Back to 2026 Abstracts
Implementation of an Artificial Intelligence-Enabled Preoperative Risk Screening Workflow to Identify High-Risk Older Adults Undergoing Urologic Surgery for Perioperative Optimization
Emma Askew, BS1, Emily Huang, MD2, Rory Maher, MS3, Patrick Purdon, PhD4, Nnamdi Onochie, BS2, Masaya Higuchi, MD1, Hiroko Kunitake, MD, MPH1, Matthew Mossanen, MD, MPH2.
1Massachusetts General Hospital, Boston, MA, USA, 2Massachusetts General Brigham, Boston, MA, USA, 3Harvard Medical School, Boston, MA, USA, 4Stanford University School of Medicine, Palo Alto, CA, USA.
BACKGROUND: 65% of urologic procedures are performed in geriatric patients, who face higher surgical risks than younger patients.
1-3 The Perioperative Optimization of Senior Health (POSH) clinic is a multidisciplinary program designed to optimize older and high-risk surgical patients through comprehensive preoperative evaluation; however, existing workflows to screen patients for referral face obstacles. We implemented an Artificial Intelligence (AI) tool to automate the screening and streamline referral of older adults with geriatric vulnerabilities at the highest risk of poor surgical outcomes to POSH clinic.
METHODS: The Flexible Surgical Set Embedding (FLEX) Score, an AI-based preoperative risk prediction algorithm using routinely available electronic health record data
4, was applied to patients scheduled for cystectomy approximately 30 days preoperatively. The model generates individualized risk percentiles (relative to the surgical population used for model calibration) and flags patients as high risk based on predefined percentile thresholds. These flagged cases undergo review by a designated surgeon for referral.
RESULTS: Between January 11 and March 31, 2026, 55 patients were screened using FLEX of which 34 were identified as high-risk. 26 were referred to POSH clinic, and 12 have completed POSH evaluation so far. Surgeon review required 30 minutes per week. The median ages of patients screened and referred were 73 and 75, respectively. The most common procedure was radical cystectomy with ileal conduit. The median length of stay was 6 days for the 43 patients who proceeded to surgery.
CONCLUSIONS: Implementation of an AI‑driven preoperative screening tool in older adults undergoing urology surgery was feasible and enabled efficient identification of those at highest risk for adverse postoperative outcomes. Future work will evaluate its impact on clinical outcomes and refine referral thresholds.
Model-generated risk percentiles (relative risk ranking) for patients stratified by referral status| Referral to POSH Clinic | Yes (n=26) | No (n=29) | |
| Category | Mean (95% CI) | Mean (95% CI) | p-value |
| Non-home discharge percentile | 88.87 (83.61, 94.13) | 65.88 (56.96, 74.80) | <0.001 |
| Readmission percentile | 57.72 (46.56, 68.88) | 45.23 (33.76, 56.70) | 0.13 |
| Mortality percentile | 81.91 (73.56, 90.26) | 58.87 (46.53, 71.21) | 0.003 |
Back to 2026 Abstracts
