From 3M Health Information Systems
Podcast Episode Transcript: Searching for the holy grail of mental health care
Gordon Moore: Hello. This is Gordon Moore, the host of the 3M “Inside Angle” podcast. Today, I am speaking today with Dr. Rebecca Resnik. She’s a licensed psychologist and founder of Rebecca Resnik and Associates, LLC, a private group psychology practice in Bethesda and Rockville, Maryland. Dr. Resnik trained in pediatric neuropsychology and psychology at the Mt. Washington Pediatric Hospital.
She conducts neuropsychological assessments, which is testing to diagnose developmental disorders, psychological disorders, and learning disabilities. She has a master’s degree in special education. And her interests include computation and linguistics and clinical psychological and the use of technology in diagnosis of psychiatric, psychological disorders. She’s the cofounder of the first computational linguistics and clinical psychology workshop held at the Association for Computational Linguistics annual conference in 2014.
And I was introduced to her by her husband, Philip Resnik, who you can hear on another podcast, and was fascinated by the intersection of her work in terms of understanding how we put the head back on the body and we think about what it means when they’re separate. So welcome, Dr. Resnik.
Dr. Rebecca Resnik: Good morning. It’s such a nice thing to be here.
Gordon: Thanks so much for agreeing to come on today. And I would like to hear more about the Cartesian dualism that has led to the separation of mental and physical healthcare.
Rebecca: Okay. So Cartesian dualism is, of course, from René Descartes, and it’s an idea that has persisted up through the centuries, even though at this point we know better that mental and physical health conditions are not separate things that respond to separate different types of tests and treatments. We know that these things exacerbate each other, go along with each other, in some ways can predict each other.
And Dr. Andrew Gerber, who was former director of the Austen Riggs Center, came up with a sort of cause, and he called it “duality’s end,” where he wanted to focus our attention as a field on this idea of really trying to eradicate Cartesian dualism in favor of care that’s incorporating both what we know about neuroscience, as well as what we know about the other systems of the brain and body.
This isn’t a new idea by any stretch of the imagination. There have been writings since the ’80s with physicians calling for changes in the fact that mental and physical healthcare are separate entities, separate systems, separate providers. In 1988, in General Hospital Psychiatry, for example, publish a position paper arguing that primary care physicians and nurse practitioners are the initial case finders of mental health problems.
And yet the studies, even back in the 80s, they knew that primary care providers were not necessarily the best people to recognize and correctly diagnose patients’ mental disorders, anymore than I, as a psychologist, am qualified to, say, determine if you have renal disorder or herpes or any other thing that’s outside of my scope of practice. So one of the things that’s always interested me professionally is—how do we break down silos across areas of specialization and foster collaboration, particularly with respect to transfer of research into clinical care and clinical practice?
Gordon: That’s a fascinating conundrum as I think about it. So there’s this generality of working with people but specialization to focus on a particular condition. And then we have practices that are separate and in different places, which means that we’ve just created the risk of poor communication, as well as the potential of problems with access to care as well. So I don’t know, necessarily, how to solve that. But it sounds like that’s kind of at the core of what you’re describing and thinking about?
Rebecca: This is a huge, complicated problem. But there’s sort of a bit of Jewish wisdom that says, “In the spirit of Tikkun Olam, we are—the problems of the world can’t be solved, but you’re not free to desist from trying.” And so I think that there are ways that we can make positive progress. But of course, we know that these are problems. Everybody understands that the status quo is simply not good enough.
And I could give you a million different examples but just to start with one that’s really interesting is that there was a study by Franklin, et al, in 2017. And one of the things that he documented in this meta-analysis of who can predict who will die by suicide, they found that there hasn’t been any improvement in 50 years. In 50 years of research. And we’re no better at figuring out who will die by suicide. As your listeners are probably familiar with, suicide rates are increasing. It’s now become one of the major causes of death in teenagers. And it’s, I believe, tenth, according to the CDC, in terms of things that adults are likely to die from.
For another example of how our system of dual care isn’t working for people, in 2017 the CDC said that over 70,000 people had died of opioid overdoses. And this is one of those classic issues, again, like suicide, that incorporates both physical healthcare providers—your physicians, your nurses, your PAs, as well as all of us in mental healthcare—psychiatrists, psychologists, counselors, social workers. And the fact is, is we are all dropping the ball in terms of really providing excellent care and using the research to improve patient outcomes.
