Precision Neuroscience Reimagined: Mental Health Research: Data Revolution

Hello. Thank you for joining me today. My name is Tina Marshall and this is Precision Neuroscience Reimagined. I’m joined today by Max Taquet, who is an academic clinical fellow from Oxford. And today we’re going to be talking about everything related to research and also clinical care within mental health and dementia. So hi, Max.

Tina Marshall: Could you tell me a little bit about your history, and then we can go onto the work that you do within the community for both research and clinical academia?

Max Taquet: Sure. So I initially graduated in engineering and started to get interested in medicine during my masters, started to develop tools for orthopedic surgery, then got interested in the brain and I did a PhD in brain imaging, and I went to Harvard for that. And then I started to become interested in the patients that were behind the images and behind the data, and I decided to study medicine, and that’s what brought me to the UK and to Oxford. And I studied medicine as a graduate entry student, and my passion for mental health and psychiatry grew during that time. And I started to collaborate with the Department of Psychiatry and then continue doing research in mental health using data scientists and adding more modalities of data, I guess, to the sort of tools that I’m using, and lately, electronic record data, which I’m sure we’re going to talk about more today.

Tina Marshall: I’m sure we will. I didn’t know that you actually started off as an engineer. So how does that translate to everything that you are doing now? I mean, what type of engineering?

Max Taquet: Well, I guess I still am an electrical engineer, which didn’t deal much with electricity as much as data and images and that sort of thing.

Tina Marshall: And the connectivity.

Max Taquet: Yes, that’s right. And I guess the way it translates is I think, as engineers, we’re trained to deal with complex systems and to understand them and without necessarily having-

Tina Marshall: Find solutions.

Max Taquet: Exactly, finding solutions without necessarily have access to having access to everything within that complex system. And I think the brain and mental health are two such very, very complex systems, probably the most complex systems you can find on the planet. And I think as an engineer, that’s the most fascinating thing you can study.

So drawn to the brain because of its complex nature. So, for you, what’s your goal in understanding? Do you have a goal in understanding how the brain works?

Max Taquet: Well, I think I have a goal in trying to see how we can help patients get better in mental health. As a clinician, that’s what I want to do. That’s my priority. And, obviously, that will require us to understand more about the brain and more about mental health, but always with the patient in the back of my mind to understand how we can make them better.

Tina Marshall: And what do you think are the biggest issues actually relating to patient care for these conditions? Because, obviously, a patient with bipolar or schizophrenia… So for everybody watching this, Max and I were talking earlier about schizophrenia and bipolar and the fact that these conditions can go up and down. So, actually, I’m going to dive straight in so that we can touch more on that subject because an area that I didn’t know was that actually, the genetics for schizophrenia touches every gene or many genes.

Max Taquet: Many genes, for sure. So it’s what we call a polygenic disorder, which means that it’s not just predicted by one gene, it’s not that one gene in your genome is faulty and that’s what causes a patient to develop schizophrenia. It’s the cumulation of a part of the genome, which predicts and increases the probability of somebody having schizophrenia or bipolar disorder, or many other psychiatric illnesses.

Tina Marshall: And so because of this, this is why it’s hard to find a cure or medicines that could work for any length of time.

Max Taquet: Well, I think it’s one illustration of how complex mental health is. It’s not just about the one single cause that causes the one single problem, and so far as we can target that cause then we can sort that problem for everyone. I think it really shows the heterogeneity, which is also presented not just in the gene but also in the brain and also in the clinical presentation of the patients. We see two patients with schizophrenia, and sometimes it’s very little in common between the two. Even clinically, even without looking at their genes, even without looking at their brain, they might be very different and, yet, the only thing we are doing at the moment is giving them the same treatment. And I think that’s where the main challenge is in terms of mental healthcare.

Do you think that because typically mental health has been an under-invested area, there hasn’t been the investment to sub-categorise those diseases? So could it actually be that even though you are seeing two patients with schizophrenia or two patients with bipolar, perhaps maybe they don’t have those conditions, or they have a subcategory within those conditions but it’s just that we don’t know that yet?

Max Taquet: Yes. I think the bottom line is, yes. I think that’s the case. I think the heterogeneity means that probably there are multiple depressions, multiple schizophrenias, and multiple bipolar disorders, and we need to better understand them from a biological point of view, but also from a psychological or social point of view, and understand that complexity and those subcategories so that we can deliver what we call precision medicine, so specific treatment for the specific individual.

