Table 3.

Overview of principal challenges to CDSS in AMS

A. Obstacles in assessing the efficacy of CDSS
Category	Challenge	Impact	Suggested solutions
Study design 8,17,29	Heterogeneous CDSS types and outcomes	Limits data pooling, weakens conclusions	Define core outcome sets; typologize CDSS
Methodological quality 29,30	Reliance on before-after studies	High bias, overestimates effect	Use cluster-RCTs or ITS with control; follow EPOC
Clinical outcome gaps 23,25	Few studies on mortality, AMR, costs	Weak insight on safety and long-term impact	Include clinical, economic, microbiological outcomes
Validation gaps 42,44	Mostly retrospective, single-site ML-CDSS	Poor generalizability, low trust	Require prospective, multi-center validation
Evaluation frameworks 31,36	Inconsistent methods/reporting	Poor comparability	Adopt CONSORT-EHEALTH, DOOR, RADAR, TIDieR

ML: machine learning; AI: artificial intelligence.