Gordon: Why do you think that is?
Rebecca: Well, in a lot of cases, I think that, you know, we have issues around cost, certainly. And we are all familiar with the idea that costs are rising. We have to get more efficient. We want to have outcome-based care. And the idea is that, eventually, we will be rewarding providers for better patient outcomes. But we all know that actually making that happen in a way that works for your general provider has been very difficult.
Certainly, with respect to cost, I don’t think anybody would argue against the idea that our primary care physicians are quite maxed out in terms of their ability to provide care. I mean, these are the best and brightest, statistically speaking, in terms of being some of the best brains and spirits we have around anywhere. And these folks are burning out like crazy. They’ve got very high rates of suicide and substance abuse. These folks are pushed as hard as we can go.
And then you have mental healthcare professionals. Mental healthcare is often reimbursed by insurance companies through separate subsidiary companies, if you’re fortunate enough to have private care. So for example, if you go to your physician to have a suspicious mole looked at, for example, that may be covered under your healthcare provider, something like Blue Cross.
But then let’s say you want to come to me as a psychologist because you’re very worried that your anxiety is interfering with your ability to sleep. Well, that care will be probably covered by a subsidiary company, just like your psychotropic medication or your dentist, your prescription care. All of these things are covered by these subsidiary companies. And part of the problem is that the reimbursement rates—despite this movement towards parity of service, our reimbursement rates for mental healthcare providers are very low.
And that creates an impediment to collaboration because not only do you have the physicians, who are taxed beyond the point where they can even be happy in their jobs anymore, never mind provide the exemplary care we all hope to get when we or our loved one go in for physical or mental healthcare, then we have all of our mental healthcare providers, who are not in a position where they can easily become part of an HMO or a medical home because the reimbursement rates are so low.
And so oftentimes, that is a major impediment—the time and the fact that the time isn’t compensated well. So bringing everybody together as a team is something we all agree would be wonderful. But then, when you try to make it work, particularly if you are in private practice, the obstacles around cost and time just become insurmountable for a lot of folks.
Gordon: I suspect that the reimbursement structure and the amounts lead to variability in terms of access and where you can find mental health professionals.
Rebecca: It’s nothing short of tragic in this country. We have to sort of great swaths of territory in the middle of our country that have two different labels and one of which is the idea that there are mental health deserts. We’ve heard of food deserts. We’ve heard of literal deserts. And if you Google CDC data on what is a mental health desert and how many Americans live in places where there just are not enough mental healthcare providers to meet the needs of, statistically, what would this population need? Are their child psychiatrists? Are there people who specialize in substance abuse disorders? Are there people who can diagnose schizophrenia?
So because those folks don’t have access, they are getting—the vast majority of their contact, when they have a mental health problem, is through their GP, their primary care physician. Now, there’s another swath of territory in this country that represents the inequality, and that is the suicide belt. So if you Google this term, “suicide plus belt USA,” something like that, you will find that there, again, is a great swath kind of in the middle of the country where you have very high rates of people dying by suicide.
And again, suicide is a preventable way of dying. And you cannot overestimate the cost in terms of untreated mental health problems and suicide and opioid addiction. And what we don’t have are the people who are the actual specialists available. Now, even if those folks are available—like, I practice in Bethesda, Maryland, and Rockville, which are some of the thickest markets for mental healthcare ever. I mean, you throw a rock, you’re going to hit a psychologist in the head in this area.
The problem is that most of us in this area, we find that the cost of doing business and attracting good staff and providing quality care, the cost of doing business makes it very hard for us to take insurance. And that’s not a matter of being greedy. It’s a matter of—just that we don’t get paid that much to start with. So even in a very thick market where you have a lot of providers, many of the people in your market space are not able to access this type of care.
So we have a lot of very preventable problems that are exacerbated by the fact that people can’t get the level of care that they need. And primary care physicians are fantastic people. I adore my primary care. But the problem is, is that we can’t expect a primary care physician to always do more. That’s not realistic. We can’t expect a primary care physician to be an expert in everything.
When they have about 15 minutes to spend with a client—they’ve got to do the exam, they’ve got to talk to the person, and they’ve got to fill out the EMR documents while they’re seeing you. And that’s if they’re on time, which as you know— I mean, Gordon, you were a primary care physician. You probably had—what?—six clients an hour some days?