Tina Marshall: And so tell me about the work that you’re doing around this. So you sit with both hats. So you sit with the research hat on and with the clinician hat. So can you tell us more about what you’re doing in the research field at the moment to help with that, to help with understanding, the sub-categorisation of these conditions?

Max Taquet: Yes. So the sort of things we do is to try and capture mental illness across different diagnoses and not just focus on one specific diagnosis, but looking across diagnoses to find some commonalities and some differences between patients across diagnoses, and we do that by using different modalities of data. So brain imaging is one, mobile phone data is another, and then electronic health data is a third one, which I think has a huge untapped potential.

Tina Marshall: So we’re talking about the diagnosis for patients. So my understanding, and do correct me if I’m wrong, is that in the UK, and this is where you can sit with your clinician and research a hat on, we don’t like to diagnose patients too soon to fit them into the box. So, actually, where we do keep all of the information of potentially your diagnosis without giving them the structured code for the ICD 10s, but you would put that into the free text. Is that right? Is it right that we don’t like to diagnose too soon?

Max Taquet: Well, I think we always keep sort of an inquisitive mind when we see a patient. We know that sometimes things down the line do not look like what they appeared at first, and so we don’t want to jump the gun, so to speak, to say, “Well, this is your diagnosis, off you go with that treatment,” because we know that things can be a bit more complex than that, sometimes a lot more complex than that. And so it’s true that sometimes we use what we call a differential diagnosis, which is the case in the rest of medicine as well, where we say, “Well, it might be that and we think the most likely diagnosis is that one. However, we think it might also be that or that or that,” and the only way to find out is to try the treatment, maybe wait, and then see whether things get better.

Tina Marshall: And how long does it typically take to actually diagnose these conditions?

Max Taquet: So it varies. It varies hugely from disorder to disorder, so it’s impossible to take to put a single number to that. But I guess just to take the example of depression, I mean, by definition, depression means that the patient has had symptoms for at least two weeks. So under two weeks, we wouldn’t make that diagnosis. However, it’s often the case that patients would wait for a long time before seeing a clinician, in part because they need to recognize the symptoms, and the need to see a physician, but also because the system is under pressure and et cetera, et cetera, et cetera.

Max Taquet: So sometimes it takes a lot longer. And then sometimes even when we see them, we realize that actually they don’t fit a nice diagnostic label. They don’t just fit the nice textbook definition of depression and they have maybe some symptoms of depression but also some symptoms of anxiety, and then we have a whole bunch of work around that. And there’s been new diagnoses being created such as mixed anxiety and mood disorder, et cetera, to sort of try to capture that heterogeneity, but I think this is insufficient in understanding the much vaster complexity of mental illness.

So if we go into your clinical work, how are you able to apply the research that you’re doing to the clinical work that you’re doing?

Max Taquet: Well, I think what’s been most informative in terms of my research into my clinical practice is that recognition that diagnoses sometimes do not make much sense, or at least they’re not necessarily helpful. We still make them because it’s important. It’s also important for funding-

Tina Marshall: And also the patients.

Max Taquet: And also the patients. The patients want to have a diagnosis.

Tina Marshall: The patients want to know, “What’s wrong with me? Give me a diagnosis.”

Max Taquet: That’s right. But at the same time, I also know that if a patient does not really fit a nice diagnostic label, we can still help them, and that’s because we know that, from a biological point of view and from a clinical point of view, those diagnostic labels are limited in what they can tell us about the actual illness and what the patient is actually experiencing.

Tina Marshall: So, really, you kind of sit there as the detective working out the symptoms that are related to the patient and then developing a care plan around all of that?

Max Taquet: That’s right. And, actually, psychologists are very good at that. So psychologists do not… I mean, they also sometimes work with diagnoses when it’s helpful, but usually, they work with what they call formulation, which is instead of just having one diagnostic label, one word, they will have an entire paragraph describing what the patient is experiencing in terms of what’s been causing it, what’s been maybe predisposing them to how they are at the moment, what’s maintaining that condition, et cetera, et cetera.

