A recent study from the University of Minnesota School of Public Health collected answers from over 2,400 hospitals in the United States. It showed that about 65% of hospitals are using AI-based prediction models. These models help with several healthcare tasks such as:
AI prediction models can help doctors and nurses make better choices and improve patient care. For example, by predicting which patients in the hospital might get worse, staff can use resources better and prevent problems. Also, finding high-risk outpatients early can help avoid emergency visits.
Even though many hospitals use AI models, how they check for accuracy and bias varies a lot. The University of Minnesota study found that 61% of hospitals check if their AI models are accurate. This means these hospitals try to make sure the AI gives correct predictions. But only 44% check if their AI models are fair and free from bias.
This difference causes issues:
AI models that are not well checked could increase health care inequalities. Paige Nong, the lead author of the study, says hospitals with fewer resources often buy ready-made AI tools that are not made for their patients. These might not fit the needs of small or rural hospitals.
The University of Minnesota study found a digital divide between hospitals with different budgets. Hospitals with more money and academic centers often create their own AI models that fit their patients better. They can also check these models for accuracy and bias more carefully. On the other hand, hospitals with fewer resources usually buy standard AI products that lack local testing and adjustments.
This gap can cause problems:
Closing this divide is important to make care fair across all hospitals in the U.S.
Experts say policies are needed to help all healthcare places evaluate AI models well. These policies could offer:
These actions would help small and rural hospitals use AI safely. They would also set common standards to protect patients no matter where they get care.
Using AI in healthcare brings ethical and legal questions too. A review published by Elsevier Ltd explains some:
These challenges show why it is important to check AI not just for tech quality but also for ethics and rules.
AI is also used outside of clinical work. It can help with tasks like answering phones and talking with patients. Companies such as Simbo AI use conversation AI to handle phone calls and scheduling. This can reduce the work for front-office staff.
For medical practice managers, owners, and IT staff, AI phone automation offers benefits like:
Still, these AI systems must be checked to make sure they correctly understand patients’ requests and treat all callers fairly, including those with different languages and backgrounds.
Healthcare leaders who use AI tools need to balance new technology with patient safety and fairness. Checking AI models—for clinical or office tasks—should happen regularly. Some steps they can take are:
Following these steps can help hospitals and clinics of all types use AI to improve health without making care less fair.
Use of AI in U.S. hospitals is growing fast. More than half now use AI prediction models for medical decisions and office work. But many do not do enough checking for accuracy and fairness. This lack of review might harm patient safety and fair care. Richer hospitals often build and test their own AI models. Poorer hospitals tend to use general products. This shows a gap that needs policies, financial help, and stronger rules to fix.
At the same time, AI automation for phone answering and scheduling offers real chances to improve healthcare offices. Administrators should also focus on checking these AI tools for accuracy and fairness.
With ongoing evaluation, rules about ethics, and better access to support, healthcare leaders and IT managers across the United States can work toward a fairer and more efficient system. AI can then serve all patients well, no matter where they live or get care.
A study from the University of Minnesota analyzed the use of AI-assisted predictive models in U.S. hospitals, focusing on their adoption, use, evaluation capacity, and biases.
Approximately 65% of U.S. hospitals reported using AI-assisted predictive models for tasks like predicting inpatient health trajectories, identifying high-risk outpatients, and facilitating scheduling.
61% of hospitals reported evaluating their predictive models for accuracy.
Only 44% of hospitals conducted evaluations for bias in their AI-assisted predictive models.
Better-funded hospitals are more likely to evaluate their AI models for both accuracy and bias than those using external models.
The study highlighted a digital divide between financially robust hospitals that can design and evaluate their models, and under-resourced hospitals that purchase off-the-shelf models.
The digital divide poses risks to equitable treatment and patient safety, as models may not be tailored to the unique needs of different patient populations.
Researchers emphasize the need for policies promoting fair AI use, which could include financial incentives, technical support, and enhanced regulatory oversight.
AI models were primarily used for predicting health trajectories, identifying high-risk outpatients, and scheduling appointments.
Paige Nong, the study’s lead author, points out that under-resourced hospitals face challenges evaluating AI models, which could compromise patient safety and equitable treatment.
Simbo is using AI agents to automate all phone related workflows. Connect now to know more.
Schedule AI Agents Demo Now© Since 2019, Simbo AI.