Gordon: Yeah. Yeah. Exactly.
Rebecca: We can’t expect these people to do things that take a lot of time, like trying to tease apart what a patient is saying to get to the fact that they’re actually presenting with depression and not just the thing that they came in for treatment for, which may be something like they’re having trouble sleeping or they’re having back pain. So what we desperately need is a way to supplement primary care so that we figure out who are the people who need mental healthcare and figure out ways that we can get them in front of a provider.
Gordon: Yeah. And that, to me, makes me think about this whole supply-demand conundrum which you pose, which I think is the lived reality of most people in the United States, where access to these services is constrained for financial reasons and then the actual presence or absence of individuals who can support that care. So now we have this issue. We know that mental health, substance abuse issues are huge confounding variables in terms of clinical outcomes and cost.
So we have a pressure to do better, a paucity of resources with which to do this. And then, even confounding that, I think, is the ability of people, like me, in primary care to be highly accurate in terms of knowing the diagnosis. If it’s a heart condition and I can know this by hooking a person up to a device or a monitor that can define that thing, it’s pretty clear. But with mental health and substance abuse, I don’t have a device that sorts the differential diagnosis and tells me what’s going on. I have words. And words can be used variably, and I think that may be part of the source of the problem. Is that your experience?
Rebecca: I agree completely, particularly with respect to making a mental healthcare diagnosis. One of the things I love to listen to is actually a psychiatry resident podcast. And part of the reason I like that is because when you listen to these folks, what they’re trying to do is hone in on the signal and the noise. And patients are very noisy. We’re thinking about information now. I mean, yes, they’re literally noisy sometimes, but we’re talking about being able to pick up on the signal and ignore the stuff that isn’t particularly relevant.
So if a patient comes in, again, they’re generally not coming in and saying, “Hey, doc, I’ve been feeling really depressed lately,” particularly a male client. They are much more likely to say something like—you know, that they’re tired. They’re angry. Things aren’t going well at work. They’re having back pain. A woman may come in, and statistically speaking, women are more likely to endorse actually feeling depressed and anxious.
But there are a lot of confounds. So let’s say that woman had a baby, say, six weeks ago. Well, this is prime time when we want to be thinking about—does this woman meet diagnostic criteria for a postpartum depression? What we know is that a lot of physicians, because they are so busy focusing on the signal that is cuing them in about is there a physical problem, that they may be missing the signal in the noise that tells them, “This could be a mental health problem.”
And again, we can’t expect our physicians to do this and have expertise in all areas. So your physician is listening to language. Okay. Let’s assume that you speak the same language I do and we come from the same cultural background. That’s a tremendous advantage. But as you know from primary care, many of your patients are going to come in and they’re going to have different ways of describing what’s happening to them if they came from another country, different culture, different part of the country.
There will certainly be gender issues in how people describe their symptoms and their experiences. So a lot of times that signal is something that we just plain old miss. Whereas, another practitioner who could be listening to that same thing, who is more sensitive to that type of signal, actually listening for things in their mind, like, “Ah, this person’s having sleep problems. Ah, this person’s having gastric distress. Ah, this person’s having difficulty breathing,” then that person is going to be a value-add in terms of diagnosing what’s actually going on.
We also, too, have many, many, many research studies that are based on this idea of a patient who has only one problem. This isn’t a new thing. When you do a study, you can’t have a patient who has, like, four or five different conditions. That’s too many confounds. And so you try to strip away those confounds and find your pure cases. So “This is a person with just anxiety.” Well, we know in the clinical world you don’t get that magical patient who only has just anxiety or just an ulcer or just back pain, just diabetes.
We always have other issues around, and it’s those other issues around the presenting problem that can be make or break in terms of—is this patient going to have a positive outcome or a negative outcome? And, of course, a negative outcome is going to mean more suffering and likely more expense on down the road. So when we’re thinking about patient language, not only do we have many barriers to detecting the signal in the noise, but the other piece of that is that our patients may not actually be willing to describe to us or able to describe to us what’s happening.
So let’s take for example a soldier comes into your office for treatment. That soldier is probably more likely than a member of the general population to under-report or even try to hide their mental health symptoms because the cost to that soldier may be that they would let down their comrades if they’re not allowed to continue with their mission. Or perhaps they won’t get promoted because they’re having symptoms of PTSD.
