Abstract
BACKGROUND: Urinary stones are a common condition with increasing prevalence worldwide. Predicting the type of urinary stones is essential for guiding treatment, yet complex imaging models are not always accessible. This study aims to identify simpler clinical predictors and explore the risk factors for infected urinary stones using statistical and receiver operating characteristic (ROC) analyses. METHODS: A retrospective analysis was conducted on 1,067 patients with urinary stones who underwent surgical treatment between 2018 and 2023. Patients were classified into infected and non-infected stone groups based on stone composition. Logistic regression analysis adjusted for age, gender, body mass index (BMI), and serum electrolytes was performed to identify significant predictors, with sensitivity analyses using inverse probability weighting (IPW) to address cohort imbalance. The predictive performance of key factors was assessed using ROC curves. RESULTS: Of the total cohort, 686 (64.3%) had infectious stones, and 381 (35.7%) had non-infectious stones. Infectious stones were more common in females, younger patients, and those with ureteral stones. Preoperative urine cultures revealed Proteus mirabilis (24.3%) and Escherichia coli (18.2%) as predominant pathogens. Multivariate analysis identified three independent predictors for infectious stones: alkaline urine pH [hazard ratio (HR) 2.54, 95% confidence interval (CI): 1.33-4.88, P<0.001], ureteral stone location (HR 5.60, 95% CI: 2.38-13.17, P<0.001), and absence of diabetes mellitus (HR 4.74, 95% CI: 1.50-15.03, P=0.01). Sensitivity analyses confirmed robustness (adjusted HRs: 2.49, 5.58 and 4.65, respectively). Among these, ureteral stone location had the best predictive performance [area under the curve (AUC) =0.782, sensitivity 84.0%, specificity 71.5%], followed by urine pH (AUC =0.766, sensitivity 79.3%, specificity 68.7%), while diabetes status showed weaker predictive ability (AUC =0.623). CONCLUSIONS: Alkaline urine pH and ureteral stone location are strong predictors of infected urinary stones, while diabetes status is less predictive. These findings highlight the importance of integrating simple clinical parameters to improve the preoperative assessment and management of patients with urinary stones, particularly in resource-limited settings.