Evaluating the impact of Artificial Intelligence (AI) assistance in cardiovascular disease risk assessment for resource-constrained settings: a randomised controlled study with general practitioners in Indonesia

We perform a three-way within-subject randomised design, where doctors as participants completed clinical vignettes or patient cases. For each patient case, participants asses 10-year atherosclerotic cardiovascular disease (ASCVD) risk and made management decisions using either a conceptual prototype of AI-based clinical decision support (CDS), automated CDS, or no decision support. This online study may take approximately 1 hour. Participants are presented with nine hypothetical patient cases in an outpatient setting. each case present a brief patient history, physical examination, and simple laboratory results that are commonly available in resource-constrained clinical settings. The trial set up and the CDS interface was designed to emulate access on a mobile device which is a likely real-world implementation of such tools in resource-constrained setting The study is simulated using Gorilla.sc, a platform for conducting online experiments, which is customised to display patient cases and the CDS interventions. Allocation of cases to the different conditions of CDS quality, and the order of presentation are randomised. All randomisations are set equal (1:1) and performed by the Gorilla software. Participants self-enrolled in the online experiment. The study URL is provided in the invitation and participants are asked to access the study from a laptop or desktop computer with an internet connection at a convenient time and place. Participants are instructed to complete all cases to the best of their clinical judgment in a single, uninterrupted session and no time limits were imposed. Participant ID and time are automatically monitored by the software. This is a within-subject cross-over trial. While participants could distinguish between the control (no decision support) and intervention conditions, they were blinded to the nature of the intervention. The Automated CDS and AI-based CDS are presented as “Risk Calculator 1” and “Risk Calculator 2” respectively, without disclosure of which system incorporated AI. This approach is designed to minimise placebo effects that could influence clinician performance. No formal washout period was required, as the interventions are software-based and delivered in a blinded, counterbalanced sequence to mitigate carryover effects.