Machine Learning-Based Operative Complexity Stratification Reduces Theatre Inefficiency In Cataract Surgery: A 12-Week Prospective Interventional Study

To evaluate whether artificial intelligence (AI)-assisted preoperative surgical complexity
prediction improves operating theatre utilization and case allocation efficiency in a high-
volume cataract surgical centre.

Tertiary ophthalmic surgical centre performing approximately 150 cataract procedures per
month within an integrated outpatient and surgical workflow environment.

A 12-week prospective interventional study comprised a 6-week baseline phase using conventional scheduling and a 6-week intervention phase using a supervised ML regression model trained on two years of operative data. The model predicted individual case duration and prolonged surgery probability from seven preoperative inputs: nuclear sclerosis grade, pupil size, pseudoexfoliation status, axial length, diabetes mellitus, prior ocular surgery, and surgeon identity. Validated against held-out data (MAE 3.9 min, R² 0.72), it prospectively allocated cases to standard or extended slots. Primary outcomes were theatre utilization and mean turnover time.

Theatre utilization improved from 71.8% to 84.6% (p < 0.001). Mean turnover time
decreased from 19.2 ± 4.6 to 13.1 ± 3.2 minutes (p < 0.001). First-case start delay was
reduced from 22.4 to 9.6 minutes. Unplanned schedule overruns decreased by 37%. No
statistically significant difference in intraoperative complication rates or postoperative best-
corrected visual acuity (BCVA) outcomes was identified between phases.

AI-assisted preoperative complexity prediction significantly improved operating theatre
efficiency without compromising surgical safety or patient outcomes. Prospective integration
of validated predictive modelling into cataract list scheduling represents a scalable,
resource-neutral strategy to optimize theatre utilization in contemporary ophthalmic surgical
practice.