Max Taquet: And so by having and representing that complexity in plain text, they’re able sometimes to find different entry points to treatment, and those entry points might be biological. So it might be that we can give that patient a drug, but it might also be psychological or even social. And we know that sometimes patients, one of the maintaining factors, just to give you one example, of depression might be that socially, they’re struggling. They might be in debt or they might have no fixed home, or they might be homeless, and all of these things are contributing to their depression. And ignoring those aspects is certainly not helpful for patients, and we need to take those into account when we treat patients.

Tina Marshall: So making sure that we understand the context around the patient and, of course, that no one patient is the same.

Max Taquet: Absolutely.

Tina Marshall: So tell me more about the data that you use. So you touched on earlier that you’ve moved into data science to help with that.

Could you explain more about how and why you think that data science could help unlock some of these challenges that you’re facing?

Max Taquet: Sure. So the only way to really capture that complexity that we’ve spoken about is to have sort of large and deep data sets. So, why large? Because, well, the heterogeneity means that it’s very difficult to have two patients alike. You need to look at large samples. That gives you more chance of finding similarities between patients, which is obviously the gist of science. We need to have some similarities to be able to make some inferences, but they also need to be deep in the sense that we need to understand the depth of the complexity.

So we can’t just say, “Well, I have a large data set of patients, and all I know about them is their diagnosis,” because obviously then we are limited in terms of what we can learn from those patients’ experiences. So we want to have those two features, which are sometimes difficult to find together in the same sort of data. But, nevertheless, I think that’s the aim. So we try to have some biological data, genetic data, and brain imaging data. The mobile phone data is helpful in understanding the patient experience on a day-to-day basis, and then the electronic health record data and, in particular, electronic health record data that does not just contain diagnosis and treatment but also individual experiences, symptoms, social factors, et cetera, are key to understand that complexity and really capture it.

Let’s go into the data that you’re collecting from a patient’s mobile phone. What data is useful for that? I mean, 90% of us have smartphones. So what data do you capture from there that you find useful?

Max Taquet: So the sort of data that I’ve been looking at is what we call ecological momentary assessment, where we simply ask patients to answer a bunch of questions every day, sometimes even multiple times a day. And so the type of questions we would ask those participants is, how are you currently feeling? And then they have a slider to answer from zero to 100. What are you currently doing?

What is compliance like?

Max Taquet Well, so that’s the trick. So that’s difficult. So compliance is not too bad at the beginning. People are excited. They have this new app, and most of that research done was in 2012, 2013, and 2014 when having an app and answering this question was exciting. Not so much anymore, I think. So in the beginning, compliance is quite good, and then you see the inevitable difficulty with that sort of data, which is the retention rate is absolutely dismissal. So it’s very difficult to keep people engaged in that kind of study for very long periods of time. So you can learn quite a lot, but it’s difficult to scale that up to many people, including, in particular, people with severe mental illnesses.

Can you learn a lot in the short period of time that they are engaged?

Max Taquet Yes, you can. So one thing, for instance, that we discovered using that kind of data is a specific relationship between what people were doing and how people were feeling. And so, essentially, the relationship that we found, in a nutshell, is when people feel rather bad, they then tend to engage in activities that will improve their mood, and that’s quite intuitive. If you feel rather down, you feel rather sad, you will say, see a friend or maybe go out for a walk because you want to restore that mood. The surprising aspect was that when people were feeling rather good, they then tended to engage in activities that would reduce their mood. Now, the interpretation of that is not that people actively want to decrease their mood, but rather that they use mood as a resource pretty much like money. If you don’t have enough money, well, you’re going to do something to-

Tina Marshall: Go do something they don’t like doing.

Max Taquet: Exactly. But there are things that you need to do which will decrease your mood. You need to do all the chores at home, you need to do a lot of things that might not increase your mood in the short-term but are likely to contribute to a good and healthy mood in the long term. And it turns out that it looks like people tend to engage in those activities with longer payoffs at a time when their mood is rather good when they can spend it out. Now, the interesting thing is, that we count that principle mood hemostasis because it’s a bit like the homeostatic principle that we have in the body to regulate our temperature, for instance. But the interesting thing that we found was that people with depression tend to have lower mood homeostasis and that perhaps might explain, at least in some people, why they have depression in the first place, which is that mood regulation, or that mood regulatory principle, is absent or weaker in people with depression.

Tina Marshall: And that’s genetic. Is that genetic?

Max Taquet: It might be principles genetic. We do not have genetic data for those participants. We couldn’t test it, but it might be that this is in part genetically driven.

