Skip to main content

Transforming the understanding
and treatment of mental illnesses.

Celebrating 75 Years! Learn More >>

 Archived Content

The National Institute of Mental Health archives materials that are over 4 years old and no longer being updated. The content on this page is provided for historical reference purposes only and may not reflect current knowledge or information.

Social Disconnection and Late-Life Suicide: Mechanisms, Treatment Targets, and Interventions - Day One, Part Two


Marty: There's Yeates. I'm glad that Yeates is here. I tend to summarize things in three words and Yeates can explain them really lovely. I think I'll say that the bottom line or the quick summary of what was going on this morning. I think it was wonderful to hear so much different evidence showing how well older adults do as they age, in terms of their mental wellbeing in general.

Yet at the same time all the different types of pressures, being the physical challenges, the social challenges, the structural challenges. They get put on people with age which make it harder or makes greater demands on their emotional resources to be able to stay connected or stay mentally well. Or physical challenges that they have to find ways to compensate and those things are the things that increase their risk for suicide. Yeates, why don't you put all that into great and beautiful words?

Yeates: All right. I think you did that. Just to track it through very briefly and I'll try to keep this short. My own take on the continuities that follow through Juliana and Laura and Robert and Amy's talk, also George's comments as you said, that it is clear there's just a lot of evidence that social connectedness has a powerful impact on health in a variety of ways. I thought it was fascinating and critically important to understand that that's probably multi-determined, that it's played out in biological fashion through neural networks, and fundamental biological processes bound through inflammatory processes and exotoxin and even the gut. All of that's involved as well as psychological and social factors being from not just through the biological, the bio-psycho-socio models certainly is very culpable and it's at work here.

That pauses challenges not the least of which is as Julianne noted, the vulnerability that we have to look at risk rather than protective processes, a conflation of socio-connectedness in all its complexity with other concepts and our still limited ability to I think reliably and validly measure some of these constructs and then to carry that across different studies, which is going to be necessary I would argue for the progress of the science in this area.

Both Laura and Amy showed, I think, beautifully about how developed mental processes aging, and development and life course perspective brings theory to help us to sort through some of these aspects of the relationships here, and those lend themselves to testing. As you're saying aging is okay. In fact, it's really pretty good, but it's critically valuable and important to understand that it's only good if the person has these other assets available to them.

Which gets us back to this notion of interactions and moderating effects. In my understanding of the science of this stuff over time, has always been that when we can look at interactions, we're only getting a very powerful window potentially into mechanism, by looking at moderators, teasing those things a part. But that that poses significant challenges.

Again, if you can measure things properly and extend that standardization of terminology and measurement across different studies, then there can be real missed opportunities there. Robert's work is really helpful showing that by examining relationship between suicide and social connectedness in the Black population, for example, really is necessary in pointing us to an appreciation of the importance of socio-cultural context.

Another one of these layers that we need to peel away and understand that the genesis of the problems that we're trying to sort through here, such as for example in the African American carry the salients of negative comments over positive, I think it's fascinating. The solutions, for example, in church or faith communities are if not culture-specific, then they are at least dependent on cultural context and factors.

Then Amy back to the notion of theory and life goals development and testable models and aiming to where George ended the conversation too about the notion again that these processes or social connectedness and their impact on health, are likely functioning with both direct and indirect effects. All of that I think is a really strong foundation for thinking about mechanisms and the challenges we face in their study. What we're about with this workshop is examining that process of scientific development from an experimental therapeutics paradigm, if you will, from looking at mechanisms to identifying targets, and then applying those targets to the development of effective preventive interventions, so that brings us to targets.

Marty: I'll put one more comment in there, I think one of our first slides was to me really very compelling, which is that we're not all in the same boat. We may be facing the same riling ocean, but we're on different types of boats. We're also coming at this from different training backgrounds of theoretical backgrounds, and I think PHQ started by giving us the language, which I think is important, and that's equally true with measurement. Jovier reminded us that the Phoenix has at least some kind of measures that we can use to help integrate our different ways of thinking in our science, but there's a lot of words out there, a lot of constructs out there that have yet to be having good measures or even definitions for, and that's another place in it as we do this work is we can identify places where we need to find a better way to communicate them. There you go.

As Yeates was just saying, it's time for us to move away to the next part of the therapy model of experimental model to targets, and to do that, we have five speakers this afternoon. The first section is going to be predictors of suicide ideation, neurobiology, big data, and potential roles for social disconnection, and this is going to be Maria Cuando, Yu Li, an Amy Byers. I'll let Maria. Are you here? If you could start by putting up your slides, and everybody's introducing yourself very briefly. Is she here?

Maria: Yes, I'm here. Thank you very much. I really appreciate the opportunity to talk with all of you today, and I am aware that we're a little bit early, but if you'd like me to start, I'm happy to do it.

Marty: Yes, please do.

Maria: Okay, terrific. Well, first of all, I think that one of the things that was really great about having the opportunity to take a look at this literature is that it allowed me to learn a lot, and that's always great fun. What I found was that while I started out not really having a clear picture of what I would be talking about today, what I found, as I reviewed the literature is that in fact, there's a tremendous amount of convergent cross-species neurobiology that is quite similar between the impact of social isolation, and suicide.

So, in the next 15 minutes, I'm going to try to show you some of the evidence for that, and of course, the data is not perfect, there are lots of holes in it, but I think they're-- it's a very interesting and stimulating way of thinking about it, and this type of information may have been presented, this was from an article that was looking at the impact of loneliness on cardiovascular health.

As you can see, in the United States, by the time people are about 80, close to 20% of them report feeling lonely, so that's a very impressive proportion of individuals. Interestingly, what we see is that the loneliness itself is adaptive, because certainly, from an evolutionary perspective, connection with others ensures safety and survival, and if you're isolated, then you need to self-protect, and of course, those types of needs trigger things like implicit hypervigilance for social threat, anxiety, and hostility.

It creates this negative feedback, if you will, of perceived social isolation, hypervigilance for social stress, attentional confirmatory memory biases, preferential attention to negative feedback from the environment, which elicits behavioral confirmation processes, and then more negative displays and social interactions and effect, which of course, have an impact on the social environment, potentially increasing the isolation of the individual.

I'm showing you this slide, which is from a project that I've been working on with Dr. Barbara Stanley for a few years, in which we developed two different subtypes of suicidal behavior with distinct biosignatures, and I'm using this if you will, as a roadmap for us to think about what types of things have been implicated in suicidal behavior that have and suicidal ideation that have clear biological underpinnings.

What you can see here is that we described two types, one of the impulsive suicide attempts, and another one of planned more lethal suicide attempts, and then the impulsive subtypes, you have HPA reactivity, so cortisol production, increased cortisol production in the face of challenge, reactive aggression, a very important parameter, and poor emotional, cognitive control. With this constellation, a life event is likely to precipitate very large increases in suicidal ideation, which can result in impulsive suicide acts.

