The direct-to-consumer virtual care organization Hims & Hers Health revealed the introduction of a new AI-powered service, MedMatch, which furnishes healthcare providers with anonymized data points acquired from the company’s client database, intended to aid mental health experts in identifying suitable treatments for patients.
Dr. Patrick Carroll, chief medical officer at Hims & Hers, had a discussion with Our blogNews about the company’s internally constructed EMR and how its customer datasets drive MedMatch.
Our blogNews: Can you elaborate on MedMatch and its functionality?
Dr. Patrick Carroll: I joined Hims & Hers in June of 2019. I came from a relatively conventional healthcare background. I served as the chief medical officer at Walgreens for five years, and before that, I worked as a primary care physician for 30 years. I headed some health systems and led a CMO, as well as several large, multispecialty groups. My expertise lies in primary care. When I came aboard, I observed the uniqueness of what Hims & Hers was accomplishing, even back in 2019, in leveraging technology in a virtual care setting to deliver high-quality personalized healthcare, not only addressing the access challenge, but also tailoring care to individuals.
I’ll talk about MedMatch. However, it is essential to first delve into our EMR, which is truly innovative and markedly distinct from the six different EMRs I encountered in my career, including Cerner and Epic, and the cumbersome nature of excessive clicking.
MedMatch is remarkable, and the reason we can achieve this, with mental health as the focus of the beta test, but with the intent to expand to all our areas of expertise, is that, unlike an environment I practiced in previously, such as primary care, where I would see a patient as a family physician. They would present a concern or a follow-up related to conditions like hypertension or diabetes. I would engage with them, review their medical history, peruse my EMR, which at the time was an iteration of Epic, and then make decisions regarding medication adjustments.
I would then issue a prescription, which they would take to a CVS, Walgreens, or a local pharmacy, and have it filled. I had no visibility on their adherence to the medication. It did not constitute a tailored system… and I might see them again in three to six, or even 12 months. Thus, I remained uninformed about their interim progress. I was unaware of their medication adherence, and I only had access to half of the data points related to pharmacy, medication, and check-ins, unless a major issue prompted them to follow up with me.
What we have at Hims & Hers is genuinely distinctive, as we operate as a fully vertical health system. When an individual comes to our platform with a specific health concern, we provide them with comprehensive information. This means that if they arrive seeking support for mental health, sexual dysfunction, or hair loss, they can access extensive material through our search engine optimization efforts and online content.
They then progress through a structured workflow, where all questions are meticulously designed to gather essential information. These workflows and protocols were formulated not only by our in-house physicians but also by national experts. Every crucial question is asked, and upon their responses, we make determinations regarding their suitability for our platform. If they pass this screening and appear to be candidates for the recommended medications, their visit is scheduled or set up for a video consultation, which is usually asynchronous.
Subsequently, through the interaction between the customer, the patient, and one of our 600 providers across all 50 states, a decision is mutually reached on their eligibility for the medication and its potential benefits. We meticulously explain the medication’s side effects and when to anticipate improvements, as well as the indications for follow-up with us. Yet, we do not leave this to chance. The prescription is mainly fulfilled by our proprietary pharmacies, handling approximately 80% of our prescriptions. Consequently, we have complete visibility on medication adherence, a privilege I did not have as a primary care physician.
Moreover, depending on the condition, we conduct regular outreach, with the frequency tailored to the specific ailment. For instance, in mental health, participants undergo routine check-ins via our program to record GAD-7s [General Anxiety Disorder 7 questionnaires] and PHQ-9s [Patient Health Questionnaires]. They complete these assessments initially upon entering the mental health platform, and we then monitor their progress over the subsequent one, two, or three years, assessing whether they are showing improvement. This information is relayed to the provider, who can then decide on medication adjustments or whether an earlier follow-up or an escalation is necessary.
In essence, we possess millions of data points stemming from a fully vertical system. Our insight spans the patient’s demographics, medical history, medications, prior experiences related to mental health, and any prior encounters with side effects from SSRIs [selective serotonin reuptake inhibitors] or SNRIs [serotonin and norepinephrine reuptake inhibitors]. We have documentation derived from these data points, eventually resulting in the generation of a prescription and routine structured check-ins. Furthermore, we oversee pharmacy adherence, since we manage the pharmacy fulfillment.
