AI-Assisted Cardiovascular Risk Assessment by General Practitioners in Resource-Constrained Indonesian Settings Using...

Area of Science:

• Digital health and Artificial Intelligence (AI)
• Cardiovascular disease prevention
• Health informatics

Background:

• Preventive strategies integrating digital health and AI show promise for mitigating atherosclerotic cardiovascular disease (ASCVD).
• AI-enabled clinical decision support (CDS) offers patient-specific insights beyond traditional risk factors.
• The efficacy of AI-CDS in resource-constrained settings for ASCVD risk management is under-explored.

Purpose of the Study:

• To evaluate the impact of AI-based CDS on 10-year ASCVD risk assessment and management in primary prevention.
• To compare AI-based CDS with automated CDS and no decision support in a primary care context.

Main Methods:

• A 3-way, within-subject randomized controlled study involving 102 doctors in Indonesia.
• Doctors assessed 10-year ASCVD risk and management decisions for 9 clinical vignettes under three conditions: AI-based CDS, automated CDS, or no support.
• Outcomes included correct risk assessment, patient management (aspirin, statin, antihypertensive prescriptions, referrals), decision-making time, and perceived AI utility.

Main Results:

• AI-based CDS improved ASCVD risk assessment by 27% and statin prescription by 29% compared to no support.
• AI-assisted decision-making was significantly more accurate for risk assessment and statin prescription, with a number needed to treat of 3.7 for correct risk classification.
• AI-assisted cases required less decision-making time (63.6s vs 72.8s), while improvements in aspirin/antihypertensive prescriptions and referral decisions were not statistically significant.

Conclusions:

• AI-based CDS demonstrates significant potential for improving ASCVD risk assessment and statin prescribing in resource-constrained primary care settings.
• Reduced decision-making time with AI-CDS highlights its value for efficient healthcare resource utilization.
• Further research is warranted to confirm the real-world applicability of these findings in low-resource environments.