Back to 2025 Abstracts
Automated Abstraction Algorithm for Radical Cystectomy Surgical Outcomes
Peter S. Palencia, BS1, Maximilian Rabil, MD
1, Michael Jalfon, MD
1, Dylan Heckscher, MD
1, Rhys Richmond, BS
2, Victoria Kong, BS
2, Aleksandra Golos, BS
2, Adam Chess, MPH
2, Michael S. Leapman, MD, MSHS
1, Jaime A. Cavallo, MD, MSHS
1.
1Department of Urology, Yale School of Medicine, New Haven, CT, USA,
2Yale University School of Medicine, New Haven, CT, USA.
BACKGROUND: Cystectomy is a potential morbid procedure that presents a prime opportunity for quality improvement and patient safety initiatives. We hypothesized that a novel electronic medical record (EMR)-based automated algorithm would demonstrate >90% sensitivity and specificity with significant inter-rater reliability with institutional National Surgical Quality Improvement Program (NSQIP) data abstraction.
METHODS: An EMR-based algorithm was developed to automatically extract surgical outcomes and quality metrics for cystectomy cases between January 2013 and March 2024 by urologists in our health system. Surgical outcomes were extracted using CPT/ICD-10 codes and EMR-based variables. Algorithm sensitivity and specificity was calculated and compared to the performance of NSQIP abstractors. Inter-rater reliability (IRR) between the dashboard and NSQIP data was analyzed with Cohen’s kappa statistic.
RESULTS: 623 cases were mutually tracked. Sensitivity of the algorithm was ≥90% for all outcomes; specificity was ≥96% for all outcomes. IRR was highest for mortality and stroke (k=1.00, CI: 1.00-1.00) and lowest for ureteral obstruction, anastomotic leak, and rectal injury (k=0.00, CI: 0.00-0.00). IRR was slight for renal insufficiency (k=0.13, CI: 0.06-0.19) and pneumonia (k=0.19, CI: 0.03-0.35); fair for unplanned intubations (k=0.24; CI: 0.04-0.44), prolonged NPO (k=0.26, CI: 0.20-0.33), sepsis (k=0.28, CI: 0.19-0.37), and urinary leak (k=0.28, CI: 0.15-0.40); moderate for UTI (k=0.46, CI: 0.38-0.53), dialysis (k=0.58, CI:0.40-0.77), and return to OR (k=0.60, CI: 0.51-0.69); and substantial for cardiac complications (k=0.65, CI: 0.49-0.81), prolonged vent (k=0.76, CI: 0.62-0.89), C. difficile infection (k=0.85, CI: 0.76-0.93), and readmission (k=0.89, CI:0.85-0.93).
CONCLUSIONS: This EMR-based algorithm for cystectomy outcomes and quality metrics achieves at least 90% sensitivity and specificity for all variables and matches or exceeds the sensitivity and specificity of NSQIP abstraction. This algorithm demonstrates how quality improvement data can be made instantaneously accessible and automated, with potential to reduce the cost and resource burden of manual data collection in national quality improvement programs.
Back to 2025 Abstracts
