(116) Integration of artificial intelligence in clinical decision-making at leading health systems

Background: : To engage effectively with health systems adopting AI in clinical decision-making, manufacturers need a clear understanding of current capabilities and integration levels. This insight can inform how they structure evidence, deliver data, and collaborate with AI-enabled providers.

Objective: : To assess how health systems are using AI platforms, the degree of integration with clinical systems, and perceived benefits from using AI in clinical decision making.

Methods: : Precision AQ conducted a survey of 25 health system decision-makers in March 2025. Participants were selected from a database and screened based on their organization's use of AI in clinical decision-making, either extensively or in pilot stages, and their involvement in AI policy.

Results: : When screening potential participants, two-thirds reported their organizations are either extensively using or piloting AI for clinical decision-making. Among respondents completing the survey, all confirmed their organizations use AI in clinical decision-making, either extensively (76%) or in a pilot capacity (24%). Most health systems (72%) report using commercial AI platforms such as Tempus or Epic, while fewer developed platforms in-house (16%) or partnered with third-party vendors (8%). Few organizations (28%) have achieved full integration of their AI solutions with their electronic health records (EHRs), while 72% report partial integration. No respondents selected “no integration.” Perceived benefits of using AI in clinical decision making included improved adherence to treatment guidelines (84% of respondents), reduced clinician workload (80%), improved patient outcomes (80%), and reduced time to treatment decisions (72%). When asked how AI decisions are aligned with clinical guidelines, 56% of respondents indicated that systems are regularly updated to reflect changes in guidelines, while 24% rely on clinician validation of AI recommendations. Pharmaceutical manufacturers are not widely perceived as influential in shaping AI-driven drug selection. Only 8% of respondents believed biopharma has a significant influence, while 52% said they have some influence, 32% said minimal, and 8% said none.

Conclusions: : Health systems have rapidly adopted commercial AI platforms with high expectations for improved care delivery, but integration with EHRs and alignment with clinical guidelines remain incomplete. It will be essential for manufacturers, clinicians, and payers to better understand how these AI-driven workflows evolve over time, shape patient care and treatment decisions, and impact clinical outcomes and costs.

Sponsorship:: Precision AQ