PD63-08 MULTI-INSTITUTIONAL STUDY: AUTOMATED PERFORMANCE METRICS TO PREDICT CONTINENCE RECOVERY AFTER ROBOTIC RADICAL PROSTATECTOMY UTILIZING MACHINE LEARNING

You have accessJournal of UrologyProstate Cancer: Localized: Surgical Therapy V (PD63)1 Apr 2020PD63-08 MULTI-INSTITUTIONAL STUDY: AUTOMATED PERFORMANCE METRICS TO PREDICT CONTINENCE RECOVERY AFTER ROBOTIC RADICAL PROSTATECTOMY UTILIZING MACHINE LEARNING Andrew Hung*, Runzhuo Ma, Steven Cen, Jessica Nguyen, Ryan Hakim, Swetha Rajkumar, Katarina Urbanova, Joern Witt, Inderbir Gill, Brian Miles, and Christian Wagner Andrew Hung*Andrew Hung* More articles by this author , Runzhuo MaRunzhuo Ma More articles by this author , Steven CenSteven Cen More articles by this author , Jessica NguyenJessica Nguyen More articles by this author , Ryan HakimRyan Hakim More articles by this author , Swetha RajkumarSwetha Rajkumar More articles by this author , Katarina UrbanovaKatarina Urbanova More articles by this author , Joern WittJoern Witt More articles by this author , Inderbir GillInderbir Gill More articles by this author , Brian MilesBrian Miles More articles by this author , and Christian WagnerChristian Wagner More articles by this author View All Author Informationhttps://doi.org/10.1097/JU.0000000000000980.08AboutPDF ToolsAdd to favoritesDownload CitationsTrack CitationsPermissionsReprints ShareFacebookLinked InTwitterEmail Abstract INTRODUCTION AND OBJECTIVE: Previously, our group demonstrated with machine learning that automated performance metrics (APMs; kinematic and events data) during robot-assisted radical prostatectomy (RARP) can predict continence recovery from a single institution. Herein, we update our results with multi-institutional data – APMs and clinicopathological data (CPD) to predict 3-mo and 6-mo continence. METHODS: APMs and CPD were collected for RARPs performed at three high-volume robotic centers. The RARP was segmented into 12 steps, and for each step, 41 validated APMs were reported. The predictive models were trained with three data sets: 1) 492 APMs; 2) 16 CPDs (i.e. prostate volume, Gleason score); 3) APMs + CPDs. We utilized a Random Forest model (800 trees) to predict continence recovery (no pads or 1 safety pad) at 3 and 6 months after surgery. The prediction accuracy was estimated through a 10-fold cross-validation process. Area under the curve (AUC) and Standard Error (SE) was used to estimate prediction accuracy. Out of bag Gini index was used to rank the variables of importance. SAS 9.4 was used for all data analyses. RESULTS: 196 RARP from the three institutions were included. Collectively, median patient age was 65 yr (49-83 yr), PSA was 7.3 ng/mL (0.53-96 ng/mL), and pathologic Gleason score was 7 (6 - 10). 21 attending surgeons were included in this study. Median RARP experience of these surgeons was 478 cases (117-4970 cases). Median continence recovery time for the entire cohort was 94 days (16-553 days). 102/196 (52.0%) were continent by 3 mo after surgery, and 129/196 (65.8%) after 6 mo. CPD alone produced an AUC 0.67 SE 0.04 for 3 mo prediction, while 6 mo prediction decreased to 0.63 SE 0.06. APMs alone produced an AUC 0.72 SE 0.04 for 3 mo, while 6 mo prediction was 0.64 SE 0.05. Finally, the combined APMs + CPD dataset produced same accuracy as APMs alone for both 3 mo (AUC 0.73 SE 0.04) and 6 mo (AUC 0.64 SE 0.05). In the top 10 ranked important features to predict 3 mo continence, APMs for certain steps were highlighted, including apical dissection (5/10 top features) and right pedicle dissection (3/10). Feature ranking for 6 mo continence again revealed APMs during apical dissection as important predictors (5/10). Notably, no patient factors were in the top 50 for either 3 or 6 mo. CONCLUSIONS: In this multi-institution study, APMs (surgeon factors) appear to be stronger predictors of continence recovery than patient factors (CPD). Metrics during the apical dissection and pedicle dissection appear to be important in determining continence recovery. Source of Funding: This study was funded in part by an Intuitive Surgical Clinical Grant; Intuitive Surgical provided the systems data recorder. © 2020 by American Urological Association Education and Research, Inc.FiguresReferencesRelatedDetails Volume 203Issue Supplement 4April 2020Page: e1295-e1296 Advertisement Copyright & Permissions© 2020 by American Urological Association Education and Research, Inc.MetricsAuthor Information Andrew Hung* More articles by this author Runzhuo Ma More articles by this author Steven Cen More articles by this author Jessica Nguyen More articles by this author Ryan Hakim More articles by this author Swetha Rajkumar More articles by this author Katarina Urbanova More articles by this author Joern Witt More articles by this author Inderbir Gill More articles by this author Brian Miles More articles by this author Christian Wagner More articles by this author Expand All Advertisement PDF downloadLoading ...