MP40-04 MACHINE LEARNING BASED EVALUATION OF SEPSIS RISK AFTER ENDOUROLOGIC SURGERY

Le résumé fourni par la source

You have accessJournal of UrologyStone Disease: Epidemiology & Evaluation I (MP40)1 May 2024MP40-04 MACHINE LEARNING BASED EVALUATION OF SEPSIS RISK AFTER ENDOUROLOGIC SURGERY Justin Lee, Richard Berman, Adithya Balasubramanian, Hriday Bhambhvani, and Ojas Shah Justin LeeJustin Lee , Richard BermanRichard Berman , Adithya BalasubramanianAdithya Balasubramanian , Hriday BhambhvaniHriday Bhambhvani , and Ojas ShahOjas Shah View All Author Informationhttps://doi.org/10.1097/01.JU.0001008788.18007.3f.04AboutPDF ToolsAdd to favoritesDownload CitationsTrack CitationsPermissionsReprints ShareFacebookLinked InTwitterEmail Abstract INTRODUCTION AND OBJECTIVE: Urosepsis after kidney stone surgery is associated with considerable morbidity and mortality. Limited research examines the use of hemoglobin A1c (HbA1c) to predict postoperative sepsis after endourologic procedures. We developed a machine learning model to predict postoperative sepsis based on demographic and clinical data and better identify patients requiring preoperative optimization. METHODS: Patients undergoing ureteroscopy, shockwave lithotripsy, or percutaneous nephrolithotomy from January 2020 to June 2023 at a tertiary care medical center were identified. Those with missing HbA1c values within 90 days of surgery were excluded. Postoperative sepsis (binary outcome) was defined as Systemic Inflammatory Response Syndrome scores≥2 following surgery. A random forest (RF) model developed with a 70/30 train-test data split and 5-fold cross-validation on HbA1c laboratory values, preoperative urine culture positivity, preoperative stent or nephrostomy tube placement, preoperative antibiotic treatment, stone size, stone location, surgery type, length of surgery, Charlson Comorbidity Index (CCI), and demographic variables. RESULTS: 1,946 patients underwent stone procedures and 382 patients had requisite clinical data including HbA1c laboratory values within 90 days of surgery. The mean age was 60 (± 15) years, mean pre-operative CCI was 3.19 (± 2.76) and 31% of patients had HbA1c≥6.5%. 58/382 (20.5%) had postoperative sepsis. For the model, area under the ROC curve was 0.79 (95% CI: 0.66-0.90) (Figure 1). Confusion matrices were calculated at various thresholds and using a 0.3 cutoff, the model had a sensitivity of 0.47, specificity of 0.89, PPV of 0.42, and NPV of 0.91. Variables contributing most to the model's performance, in descending order, were HbA1c, preoperative hemoglobin, and stone size, established by permutation importance. CONCLUSIONS: RF modeling performed well in predicting sepsis after endourologic kidney stone surgery and identified HbA1c as the most important clinical predictor. Further model development will help guide preoperative surgical optimization and planning. Download PPT Source of Funding: NIDDK 5T35DK093430 © 2024 by American Urological Association Education and Research, Inc.FiguresReferencesRelatedDetails Volume 211Issue 5SMay 2024Page: e664 Advertisement Copyright & Permissions© 2024 by American Urological Association Education and Research, Inc.Metrics Author Information Justin Lee More articles by this author Richard Berman More articles by this author Adithya Balasubramanian More articles by this author Hriday Bhambhvani More articles by this author Ojas Shah More articles by this author Expand All Advertisement PDF downloadLoading ...

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.