Direct-to-consumer virtual care company Hims & Hers Health announced last week the launch of a new AI-enabled offering, MedMatch, which provides healthcare providers with anonymized data points generated from the company's customer database, aimed at helping mental health professionals identify suitable treatments for patients.

Dr. Patrick Carroll, chief medical officer at Hims & Hers, sat down with MobiHealthNews to discuss the company's internally built EMR and how its customer datasets power MedMatch. 

MobiHealthNews: Tell me about MedMatch and how it works. 

Dr. Patrick Carroll: I came to Hims & Hers in June of 2019. I came from somewhat more of a traditional healthcare background. I was the chief medical officer at Walgreens for five years, and then, prior to that, a primary care physician for 30 years. I led some health systems. I led a CMO. but also some large, multispecialty groups. My DNA is primary care. So, I just saw that what Hims & Hers was doing even back in 2019 is really unique, you know, how you leverage technology in a virtual care environment to deliver really high-quality care and not only solve the access issue but really to get down to personalized healthcare for folks.

I'll talk about MedMatch. But we should probably spend a few minutes talking about our EMR, which is really innovative and very different from the six different EMRs I was on in my career, including Cerner and Epic, and the tyranny of the million clicks.

MedMatch is fascinating, and the reason we're able to do this, and the beta test is in mental health, but we're going to be able to do it for each one of my verticals, is that, unlike the world I practiced in, in primary care, I would see a patient as a family physician. They would come in with a complaint or a follow-up for something like a hypertension or diabetes. I would talk to them. I'd look at their past history, you know, review my EMR, which at the time was a version of Epic, and then would make decisions in terms of medication modifications. 

And I would generate a prescription for them, and that prescription would go to a CVS, or Walgreens, or a local pharmacy, and they'd get it filled. I had no idea whether they were adhering to the medication or not. So it wasn't really a verticalized system ... and then I might see them back in three to six, or even 12 months. So, I didn't know what was going on in the interim. I didn't know whether they were adhering to medication, and I was missing half of the data points around pharmacy, medication and check-ins. Unless they had a major issue, they didn't follow up with me. 

So what we have at Hims & Hers is truly unique in that we're an entirely vertical health system. So, someone actually comes onto our platform with a specific health issue. We provide content. In other words, if they come on, and it's mental health, or they come on and they have sexual dysfunction, or they have hair loss, they can read everything about it through a lot of our search engine optimization initiatives and the content we put online.

And then they come on, and they actually go through a workflow that is very structured, and all of the questions are designed to get the essential information for them. Those workflows and protocols were built not only by our own physicians, but by national experts. And so every essential question is asked, and they answer, and then we make decisions based on whether they're appropriate for our platform or not. If they get through that screening, if it looks like they may benefit from those medications, that visit gets put in the queue if it's asynchronous, which most of our visits are, or the video visit gets set up asynchronous.

So then this interaction between the customer, the patient and the provider, we have over 600 providers in all 50 states, and then a decision is made through that communication from the customer, the patient and the provider whether they would qualify for the medication and whether it would be beneficial for them, and then all the side effects are going through in terms of the medication, and then all the information about when to expect improvement, you know, the indication of when to follow-up with us. But we just don't leave that to chance. The prescription actually gets sent to, you know, 80% of our scripts now are filled by our proprietary, our own pharmacies. And so we have full visibility on medication adherence, which I never had as a primary care physician.

In addition, on a regular basis, depending on the condition, the frequency is determined by the condition. We do regular outreach. For example, on mental health, they get regular check-ins through our program to document GAD-7s [General Anxiety Disorder 7 questionnaires] and PHQ-9s [Patient Health Questionnaires]. So they do that initially when they come on the mental health platform, but then we're able to see through their entire journey over the next year, two years, three years, whether they're improving or not, and that information gets back to the provider, and then they can make decisions on medication adjustments or whether a sooner follow-up is needed or any form of escalation that needs to occur.

But basically, what we do is we have millions of data points that come from a fully vertical system. So we have insight into the patient's demographics, past medical history and medications. For example, on mental health, prior experience, and side effects with SSRIs [selective serotonin reuptake inhibitors] or SNRIs [serotonin and norepinephrine reuptake inhibitors] in the past. And then we have the documentation that comes from that, those data points that come from that, the prescription that is generated, and then the follow-up in terms of efficacy through regular structured check-ins. And then we also have the pharmacy adherence since we do the pharmacy fulfillment.