In contrast, we described another more sustained suicidal ideation that occurs in individuals who actually have very little aggression and good cognitive control, but have a low serotonergic tone, and in the context of depressive episodes, develop this sustained suicidal ideation, and then the precipitant for the suicidal ideation is often more difficult to identify, and these individuals tend to make more planned lethal suicide attempts.

Keeping in mind those parameters that are relevant to suicidal behavior and looking at these two different, these are very busy slides, but it's just to remind me to tell you, this is from the cardiovascular literature, biological impact of the-- of social isolation, and this is from suicide literature, biological underpinnings of suicide, and this is in suicide death. What you can see very quickly, first of all, that the behavioral health changes that are observed even in cardiovascular studies are linked to what we see as risk factors for suicidal behavior, things like smoking, things like sleep disruption, but also the biological systems are very similar.

Here you have ACTH in the adrenal gland, and epinephrine, and cortisol, which of course is very critical, depicted here, in suicidal behavior, you also have the impact of cytokines and pro-inflammatory genes that you heard about this morning, and here again, in the suicide, neurobiology, cytokines have an important role. Similarly, you have the sympathetic nervous system with its activation of inflammation and also with its activation of noradrenergic processes, again, also very important in the neurobiology of suicide.

The overlap was really pretty impressive in my mind, and this lovely review details across the lifespan, some of the health-relevant consequences of loneliness by age group. You can see that in the group that's 50 or older, you see depressive symptoms, depression, pain, clearly related to suicide risk, poor sleep quality, again, related to suicide risk, and from a neurobiological perspective, the increase in cortisol response to awakening as well as pro-inflammatory processes.

One of the things that I didn't expect to see was evidence from the animal literature about the impact of social isolation on monoamines, and this is a really interesting study that was conducted in mice that were isolated immediately after being weaned. What they observed was twofold. One was that the mice that were isolated were more aggressive in a resident intruder paradigm, and two, that they also had very broad downregulation of transcription of serotonin receptors across seven different receptors.

You may recall that in that slide I showed you about the model, the serotonin 1A receptor was very prominent in the more sustained suicidal ideation with more planned suicide attempts. This was really interesting, and I'll just note that many of the animal studies use isolation immediately after weaning or in the case of fruit flies, immediately after hatching. It's not exactly the same thing, of course, as what we're talking about, which is isolation or social disconnection during later life, but I think it's informative, nonetheless.

In this other study conducted in hamsters, male hamsters, these hamsters were isolated. The timing of the isolation is not noted, but from my reading of the literature, and I'm only presenting two of the many, many studies that are out there showing this type of effect, most of them use post-weaning isolation, but what they found is that the vasodepressor 1A receptor in the dorsal raphe nucleus, was lower in isolated male hamsters, and you recall that

the dorsal raphe nucleus is essentially the home of the vast majority of serotonin cell bodies, so that it's from the dorsal raphe nucleus that the serotonin system emanates if you will, two other structures in the brain including the prefrontal cortex. There is a closer interaction between the vasopressin 1A receptor and serotonergic neuron activity. Activating this receptor increases neuronal activity, serotonin neuronal activity. Interestingly, of course, if you have a reduction of serotonin activity -this is been shown both in humans and in animals- then, that decrease, in serotonin release, is often associated with increase aggression.

Again, going back to this idea that the relationship between aggression and serotonin is importantly related to suicide risk and clearly, delineated here in this model of social isolation. The other thing that was of interest is that there was a clear increase in serotonin 1A receptors in the bed nucleus of the stria terminalis, which is a very important interface between the affected forebrain, including amygdalin, ventral hippocampus, and medial prefrontal cortex. All structures are very critical for mood and other types of regulation. This bed nucleus of the stria terminalis is an interface between that affected forebrain, the hypovolemic, and brainstem areas that are so critical to neuroendocrine, autonomic, and behavioral responses to threats.

Social isolation, as I showed in one those initial slides, also has an important impact on the HPA axis. These are human studies done in-- some of them in older individuals. This is from a large cohort in the UK, more than 2,000 individuals aged 60 to 64. They looked at the relationship between three different cortex or measures, and different indicators if you will, of social isolation. What they found is that, for example, those individuals who are newly living alone tended to have greater nighttime cortisol, and greater diurnal slope in the secretion of cortisol.

You can see here also that people who are single, never married had higher cortisol at night, and that those who have been widowed less than three years had rather higher cortisol at night and also, higher diurnal slopes.

You can see that this impact on all these different measures is manifest in individuals who may be vulnerable to been more isolated by virtue of their marital status or their living arrangements. This is another interesting study also in healthy individuals. This is over 200 healthy middle-aged men and women where they measured social isolation of the close person's questionnaire and found that it was related to three different things. Importantly, in terms of the hypothalamic-pituitary-adrenal axis which, of course, is the driver of cortisol secretion, that those who reported high social isolation, really, had a much greater impact secretion of cortisol in the context of a psychosocial stress that was administered in the lab.

You can see very easily that the area under the curve, the cortisol, in this group, is much greater. They also had other signs of greater distress, physiological signs of distress, such as greater increase in systolic blood pressure that failed to go down as quickly as it could be. In the men, also, a cholesterol response to the stressor. Interestingly, this study was geared more towards looking at cardiovascular consequences of social isolation but, of course, clearly quite relevant to the biological factors involved in suicide risk. Social isolation also had an impact on hippocampal function and instruction. These are from animal studies.

Interestingly, in adult female [unintelligible 02:34:34] is a type of South American rodent. What they found was that after six months of social isolation, the rodents showed normal auditory-cued fear memory. For example, a tone paired with the foot shock, they were able to remember that quite well. But didn't do as well is the foot shock was been paired with something like the color of the cage where the paradigm been implemented. Of course, this contextual theory memory is hippocampal-dependent. You really need hippocampal input to be able to register that type of memory. Interestingly, the social isolation reduced hippocampal CA1 subfield volume without affecting the CA3 sub-region or the total hippocampus volume.

As you know, major depression CA1 volume has been implicated. It's a predictor of illness duration. There is also some interesting study in the ongoing studies looking at this in suicidal individuals as well. This is of interest, this study, which is in mice, because the long-term potentiation is a reflection of synaptic plasticity. As we've been discussing, social isolation reduces learning and memory. One that they found in this study that was very interesting is that this long-term potentiation was reversible if you gave a glucocorticoid antagonist to the individual returning the long-term potentiation to normal and therefore, again, increasing synaptic plasticity.

When the social isolation itself increased the corticosterone levels and impaired the long-term potentiation in hippocampus. The same effects could be induced with chronic corticosterone. You can induce it with external corticosterone and rescue it with an antagonist. This was completely parallel with what was observed secondary to social isolation. I will finish with just a couple of comments about the impact of social isolation on the neurobiology of a few clinical features that are observed in suicidal behavior. Here, again, we are talking about aggression. This is in a rodent study in which they, again, in the post-weaning phase isolated these rodents. What they observed is that they had persistent fears.

