al combined clinical (i.e., APACHE-II and BISAP scores) and radiologic (i.e., Balthazar grade and EPIC score) scoring systems by classification tree analysis for predicting SAP (Choi et al, 2018). Xu et al reported that adaptive boosting algorithm (AdaBoost) could predict development of multiple organ failure, complicated by moderately severe or severe AP (Xu et al, 2021). However, the above models were limited due to lack of individualized prediction on the test sample.…”

Predicting severity of acute pancreatitis: Emerging role of artificial intelligence

“…There have been several studies which have evaluated the utility of AI in acute pancreatitis for predicting severity, organ failure, mortality, venous thromboembolism, etc 4 . Table 1 shows data from eight studies of which five have evaluated the role of AI in predicting severe acute pancreatitis and three for predicting organ failure (which also denotes a severe disease) 5–12 . These studies differ in the methodologies and algorithms used and some have applied more than one AI technique to arrive at the best model.…”