Health systems around the country are piloting and rolling out AI projects to improve efficiencies and, ultimately, patient care. But how many of these initiatives are actually driving a return on investment?
Becker’s reached out to CIOs and AI chiefs around the county to ask: Which AI project has created the most ROI for your health system? Here are their responses.
Pam Austin. Senior Vice President and CIO of Ballad Health (Johnson City, Tenn.): Ballad Health’s employed physicians, Ballad Health Medical Associates, have seen great success with the implementation of the DAX Copilot.
As with all technology, our DAX scribe is evolving and has advanced from limited AI with a human clinical data specialist completing and placing the note, to a generative AI-driven scribe tool with nearly immediate note turnaround.
Transitioning to DAX Copilot offered us a myriad of advantages that align with our commitment to efficiency, innovation and superior patient care. Reducing the time spent on each patient’s encounter improves documentation quality, lessens burnout and enhances overall job satisfaction for our physicians and clinicians. Here are some examples:
1. Immediate note creation: DAX Copilot enables instant note creation, significantly reducing documentation time and allowing us to focus more on patient care.
2. Editing tools: With DAX Copilot, we have access to advanced editing tools that enhance the quality and accuracy of our documentation.
3. Deep integration into Epic: Seamless integration into Epic ensures smooth interoperability and enhances our overall workflow efficiency.
4. Generative AI capabilities: Leveraging generative AI, DAX Copilot empowers us to manipulate data in conversations and charts, offering deeper insights and facilitating more informed decision-making.
Additionally, we anticipate this service will be adaptable for certain inpatient clinicians this calendar year and become compatible with Android soon after.
William Carracino, MD. Chief Digital Health Executive and Chief Medical Informatics Officer of Lee Health (Fort Myers, Fla.): Ambient dictation is hitting on all cylinders: It is provider satisfaction, it is patient engagement, and it is financial ROI.
It’s something we’ve struggled with since time immemorial. Many, many years ago, when I was in ambulatory practice, I started dictating in front of the patients, and the patients absolutely loved it because they got to hear it. As we transitioned to the computer and our backs turned to the patients, there was a disconnect. And most docs are not as good of typists as they are chatters.
We have almost 300 doctors on it so far, and we are starting to see ROI, and doctors are adding more patients during the day.
The time savings have ranged between seven and 15 minutes, depending on specialty. We have some folks who are approaching up to 15 to 20 minutes of saved time, and that adds an extra follow-up.
Reshma Gupta, MD. Chief of Population Health and Accountable Care at UC Davis Health (Sacramento, Calif.): We developed an AI model to look at who in our patient population has a risk of hospitalization or an ED visit in the next 12 months. We then hand that list over to our care management staff that includes nurses, social workers, pharmacists and others to do proactive calls and outreach to those patients, identify their social needs, identify any symptoms they’re having, and to try to keep them healthier at home.
Our internally developed AI model outperforms a number of other out-of-the-box AI models purchased externally when tested head to head. Even a slight increase in performance metrics of these models has a huge impact on the efficiency of how our staff and resources are deployed. A better AI model is more likely to identify patients who truly would have ended up hospitalized, readmitted or in the ED to target interventions and reduce unnecessary costs.
It works better than what we used before, and is able to reduce the hospitalization rate within that program by 5% or 10%. That improvement really adds up to ROI.
John Higgins. Vice President of Talent Management at Essentia Health (Duluth, Minn.): We have an AI tool that sits on our career website that shows up in the form of a virtual recruiter, or chatbot. It will assist you with finding the right job. It can help you complete the employment application itself, and even go as far as asking some screening questions to help identify if you’re a good fit for the position or not.
Once a candidate has applied and has been qualified by the recruiter for the opportunity, the tool can assist with scheduling interviews. So we’re able to leverage the tool to send out a note to a candidate that allows them to self-schedule their own time. It sends out itineraries for their interview. You know where to report. You know where to show up. It confirms those meetings on leaders’ calendars.
Historically, we had dedicated interview schedulers who all they did all day long was schedule those interviews. We went from an interview scheduling time frame of an average, historically, of a little over three days to 28 minutes. The ROI has been pretty incredible.