For example, if the rodent had been isolated, they would learn the tone paired fear. The fear will remain for a longer period of time than in non-isolated rodents. These are learned fear reactions. Then, in the innate fear reaction, they portrayed these shadows that will mimic the flight of a predator swooping down on the rodent. In this case, also, they found that the fear persisted in the isolated rodents for much longer than in the ones that were socially weaned. These isolated rodents also manifest increase aggression. I told you about this from another experiment where the resident intruder paradigm demonstrated significantly greater aggression in these isolated rodents.

Interestingly, the mechanism was determined to be related to the type two neuro [unintelligible 02:38:17] peptide system which is very critical to acute fear memory consolidation. If they infuse an antagonist, NK3 antagonist which shuts down the system, they can reverse the social isolation-induced aggression and fear. I thought that was a very elegant way of demonstrating this association. Finally, I think that there's been, in the last five to seven years, a lot of attention paid to the impact of sleep on suicide risk. Social isolation also has a tremendous impact on sleep, resulting in- I haven't shown you all of these studies.

For example, in fruit flies, it's been demonstrated quite clearly that social isolation disrupts sleep, not only in terms of decreasing the total amount of sleep which you can see here. These are also in fruit flies. But also, other experiments have demonstrated that the fruit flies also have very fragmented sleep. It's not efficient sleep. This acute sleep loss that's induced by social isolation starts a process in the brain in general, that's related to the unfolding protein response. One way that I think about the unfolding protein response, this is the system that's responsible for if you will, cleaning up synapses, cleaning up extra mixed folded proteins that haven't been folded at all.

There are different ways of detecting that this process is going on, including with these endoplasmic reticular chaperones. What they found in these experiments where the socially isolated flies had greater levels of the BiP chaperone, but interestingly they could manipulate them pharmacologically. For example, if they took a socially isolated fly and applied caffeine, they would see further increases in the BiP levels. In contrast, if they gave Zolpidem, which as you know is a sleep-inducing drug, they would see a reversal and a lowering of the levels of BiP. When they genetically manipulated the flies to have increased sleep or decreased sleep, they also saw the same changes in BiP levels.

One of the interesting things about this model is that they were really thinking about, is it that people who are socially isolated have sleep impairments because they have decreased synaptic connections and therefore, they don't need to do pruning and cleaning up? As I was mentioning, so they don't need sleep. Or is it that this is really a cellular stress model where the social isolation reduces the sleep, and that reduced sleep impacts the unfolding protein response which suppresses a global protein translation, and limits the synaptic proteins? Critically, if the UPR process goes on for too long, it results in cell death. It's a really critical pathway for maintaining healthy neuronal functioning.

With that, I'm just going to show you the slide again. I hope that I've been able to illustrate very briefly the close correlation between the physiological impact of social isolation on multiple systems in the cerebral cortex and other subcortical regions, and those that are implied also, or implicated also in suicide biology. With that, I will stop, and thank you for your interest.

Marty: Thank you, Dr. Oquendo. This was great. You did us all a favor by covering such a large amount of complex information. It's very helpful. Thank you. We're next going to turn to Dr. Lee. Yu Lee are you there? Can you put up your slides?

Yeates: Do we have Yu here? I'm not sure we do. If not, we could come back to him if Amy would be willing to step in.

Amy: Sure. Sure

Marty: Amy always steps in.

Amy: I’m happy to.


Yeates: Thank you.

Marty: Amy, why don't you go now, introduce yourself, and then we'll come back to Dr. Lee?

Amy: I'll go ahead and share my slides.

Amy: Thank you, this is an honor to speak today. I'm going to be talking about late-life suicide, veterans, and large-scale data. Where does social disconnection fit in? This has been an incredible challenge for me to think about this, which is wonderful. I think it's so important. My name is Amy Byers. I'm a professor of psychiatry and behavioral sciences. I have a joint appointment with the Department of Medicine and Division of Geriatrics at the University of California, San Francisco.

I'm also a Research Health Science Specialist with the San Francisco VA Healthcare System. I will say really quick that I just want to acknowledge my funders, NIMH and VA. I also need to acknowledge that the opinions, findings, conclusions, or recommendations that are expressed here today by me and with my coauthors are not necessarily reflected views of the US government, or the US Department of Veterans Affairs.

First off -I'm just setting a stage here- I'm going to talk to you about the current state of my research with respect to potential risk factors for late-life suicide among older veterans, but I use big data. I use VA national data. I'm going to start with setting the stage for you that older veterans, and I'm talking sometimes older veterans, I'm talking 50 and older or 65 and older, but 50 and older is over 70% of the veteran population. Think about that. By the year 2025, veterans 65 and older will be approximately 50% of the veteran population. Then, when we're talking death by suicide, approximately 70% of veteran suicides are in older veterans.

Then, I will discuss, where does social disconnection potentially fit in here, and go through what I think is the potential role of social disconnection. I'd love input and your help in thinking this through. This slide, it's a set up to show you the distribution of the veteran population. I don't think folks are always aware of what the veteran population looks like. There's four main quadrants here. The leftmost upper quadrant is the veteran population, all veterans. I created this slide with my research associate. The X-axis is age in years going from about 20 up to 80 plus. The Y-axis is the proportion of the veteran population.

Then, the different colors of the bars, are progression with times. The darkest is actually year 2010 then, the middle color purple is actually year 2015, and the lavender color is projected for 2025. I just want to give you an idea. The veteran population is an older population. Now, look to the right, that graph there is veteran suicide rates per 100,000 by age group. On the X-axis, this is broken down to four age groups, there's 18 to 34, 35 to 54, 55 to 74, and 75 and older. When you look at this and take into account the Y-axis rate per 100,000, it looks like those 18 to 34 have the highest rate, which they do. They have the highest rate. This was in 2017, which is our most recent data from the suicide data report that the VA gives out, that produces every year.

Now, look at the bottom, left hand, lower quadrant. That's veteran suicide count by age group. The X-axis is age like before, age groups, and then it's number of suicides. Where the largest number of suicide counts remained to be highest in those 55 years and older. I just want to stay here, and we really think it's important to think about the dynamics of how these populations change. We should think about our denominators, and with time veterans are aging into the over 55 group. I do think in part the rates, we may think there's a decline or there is a decline. You see a decline as so data shows. It's very well that the counts are staying the same, or actually they're increasing.

I also want to show you here briefly the VA national data, the data linkages. The VHA, the Veterans Health Administration is the largest integrated healthcare system in the United States. It actually makes nice data to study because of that aspect. There's a richness of data. There's inpatient-outpatient data, there's pharmacy data. We can link data to Medicare data inpatient, outpatient pharmacy data, the national suicide data for the VA is, I think, like none other. For the most part, in terms of electronic health record data. There's the SPAN database, which is the Suicide Prevention Applications Network. It's a suicide attempt data, mostly nonfatal, actually has the date of the attempt. We actually have an event date.

