Solving the comorbidity problem of mental health care through technology
Founder Series with Malekeh Amini of Trayt
Hack Mental Health had the privilege of interviewing Malekeh Amini, the CEO of Trayt. Trayt is a startup working to uncover the connection between the brain, body, environment, and behavior to find a solution to the comorbidity problem within mental health care. Part one will introduce how Malekeh got started with Trayt and the technologies they are leveraging to solve the comorbidity problem, while part two of this article will focus on the challenges and how different communities like Hack Mental Health could work together to solve the problems together. This article is part one of the two-article series.
Q) To start off, could you tell us a little bit about yourself and how you got to start Trayt?
I studied engineering at USC, which lead to me being in the automation, data and software side of engineering. Then I went to Harvard, got my MBA and shifted careers into healthcare by joining the Boston consulting group, who are a thought leader in that space. There I ended up doing a lot of different programs, both for pharmaceutical engagements in medical systems of payers, provider groups and all kinds of different work in strategy and operations in the healthcare space.
But I took a nonlinear path to get to where I am. After doing many years of consulting work, I had my five-year-old son diagnosed with a series of neurological disorders that also had psychiatric comorbidities and other medical conditions that accompanied them. Initially, I looked at that situation and continuously asked myself, why is this case so complex? And after doing some research, I realized that in fact, he is not an unusual case and that this is the norm, not an exception.
Despite the complexity, the current standard of care is using an oversimplification to really diagnose and treat them. And so that’s when I started to take a more active hands-on role in this. The other thing that really drove me was that everything that I had done in healthcare, even though it was called patient-centered, actually was targeting the corporate profitability of our clients. And even though they were called patient-centered, they had never touched a patient nor addressed the patient outcomes problem.
“The other thing that really drove me was that everything that I had done in healthcare, even though it was called patient-centered, actually was targeting the corporate profitability of our clients.”
Not only that, they also removed the incentive from the physicians and the clinicians that needed to take care of those patients. So even though it was called patient-centered, it had created a system that directly hurt the patients and the physicians who were treating them, which made me not only angry at the system but also at myself for really making a negative impact on patients’ lives.
So the approach I took was to look at these patients holistically, align all of the key decision-makers (patients and physicians) and create a system that economically benefits all while addressing the patient outcomes. Then everyone will benefit and we don’t need to take a backdoor approach of focusing on corporate profitability.
So that’s how I got more hands-on in this space to start Trayt. I started the company five years ago and moved to the Bay Area where the ecosystem was here to start a company from scratch.
Q) So one area that I would like to know more about is the care that you talked about. Psychiatric disorders are known to be difficult due to comorbidity, which makes narrowing down the problem difficult. What lead you to find the solution to be fully responsive to this challenge?
The standard of care today is based on relying on oversimplifying the symptoms in order to almost force a diagnosis of the patients because we have a diagnosis driven system of care. As all reimbursements need to have a diagnostic code, it makes it really difficult to narrow down the problem if you’re oversimplifying things to the point of not being able to distinguish between different disorders.
This problem not only impacts patient care, but also drug development and research. If you have broad diagnostic criteria that are oversimplified, then you are not able to recruit targeted patients who are responding to a specific treatment that can provide the types of efficacy data that would help get a drug through the clinical trial process. So it really was a significant problem across the system of care and all of the aspects of mental health, possibly explaining the poor performance of clinical trials in the brain space.
It’s significantly worse than other clinical trials — less than 60% adherence, 30% attrition from the clinical trials, and then 90% failure rates. So we know that this mechanism of oversimplification and broadening the diagnostic criteria are not working in clinical care nor in developing any kind of treatment.
“It’s significantly worse than other clinical trials — less than 60% adherence, 30% attrition from the clinical trials, and then 90% failure rates”
So our approach to solving this problem is the result of years of collaborative research with some of the thought-leading academic institutions. We created a sophisticated outcome system that captures the complex set of factors that impact behavioral symptoms in order to treat the patient in a holistic way, which includes psychiatric and non-psychiatric comorbidities such as adverse childhood experiences, social determinants of health and some other environmental factors that frankly impacts the drivers of behavioral symptoms.
