User Research: Data-Driven Healthcare

Sydnor [00:00:04] We brought together experts from different areas of WillowTree and spent time together in a room really understanding: what are core problems that patients, care teams, or providers face?

Sydnor [00:00:16] Kristen, our awesome researcher, took the early concepts from that first working session and did really thoughtful user research.

Kristen [00:00:26] We designed a survey to understand both patients' experiences during and after a recent emergency healthcare visit to ask them about jobs that they had -- so tasks that they needed to accomplish while they were in the hospital visit and after. Were those important and how satisfied were they with being able to solve those jobs or perform those tasks? We found that in the survey, a lot of people are just asking questions to the internet writ large and getting responses that may or may not be vetted with actual medical professionals. People just basically want to be able to ask questions with answers that are doctor-backed. And so there's this big need, this big opportunity to fill that gap. The data is there. There's no reason why it can't be shared in the right places.

Kristen [00:01:12] When we came into this presentation, the outcome that we were all aiming for was to narrow the scope.

Kristen [00:01:20] The first thing I wanted to do is get an actual account of an end-to-end user journey of a patient's emergency healthcare visit. And it just so happened that Andrew was able to give me a firsthand account. He had just been in a hospital experience.

Andrew [00:01:34] So I was keeping track of medications, my medication schedule, the what mixes, what can't mix. I was keeping track of limitations on what can I pick up, when can I drive, those types of things. The other thing I was worried about was insurance and how am I gonna pay for this, who's gonna pay, how much do I owe, what is all that? So when it comes to followup questions and what you have, it's more of a what-if analysis rather than it is, this piece of paper says take this twice a day.

Andrew [00:01:59] I don't wanna see what Reddit has to say about this concept. I want to see what the Mayo Clinic has to say about it, I want to see the World Health Organization says about it. I want see what real sources have to say about it.

Kristen [00:02:14] We think that focusing on aftercare resources will impact re-hospitalization rates, which is a primary measure in hospital quality.

Andrew [00:02:23] Yeah, there's your ROI, that's how we prove out the ROI of this.

Sydnor [00:02:27] There's aligned incentives between patients, physicians, and even the health insurance companies that are providing payment for that care where all parties are incentivized to make sure that that patient is receiving the appropriate amount of care for their needs, but also then not coming back into the hospital. People often don't know if they're experiencing symptoms when to call for help. Some people won't seek the care that they really need and some people will immediately go to the emergency room and that's not good for anyone.

Kristen [00:02:58] So this all culminates into these three primary opportunities that we're pulling out. The first one: Live In-Room Transcripts and AI Note-Taking. We've said throughout this that we want to focus on post-visit, but this is actually essential to being able to reference that doctor-provided information. The second is a Conversational Post-Visit AI Assistant. The ability to search and ask questions about personalized discharge information and do personal research with the doctor-backed references related to their condition is huge and it's very needed. And the third is sort of a smaller piece of the pie, but something that was highly ranked and mentioned in open-ended responses, which is this Personalized AI Medication Manager. So managing medications holistically with personalized schedules, special instructions, trackers, and identifying any adverse reactions with other medications or dietary restrictions. That obviously also means that we would need to incorporate their personal data and so maybe isn't part of the MVP of this one.

Christopher [00:03:54] The Live In-Room Transcripts and AI Note-Taking: using any of the speech-to-text models, being able to really incorporate any of that. And in terms of the Conversational Post-Visit AI Assistant, I mean, this sounds more or less like some sort of RAG setup with, we're just gonna put in approved documents and we can sprinkle in some personalized, specific details about the patient themselves. For the AI Medication Manager, I still have concerns with specifically generative AI being involved in the AI Medication Manager.

Andrew [00:04:31] We're going to save the AI-powered medication schedules. We're gonna save that for now and we're really going to focus on the conversational AI after discharge. And so we're going start building a prototype that brings together the data sources that support that, including the transcriptions from the doctors and patients' notes. We're are going to include clinically-approved documentations. And then the third is patient records. And so from the short term, we're probably going to fabricate some of the patient records and some of their conversations with doctors, just so we're not impeded by regulations and things while we build this proof of concept.

Sydnor [00:05:03] For purposes of this exercise, we're gonna create something that's likely to be a standalone app. But realistically, we know this would need to be embedded into existing workflows for a clinician or how a patient would want to navigate their own care experience. And now our engineering teams and our design teams are gonna go figure out how to make it happen. So excited to see what's ahead for this solution.

‍