Then, there's the national Mortality Data Repository which is the MDR data. That provides cost specific mortality data. It's data that comes from the CDC, causes a big national death index, and we can get suicide data information for that. For this slide, I'm honing in on potential risk factors, or what I like to call indicators, markers, or risks for suicide among older veterans. The VA is very concerned about suicide prevention. It's a humongous priority. The majority of veterans are screened on depression, post-traumatic stress disorder, suicidal ideation, dramatic brain injury, and alcohol use. But I really think when we're talking late-life, the reality is that late-life suicide is, at least in my work I do, I really think it's so much more complicated.

There's a complexity of factors, especially at the intersection of physical and mental health. In these last four bullets here, I'm going to touch on some of the work that I've done, and then talk about, at the end, where I think social disconnection might come in. First off, I've done some work with my colleague Dr. Ruth Morin on comorbidity profiles, doing some latent class analyses, and taking in too a long litany of medical and psychiatric diagnoses based on ICD codes, and seeing how they fall together, and those who attempt suicide, older veterans who were last seen in primary care, because as we know, most older adults, including older veterans use primary care. In fact, it's been documented that up to 70% of older adults had primary care visit within one month of an attempt.

I also want to talk about some profound life transitions. I think these are important to study and to continue to study when it comes to risk of suicide. With the work I'm doing, that's funded by NIMH, and with my colleague, Dr. Lisa Barry from the University of Connecticut, we're looking at a recent release of older adults going from prison to community, and risks associated. There's also vaunting diagnoses that happen in late life, like a dementia diagnosis. I'm going to talk to you about some medication use, give you some ideas, and taking into account type and potentially polypharmacy. This slide here sets the stage for- this is a paper that was published in JAGS with my colleague, Dr. Ruth Morin, and coauthors.

It shows you the distribution of comorbidity type, or you can think of it as classes or clusters based on latent class analysis. This is in 2,131 veterans, 65 years and older, who attempted suicide between 2012 and 2014, and had a last visit of primary care. There are five main clusters here or classes here. Let me orient, teach you. The X-axis is the litany of medical and psych disorders. We're starting with a chunk of cardiovascular then, there's TBI, and then you'll see chronic pain, sleep disorders, depression, and then from depression on, you'll see anxiety disorders, and then even substance use disorders and schizophrenia noted there. Then, the Y-axis is the probability of having a diagnosis.

Let me just tell you briefly about the five profiles that we found based on analysis. There's minimal comorbidity, which is the green line above. That's about 23% of the sample. There's chronic pain, osteoarthritis, the yellow line there, which is about 30% of the population, then there's the purple line, which is depression chronic pain, which is about 23% of the sample, and then there's depression medical comorbidity, about 17%. That's the red line. We got the blue line, it's high comorbidity, and it's really high in terms of percentages, probability of psych disorders there. That's about 7%. This slide is looking at the same sample of individuals but breaking it down by lethality means by cluster.

The top row shows you the total sample, and then we have the five clusters. I just want to touch on a few things here, mainly focused on the highlighted parts. In terms of prior ideation, at least that was documented in the medical record, you're seeing about 68% for the high comorbidity group, and it's really small for the minimal comorbidity and chronic pain osteoarthritis. Again, this is the intersection of clinical care. You can interpret that how you interpret it. In terms of death by suicide or fatal attempts, 61% of the total sample died by suicide. Their attempt was fatal. You'll see that it was 86% in the chronic pain osteoarthritis group, and it was 73% in the minimal comorbidity group. It went down to 10% in the high comorbidity group, which is interesting.

Firearms were by far the highest in terms of method for death by suicide, with 83% of the total sample. It went down to 47% for the high comorbidity group, and then overdose was the highest for the high comorbidity group of 20%. I do want to note down here that 20% of all fatal attempts occurred within seven days of primary care visit, and 42% of all nonfatal attempts occurred within seven days of primary care visit. Dr. Morin and I said, "Okay, let's see what happens if we look at last visit with mental health." This is what this slide shows you. The left is pretty much of what I've already talked to you about from the previous slides.

Then now, let's think about last visit mental health. Among suicide attempters, that was their last visit, 65 and older. Similar criteria, 2012, 2014.

You're seeing that with the comorbidity profiles, depression stands out in every single one of them, including the minimal comorbidity, the chronic pain osteoarthritis, where in the last visit primary care over 50% of those individuals did not have a depression diagnosis or minimal. You'll notice that 61% in the last visit with primary care had a fatal attempt, where it was 23% in last visit mental health services. 73% fatal attempts in the minimal comorbidity group for last primary care versus 32% in last visit with mental health. Firearms were used in 83% of fatal attempts in last visit primary care, and it was 65% in last visit mental health.

I do want to touch on briefly about the profound life transitions and how important I do think transitions are in late life. We see here that this is some work done with Dr. Lisa Berry, and it's showing cumulative incidence of suicide attempt. The red line is reentry and showing. Ultimately what we found is that there was a 3.5-fold increase to risk of a suicide attempt for those who reentered in late life versus those who were never incarcerated. Both of these were published in the American Journal of Psychiatry. You're seeing in this graph, these higher accounts of suicidal behavior clustered into subtypes.

One of the great things about using this big data is we can look at these groups like dementia subtypes and have over 56,000 individuals to look at. On the X-axis is the dementia type. You're seeing here that the frontotemporal dementia, as well vascular dementia and mixed dementia had the highest occurrence of suicidal behavior. Here I'm touching on some work with my colleague, Dr. Kerri Gibson. This is a recent in-press publication coming out soon in medical care.

This is menopausal hormone therapy as a risk factor for suicide deaths in a national sample of midlife and older women veterans. We had nearly 300,000 women veterans, 50 and older. What I want you to really focus on is that last model for that bottom row, and seeing that when we adjust it for everything in hormone therapy in there, adjust it for demographics, medical diagnoses, psychiatric diagnoses, substance use disorders and psychoactive medications, we still found a two-fold increase risk of death by suicide in these older women.

Now, I wanted to bring it all together in this slide and think about the potential role of social disconnection. In the first two columns in this table, pretty much I'm just recapping some of the stuff I've already talked about in each of those four different studies I mentioned. In the last column I want to go through with you is thinking about, can a comorbidity profile work? How might we think about the potential role of social disconnection? This is the intersection of clinical care, and in terms of the data I'm researching and looking at.

I think it's important that clinicians trust their internal compass when it comes to their patients and social disconnection. I've talked to clinical colleagues. They do note social isolation. They do notice in their patients' loneliness. They do notice in their patients' hopelessness. They are concerned, and they do care, and they do want to collaborate more and think through this. I think that potential the mental health work that we did with looking at last visit in mental health may suggest that with a greater interaction with mental health services, there's greater social supportive services, potentially, that are having a real impact on individuals at least with failed attempts.

