Clinical management implications from 33 579 urinary stones: novel patterns in composition, comorbidities, seasonal variation, and machine learning-based urosepsis prediction
Zhenglin Chang, Xiujing Han, Han Zhao, Hezhaoyang Huang, Haiyan Liu, Xiaocong Pan, Haojie Wu, Weihong Lin, Ming Lei, Guohua Zeng, Baoqing Sun
Abstract
BACKGROUND: While urolithiasis affects up to 14.8% of the global population, comprehensive analyses of stone composition patterns in relation to multiple comorbidities, seasonal variations, and infection risk remain limited. METHODS: We analyzed 33 579 urinary stones collected from 2014 to 2024 in Southern China. Stone compositions were categorized and examined in relation to patient demographics, 30+ comorbidities, seasonal patterns, and infection risk factors. Machine learning algorithms were employed to develop a predictive model for urosepsis. RESULTS: Analysis of 33 579 urinary stones demonstrated that calcium-containing stones remained predominant, followed by infection stones and uric acid stones. The study revealed significant gender disparities, with women exhibiting higher rates of infection stones, urinary tract infection (UTI), systemic conditions (connective tissue disease, anemia), and endocrine disorders (diabetes, thyroid dysfunction), while men showed higher prevalence of calcium stones, metabolic disorders (hyperuricemia, gout), and cardiovascular conditions. Anatomical distribution analysis revealed distinctive patterns: calcium stones were most frequent in ureteral stones, uric acid stones showed the highest proportion in bladder stones, and infection stones were most prevalent in posterior urethral stones and kidney stones. Location-specific comorbidity profiles emerged, with bladder stones showing unique associations with age-related conditions. Seasonal analysis demonstrated summer peaks for infection stones and UTI, while cardiovascular comorbidities peaked in winter, and metabolic disorders showed fall predominance. Age-stratified analysis uncovered distinct patterns, including increased uric acid stones in elderly males and peak cystine stones in adolescents. The first machine learning-based urosepsis prediction model achieved robust performance by integrating multiple clinical parameters, identifying key risk factors including gender, infection status, anemia, stone characteristics, and anatomical abnormalities. CONCLUSIONS: This study presents the largest comprehensive analysis of urinary stone compositions in Asia, revealing previously unreported patterns in comorbidity associations, seasonal variations, and location-specific characteristics. The novel machine learning-based prediction model for urosepsis risk represents a significant advance in clinical risk stratification for urolithiasis patients.