AI-augmented clinical decision in paediatric appendicitis: can an AI-generated model improve trainees' diagnostic...

Accurate diagnosis of paediatric appendicitis remains a challenge due to its diverse clinical presentations and reliance on subjective assessments. The integration of artificial intelligence (AI) with an expert's ''clinical sense'' has the potential to improve diagnostic accuracy. In this study, we aimed to evaluate the effectiveness of the Artificial Intelligence Pediatric Appendicitis Decision-tree (AiPAD) model in enhancing the diagnostic capabilities of trainees and compare their performance with that of an expert supervisor. Between March 2019 and October 2022, we included paediatric patients aged 0-12 years who were referred for suspected appendicitis. Trainees collected clinical findings using five predefined parameters before ordering any imaging studies. The AiPAD model, which was blinded to the surgical team, made predictions from the supervisor's and trainees' findings independently. The diagnosis verdicts of the supervisor and the trainees were statistically evaluated in comparison to the prediction of the AI model, taking into account the revealed correct diagnosis. A total of 136 cases were included, comprising 58 cases of acute appendicitis (AA) and 78 cases of non-appendicitis (NA). The supervisor's correct verdict showed 91% accuracy compared to an average of 70% for trainees. However, if trainees were enabled with AiPAD, their accuracy would improve significantly to an average of 97%. Significantly, a strong association was observed between the expert's clinical sense and the predictions generated by AiPAD.

The utilisation of the AiPAD model in diagnosing paediatric appendicitis has significant potential to improve trainees' diagnostic accuracy, approaching the level of an expert supervisor. This hybrid approach combining AI and expert knowledge holds promise for enhancing diagnostic capabilities, reducing medical errors and improving patient outcomes.

• Sharpening clinical judgement for pediatric appendicitis takes time and seasoned exposure. Traditional training leaves junior doctors yearning for a faster path to diagnostic mastery.

• AI-generated models unlock the secrets of expert intuition, crafting an explicit guide for juniors to rapidly elevate their diagnostic skills. This leapfrog advancement empowers young doctors, democratizing medical expertise and paving the way for brighter outcomes in clinical training.