Tina Marshall: That’s very interesting.

So then let’s touch on, what data we have that can be absolutely helpful. So we were talking about the EHR data and having really deep longitudinal EHR data. What can you do with that?

Max Taquet: So what we can do with that is identify specific patterns or clusters of participants which might defer from others. And those clusters can be driven in part by genetic and biological information, so genes, the brain, but it might also be driven by psychological factors and social factors. So understanding the individual symptoms and their trajectory might be extremely informative. Understanding the social determinants, we were talking about homelessness earlier, which might be an important factor, as well as other social factors. And so doing those clustering might be very informative in making predictions about specific patients’ trajectories. So just to give you a specific example, there’s a thing that we call treatment-resistant depression, which is when you’ve tried to treat depression in somebody, and you’ve tried more than once with more than one kind of treatment, and it didn’t work. And so the patient after those attempts is still depressed. Now, it would be very good to be able to predict who is going to develop treatment-resistant depression, so that you don’t have to go through the hurdle and the pain of those patients-

Is this for any patient? So the treatment-resistant depression is this for patients… So the medication doesn’t work at all or the medication works and then six, nine months down the line, then it stops working? Or 18 months after they first started on a particular medication, it stopped working?

Max Taquet: Typically, it means it doesn’t work at all. So it means we’ve tried those treatments and they are still depressed without any sign of remission, so any sign that they’re getting better. And so rather than exposing those patients to those treatments which are not working for them, it would be good to predict that and be able to explore other avenues of treatments. And for research, that’s particularly key because that means that if we were able to recruit those participants even before they were exposed to specific treatments, we might be able to design and define and identify novel treatments that might work better specifically for them. Again, sort of touches upon the idea of precision medicine.

Tina Marshall: That sounds amazing.

As we’re talking about the patients, let me ask you, for the patients that you see and that you work with, how do they feel about clinical trials?

Max Taquet: Well, I think it varies. I think most patients are keen to find ways to get better and for the rest of the community of patients to get better. And so, often, they are keen to help develop the science. Obviously, there are some patients who prefer not to be exposed to things that they, not they’re not sure are going to work for them.

Tina Marshall: But everyone’s quite positive?

Max Taquet: But most people are actually quite positive. Yeah.

Tina Marshall: And then I have another question, actually, regarding the differential diagnosis that we were talking about, where clinicians are writing down their differential diagnosis before it gets to the confirmed diagnosis.

How often is that the same? How often is that right, that differential diagnosis? Is there any research or any stats that’s been done on that to say, actually, that, “We don’t want to diagnose them too soon, we’re still exploring? However, I know that probably 80% of my differential diagnosis will end up being the actual diagnosis.”

Max Taquet: Well, I think it’s fair to say that because the diagnostic categories that we use are so broad that, actually, the likelihood of being completely wrong is relatively small. It will obviously become more-

Tina Marshall: It’s almost a real shame actually, isn’t it, if they’re so broad?

Max Taquet: Well, exactly. You can’t really get it wrong. But there are some studies where they’ve asked two clinicians independently to diagnose the same patient. And the findings there were very interesting and perhaps surprising, which is that if the clinicians were seeing the same patient at the same time and maybe one of them was just observing, or sometimes it’s video recorded and they watched the video later, they tended to agree quite well in terms of the diagnosis. Whereas, if the same patient was seen by two clinicians independently, as in the two clinicians were doing their own and conducting their own interview of the patient, they tended to disagree a lot more.

Tina Marshall: So even if it was 1:00 in the morning, 1:00 in the afternoon, the same patient, they would still disagree?

Max Taquet: The diagnosis is not right.

Tina Marshall: That’s interesting.

Max Taquet: And that has a profound impact in terms of research, but also in terms of treatment for those patients.

Tina Marshall: Absolutely. And I suppose, because we’re dealing with the brain, is that type of scenario related particularly to mental health?

Max Taquet: Yes, a lot more. There are some instances in the rest of medicine where this is also the case, that there is some uncertainty about the actual diagnosis. But I think it’s fair to say that this is a problem that’s much more serious and prevalent in mental health.

Do health scores help with anything with things like that? I mean, as a clinician, if a patient has been handed over to you from another clinician, perhaps you’re reading through the notes, and you’re not entirely sure that you agree with their differential diagnosis. But then would any health school data help you have a better understanding of what’s going on with the patient or give you a clearer view?