We also created a proprietary measurement scale that was much more patient-centered. It spoke to how a symptom manifested itself in a behavior in order to not only understand each symptom better, but also to provide a higher resolution dataset that described the patient holistically in a way that could drive a treatment path or a treatment decision. And so that’s how we created a system of care based on the data that we’re collecting.
Q) That’s great! Could you tell us more about the technology that Trayt leverages to solve these novel problems and do what was not possible before in psychiatric treatment?
The proprietary aspect of Trayt really is in its content and in capturing the right data in order to holistically see psychiatric patients. So our technologies are simple, while our content and creation of insights out of these larger datasets are the complex parts of our application sets. So one of the biggest challenges in not only psychiatric health, but also in any kind of healthcare system, is the challenge of measuring and making sure that you secure patient privacy.
Data security technology in the past would have had a requirement to be physically on a server in a hospital, which eliminated the possibility of connecting these technologies to each other. This disabled the ability to analyze and create insights from the data across all of these different hospitals and technologies. So what has made our job much easier are the tools and the security layers and their privacy measures that follow the guidelines of HIPAA, but provided on AWS, for example.
The other component that I think, especially in psychiatry, has really been effective is the existence of smartphones. So psychiatry has traditionally been a lot of diary on pen and paper. Even the actual assessments, outcome measures, and diagnostic paths have always been a pen and paper, forming an unstructured data.
So the ability to engage the patients to submit that data through their smartphones and create a digital path not only makes that data available in real-time to clinicians, but also enables them to look at longitudinal data in a web-based portal. When it was pen and paper, the data were never tabulated, which made it impossible to extract any efficacy information nor understand whether the patients were progressing.
In psychiatry, this makes a significantly higher difference. For one, you reduce the time of visits by 30%, which used to be for administrative and paper filling in the clinic and ad hoc decision making that was just done in the clinic. So now we’ve eliminated that with this technology.
Now, the patient immediately clicks on the link on their smartphone, fills out the questionnaire in a very user-friendly way and submits the questionnaire. So when they come into their next visit, that information has already been submitted, reviewed, and the session is based on what do we do next? What does this information tell us and how do I look at the pattern of progress based on the pieces of data that I’ve collected over time? This technology enabled us to improve costs by 30% per patient, so it’s a significant impact.
The other really important component is the ability to create a cloud-based platform. The advancements in computational tools and machine learning are now really helping us connect these dots between psychiatric symptoms across psychiatric disorders.
They can identify those patterns of environmental factors and medical conditions to fully understand the patient’s condition. And I think 15 years or 20 years ago, this would have been very difficult to actually do.
Q) I would like to look more into what you talked about when you mentioned the cloud-based platform and the advancement of computing tools with machine learning that wasn’t available before. How are you leveraging those technologies to better treat the patient?
Sure! So one of the biggest problems right now in clinical research and in providing patient care is the lack of consistency in the data that has been collected, with a lot of scattered data points still sitting in the binders. The way we have changed this is by using some of the technologies to collect consistent data.
We are now the leading platform that collects consistent comorbidity and comprehensive multifactorial data on every patient that goes through our hospitals in psychiatry. Then using AI, we identify those patterns of comorbidity with which co-occurring conditions or symptoms occur and pinpoint what type of treatments have been proven to respond based on a subtyped profile, rather than using a broad diagnosis response to a treatment that we know cannot be tracked.
For example, by treating for a subtype in depression, we can utilize a particular type of therapy that is more effective and targeted towards treating patients of that subtype.
Then for machine learning, honestly, the bigger this dataset gets, the more refined those algorithms get. So you can now effectively create predictive models when a patient comes in and based on the initial intake, you can immediately say this is that subtype, this is the treatment that they respond to, and that patient goes off with a much higher chance of getting the right treatment at the point of care the first time they go in.
Want to learn more? Subscribe to HackMentalHealth’s listserv to get notified when the second part of this article goes live!
Originally published at https://www.hackmentalhealth.care on December 23, 2019.
