AB019. Bridging the gap: enhancing drug allergy knowledge and management confidence in non-allergist physicians with ADAPT-2—a multinational, AI-assisted, educational initiative

Background: Mislabeled drug allergy (DA) is a global public health challenge associated with emerging antimicrobial resistance, necessitating not only efforts from allergists but also multidisciplinary collaborations with non-allergist physicians. A randomized crossover trial, Advances in Drug Allergy & Penicillin Testing (ADAPT; NCT06399601), was first conducted in late 2023 and demonstrated promising outcomes in improving non-allergist physicians’ DA knowledge and confidence through an intensive educational programme. Following ADAPT’s success, we commenced a multinational phase in 2024—the ADAPT-2. This report aims to report the effectiveness of this initiative on improving participants’ DA knowledge and management confidence.

Methods: Non-allergist physicians were recruited from four localities, namely Australia, Hong Kong, mainland China, and Sri Lanka, and received a 1-day intensive training in DA. The training programme in mainland China was translated into Mandarin with artificial intelligence (AI) assistance. Participants’ DA knowledge, confidence, and practice were evaluated by structured surveys (scored on a 0–100 scale) before and after the course. Subgroup analysis was performed between physicians from advanced economies (AEs) and emerging economies (EEs).

Results: A total of 181 physicians [13 (7.2%) from Australia, 25 (13.8%) from Hong Kong, 98 (54.1%) from mainland China, and 45 (24.9%) from Sri Lanka] were included. At baseline, physicians’ knowledge and confidence level for DA were unsatisfactory (53.5±17.2 and 47.5±22.7, respectively). In particular, physicians from EEs (mainland China and Sri Lanka) showed significantly lower scores than those from AEs (Australia and Hong Kong) in DA knowledge (49.1±15.5 vs. 70.1±12.7, P<0.001), but not in confidence (46.8±23.2 vs. 50.2±20.9, P=0.41). After ADAPT-2, there were significant improvements in DA knowledge (53.5±17.2 to 72.5±16.0, P<0.001) and confidence (47.5±22.7 to 71.2±17.6, P<0.001). AI-assisted translation was non-inferior to the native programme in effectiveness (P=0.65) and had high satisfaction (98% somewhat clear or better).

Conclusions: Lack of DA knowledge among non-specialists remains ubiquitous with significant disparities between AEs and EEs. An intensive, structured, AI-assisted training is effective to significantly enhance participants’ knowledge levels.

Keywords: Artificial intelligence (AI); delabeling; disparities; drug allergy (DA); education