I think it's imperative, absolutely imperative- [crosstalk] I think that's especially imperative in those who are socially isolated, and to ask about other lethal means, even drug use and medication use. In the reentry study, I think it's just so important to increase supportive services at the point of transition. Prisoners are being released late at night. They may or may not have individuals to go to. They may be very alone on the outside, more so than not, especially in later life. Some of these persons have been in prison for 30 years and now they're going out into community. They're also being separated from a social network that they had in prison, and that may impact their risk. [unintelligible 03:00:05] dementia patients I think it's-- when you have a diagnosis of dementia, or even think about diagnosis of MCI: mild cognitive impairment, feeling alone and hopeless. I think it's imperative to increase supportive services at the point of diagnosis. We do it for cancer, have supportive services, and I realize cancer is often a more terminal disease. I think dementia patients feel a sense of hopelessness and engage their caregivers more.

In terms of medications like menopausal hormone therapy, many of these women we noticed had persistent hormone therapy use. They were trying different things and may very well been feeling alone and hopeless as navigating these menopausal symptoms. Probably really challenging. Increasing supportive services again, and potentially monitoring, which I do think would be important with these certain unique medications that we may prescribe later in life.

To sum it up for you, social disconnection at the patient level in big data is very difficult to measure, but I do think that my work has shown there are indicators potentially of something else going on. I think that is potentially related to social isolation, loneliness, hopelessness. I think suicide prevention is everyone's business and so it's growing old. I think managing the complexity of medical and psychological disorders is really important. I do think it's important to do in the context also of assessing for social disconnection.

I think it is incredibly important that we entrust more, empower more my clinical colleagues, internal clinical instincts, and heuristics. I think these are powerful suicide prevention tools, especially when it comes to assessing such things as social isolation and loneliness. In terms of future directions as I mentioned before, I really want to engage more of my clinical colleagues to inform my research. That would help to address social disconnection and late-life suicide.

I challenge all of us to think about in this time of COVID. What does it mean now with telehealth in older adults? I've talked to some clinical colleagues who do telehealth most of the time, and they don't know. They're like, "I'm trying to assess suicide risk" and they have firearms in the house. They're socially isolated. There's nobody else around. I'm scared. What do I do? What do I say to them? That's some food for thought there. I just want to thank you for your time, thank my funders, thank my institute, thank my coauthors, thank my clinical colleagues. I want to thank the men and women veterans who entrusted their care to the VA and made these studies possible. Thank you so much.

Jovier: Thank you.

Marty: Thank you, Amy. Like Dr. Oquendo, you gave us a great richness of information and from different perspectives and various thoughtfulness. It's looking forward to the discussion. Dr. Yu Li is unable to make it today suddenly, which is unfortunate. What we're going to do is, is Nancy Donovan ready to go forward. Jovier?

Nancy Donovan: I'm here, you want me to go forward?

Marty: If you could Nancy. What we thought you would do is that we would let you that you talk now, then we take our break, then Dr. Deroski will talk and we'll have all round table.

Nancy: Sure.

Marty: Okay, great. Thank you.

Jovier: Thank you, Nancy. Thank you.

Marty: Alright. Yeates is supposed to say your name but I'm going to say your name. We get to hear from Nancy Donovan who is going to introduce herself.

Yeates: I was hoping I get to do that.

Marty: Okay, you can say her name too.

Yeates: No. Go ahead. Nice to meet you, Nancy, we’ve not met before. Nancy Donovan thank you. If you want to introduce yourself then carry on. We appreciate your flexibility here.

Nancy: Sure. Hello, thank you for the opportunity to present today. I'm Nancy Donovan, I'm Director of Geriatric Psychiatry at Brigham and Women's Hospital. I'm an associate researcher at Massachusetts General Hospital Harvard Medical School. I'm going to be talking about social disconnection in older adults with Alzheimer's disease and other neurodegenerative conditions. I'd like to start out by acknowledging the National Academy of Science, Engineering, and Medicine committee report on Social Isolation and Loneliness in Older Adults. I'm going to be referencing this report. I was a member of the committee and worked on this section.

I'm a geriatric psychiatrist and an Alzheimer's disease researcher. The objective of my talk is to describe changes in social function that are observed in individuals with neurodegenerative disorders such as Alzheimer's disease. Changes that become more severe at dementia stage impairment. Currently, 5.8 million older adults are diagnosed with Alzheimer's disease, and this number is expected to double over the next 20 years. This number does not account for individuals affected by Alzheimer's disease at earlier stages of impairments such as mild cognitive impairment. It also doesn't include the number of individuals affected by other forms of dementia. My comments today will concern a very sizeable proportion of older Americans.

Today there's been very little research focusing specifically on social impairments in Alzheimer's disease and neurodegenerative disorders more broadly. There's also been very little research on the emotional consequences and quality of life impacts resulting from social disconnection and dementia. I'm going to begin with a review of the literature pertaining to older adults with dementia. Then I'm going to follow with perspectives from our own research suggesting that loneliness and changes in social engagement may be present even in very early Alzheimer's disease among individuals who have biomarker evidence of the disease but no overt cognitive impairment.

Social withdrawal and other changes in social function are recognized features of neurodegenerative disorders such as Alzheimer's disease, frontotemporal dementia, Parkinson's disease and Lewy body dementia. These changes can also arise and create dementia stages when neurocognitive deficits are less severe or not yet apparent. Primary deficits and core cognitive domains such as memory, reasoning, and executive function or language skills can degrade social function, they do degrade social function with increasing cognitive impairment, disengagement from community groups and reduction in social ties, parallel functional losses, and other aspects of daily living. This global cognitive decline culminates in severe deficits in comprehension, communication, and interpersonal function in late dementia.

Neurodegenerative disorders may also give rise to specific impairments in social processes such as interpreting information from faces, recognizing the thoughts and feelings of others, displaying and experiencing empathy, regulating emotional processes and behaving within social norms. A classic example of social cognitive impairment and dementia would be the Capgras misidentification syndrome, in which a person with dementia fails to recognize or distorts the identity of a spouse or a close family member or friend.

Altered neuropsychiatric function may also impair social function and neurodegenerative disorders. For example, social withdrawal may be part of a broader apathy syndrome. A neuropsychiatric state which is common across dementia types. Misperceptions or overestimation of social threat predisposed to symptoms of agitation and aggression in persons with dementia and anxiety or depressive symptoms or syndromes may underlie or contribute to social avoidance or withdrawal.