And so you do have many people, for example soldiers, people who, say, have postpartum depression, and they’re worried that if they appear to you as an unfit parent, you might take their baby away or not allow them to care for their own child. You have many people who are afraid that their physician or their psychiatrist or psychologist is going to hospitalize them, not in the general hospital but in terms of a psychiatric hospital. So you have your people who are deliberately going to try to hide their symptoms and for good reasons.
But then you have other people who actually aren’t capable of really telling you what’s going on. So for example, let’s say a schizophrenic comes into your office and they’re having a lot of negative symptoms. You may look at that person and think that they’re very phlegmatic, that they’re really pretty withdrawn, they don’t have a lot of energy. They can’t tell you what’s going on. Or perhaps a person with an autism spectrum disorder may not really be able to communicate to you about their anxiety and their depression because they don’t have the insight to think about it in a concrete way and explain it to you in a way that you’re going to pick up on that signal in the noise.
And so what we really need is some way to get around this dependence on patients coming in and saying those magic phrases that we’ve been trained to listen for or coming in and reporting their symptoms in a way that will grab their physician’s attention in that magic 15 minutes, and then we’ll know exactly what to do because diagnosis of mental health disorders is a very thorny, Gordian knot type of issue. And even psychiatrists who created the DSM fought over these things and get them wrong at too high a rate.
Gordon: Yeah. It reminds me. I remember reading a study by a psychiatric researcher using a structured interview that was videotaped and then forwarded to a psychiatrist for evaluation. And I suspect that what maybe you’re getting at is that having some kind of structured instrument to elicit certain features could lead a mental health professional to a very quick diagnosis that may be a little bit out of reach of a PCP. Is that where you’re going? Is that something that’s been developed and can be used in practice?
Rebecca: Oh, sure. Yeah. And we do use those. Now, for me, as a mental healthcare provider, my time is less expensive than yours. So the fact is, is having me spend an hour with a patient versus you spend an hour with a patient is pretty cost-effective. I’m just cheaper than you are because I didn’t go to medical school. And when someone comes to a mental healthcare provider, what mental healthcare providers—psychiatrists, psychologists—we are trained to go through the clinical interview process.
That’s a language elicitation task where we’re asking them to describe things like vegetative functions. Are you eating? Are you sleeping? Psychosocial stressors—did your husband just leave you? Did you just lose your job? How much are you drinking? Of course, we know they always lie about that. We all do. And you’re going through looking, again, for the signal in the noise because it’s a pattern recognition task.
Right now we are very dependent on descriptive instead of explanatory theory. And one of the great examples of our descriptive theory in action is the Diagnostic and Statistical Manual, 5th Edition, of the American Psychiatric Association, what we call the DSM. And the DSM has—for every agreed upon diagnosis, there is a list of symptoms that you’re looking for. And you, ideally, go through this symptom count where you say, “Okay. This person has three of these symptoms, one from this group, one from this group. Okay. This looks like obsessive-compulsive disorder. This looks like schizophrenia.”
What generally happens over time is that as people become more experts, they use heuristics less. And that reflects the work of Dr. Gerd Gigerenzer, out of the Max Planck Institute. When you get to be an expert, you’re able to pick up on things—that signal in the noise, that pattern recognition. You become a real expert at it. But here’s the problem. We know that collecting some kind of hard data from a person, some kind of quantitative measure, actually improves our diagnostic validity and improves our outcomes.
The American Academy of Peds, for example, recommends that physicians give people a rating scale because you run into problems like—a physician will say, “Is your child talking yet?” And the parent will say, “Oh, yes. He says long, long sentences.” Well, that may be a good sign that this child’s extremely bright. It might be a sign that the child has an autism spectrum disorder and is reciting passages from their favorite cartoon show. And what the physician is hearing is, ah, language is established, even precocious. Things are good.
But if you actually had that person fill out a rating scale, you would find that the symptoms of an autism spectrum disorder are actually present, and you missed that signal in the noise. What you also want to have people do a lot of times is fill out rating scales. Again, when you fill out a rating scale, it’s a factor analysis structure. And what you’re doing is comparing that person’s responses with the responses of, hopefully, thousands of other people who have some kind of demographic similarity. So for me, it would be 45-year-old, Caucasian women with advanced degrees.
