Krisda Chaiyachati, MSHP ’17, Penn Center for Innovation

The Pulse: What is your professional background and area of health care interest?

I’m clinically trained as a general internist or primary care doctor. I am also an Assistant Professor of Medicine at the Perelman School of Medicine, a Clinical Innovation Manager at Penn Medicine’s Center for Health Care Innovation, and the Medical Director for Penn Medicine OnDemand Virtual Care.

The Pulse: Can you provide an overview of the Penn Center for Health Care Innovation?

The mission of the Penn Medicine’s Center for Health Care Innovation is to reimagine health care delivery for dramatically better value, patient outcomes, and experience for Penn Medicine patients and serve as a national model. What differentiates our Center from many innovation centers is that it is embedded within the operations of a health system. We focus on identifying tools and processes that Penn needs to be a trendsetter and improve healthcare delivery. Then we apply design-thinking to reach those goals.

There are three main arms of the Center. First, there is the Nudge Unit, which consists of a design team that incorporates principles from behavioral economics to “nudge” patients and clinicians towards higher value decision-making. For example, one initiative has been to change the display defaults in our electronic health record to incentivize greater clinician prescribing of generic drugs. This initiative saved us more than $30 million in two-and-a-half years.

Second, we have the Center for Digital Health, which is as an incubator for digital health ideas and solutions. Here, researchers have studied everything from whether social media can predict mental health conditions to the use of crowdsourcing to map out AED machines across the state.

Finally, there is the Acceleration Lab, which leverages our own faculty and staff’s ideas to design and test solutions. This is our bread and butter.

The Pulse: What is an example of a project coming out of the Acceleration Lab, and can you describe how it came about?

One example would be Penn Medicine’s OnDemand Virtual Care Program, which I oversee.

We first noticed that Penn Medicine employees were visiting the emergency department (ED) far more than we might have expected. Around one in five employees visited the ED each year, and a significant proportion of those visits were considered avoidable.

Employees are a unique consideration for a self-insured health system like Penn Medicine. Self-insured employers, whether they are health care providers or not, take on the financial risk for employee’s healthcare spend. So, these employers typically have programs to control costs such as wellness programs and incentives to encourage primary care visits. As a health system, it’s a win-win to have convenient alternatives to an ED to take care better care of our employees and decrease congestion in our waiting rooms.

We had a strong motivation to reduce avoidable ED visits: improve the happiness of our employees and reduce unnecessary utilization. So, we developed and piloted a free telemedicine clinic where employees could connect to an NP or MD virtually, receive medical advice, a prescription when clinically indicated, and coordinate follow-up appointments if necessary.

The results were promising – we saw a sizeable reduction in the use of EDs in the pilot and reduced spending. Because of that, we’ve turned the pilot program into a fully scaled practice serving all 60,000 covered lives under our insurance plan and we transitioned the clinic to be public facing since January 2020.

The Pulse: Can you speak a bit about the economics of telemedicine vs. ED for a health system?

The economics, at face value, seem straight forward: the cost for each telemedicine visit is much cheaper than say the emergency department. You don’t have the overhead of office space, imaging, or equipment. The beauty of the program is the convenience and accessibility However, this is where the economics can get tricky – it may be so convenient, and the price may be so affordable for patients that they call when they didn’t actually need health care. In other words, there is a risk of visits that wouldn’t have occurred otherwise. What we want is for telemedicine to substitute for in-person care, especially the ED. In real life, though, you are going to have a mix of both overuse and substitutions. For a program to be financially viable, you just want to make sure the quantity and cost of unnecessary don’t outweigh the benefits of substitutions.

To date, we’ve measured substantial cost savings and seen more examples of substitution. That said, to date, our user base has been employees of a healthcare system. They may be relatively savvier and more knowledgeable about how and when to use healthcare compared with non-health care workers. As we open this to the public, we will continue evaluating whether telemedicine, the way we offer it, is keeping costs down.

The Pulse: What have been the biggest challenges in developing and scaling this program?

From a clinical standpoint, the biggest challenge is learning how to provide effective virtual care. What are patients’ expectations? What should be their experience? How do we transition and train providers, who are used to in-person care, on best practices for providing virtual care? There is the technical build challenge – how do we keep this operating, so the patients’ experience is seamless and high-quality. Finally, we need to make sure the economics are attractive for insurers and patients. We are constantly combing through the claims and cost analyses to understand the program’s value. Ultimately, I personally want to know if this improves access to care for patients who were struggling to get it previously.

The Pulse: Before we run out of time, I want to talk a bit about your chatbot initiative because it sounds interesting.

The chatbot fits the same theme – we want patients to access high value care without any sort of friction. We are designing a chatbot as an extension of our clinical teams. It should be able to receive patients’ text messages and answer questions right away or efficiently triage messages to a clinical provider. Hopefully for many patients, this saves them a trip to see a provider or create long waits before easy questions can be answered. For providers, we hope it can reduce the burden on providers by saving them hundreds of five-minute phone calls, emails, patient portal responses, or text messages. The chatbot can be a critical member of the care team, helping providers truly manage populations while funneling high-needs patients their way.

At the same time, we think we can leverage the chatbot as a tool to gather data, process information, and present it efficiently to providers. We are trying to develop this technology one disease at a time. The clinical case I’m working on is hypertension treatment and management. We hope that the lessons we learn for conditions like hypertension will be valuable and transferrable to other conditions as we scale the application of artificial intelligence into how Penn Medicine delivers care in the future.

The Pulse: What stage are you in the development of this technology?

We are currently doing the technical build of the chatbot for hypertension. During our first phase, we have human beings pretending to be chatbots, to see if and how people engage with the tool. Then we plan to phase out the human being by building in components that are completely automated and responses are chosen by the chatbot. This is just to the beginning and I look forward to reporting more in the near future.

The Pulse: Thank you very much for your time!

Interviewed by Neha Srivastava, December 2019