There are very limited data regarding the prevalence of loneliness in dementia. A single population-based study of 589 Swedish older adults found that loneliness was more prevalent in those with dementia as compared to those without. A small number of studies have examined the experience of loneliness in persons with dementia. For instance, in a semi-structured interview of 70 persons with early-stage dementia in their caregiver, participants with dementia spoke of the experience of relationship loss and the importance of having meaningful relationships and people to converse with. Loss of memory function was acknowledged as a barrier to social relationships and caregivers tended to perceive loneliness from observed emotions and behaviors, even if loneliness was not explicitly expressed by the person with dementia.

Other research in dementia care has advanced the concept of unmet needs among persons with dementia who are unable to communicate or fulfill their own basic needs. A scoping review of 41 studies examined self-reported needs and experiences of persons with dementia living in nursing homes. Thematic analysis identified meaningful relationships and support for grief and loss as two of the eight main concerns expressed by persons with dementia. Fear of loneliness was expressed particularly among residents with advanced dementia. Many of these participants wrongly believed that family had not visited them, even when they had.

In a separate study, 70% of 89 nursing home residents with dementia reported feelings of loneliness. Loneliness and social isolation are not currently assessed in standard dementia care, and the established instruments that we use for measuring these constructs in the general population have not been validated for persons in dementia and, indeed, they may not be appropriate. Therefore, there remain fundamental gaps in our understanding of social disconnection and dementia, how it emerges and evolves, its neural underpinnings, its prevalence and impacts, and how best to ameliorate this important area of deficit.

My own research examines Alzheimer's disease as a possible etiology of neurobehavioral changes in cognitively normal older adults. This work has included studies of depression and anxiety. However, we've also looked at other types of neurobehavioral signs or symptoms such as loneliness and changes in social engagement. We use PET imaging to examine very early accumulation of Alzheimer's pathologies amyloid-beta and tau in relation to neurobehavioral markers in a large cohort of older adults who are without cognitive impairment.

The literature as a whole has focused on social isolation and loneliness as predictors or antecedent risk factors of cognitive decline and dementia. Our research has considered the reciprocal possibility that social function and social socio-emotional wellbeing may begin to subtly change in cognitively normal older adults at the preclinical stage prior to the onset of mild cognitive impairment or dementia. In that sense, a loneliness and social disengagement may be analogous to early incremental declines in cognition such as memory and executive function that we recognize during the pre-clinical phase as Harbinger is a future cognitive impairment.

A recent publication from our group supports this idea. In a three-year study of cognitively normal older adults, we found that there were declines and level of social engagement among these participants with high brain amyloid-beta that would be the group in the red. There was no change in social engagement over three years in these participants with low amyloid-beta, the group in blue.

We also found that among participants with higher amyloid-beta, those with low social engagement at baseline on the right side of the figure, those with low social engagement had accelerated cognitive decline compared to those with higher baseline social engagement. Together these findings suggest that cognition and social engagement may decline in tandem and be measurable, even during a preclinical stage of Alzheimer's disease.

We've also performed biomarker studies looking at the cross-sectional association of brain amyloid burden measured by PET with loneliness and cognitively normal older adults. The goal of this study was to evaluate loneliness as a potential neurobehavioral marker in very early Alzheimer's disease. We found that higher amyloid-beta was associated with greater loneliness in an unadjusted analysis and also adjusting for numerous relevant covariates.

In dichotomous groupings, to examine the strength of this association, we found that participants with high amyloid-beta had a 7.5-fold higher odds of being lonely rather than non-lonely compared to the low amyloid-beta group. This association of amyloid-beta and loneliness was also stronger in carriers of the APOE4 [unintelligible 03:14:18], which is a major genetic risk factor for Alzheimer's disease. More recently, we've studied tau accumulation and loneliness and cognitively normal older adults. The accumulation of tau pathology occurs in a very predictable sequence in aging older adults and also across stages of Alzheimer's disease. We focused on tau accumulation into pre-specified regions of interest with a special interest in the entorhinal cortex.

The entorhinal cortex is the site of tau accumulation and typical aging, and also in the initial transition from typical aging to early Alzheimer's disease. This is a region where we see increased tau accumulation at the preclinical stage of Alzheimer's disease. Functionally, the entorhinal cortex is a gateway between the hippocampus and neocortex, and its function is critical for episodic and semantic memory processes. This includes processing of social and emotional information.

We found that higher tau accumulation in the right entorhinal cortex was associated with greater loneliness adjusting for multiple comparisons and multiple confounders. We also performed whole-brain maps, correlating loneliness score with tau PET signal. Here you can see positive correlations of loneliness with tau signal in the entorhinal cortex as what would be predicted by our main models. There were also clusters of association with tau signal and the right fusiform gyrus. In this exploratory map, it was unexpected, but it was intriguing as the fusiform gyrus is a region known to be important for face processing.

Prior FMRI studies of emotional face processing and face recognition tasks have shown that older adults perform more poorly than younger adults. Unlike those who are younger, older adults failed to activate right temporal and limbic regions, including the fusiform gyrus during these tasks. It's possible that the accumulation of tau pathology in the right entorhinal cortex and right fusiform gyrus may be associated with deficiencies in the retrieval of semantic information from faces or other deficiencies in face processing. This may be detectable as early as the preclinical stage of Alzheimer's disease.

Our work then provides preliminary evidence that cognitively normal older adults may experience subtle, gradual declines in social connection during preclinical Alzheimer's disease. Much more research is needed to understand the social neuroscience of neurodegenerative disorders and their impacts on social connection and potential suicide risk across the full continuum of preclinical to dementia stages of these diseases. I just want to thank my collaborators and thank you for your attention.

Yeates: Thank you very much, Dr. Donovan. It's such an important and rich area for so many different reasons. I think, right, Marty? At this point, we're going to go to the break. Is that the plan?

Marty: That's the plan.

Yeates: Good. We have until 3:00. Why don't we reconvene then at which point we'll ask Dr. Dombrowski to step to the podium? Then follow that with the roundtable discussion. Thank you. We'll you at 3:00.


Jovier: Good afternoon. This is Jovier Evans with a bit of brief housekeeping. I sent out a chat message to everyone that since we lost one presenter, we will have an altered schedule for the rest of the afternoon. Dr. Dombrowski is going to present from 3:00 to 3:15, which will be followed by the roundtable session from 3:15 to about 3:45, and then a wrap-up. You'll get some time back in your afternoon today. I want to thank you all for your continued persistence as we navigate Zoom in the internet. With that said, I'll turn it back over to our co-chairs.

Yeates: Okay, thank you Jovier. Welcome back, everybody. With that information we'll proceed, our next speaker and this only speaker after which we'll transition directly to the round table discussion, will be Dr. Alex Dombrowski. Alex, are you ready to go?

Alex: Yes, can you see my screen?

Yeates: I don't see slides.

Alex: Just a second. How about now?

Yeates: Yes.