And the nice thing about those things is that patients will often endorse symptoms, behaviors, and problems on rating scales that they don’t necessarily in your clinical interview. The problem with rating scales, though—again, you have the cultural, you have the linguistic, and you have the insight problems that can lead people to over- or under-report in their responses to rating scales. I give rating scales all the time, and one of the big problems is that the people who design the rating scales are expressing things in terms of their culture and their understanding.
And what my patients often do is say, “What do they mean by that question? What do they mean by am I sad most of the time? I get sad every now and then, but does that mean—what does it mean I’m sad most of the time?” It can be very difficult for them to understand what the question is actually going for. And so combining both the natural language and the data together is sort of one of those things where you get this nice synergy.
And when things line up around a particular pattern, support a particular hypothesis, you’ve got convergent validity, and that’s what you’re after. The Diagnostic and Statistical Manual was very publicly rejected by Dr. Tom Insel, who was head of the NIMH back in 2013. I sort of joke about this as the breakup letter. It was a pretty bold and very public rejection. But Dr. Insel had a point.
The DSM, these diagnostic categories that are supposed to be used by all mental healthcare providers, they didn’t have sufficient reliability. So if a patient came to see me, I might diagnose that patient with an anxiety disorder. That person might go to a colleague of mine and be diagnosed with a dysthymic disorder. That person might go to another colleague and be diagnosed with a depressive disorder, might go to another colleague and be diagnosed with a substance abuse disorder.
And this isn’t good enough. So while we’re stuck in this sort of descriptive way of understanding mental health disorders, what we’re finding that we really need is something better. So one of the things that’s got a lot of people—such as me; my husband, Dr. Philip Resnik, who’s a professor of computer science and linguistics at the University of Maryland; and other researchers in the area—we’re very excited about the potential to help inform mental health diagnosis, particularly with an eye towards—let’s find that signal in the noise.
Let’s find these people who are suffering and who need mental healthcare or psychoeducation or substance abuse treatment. Let’s find these people because we’re not finding them now. We’re not meeting their needs. And, of course, we know that status quo is not good enough. It’s leading to a whole bunch of bad outcomes that are actually preventable. There is also a new field of computational psychiatry. I wanted to mention that too.
Dr. Alan Anticevic up at Yale has published a book introducing the field of computational psychiatry. It was such a pleasure to meet him. There are many people who are coming to this conclusion that we can bring algorithms to bear on this task of finding signal in the noise of patient data, their behavior, their language, their symptom presentation. And so what I often talk about is that we are searching for a holy grail. And what we really need is an intervention that is cheap, fast, and valid, that we can have primary care practices using to screen for mental health problems.
The important thing is that if someone goes to a primary care physician, we need a tool that they can use that can be administered by somebody, say, on the level of a nurse or maybe even a tech, because we know the physician has no extra time. So having had two children—I am a G2P2, to use your parlance—I have given many, many urine samples in my time. All of the women out their listening who have had children will recognize that this is part of women’s healthcare.
And it made me think about how much information, how much signal physicians get out of something that really is not that intrusive. You can find out all sorts of things about a person from their urine sample. Are they doing drugs? Are they pregnant? Do they have diabetes? And what we need in mental healthcare and in general practice is we need the equivalent of that. The human brain is much more complicated than a lot of our systems, and so we don’t understand it as well.
And the way that we can get information about how our brain is functioning— right now, our best bet is language because we can’t put everyone under an MRI. We can’t do a PET scan on everyone. And even if we could, we can’t learn things like—questions we need answered, like is this person at risk of dying by suicide? I can’t put you under a scan and find that out about you. So what can we do?
We can get patient language, and we can have really smart, wonderful technology types of people—like my husband, for example, who I think is wonderful—the type of people who specialize in teaching computers to analyze language, to find that signal the human beings could be interested in. So imagine you go to your primary care. And for me—probably not for you, Gordon. You probably don’t have to leave urine samples.
But imagine that as you go and they say, “Okay. Please leave your urine sample,” or “Please sit in the chair for the phlebotomist. We’re going to do a blood draw,” imagine that you also had to leave a language sample. So let’s say you had a structured prompt, something very open-ended, like, “Describe the biggest stressor in your life right now,” or “Describe the biggest problem in your life right now,” or “Describe what the past month of your life has been like.”