With these myriad data points, one can envision the treasure trove they represent for individuals involved in machine learning or AI. This data enables the identification of specific SSRIs for mental health, the focus of our beta test, best suited for a particular patient, based on the responses, background, demographic, and previous experiences with side effects, indicating which medication is most likely to be effective for that individual.
Ultimately, we provide the recommendation to the provider, who then makes the decision on which medication to prescribe.
We are currently in the initial stages, but we are observing some very positive signs, and we can then apply that to each of our verticals to personalize and identify specific medications and even dosages that will suit that patient based on all the data points we possess.
MHN: The AI utilizes data from your existing clients. What specific kind of data regarding mental health will aid care providers in making more informed decisions?
Carroll: What we have established is a way to provide them with real-time visibility for that particular patient, comparing it to all of our datasets, which medication would be the most probable to work for that patient. It then suggests that medication. Again, the provider, for various reasons, can choose to take a different approach, which is acceptable. Nevertheless, it essentially offers them a clue as to which medication is likely to work based on all of the anonymized data they are analyzing, and we will present that directly to our providers.
Currently, we are implementing this with providers – testing it. We have received highly positive feedback on an iterative model, although, as with all these AI models, they improve over time. The key point to bear in mind is that we don’t dictate to the provider, “You must prescribe this medication.” We simply provide them with guidance. It’s almost like … not cheating on the test, but at least informing them that this is more likely to be effective than the alternatives.
MHN: Does the company intend to introduce MedMatch across its entire platform at some stage beyond mental health? Is there an expectation of when this might occur?
Carroll: I don’t have a concrete timeline for that. Our aim is to perfect mental health with MedMatch because we consider it as one of our most diverse verticals in terms of medication, to be truthful.
It is also the most challenging decision for providers, whether they are utilizing an SSRI, as opposed to an SNRI or Bupropion, and it’s more intricate, as you are aware, 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 complex. Thus, what we are considering is which [vertical] can provide the most benefit for our customers the fastest and which one warrants giving the most guidance to our providers.
MHN: Is there anything else you would like to address that we have not covered yet?
Carroll: I believe a related aspect to what we are doing on this technological front and is highly underrated is that we developed our exclusive EMR. Hence, when the company commenced, we had the option to opt for something off-the-shelf as opposed to building our own, but as you can observe, what we do is quite unique. It is not solely virtual, but a significant portion is asynchronous, and some is synchronous. It is a platform specifically designed for the interaction between the customer and the provider. We are free from the constraints of solely conducting billing and coding, as traditionally EMRs do. Therefore, with our EMR, we have managed to construct guidelines that enable us to truly deliver an exceptional customer, as well as provider experience.
Another aspect of the EMR that people fail to realize is that when you construct something internally like this, we can more robustly track quality, for what happens is that we can identify providers who may not be adhering to guidelines, as it is all structured in our EMR, and we are capable of grading them. Last year, we completed over 50,000 encounter evaluations based on their interactions with patients, including, are they prescribing the appropriate medication, are they creating a coherent note in that chart, are they offering appropriate follow-up?
While leading large medical groups, I never had an EMR that could truly integrate quality as part of it, and then offer an experience that is well-received by both our clients and our providers, as they are not dedicating the majority of their time to handling billing and coding documentation. Instead, they are focusing on adhering to guidelines and ensuring they meet the quality metrics we have defined.
As the company expands and we introduce new services, we are able to seamlessly integrate those, and we have substantial flexibility to accomplish that. Naturally, we must also recruit a significant number of engineers for this, as you can imagine. However, instead of having to approach one of the established vendors and request modifications, I honestly don’t think they have thus far established an EMR that truly functions effectively in the digital, virtual health, asynchronous world. Thus, this constitutes some of the stimulating work we undertake with our EMR. And MedMatch is essentially an extension of leveraging our technological expertise to collect those data points to deliver improved care.