Alex: All right. I am Alex Dombrowski from the University of Pittsburgh. I am a geriatric psychiatrist and I direct the decision neuroscience and psychopathology lab. I would like to thank Jovier, Liz, Marty, and Yates for organizing this conversation. Liz asked for high-level comments and I will make two. One about the life course of suicidal behavior and psychopathology and another about the value of formal mathematical theory in elucidating the suicidal diathesis.

I would like to acknowledge the members of our lab collaborators and the funding agencies, the MAMH and AFSP as well as people who participated in our studies.

Karen Horney in Neurosis and Human Growth, said that you cannot teach an acorn to grow into an oak tree, but given the right conditions it will. In the same way, a human will grow into a mature person given the right conditions. In our studies of older depressed people in the suicidal crisis that [unintelligible 03:21:59] and I have been conducting in Pittsburgh, we see many people whose growth has been disrupted early in life, just like this Japanese Oak. They present with early-onset suicidal behavior, personality pathology, cluster B traits, introversion, neuroticism, which are paralleled by impulsive behaviors and poor real-life decision outcomes.

We also see more social transmission of suicidal behavior in these folks. I know that this group somewhat resembles Maria or Kendo's impulsive subtype. We do not see these features in late-onset cases. The puzzle is what is the nature of the diathesis. Following George [unintelligible 03:22:51] we might surmise that it is neurocognitive as we have also just heard from Nancy. How do we understand the nature of this diathesis? Should we start the consortium to map the structural and functional correlates of late-onset suicidal behavior?

Unfortunately, we're learning that the value of data-driven approaches in elucidating psychopathology has been limited as we have recently heard from the ABCT consortium, for example, and are good statistical, but also epistemic reasons for that. As I was recently reminded by Sam Gershman, Platt said half a century ago that certain systematic methods of scientific thinking may produce much more rapid progress than others. By that, Platt meant that in fields like molecular biology, people devise strong hypothesis, design experiments to adjudicate between them, and regardless of the outcome, they feel gained knowledge and advanced.

By contrast in fields where people just set out to measure a lot of things and hope that the theory will emerge out of the observations, the progress is often slower. I think that still remains the case. For suicidal behavior, a good place to start are the current psychological theories. The theories of entrapment, escape, the interpersonal theory, the motivational volitional theory, and Klonsky's Three-Step theory. All of these theories have one notion in common and that is suicide is an escape from aversive internal states. The three former theories also postulate a factor that accelerates the transition from contemplating suicide to suicidal action. It's often referred to as the capability for suicide.

These theories give us a rich heuristic framework for thinking about suicide. However, recently people have criticized their limitations such as empirically limited predictive power and vague predictions. In part, this is due to what Neil called the crud factor, which is a novel construct. Let's say hopelessness is not very distinct from previously used constructs describing psychopathologies such as depression. I think that's something that Julianne Yates and Marty alluded to this morning. To that, I will add that this theory is focused narrowly on suicidal behavior as a phenomenon to be explained, and also as a source of insight.

In other words, they're not strongly connected with basic science and they do not rely on insights about human behavior in general. Finally, they use natural language constructs that do not give us an account of lower-level factors. Let's say at the level of neural systems that interact to produce suicidal behavior. Well, what is the alternative? Here's one, the realization that the brain and computing machines have much in common may suggest new and valid approaches to psychopathology and even to psychiatry. That was Norbert Wiener in 1948.

Indeed, wherever we look at the brain, the neurons do not think or feel. All they do is combine some inputs, perform some function on them, and produce one output. The same dynamics are instantiated in artificial neural networks. One question is what function do they perform? A really good account of this function of neural computation and also of how an agent can use environmental inputs to produce adaptive behavior is offered by reinforcement learning.

In reinforcement learning, the agent finds himself in a stage S, where it can follow the policy pie to choose between a set of actions, A, and each of these actions can lead to a reward R which could also be a negative of punishment with a probability P and then they will move on to S prime, the next state. The agent relies on the expected value or the expected reward of the actions to make their choices. That's the policy. This expected value is not simply a fiction, but something that the human and the primate and rodent brain has been shown to compute in the region that in humans is called the ventromedial prefrontal cortex.

Interestingly Amy told us about the high rate of suicide in FT, for the frontal temporal dementia. This was, of course, the region that's affected early in FTD. The problem, however, is that these expected values are not known. The world is uncertain, so the agent has to learn them, and he does so in reinforcement learning by updating the value based on the difference between the actually obtained and the previously expected reward. That's the prediction error. Prediction errors are found in the dopaminergic mesostriatal system. Then they are broadcast through cortico [unintelligible 03:28:45] thalamic loops to the associative cortex, such as the frontal operculum.

There's also a reward feedback to the visual cortex, which you can think of as an attentional feedback. These are, by the way, the results of multiple human brain imaging studies from a recent net analysis. We can think of the agent as, say your Roomba robot, who's looking for dust on your floor, but we can also think of them as a person in a suicidal crisis who's trying to compare suicide to their alternative solutions. The simplest hypothesis that reinforcement learning offers, is that their policy is noisy or biased. That's the hypothesis that we tested in multiple behavioral studies. We do see disrupted choice processes in suicidal behavior.

Particularly in high lethality suicide attempters as well as disrupted value representations in the ventromedial prefrontal cortex. We have limited evidence that in the social domain the choices are biased but that's still very preliminary. A complementary hypothesis is that people are impaired in learning the values of actions learning what is the best thing to do in a crisis. That's informed by the insight that a suicidal crisis is usually triggered by a major event that produces change and uncertainty.

There is some behavioral evidence to support this disrupted learning hypothesis in attempted suicide in late life. Interestingly, we did not find disruptive prediction errors in the brain, so that's an informative negative, but what about social learning? As people have said today, both the triggers, the protective factors, and the deterrence to suicidal behaviors are often social.

We are interested in how suicidal people learn, what works with other people, and how their own actions affect others as a model for what drives the escalation of the interpersonal crisis to the suicidal crisis. To answer this question, Lena [unintelligible 03:31:15] and I used iterative economic exchanges where you are interacting with another person and you can choose to cooperate or invest your money with them or not and you don't know whether that person will return your investment, cooperate or defect, keep the money and you can see how an iterative exchanges trust can be built and disrupted.

When we looked at the neural mechanisms of learning in this situation, what we saw is that what people learn is not how good the counterpart is, but whether what they're doing with the counterpart works and that's reflected in the straight learning signal. You can see that in the striatum, the signal and prediction errors are higher for correctly predicted cooperation compared to the control conditions, but also to correctly predicted defections where a bad thing happened, the person cheated you, tried to cheat you, but you know that you did the right thing and that's a positive reward prediction error.

When we examine these signals in older suicidal people with late-life depression, we compared older suicide attempters in purple to suicide ideators, non-suicidal depressed and healthy controls and saw that all patient groups had impaired signals in the striatum, impaired learning signals, not much was seen in the thalamus and that's by the way meta-analytic a priori prediction error network or the frontal operculum but there was a disruptive learning signal in the occipital cortex, that visual reward feedback. The whole-brain analysis corroborated with these findings.