So when you have those millions of data points, you can imagine that's like a treasure trove for someone in machine learning or AI. They can actually take all that information and then identify, for example, in mental health, that's our beta test, which SSRI specifically for folks with the way they answer the questions, with their background, with their demographic, with their past experience in terms of side effects, what worked and what didn't, which SSRI is most likely to work for that specific patient.

So, it becomes very personalized. And then, what we do is we then suggest to that provider, the provider ultimately makes a decision, which medication to choose from, and we just do generic medications. So SSRIs, SNRIs, Wellbutrin. But the provider is given some guidance based on the data that we have this medication is more likely to work in a quicker manner with this patient. So that's essentially what we've created in MedMatch. 

We're in the early days of it, but we're seeing some really positive signals, and then we can actually carry that over to each one of our verticals to really personalize and identify specific medications, specific even dosages of what will work for that patient based on all the data points we have.

MHN: The AI uses data from your existing customers. Exactly what type of data in regard to mental health is going to help care providers make more informed decisions?

Carroll: What we've created is a way to give them visibility in real-time for that specific patient for their background, comparing it to all of our datasets, which medication would be the most likely to work for that patient. And so it surfaces that and makes a suggestion. Again, the provider, for various reasons, can say, "Well, okay, I get that, but I'm going to do something different," which is fine. But I think basically it gives them the partial answer to the test of which medication is going to work based on all of that anonymized data that they're looking at, and we will surface that directly to our providers.

And so we're actually doing this with providers now – testing it. And we've gotten very positive reviews on an iterative model, though, as all these AI models are, they get better with time. The key thing to remember is we don't just say to the provider, "You must prescribe this medication." We just give them a clue. It's almost like … not cheating on the test, but at least letting them know that this is more likely to work than the other.

MHN: The company plans to roll out MedMatch across its entire platform at some point beyond mental health, correct? Do you have a scope of when that might happen?

Carroll: I don't have the crystal ball on that. We definitely want to get mental health with MedMatch perfected because we actually look at that as one of all our verticals with the most variation in terms of medication, quite honestly.  

It also is the toughest decision for providers, you know, whether you're using this SSRI, versus an SNRI, versus a Bupropion, and it's more complicated, as you know than a PDE5 [Phosphodiesterase 5], which, you know, sure, there's generic Cialis, which is longer acting, generic Viagra-type medications that are shorter acting ... that's not as complicated. So I think what we're looking at is which [vertical] can have the most benefit for our customers the fastest, and which one, for our providers, is worth giving them the most guidance on.

MHN: Is there anything you want to add that we still need to cover?

Carroll: I think a corollary to what we're doing on this technology front and is really underappreciated is that we built our proprietary EMR. So when the company started, we had the choice of going to something off-the-shelf versus building our own, but as you can see, what we do is fairly unique. It's not only virtual, but a lot of it is asynchronous, but we also do synchronous, and it's a platform that is specifically designed for that interaction between the customer and the provider. And we are freed from the chains of just having to do billing and coding, because that was essentially what traditionally EMRs do. They're billing and coding machines. And so what we've been able to do is to create this EMR with built-in guidelines that allow us to really deliver a great customer, but also provider experience.

And the other part of the EMR that people don't realize is when you build something internally like that, we're able to follow quality in a lot more robust manner, because what happens is that we're able to identify providers who may not be following guidelines, because it's all structured in our EMR, and we actually are able to grade them. Last year, we did over 50,000 encounter reviews based on their interaction with patients, but also, are they prescribing the appropriate medication? Are they making a coherent note in that chart? Are they providing appropriate follow-up?

When I led large medical groups, I never had an EMR that could actually really embed quality as part of it, and then also provide an experience that is well received by both our customers and our providers, because they're not spending all their time doing billing and coding documentation. They're spending their time following the guidelines and making sure they hit the quality metrics that we've set up.

As the company develops, as we add new services, we're able to bucket those on, and we just have great flexibility to do that. We also have to hire a lot of engineers for that, as you can imagine. But other than for us to have to go back to one of the traditional vendors and say, "Can you adjust this? Can you adjust that?" Quite honestly, I don't think they have set up, to date, an EMR that really works well in the digital, virtual health, asynchronous world. So that's been some of the exciting work we do just on our EMR. And then MedMatch is just an extension of using our technology chops to collect those data points to deliver better care.