February 04, 2021
By Julie Gould
Joe Schmid, chief technology officer at SymphonyRM, discusses how the use of artificial intelligence (AI) can help influence consumer and patient behaviors to connect them to care, and provides examples of how his team has helped and is continuing to help health systems across the country with patient engagement and health outcomes with the use of various data strategies.
Let’s talk about AI in health care. How does use of AI span in health care from the clinical side to the patient engagement side? How does AI help influence consumer and patient behaviors to connect them to care?
Great question—when most people think of the usage of AI in health care, they typically gravitate towards clinical use cases like cancer detection in radiology or perhaps applications like drug discovery. Significant investments have been made in AI for a lot of “behind the glass” applications in health care that help patients once they are checked in and admitted for care. What we are seeing now in health care is the advancement and rapid adoption of leveraging clinical AI concepts for patient engagement as this has been an area of underinvestment over the past decade. The use cases related to engaging and influencing patients will have an even greater impact and can be applied much more quickly to large populations of patients.
As important as the clinically-focused use cases are, health care providers often work with pre-diagnosed patients or know they’re at risk for certain chronic conditions. Prevalent conditions like chronic kidney disease, COPD, heart disease, and others can be immensely improved by helping patients take actions on everything from early screening to lifestyle changes, like diet and exercise. Using AI, we can help predict what messaging, channel(s), and timing are most effective and relevant to engage and influence patients. For instance, engaging a 39-year-old mother of three is very different than engaging a 65-year-old man. AI can use data to make personalized recommendations, measure the results, and continuously improve those recommendations to help connect patients to the care they need.
Effective solutions for patient engagement and influence will need to take into account both behavioral analytics data (models that predict what messaging, channel, time, etc. will be most effective for this patient) and clinical data (data that is in the EMR). Trying to make recommendations that engage patients without clinical awareness is, not surprisingly, destined to fail. Knowing that a patient was recently diagnosed with stage 3 chronic kidney disease or that a woman has a higher risk of breast cancer are critical in being able to provide relevant, personalized outreach that effectively engages patients and influences them to take action.
Do you have any examples of how your team has helped and is continuing to help health systems across the country with patient engagement and health outcomes with the use of various data strategies?
We’ve been very lucky in having the opportunity to help many health systems with their proactive outreach and patient engagement over the past few years. Our recent work with clients like Virtua Health combines clinical AI with behavioral analytics to proactively influence patients in multiple areas like breast cancer, bariatrics, cardiology, etc. Specifically, our machine learning models, which have been trained on hundreds of millions of EMR records, predict which patients have the highest risk in each of these areas. Our breast cancer model, for example, is able to identify that patients scoring in the top 5% have a 15X higher risk. By targeting patients by their risk levels and creating audience segments that map to relevant, customized content, we were able to help over 2000 high-risk women get scheduled for mammograms.
That’s just one example out of the thousands of outreach campaigns that we’ve run in conjunction with clients. Using modern machine learning techniques, models trained using clinical EMR data are extremely accurate in predicting risk and propensity, often producing 6 to 10 times greater accuracy than standard national models. When health systems utilize these model scores in conjunction with behavioral analytics, it produces a very powerful mechanism to optimize and deliver hyper-relevant messaging that is delivered in the right channel (email, text, phone) at the right time to maximize patient outcomes.
Moving to the popular topic of 2020, COVID-19, how does data play a critical role in vaccine distribution strategies? How does it help improve and classify a system’s patient population?
Without a doubt, COVID-19 vaccine distribution and inoculation are the top priorities for health care in 2021, starting now. The closest set of data that we can take a look at to help in this national initiative is around the successes that we’ve had with the flu vaccine. From our own data of the thousands of campaigns that we’ve run with our clients, email outreach for flu shots have some of the highest conversion rates. With an 11.3% overall conversion and a whopping 49.0% conversion for patients who open an email, our data show that patients are highly responsive to vaccine campaigns. Of course, the COVID-19 vaccine is nowhere equivalent to flu shots.
Fear and uncertainty are likely to result in some populations avoiding or delaying the vaccine, while others will be clamoring to get access as early as possible. The initial roll out has already begun and we’ve been working with our clients to lay out comprehensive campaign strategies that uses data and content to optimize communication outreach for each rollout phase.
A crucial part of that strategy is around patient education and follow-up approaches for those that are qualified but reluctant to participate. It’s important that this strategy is bilateral, ie, not just pushing educational content, but collecting the relevant data to refine the strategy. Along with educational content, health care organizations need to be collecting data from their patients on their views of the vaccine, eg are you willing to be vaccinated as soon as you’re eligible? If not, why not? Collecting this type of data will allow for more intelligent outreach, with messages targeted to different segments of the population and overcome reluctance.
A lot of health systems are finding that patients are delaying care amid the pandemic. Can you talk a little about the opportunities that are available to health systems for re-engaging patients following COVID-19?
We went through a first wave of this with our clients earlier this year. After the initial cancellation of elective procedures in the spring when the pandemic first hit, we partnered with our clients to design a two-phase outreach strategy. The first phase was focused on “schedule restoration”, ie, rescheduling canceled appointments. The audiences for these campaigns were quite simple and built by identifying patients with canceled appointments that had a cancelation reason of “COVID” or similar.
Phase two was focused on maximizing utilization of capacity. We used our AI models to identify candidates in high value service lines with available capacity that included orthopedics, bariatrics, cardiology, etc. By targeting patients with higher propensity to use those services, we were able to help our clients accelerate revenue recovery and maximize their operating capacity.
With the recent rise of COVID-19 cases, clients are again seeing patients delay care. However, with the availability of the COVID-19 vaccine, we’ll be helping clients not only manage their COVID-19 vaccine outreach and communications, but we will also be using data to help identify patients who’ve delayed care or are overdue for routine screenings. The lesson we’ve learned from COVID-19 in 2020 is that we can help health systems to quickly re-engage patients to continue on their care journey while accelerating their revenue recovery.
Is there anything else you would like to add?
Health care organizations interested in adopting AI for patient engagement no longer have to fear the “blackbox syndrome.” Modern AI-driven platforms deliver data transparency to help systems with attribution and a clear line of sight to their returns on their investments. Our mission as AI thought leaders is to help health care teams increase their visibility into their data so that they can understand how they are driving top line revenue.
Health systems can quickly set up this technology today with very little impact on their IT infrastructure and achieve substantial ROI in less than a year. It’s one of the most high-impact projects that health systems can prioritize, as we exit the COVID-19 era. Health systems have been sitting on a goldmine of data within their EMRs and there’s no better time to activate it to re-engage their health community.
About Joe Schmid
Joe Schmid is Chief Technology Officer at SymphonyRM where he is responsible for architecting and developing the AI technologies that power the nation’s leading health systems. He has spent the past decade pioneering machine learning based technologies that help financial institutions like JP Morgan Chase and American Express with fraud detection to AI-powered health care patient engagement technology for health systems that include Intermountain, HonorHealth, MultiCare, Virtua, etc. Prior to SymphonyRM, he held engineering and executive roles at Silicon Valley startups in speech recognition and mobile applications.