Peer-Reviewed Studies Tie Ambient AI to More Surgeries at Houston Methodist

Houston Methodist researchers published the first peer-reviewed evidence that ambient AI in the operating room is associated with a 7% rise in surgical case volume, about 25 additional cases per month, in a high-acuity cardiothoracic suite after deploying Apella's computer-vision platform. The study, published June 27, 2026 in the Journal of Imaging, used synthetic control with difference-in-differences estimation across 116,098 comparison cases from 11 other Houston Methodist sites. A second peer-reviewed study in Perioperative Care and Operating Room Management found that a predictive scheduling tool improved accuracy by 40% and reduced late-ending OR days by 46 percentage points across six Houston Methodist hospitals.

Healthcare https://www.unite.ai/series/healthcare/ Peer-Reviewed Studies Tie Ambient AI to More Surgeries at Houston Methodist Add Unite.AI to your preferred sources on Google https://www.google.com/preferences/source?q=unite.ai Houston Methodist researchers have published what they describe as the first peer-reviewed evidence that ambient AI in the operating room is associated with a measurable increase in surgical case volume: a 7% rise, or roughly 25 additional cases per month, in a high-acuity cardiothoracic suite after the system deployed Apella’s computer-vision platform. The study https://www.mdpi.com/2313-433X/12/7/283 , published June 27, 2026 in the Journal of Imaging , covers 5,417 surgical cases monitored over 16 months at the system’s flagship hospital. A second peer-reviewed study, published in Perioperative Care and Operating Room Management , examined a different mechanism across six Houston Methodist hospitals: a predictive tool that flags cases likely to be mis-scheduled. For flagged cases, scheduling accuracy improved 40% mean absolute error fell from 57.4 to 34.2 minutes and late-ending OR days dropped 46 percentage points, with no reduction in case volume, according to Apella’s August 5, 2026 announcement https://www.prnewswire.com/news-releases/new-peer-reviewed-research-provides-first-independent-evidence-that-ambient-ai-increases-surgical-case-volume-302843663.html of the two publications. The volume finding matters because of how it was produced. Most vendor efficiency claims rest on before-and-after comparisons that cannot separate the technology’s effect from everything else changing in a hospital system. The Houston Methodist team instead used synthetic control with difference-in-differences estimation, a quasi-experimental method from economics: they built a weighted counterfactual from 11 other Houston Methodist sites that had not deployed the platform, 116,098 comparison cases in all, and measured the intervention suite against it. The result was a statistically significant increase of approximately 25 cases per month 95% CI 8.3 to 41.0 , achieved without adding rooms, staff, or hours. “Claims about AI’s impact in surgery are often met with skepticism, and understandably so,” said Roberta Schwartz, Houston Methodist’s chief innovation officer and the paper’s senior author. “That’s why we approached this work with the same rigor as any clinical research study, using independent peer review and a methodology designed to demonstrate causation, not just correlation.” What the cameras actually measured The system the paper calls image-based AI is built from wall-mounted cameras, four per room, feeding a YOLO-based object detection model coupled with a transformer-based event detector. It identifies patients, scrubbed and unscrubbed staff, draping status, and equipment, then converts those detections into timestamped perioperative events: patient entry, draping, turnover, wheels out. The training dataset covered 137,517 surgeries across 315 operating rooms. That granularity let the suite fragment the surgical day into phases the EHR never captured. The platform inserted a “Patient Draped” event between “Anesthesia Ready” and “Case Start,” for example, turning one opaque block into two measurable ones. Live dashboards showed case progress and predicted durations across all 15 rooms, and automated text alerts told surgeons when a patient was draped or their next case was approaching. The deployment ran under an IRB-determined quality improvement framework with patient consent folded into general consent documents, video retained a maximum of 30 days and audio 7, and access to intraoperative recordings restricted to three designated roles. Where the evidence has edges The paper’s own limitations section identifies several constraints on the volume finding. The intervention was not the software alone but a bundle: the platform plus turnover-time reviews with OR management, housekeeping and anesthesia workflow reviews, and surgeon notification protocols. The authors state plainly that the estimate “cannot separately attribute the observed gains to one or the other component.” The statistical picture is also narrower than the headline. The in-space permutation test the method’s built-in check against chance returned an empirical p-value of 0.182, directionally consistent but short of conventional significance. Secondary outcomes including unplanned overtime did not reach significance, and a sensitivity analysis on total operative minutes was not significant either, meaning the gain came as more cases per unit of OR time rather than more total operating time. The authors interpret the effect as “associated with, rather than caused by” the deployment. Disclosures cut both ways. Houston Methodist Hospital holds a minority equity interest in Apella and joined the company’s Series B round in January 2026. An Apella employee conducted the original analyses from aggregated monthly data — with no access to patient-level records — and a Houston Methodist analyst independently replicated them. Apella did not fund the study and had no role in its design or the decision to publish, per the paper’s conflict-of-interest statement. The system now runs in more than 200 of the health system’s operating rooms. What happens next The authors lay out a specific research agenda: system-wide evaluation across additional Houston Methodist hospitals and other surgical specialties to test whether the cardiothoracic gains generalize; formal evaluation of turnover time and first-case on-time starts as primary endpoints; and peer-reviewed replication of a 2024 conference finding that computer vision tracked OR events more consistently than manual documentation. Ambient AI deployments in clinical settings have mostly been judged on documentation burden ambient listening tools are now reaching nursing workflows https://www.unite.ai/mount-sinai-miami-beach-extends-epics-ambient-ai-to-nurses/ and the Houston Methodist pair of studies is an early attempt to judge the operating-room variant on throughput instead. For a sector where hospital executives field constant AI pitches with thin evidence behind them, the more durable contribution may be methodological: a health system testing a vendor’s platform on its own terms, against its own control sites, and publishing the edges alongside the result.