What you need to know:
– Recent data briefs from the American Hospital Association (AHA) and ASTP/ONC use data from 2023 and 2024 to examine trends in predictive AI adoption, assessment, and governance. The findings highlight both the widespread adoption of the technology and the sustained challenges of its impartial implementation.
– The use of artificial intelligence (AI) in healthcare has seen significant growth over the past decade, and predictive AI has become an important tool for improving results and operational efficiency. Predictive AI uses machine learning to predict future events, and in hospital settings, it includes predicting patient readmission risk, facilitating scheduling, and simplifying billing procedures.
Predictive AI adoption trends
Predictive AI adoption is on the rise, with 71% of hospitals reporting use in 2024, a marked increase from 66% in 2023. The most common application of this technique is to predict health trajectories or risks for hospitalized patients. However, the fastest growing use cases are management, with a significant increase in using AI to simplify billing and stimulating scheduling.
The “digital disparity” continues while recruitment grows. The report shows that small, rural, independence, and government-owned hospitals are behind large, urban, system-owned counterparts when adopting predictive AI. For example, in 2024, 86% of multihospital system members used predictive AI, compared to only 37% of independent hospitals.
Hospitals also source AI models from various locations. In 2024, 80% of hospitals used predictive AI from electronic health records (EHR) developers, 52% developed third parties, and 50% used self-developed AI.
Evaluation and governance of predictive AI
As predictive AI becomes more common, its assessment and governance will become more important. The report reveals that most hospitals evaluate AI models for both accuracy and bias. In 2024, 82% of hospitals assessed accuracy, and 74% assessed bias. More and more hospitals are evaluating all or most models for these factors.
The responsibility for evaluating these models is often shared among multiple entities within a hospital or healthcare system. Three-quarters of hospitals reported that in 2024 multiple entities were responsible for the assessment of predictive AI, with certain committees or task forces being the most common entities (66%). Department or departmental leaders were also frequently mentioned as responsible parties (60%).
Future outlook and challenges
The findings in the report suggest an aggressive and competitive market for AI tools, particularly management tasks. However, while the use of predictive AI for management purposes is rapidly increasing, its use for clinical applications such as treatment recommendations and health monitoring remains low. This may be due to the high risk of errors associated with clinical use.
For more information about ASTP/ONC data briefs, please see Hospital trends in predictive AI use, assessment and governance, 2023-2024