Simon Nazarian. Executive Vice President and Chief Digital and Technology Officer of City of Hope (Duarte, Calif.): Our most significant financial ROI thus far has come from an in-house suite of revenue cycle optimization products, yet City of Hope’s greatest realized impact has been through the development and deployment of oncology-specific clinical decision support tools. Examples of these tools include predictive models for severe surgery complications, sepsis prediction for bone marrow transplant patients, 90-day mortality risk for palliative and supportive care, and length-of-stay prediction. These models have substantially enhanced patients’ clinical outcomes and improved patient and provider experiences.
Currently, City of Hope is working to transform cancer research and care using our oncology-specific large language model framework, HopeLLM. This multifaceted, City of Hope-developed AI technology provides comprehensive patient history summaries, matches patients to clinical trials and provides real-time access to critical information to streamline decision-making. Overall, this initiative aims to enhance patient care, accelerate research and precision medicine, alleviate administrative burdens and improve the physician experience, as well as make clinical trials more accessible, potentially saving lives. HopeLLM has significant potential to impact ROI and is part of the City of Hope in-house suite of AI-enabled products.
Evan Orenstein, MD. Chief Medical Informatics Officer and Vice President of Data and Analytics at Children’s Healthcare of Atlanta: For predictive AI, we have implemented both the Epic ED pediatric sepsis model as well as a locally developed model to detect pediatric sepsis in the emergency department, both of which have shown faster time to recognition and treatment including antibiotic and IV fluid administration. We have also shown that the Epic pediatric asthma risk-of-exacerbation model has decreased emergency visits and hospitalizations. Additionally, we have launched an automated pediatric early warning score and are developing and evaluating new models for clinical deterioration, volume forecasting, inpatient sepsis and acute kidney injury.
For generative AI, an AI scribe randomized pilot study has demonstrated substantial improvements in clinician experience by reducing administrative burdens, and we’re expanding this pilot through a larger randomized crossover trial. Additionally, we’ve introduced a secure AI chatbot for trained staff with usage growing substantially during the last several months. Our mobile app features an AI chatbot that provides digital concierge services for patients and families, and we are also seeing AI become increasingly embedded in vendor products to do things such as automate revenue cycle collection processes.
In short, the AI scribes have had the greatest impact on clinician experience, while our pediatric ED sepsis predictive model suite has had the most positive impact on quality and safety measures.
Philip Payne, PhD. Chief Health AI Officer of BJC HealthCare (St. Louis): I would say today the most measurable ROI is from ambient AI.
For those providers who have used our ambient solution for 60 days or more, around 65% have seen a reduction in documentation time of about an hour a day. And a third of those providers have actually seen a two-hour-a-day reduction in documentation workload. It has some very specific ROI implications, not only in terms of volume of patients who can be seen but also things like same-day closure rates on those encounters, which, of course, has a direct impact on revenue cycle as well.
Qualitatively, we have this amazing feedback from our providers that this is a life-changing technology. We have providers who would go home and spend 60 to 90 minutes of “pajama time” doing after-hours documentation. So that’s a dramatic change from a quality-of-life and workforce satisfaction standpoint.
And we have done some pilot work where we’ve interviewed our patients, and the patients are also very happy with the technology. Most of them describe it as game-changing, in that the providers are able to focus their attention on the patients and have a conversation with them, as opposed to sitting behind a keyboard.
Dwight Raum. Chief Digital Information Officer of Rochester (N.Y.) Regional Health: We’re still early in our AI journey — we’re in the “go-slow-now-to-go-fast-later” phase. This includes creating an AI center of excellence, running private LLM models, partnering with Rochester Institute of Technology to provide algorithmic expertise, staffing up key positions, and hiring a leader to take on innovation and AI.
We’re also using the AI center of excellence to navigate both the push and pull of initiatives. Our “push” list represents technology-driven opportunities that can be organically aligned with our problem space. Our “pull” list is strategic-plan guided, targeting specific areas and opportunities.
