OBJECTIVES: Traditional scoring systems have been widely used to predict acute pancreatitis (AP) severity but have limitations in predictive accuracy. This study investigates the use of machine learning (ML) algorithms to improve predictive accuracy in AP. METHODS: A retrospective study was conducted using data from 101 AP patients in a tertiary hospital in Türkiye. Data were preprocessed, and synthetic data were generated with Gaussian noise addition and balanced with the ADASYN algorithm, resulting in 250 cases. Supervised ML models, including random forest (RF) and XGBoost (XGB), were trained, tested, and validated against traditional clinical scores (Ranson’s, modified Glasgow, and BISAP) using area under the curve (AUC), F1 score, and recall. RESULTS: RF outperformed XGB with an AUC of 0.89, F1 score of 0.82, and recall of 0.82. BISAP showed balanced performance (AUC = 0.70, F1 = 0.44, and recall = 0.85), whereas the Glasgow criteria had the highest recall but lower precision (AUC = 0.70, F1 = 0.38, and recall = 0.95). Ranson’s admission criteria were the least effective (AUC = 0.53, F1 = 0.42, and recall = 0.39), probable because it lacked the 48^th h features. CONCLUSION: ML models, especially RF, significantly outperform traditional clinical scores in predicting adverse outcomes in AP, suggesting that integrating ML into clinical practice could improve prognostic assessments.

Background: Polymyxins are last-resort antibiotics for multidrug-resistant Gram-negative infections, but their nephrotoxicity limits clinical use. Objectives: To compare the nephrotoxicity of colistin versus polymyxin B and identify risk factors for acute kidney injury (AKI). Methods: We retrospectively analysed 258 patients (132 colistin, 126 polymyxin B) with normal to moderately impaired renal function. Machine learning (ML) algorithms predicted AKI risk, and propensity score matching was used to evaluate the risk of colistin preference over polymyxin B. Results: Colistin nephrotoxicity (defined as at least stage II KDIGO AKI) was significantly more frequent with colistin than polymyxin B in both unmatched (for colistin 31% versus for polymyxin B 19%, P = 0.04) and matched cohorts (34.4% for colistin versus 14.1% for polymyxin B, P = 0.01). ML identified lower baseline eGFR, higher uric acid, hypalbuminaemia, higher age, concomitant nephrotoxic drug use, weight, vasopressor use and polymyxin type as top predictors. Colistin preference over polymyxin B resulted in significant nephrotoxicity risk in both unmatched (OR: 2.08, 95% CI [1.14–3.79]) and matched cohorts (OR: 3.42, 95% CI [1.38–8.5]). Complete renal recovery occurred in only 41% of AKI cases within 30 days. Conclusions: Colistin demonstrates significantly higher nephrotoxicity than polymyxin B. The complex relationship between baseline renal function and AKI risk suggests careful monitoring for patients with moderate renal impairment, regardless of polymyxin choice.

Aim: Alterations in plasma lipid levels have been shown to be correlated with the severity of infections due to various pathogens such as bacteria, viruses. In this study, we aimed to evaluate the lipid metabolism changes associated with disease severity and prognosis in hospitalized COVID-19 patients during and after (post-COVID) the disease. Material and Method: Patients who were hospitalized in the COVID-19 wards between April 02, 2020, and November 20, 2020 and were then evaluated in the follow-up outpatient clinic were retrospectively searched. Results: Lipid levels were present at the admission and follow-up for 95 patients. The mean (S.D) age was 48.49 (16.4), and 49(51.6%) were male. The mean (S.D) day between the admission and the first visit in the COVID-19 follow-up outpatient clinic was 27.8 (12.8). LDL-C (p=0.044), and HDL-C (p=0.004) levels were significantly lower in the severely ill group at the admission. Total cholesterol, LDL-C, HDL-C, and triglyceride levels on follow-up were significantly higher than those levels on the admission day (p

Background Elevated serum uric acid, associated with cardiovascular conditions such as atherosclerotic heart disease, hypertension, and heart failure, can be elevated by thiazide or thiazide-like drugs (THZ), essential in hypertension management. Identifying clinical determinants affecting THZ-related uric acid elevation is critical. Methods In this retrospective cross-sectional study, we explored the clinical determinants influencing uric acid elevation related to THZ, focusing on patients where THZ was initiated or the dose escalated. A cohort of 143 patients was analyzed, collecting baseline and control uric acid levels, alongside basic biochemical studies and clinical data. Feature selection was conducted utilizing criteria based on mean squared error increase and enhancement in node purity. Four machine learning algorithms - Random Forest, Neural Network, Support Vector Machine, and Gradient Boosting regressions - were applied to pinpoint clinical influencers. Results Significant features include uncontrolled diabetes, index estimated Glomerular Filtration Rate (eGFR) level, absence of insulin, action of indapamide, and absence of statin treatment, with absence of Sodium-glucose cotransporter 2 inhibitors (SGLT2i), low dose aspirin exposure, and older age also being noteworthy. Among the applied models, the Gradient Boosting regression model outperformed the others, exhibiting the lowest Mean Absolute Error (MAE), Mean Squared Error (MSE), Root Mean Squared Error (RMSE) values, and the highest R2 value (0.779). While Random Forest and Neural Network regression models were able to fit the data adequately, the Support Vector Machine demonstrated inferior metrics. Conclusions Machine learning algorithms are adept at accurately identifying the factors linked to uric acid fluctuations caused by THZ. This proficiency aids in customizing treatments more effectively, reducing the need to unnecessarily avoid THZ, and providing guidance on its use to prevent instances where uric acid levels could become problematic.