Pre‐implementation safety evaluation of an AI decision‐support system for surgical antimicrobial prophylaxis
Joshua Andrews, Sanjat Kanjilal, Camile Kotton, Brandon Stretton, Alasdair Leslie · 2026
Read Pre‐implementation safety evaluation of an AI decision‐support system for surgical antimicrobial prophylaxis by Joshua Andrews, Sanjat Kanjilal, Camile Kotton, Brandon Stretton, Alasdair Leslie (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/PMC13458242.txt.
Description
Joshua Andrews and colleagues stress-test a large language model before letting it anywhere near surgical antibiotic decisions. Antibiotic prophylaxis is a high-risk area for digital advice because errors fuel resistance and harm patients. Their 2026 pre-launch study fed the model public guideline text at prompt time, with no retraining. It ran 20 simulated surgical scenarios past two local clinicians, who scored agent, dose, timing, re-dosing, and citation.
Scores reached 4 out of 5 domains, with dosing fully accurate. Timing sagged to 70 percent and naming the right guideline to just 45 percent. Two scenarios, 10 percent, carried real clinical risk: one dropped required anaerobic coverage, and another failed to re-dose during a long operation, though the small test count makes that rate shaky. The team concludes guideline-based AI under hospital control is feasible to build, while the stewardship risks found here bar clinical use until further testing closes them.
Catalog record
| CREATOR | Joshua Andrews, Sanjat Kanjilal, Camile Kotton, Brandon Stretton, Alasdair Leslie |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 29893 |
| LICENSE | https://creativecommons.org/licenses/by/4.0 |
| RIGHTS | CC BY |
| SOURCE | Europe PMC |