PURPOSE: We developed a computer model to predict the outcome and the duration until passage of ureteral/renal calculi. MATERIALS AND METHODS: A retrospective, randomized study was performed of the outcome in 301 patients presenting to the emergency room for renal colic. Presenting characteristics of those diagnosed with a single calculus by computerized tomography were recorded for analysis. Predictors of stone passage and passage duration were identified and then used to create a logistic regression model. The algorithm was trained on 141 randomly selected patients and then tested on a separate 160 patients. Model accuracy was compared to predictions from 10 experienced urologists and 9 urology residents in 77 randomly selected patients. The model was tested further in 30 randomly selected patients at a private hospital to assess its general applicability. RESULTS: The model prediction accuracy in 160 patients was 86.3% for passage and 87.3% for duration (less or greater than 2 weeks). In the comparison group the model, the 10 experienced urologists and the 9 urology residents had an overall prediction accuracy of 88.3%, 70.5% (p = 0.006) and 72% (p = 0.007) for passage, and 87.1%, 71.6% (p = 0.007) and 81% (p = 0.075) for duration, respectively. Prediction accuracy was 93.3% for passage and 90.3% for duration when tested at a private hospital. CONCLUSIONS: Our model provides outcome and duration of passage predictions for patients presenting acutely in the emergency room with a single ureteral/renal calculus. It performs better than experienced urologists and urology residents. It can be applied to a private practice setting with equal accuracy.
RCT Entities:
PURPOSE: We developed a computer model to predict the outcome and the duration until passage of ureteral/renal calculi. MATERIALS AND METHODS: A retrospective, randomized study was performed of the outcome in 301 patients presenting to the emergency room for renal colic. Presenting characteristics of those diagnosed with a single calculus by computerized tomography were recorded for analysis. Predictors of stone passage and passage duration were identified and then used to create a logistic regression model. The algorithm was trained on 141 randomly selected patients and then tested on a separate 160 patients. Model accuracy was compared to predictions from 10 experienced urologists and 9 urology residents in 77 randomly selected patients. The model was tested further in 30 randomly selected patients at a private hospital to assess its general applicability. RESULTS: The model prediction accuracy in 160 patients was 86.3% for passage and 87.3% for duration (less or greater than 2 weeks). In the comparison group the model, the 10 experienced urologists and the 9 urology residents had an overall prediction accuracy of 88.3%, 70.5% (p = 0.006) and 72% (p = 0.007) for passage, and 87.1%, 71.6% (p = 0.007) and 81% (p = 0.075) for duration, respectively. Prediction accuracy was 93.3% for passage and 90.3% for duration when tested at a private hospital. CONCLUSIONS: Our model provides outcome and duration of passage predictions for patients presenting acutely in the emergency room with a single ureteral/renal calculus. It performs better than experienced urologists and urology residents. It can be applied to a private practice setting with equal accuracy.
Authors: Peter Alexander Massaro; Avinash Kanji; Paul Atkinson; Ryan Pawsey; Tom Whelan Journal: Can Urol Assoc J Date: 2017 Mar-Apr Impact factor: 1.862