Artificial-Intelligence powered surgical risk score improves postoperative outcomes at high- volume cancer hospital

Abstract

<title>Abstract</title> A real-time, automated artificial intelligence powered preoperative surgical risk score (SRS) that classifies patients as “high” or “low” risk for postoperative complications (POC) was integrated into the clinical electronic medical record. To evaluate POC after integration (‘Go-Live’), we studied patients undergoing unplanned inpatient surgery who were categorized as “SRS-exposed” or “SRS-Non-exposed” with exposure status based on whether the patient’s SRS was made available (exposed) to the surgeon preoperatively. 1,969 inpatients were included. In the 1,486 SRS-Non-Exposed patients, 849 were pre-‘Go-Live’ and 637 were post-‘Go-Live’. In the 483 SRS-exposed patients, 260 were pre-‘Go-Live’ and 223 were post-‘Go-Live’. The pre- and post-‘Go-Live’ POC rates in the SRS-Non-exposed group were 37.2% and 39.4% ( <italic>p=0.81</italic> ) while in the SRS-exposed group, the rates were 36.4% to 29.6% ( <italic>p=0.06</italic> ). High-risk patients had decreased rates of POC, 58.4% to 46.0% ( <italic>p=0.03</italic> ); return to OR, 17.6% to 5.5% ( <italic>p=0.01</italic> ); and length of hospital stay,19.6 to 17.5 days ( <italic>p=0.05</italic> ).

Ähnliche Arbeiten