That’s a language elicitation task, just like a urine sample. We’re going to get information about what your brain is doing. This product of your brain is language. And let’s say we can record that, have some kind of software that goes through and looks for things like key words, the amount of language that you produced, what kinds of adjectives you have, how many pronouns you used, and come out with a printout, just like your urinalysis pages or your pages where they have all the lists of, like, what your cholesterol is.
Well, let’s say you had that for language. And you could have the nurse or one of the other support practitioners bring that document to the physician. So before the physician even goes in the room, they have a document that says something like, “This person is at very high risk for suicide,” or “This person is at very high risk for substance abuse.” “This person is at very high risk for schizophrenia, depression.”
Let’s say we could only catch, like, 10 percent of extra people. When we’ve got about 26 percent of adults, at any given time, who are meeting diagnostic criteria for some kind of mental health problem, even if we could only catch an extra 10 percent of people, that would be a huge increase in the amount of understanding and improved diagnosis and even, hopefully, the amount of suffering that we could reduce with people.
Then we would have a better idea of who needs something like psychoeducation, who needs substance abuse treatment, who might need therapy. Unfortunately, in this country, with the master-builder model of the primary care provider—and I’m taking that idea from Dr. Atul Gawande’s wonderful book The Checklist Manifesto. He argued that the primary care physician has come up through our culture as kind of this master builder who has to be an expert on everything.
He’s arguing that, again, we can’t expect that to be, and we don’t want to try that because we know the outcomes are not good. So let’s say we help that primary care physician to detect some signal that they might’ve missed and get people to the types of providers who could actually really help things. Like, let’s say you need some stress reduction psychoeducation so that you can manage your symptoms of Crohn disease better.
Let’s say that you need somebody to work with you around medication adherence. Let’s say you need somebody to help you stop beating the people in your family. There’s many, many avenues where that physician can serve as gatekeeper but pointing people in the right direction, towards the people who do have that kind of expertise.
Gordon: You really sum this up nicely in a sense. You pose this problem of supply and demand and inter-rater reliability issues and language. And you’re positing sort of a computational approach to solve that. Does that exist yet, or is this just an idea?
Rebecca: It’s not just an idea. There was the work of Pennebaker, I believe who is in Texas. And I actually got to meet him, which was really cool. What he did was he took writing samples, language samples from college students. And those samples that were taken through the years have been put together into a corpora called LEWC. And this is one of the things that language researchers, like my husband, like the other members of our team in the computational linguistics and clinical psychology community—not me. I’m not the tech person. But I love working with people who understand tech because they’re very exciting, cool, creative people.
Those folks have been using this LEWC database for ages. And part of the problem is that, you know, they’re using little prompts from college students. And some people, like Dr. Glen Coppersmith and Dr. Tony Wood at Qntfy, are mining social media data. They’re using Twitter data. What we really need is to have more of this data available so that the computer systems can train on it because computer systems need a lot of data.
And what we don’t have in all of these medical records all over the country, in file cabinets and at Iron Mountain—we have piles and piles and piles of paper. There’s so much that could be learned. And yet researchers have a very hard time working with medical records to mine data. It’s a nasty, thorny problem. Never mind the fact that every EMR has all of the data in a completely different form.
But the good news is that when you get a whole bunch of data in front of these folks, they get excited and they rip through it. And they can find all sorts of really excellent stuff. Like, for example, there’s been work out of Qntfy—my friend. Dr. Glen Coppersmith, he has been able to get preliminary results showing that their algorithm can use social media data and predict who will die by suicide at five times the rate that a clinician can do it.
That is really exciting, especially when you consider that an algorithm and software are so much cheaper than an actual clinician. And if we can find something that’s cheaper and more accurate and could save lives, that’s really exciting.
Gordon: Wow. That is. That is getting right at the crux of the supply-demand and speeding things up. Well, Dr. Resnik, I want to thank you for your time today.
Rebecca: It was so nice talking to you.
Gordon: Yeah. This was a terrify discussion. And so I think it’ll be very useful to our folks who are listening and thinking about how to apply this in policy and in working on clinical care.
Rebecca: I hope so. I’ve had the chance to listen to your podcast, and I’ve really been excited by the ideas and the thoughtfulness of your guests. It’s a wonderful podcast.
Gordon: Thanks so much.