Managing predictive AI as hospitals scale adoption

Hospitals increasingly use predictive AI — but must also manage it well

In recent years, there has been a meaningful spike in hospitals’ use of predictive AI, which uses algorithms and machine learning to analyze electronic health record (EHR) data — including real-time data, testing results, and historical health data — to predict serious health outcomes, so they can be addressed before they become a problem.

This doesn’t necessarily reflect hospitals’ active pursuit of technology platforms or standalone products with predictive AI. Rather, the change has largely come about because predictive models have increasingly been embedded into the EHR systems that hospitals are already using to manage patient care and run operations.

In 2024, 71% of hospitals reported using predictive AI, up from 66% in 2023, according to federal hospital data published in September 2025 by the federal Assistant Secretary for Technology Policy (ASTP).¹ ASTP is a division of the Office of the Secretary for the U.S. Department of Health and Human Services. More than 2,000 U.S. hospitals participate in the research each year.

EHR platforms are the most common sources of predictive AI. In 2024, 80% of hospitals used EHR-embedded predictive AI. By comparison, 52% of hospitals reported using predictive models outside of their EHR systems — developed by a third party. And 50% developed their own predictive AI model for internal usage.¹ (The numbers total more than 100% because some hospitals use more than one system.)

These figures indicate that the usage of EHR-embedded predictive AI has moved beyond the experimental phase and into the mainstream. Because predictive models are becoming part of hospitals’ core operational infrastructure, hospital governance of these models can no longer be treated as optional. Hospital leadership must evaluate and manage AI usage to determine if it meets the hospitals' guidelines.

Why governance is necessary for EHR-integrated predictive AI

As EHR developers have built more predictive AI into their platforms in recent years, hospitals’ usage of healthcare AI has been shifting from pilot to practice.

Hospitals aren’t passively using predictive AI; they’re integrating it into operational workflows, analyzing patient data to forecast health trajectories, prevent adverse events and predict readmission risks. Formal oversight can help healthcare organizations evaluate performance, monitor risk, and clarify accountability, establishing precedents to guide how they address future challenges.

Many hospitals are assigning committees and leaders for oversight of predictive AI. They’re monitoring the release of AI-enabled features after they’ve been released within EHR platforms. They’re capturing data and looking for trends to improve patient safety and other metrics. They’re evaluating their predictive models for accuracy and bias, to promote health equity. And they’re using predictive AI as part of their core operational infrastructure to streamline revenue management, monitor staffing needs, and reduce clinician burden.

How EHR-integrated predictive AI can ease clinician burden, improving operational efficiency

Predictive AI has the potential to streamline tasks for clinicians and hospital staff, including:

• Analyzing medical imaging to detect abnormalities faster and with greater precision than humans.
• Detecting early signs of disease before signs or symptoms arise that would be noticed by humans.
• Predicting inpatient health trajectories by assessing their risk of deterioration (such as sepsis onset), enabling clinicians to intervene early.
• Assessing patient readmission risk to provide at-risk patients with more effective discharge plans and additional follow-up appointments, as needed.
• Automating tasks related to scheduling and billing, such as predicting patient flow and minimizing the risk of no-show appointments, to give staff more time for patient-facing tasks or other needs.

Data shows that EHR-embedded predictive AI is already streamlining administrative tasks and relieving some clinician burden in hospital settings.

The usage of AI to automate billing procedures jumped significantly from 2023 to 2024. In 2023, 36% of hospitals reported using predictive models to automate billing, according to ASTP’s 2025 report. By 2024, 61% of practices had adopted this strategy. Similarly, the adoption of predictive AI usage to facilitate scheduling rose from 51% in 2023 to 67% in 2024.¹

Physicians are growing more comfortable with predictive AI each year, according to data from athenahealth's 2026 Physician Sentiment Survey, which polled more than 1,000 clinicians nationwide.² In 2026, clinicians reported* that their usage of AI had increased, including tasks that are performed by predictive AI, like clinical usage (such as stratifying risk and forecasting patient outcomes), administrative usage (such as optimizing patient scheduling) and revenue cycle management.

Formal evaluation and governance can affect predictive AI adoption rates

Each hospital that adopts predictive AI technology has a responsibility to evaluate its usage and define an oversight process. The willingness of hospitals to implement good AI-related governance can directly impact how willing and proficient clinicians and staff will be to use predictive AI.

Hospitals that evaluate predictive AI models for accuracy, bias and post-implementation performance, making changes as needed, are more likely to encourage its usage than hospitals that have yet to assess the value of predictive AI.³

A hospital’s culture can also encourage clinicians and support staff to use predictive AI. When this happens, hospitals are more likely to have strong AI adoption rates.³

EHR-integrated AI adoption rates are not uniform nationwide

Several factors can influence whether a hospital uses EHR-embedded predictive AI, including the size of a hospital's operating budget and its information technology infrastructure.

Hospitals that are large, located in urban settings, or system-affiliated tend to have greater access to EHR-embedded predictive AI. Smaller, rural, or independent hospitals are less likely to have access.

Data from ASTP’s 2025 report demonstrates this disparity.¹ In 2024:

• Only 59% of small hospitals used predictive AI, compared to 96% of large hospitals.
• Only 56% of rural hospitals used predictive AI, compared to 81% of hospitals in urban settings.
• Only 37% of independent hospitals had predictive AI, compared to 86% of hospitals that were part of a multi-hospital health system.

Smaller, rural, and independent hospitals are more likely to have smaller budgets and outdated computer systems, placing them on the wrong side of the digital divide. Lack of access to EHR-embedded predictive AI in these locations is an imbalance of health equity that may affect patient outcomes and operational efficiency. Without predictive AI to identify patients at risk of hospital readmission or to handle time-consuming tasks like patient scheduling, both care quality and other key performance indicators may suffer.

For wider adoption, EHR-embedded AI must be helpful without overstepping boundaries

Clinicians and support staff who embrace EHR-integrated predictive AI are fond of its time-saving features and its ability to automate tedious tasks to improve productivity levels. An AI tool like this, with its seamless integration into the EHR, seems organic for users.

As EHR providers and medical technology companies add new AI features to the EHR, they should prioritize ease of use, time saved from performing mundane tasks and the capacity to facilitate interoperability — the ability of different healthcare technology systems to exchange and use healthcare information effectively.

Interoperability is key for buy-in and trust. Clinicians will be less likely to rely on AI if data remain fragmented and tools can’t be integrated into existing workflows.

To encourage more hospitals, clinicians and support staff to adopt EHR-embedded predictive models, enhancements to EHR-integrated predictive AI should not be making decision autonomously; they should help clinicians make the decisions.

AI tools cannot, and should not, try to be autonomous or replace clinical decision-making. Human judgment and interpersonal interactions are non-negotiable elements of care for clinicians and patients.

EHR-integrated AI tools appeal to hospitals and clinicians more than standalone products

For ease of transition and greater interoperability, hospitals should adopt predictive AI tools that are built into the established EHR platform. Clinicians are more likely to use AI features that are intuitive and that can help improve their productivity and workflow.

When hospitals are ready to roll out predictive AI features, they should demonstrate their support by deploying them with guardrails, appropriate governance, and oversight.

When predictive AI is used optimally in a hospital setting, clinicians can use it to decrease time spent on documentation and increase their ability to provide high-quality, compassionate care that patients expect from their hospital.

Find out how more about athenahealth’s approach to AI, including how we work with customers on responsible implementation of AI in healthcare technology.