Artificial intelligence advancements have expanded the possibilities for digital tools to improve hospital workflows.
However, many of these technologies are still too new to have a track record of success. Becker’s reached out to four leaders to find which AI projects have garnered the best return on investment in clinical operations.
Editor’s note: Responses have been lightly edited for clarity and length.
Vi-Anne Antrum, DNP, RN. System Chief Nursing Officer at Cone Health (Greensboro, N.C.): We’ve implemented AI that interfaces directly with patients who have chronic illnesses, like hypertension or diabetes. The AI helps them engage in their own care — prompting them to schedule appointments, check blood sugar or blood pressure, and report those results. So far, we’ve outreached to thousands of patients and have seen nearly 60% schedule and attend a primary care appointment. And keep in mind — these are patients who hadn’t seen a provider in at least three years. That’s a big deal.
We’ve also seen great success with AI in the ambulatory space, particularly with ambient listening for providers. This has allowed our providers to spend less time documenting and more time talking with patients. It’s been a huge satisfier for both parties. Clinicians don’t go into medicine to document for hours, they want to take care of people. Ambient listening reduces burnout and supports well-being. I’m very excited to see that become more widely available to other clinicians.
On the leadership front, using tools like Microsoft Copilot has also been a win. Having it embedded in our suite has helped leaders take meeting notes, create action items, and reduce cognitive load. That frees them up to focus on more strategic issues instead of clerical tasks. That’s been great too.
Russell Cameron, MD. Chief Medical Information Officer and Vice President at PennHighlands Healthcare (Dubois, Pa.): [We use a] software that searches through the chart — both discrete as well as free text data — and presents an assessment and plan to the provider, with hyperlinks back to the parts of the charts where the data was found.
Return on investment — definite improvement in collections per patient which far exceeded the cost of the software. This is because of increased specificity of diagnoses, increased number of complications or comorbidities and major complications or comorbidities, which helps the coders. Our original analysis showed the hospital’s case mix index increased close to 10%, with an increase in the complications or comorbidities/major complications or comorbidities capture rate of 7%.
Measurable improvement in quality — capturing the above data has helped change our measured quality measures like “mortality rates” because the observed to expected calculations are affected by the patient’s true degree of illness. This also provides a financial ROI thru various value-based reimbursement contracts.
Efficiency — the hospitalists have calculated that this software saves them about 10 minutes per note, or one to two hours per day, per hospitalist. Adoption by our hospitalists is over 95%. Notes are more standardized, which helps other providers find documentation. Although not measured, it is felt that the software decreased physician burnout.
Nariman Heshmati, MD. Chief Physician and Operations Executive at Lee Physician Group (Fort Myers, Fla.): The one that’s had the best return right now is AI scribes. They’ve been the most reliable to implement, have broad reach and are clearly showing improvement. We’re already seeing reductions in pajama time and that’s been the most impactful so far.
We’re also exploring AI for message triage, note generation, and visit routing — deciding if a patient should be scheduled for virtual care, in-person visits or self-care. But the data so far hasn’t shown a reduction in workload or message volume, so it’s unclear if those applications will pan out.
Joseph St. Geme III, MD. Physician-in-Chief and Chair of the Department of Pediatrics at Children’s Hospital of Philadelphia: We’re in the process of implementing an ambient listening tool that is being rolled out as a series of pilots. This tool captures physician-patient/parent discussion during a patient encounter and then summarizes the exchange. I’m anticipating that this tool will have a major impact on physicians and on patients/parents, lightening the load for physicians related to documentation of the history during patient encounters and allowing physicians to engage with patients and parents, rather than focus on the computer. Application of a related tool will allow physicians to respond to patient and parent electronic communication more promptly and with less time and effort. Yet another related tool will allow much more efficient and accurate review and summary of dense patient charts, improving the process involved in consultation on patients with a long and complicated past history.