Max Taquet: Well, I think they give a good snapshot and a quantitative one about the sort of symptoms that the patient is experiencing at that moment. Certainly, for research, they’re extremely important so that we have a sense of how severely affected the patient was at one point in time, and then how much better, if at all, they get later down the line. So in that sense, it’s very important. Obviously, for randomised clinical trials, that is absolutely key, and this is the gold standard to assess response to treatment in those instances. So they are very helpful. In terms of making a diagnosis, perhaps not that much.

Max Taquet: I think they’re sometimes used as a screening tool, mostly in large epidemiological data studies when they want to assess the prevalence of one disorder in a large population. I mean, it would be very difficult to send one clinician and have a one-and-a-half-hour interview with each of them to assess whether or not they have depression, for instance. And so a simple health score can be quite helpful, but not necessarily in terms of making the diagnosis, even though it might be helpful in getting a snapshot at the beginning.

Tina Marshall: So if we’re moving on to the life sciences industry and how the industry is reacting to mental health and dementia. So, typically, we know that some organisations were shy about conducting business in the UK because the NHS is quite difficult to deal with, or could be quite difficult to deal with, just because it could be fragmented. And, obviously, even though the NHS is one brand, it’s many different organisations. But what we’re finding now is that actually the focus on mental health and dementia is really rising, and these commercial organisations and our pharmaceutical companies, do want to invest in these areas.

Where would you say the biggest pain point is? Where, in your opinion, should the investment go? What is, I guess, I don’t know, the biggest therapy area that is called is in the biggest burden to the economy at the moment?

Max Taquet: Wow, that’s a broad question. I mean, mental health, we know, is having a huge impact and burden of health conditions, and it’s been shown time and time again in large populations, in large studies covering the entire planet, really. So we know that depression is one of the leading causes of disability worldwide. That’s been shown time and time again. So, clearly, mental health is a big area of focus. I think it’s great and very encouraging to see that companies and pharmaceutical companies are coming back to mental health, at least some of them, despite sort of de-investing sort of mental health altogether for quite some time. I think they realise that this is a huge burden. There’s a new understanding of neuroscience. Everything we’ve talked about today is obviously very important in understanding that complexity, and maybe subgroups of patients that might be treated with one new drug and other groups that are well treated with existing drugs. So all of those are very encouraging even though the slow move in the landscape of treatments.

Let’s talk about electronic health records and the data that are held within the EHR. What would you say are the biggest pain points that you would have as a researcher, and let’s say that life sciences would have in being able to use that data to facilitate drug development?

Max Taquet: Well, the biggest pain point is the fact that most of the electronic health data that we have for mental health is free text. The complexity of mental health means that clinicians record their information as letters and texts, not as simple scores as we were talking about earlier.

Tina Marshall: Would you want them to record it as simple scores?

Max Taquet: Well, it would be nice to have that on top of the free text. But we can’t do without the free text because we need the free text to capture that complexity, and then we need to have clever ways of interpreting that free text to extract the relevant information from that free text.

What type of information would be the most relevant? What type of information do you want to extract, and how do you see that happening?

Max Taquet: So I think, obviously, we want to know the diagnosis because it’s important. They have their treatment, et cetera. But also we want to understand their symptoms, their individual experiences, and the trajectory of those symptoms over time. Those are critical to understanding the individual experiences at a very deep level rather than the very shallow level of saying, “This is your diagnosis.” We want to know that perhaps this is a patient who was experiencing a lot of depressed mood and anhedonia, which is difficulty enjoying things in life, at the beginning of the illness. But, actually, now what we are really realising is that they’re experiencing a lot of what we call biological symptoms of depression, such as poor sleep, poor appetite, and low energy. And being able to capture those trajectories over time I think is critical to be able to design targeted treatments for those patients.

Tina Marshall: Brilliant. Thank you very much.

Max Taquet: Thank you very much, indeed.

Tina Marshall: I hope that you enjoyed listening today or watching this episode of Precision Neuroscience Reimagined. I’ve been very lucky to be joined by some very interesting guests, and we have more to come. So please do like, subscribe, and share this. Thank you.

Listen to the latest episode of Precision Neuroscience and more here: or you can find all of the episodes on YouTube:

Newsletter Updates

Enter your email address below and subscribe to our newsletter