We sought to extend this because we really believe in replication to another sample of people with borderline personality looking at high lethality suicide attempts, which are probably close to late-life attempts in their phenomenology. We saw that in borderline high lethality attempters compared to low lethality attempters, non-suicidal borderlines, and healthy controls, the learning signals were depressed in the striatum bilaterally, not so much in the midbrain and frontal operculum and also depressed in the visual cortex. These are just the late-life depression data for comparison, that there is a disruption somewhere in this [unintelligible 03:34:01] network and this was supported by whole-brain analysis. This disrupted strata learning signals and high lethality attempters.

The conclusions from the small series of proof of concept studies are that people with depression and borderline personality vulnerable to suicidal behavior show what we call a disruptive corticosteroidal egocentric learning signals that inform them about how well their actions work with others and that may be one of the substrates that facilitated the escalation of an interpersonal crisis into a suicidal crisis.

More generally, we conclude that brain mapping by itself cannot uncover the neural mechanisms of suicidal behavior and current psychological theories offer useful heuristics but need to be formalized to bridge to neural mechanisms. One useful approach to decoding the neural mechanisms is reinforcement learning, but it's not the only useful framework. Perhaps information theory is another one as we are discovering in our studies of the hippocampus and also others. I would like to leave you with the words of Marvin Minsky, "Unless we can explain the mind in terms of things that have no thoughts or feelings of their own, we'll only have gone around in a circle." Thank you.

Yeates: Thanks, Alex. That's really terrific. At this stage, would like to hand things over to Dr. Vonetta Dotson who will chair the interactive session with the four speakers that we've had this afternoon and encourage others to please feel free to provide input through the chat box. We're happy to entertain questions there as well. Dr. Dotson.

Vanetta: Thank you. I would like to first give the caveat that my neighbor has decided to start doing some housework right before I turned on my volume and so I apologize in advance if we get a little bit of noise, but it stopped momentarily. Hopefully, it'll stay stopped, but I do appreciate the opportunity to be here. Thank you to the organizers for inviting me to be a part of this. We've heard a lot of excellent talks today and now we're going to focus on further discussing the presentations from this afternoon. If anyone in the audience has questions, please feel free to post them in the chat and I will pass them on.

I did not plan to show any slides myself. I'd rather just jump into letting the speakers further discuss some of the things that they've brought up. Maybe we can start with a more open-ended question for the speakers. If there's anything else now that you've heard each other's presentations and thought about these topics a bit, is there anything that any of you would like to add or chime in about?

Participant: Can I ask a question of Amy Buyers?

Amy: Of course.

Participant: Okay. As I was preparing for this session, one of the things that I couldn't readily find was the excess risk of suicide attributable to dementia diagnosis. Do you have a sense of how elevated the risk is in people with dementia? I did see your paper showing that the risk was highest in FTD and in vascular dementia, but I'm just wondering in terms of overall, is it a significant contributing factor?

Amy: What's interesting is there's actually a lot of mixed results and it all depends on the population, the timing, how suicide is measured. Is it ideation? Is it death? It's really hard to measure death. It's also like, where does the data come from? Actually, I have in the works as being submitted very soon, a paper on looking at dementia and MCI and risk-related and our findings are that recency of a diagnosis actually does show increased risk and that increased risk is significant, but it's more timing of the diagnosis that's important, it seems like opposed to for dementia. I think that's fairly consistent with dementia diagnosis is closer in time to react to diagnosis, the level of dementia but it's really hard to tease apart because I think time really comes into it in such an important way. It's mixed, you'll find some studies that show it's protective, it's not significant. Some studies show it's significant, so it's really hard to tease that apart.

Participant: Thank you.

Alex: If I may comment on this one, Vanetta. I'm also aware of Annette Erlingson's studies in Danish cohorts that perhaps Amy was referring to and Yates was a collaborator on those studies that show-- There was an earlier, as I understand, smaller analysis showing that hospitalized dementia patients had higher rates after discharge and more recent analysis that came out, I think in 2016, suggested that Alzheimer's disease as a whole was not a risk factor, but Huntington's was and that's something that I personally find very interesting that disorders that disrupt the limbic and paralimbic circuitry that's relevant to reward learning and decision [unintelligible 03:40:11] such as frontal temporal dementia and choice abnormalities are well documented in these patients even before in cognitive decline. In Huntington's, those disorders have elevated rates, whereas Parkinson's, or Alzheimer's as a whole does not. Not that we have not seen suicidal patients with early Alzheimer's disease.

On that note, I was wondering if I could ask you about this, Nancy, you often hear patients or even friends honestly say that if they are diagnosed with dementia, they would want to commit suicide at a certain stage. I asked colleagues at the Alzheimer's Disease Research Center in Pittsburgh, and they said they were not aware of any cases of that. Although patients do say that. I was wondering. Nancy, what your experience was with those cases?

Nancy: My experience is anecdotal really and taking care of patients. One of the things that I've noticed is that patients who themselves have had traumatic caregiving experiences, they've been a caregiver for a spouse say, or a parent with a prolonged and difficult course of dementia, or perhaps difficult like cancer diagnosis and end of life experience. Those are the ones that I worry about when they tell me that this is what they want to do.

First of all, I think that by the time a lot of patients with Alzheimer's disease come to diagnosis, they're already quite impaired. They can certainly express suicidal ideas. Of course, then we work really carefully with patients to take away anything that they can harm themselves with. I guess I do feel like there is an elevated suicide risk in many of our Alzheimer's disease patients, the ones with agitation and disinhibition and paranoia. The more neuropsychiatric burden they have, I think their risk is very high and poor judgment. For instance, when they open up car doors, when the car is moving, what is the motivation, are they trying to harm themselves?

I do feel like there is a risk earlier in the course though when people are less impaired and when they are explaining their thought processes a little bit more in a more sophisticated way. I noticed that it's the people who have some other contributing factors, either they have like a personality disorder or they've had some traumatic experience. It's usually a multi-determined at risk, I think. Does that answer your question?

I will also say that I've had patients, and for that matter, people that I've known who have said that if they had a dementia diagnosis, that they would want to end their life, and as they're experiencing the course of illness, even at the very end of their life, they still want to remain alive. I think people really change their thoughts about that across the course of their season. I think it's very influenced as well by the level of social support that they have.

Amy: It's interesting. I see Yates' comments in the chat, and I think that's really something I struggle with is I use EHR data or EMR data. Unless I use something like natural language processing to actually get at physician's, clinician's notes. I don't even know how well that's documented. I think the struggle is for me is how can we measure that to study it, to see if it matters, makes a difference, or see if it's having an impact on those individuals.

I know like when individuals get a certain diagnosis that oftentimes there's family around, there's caregivers around, there might be support structures that they'd have, but then at the clinical level, what supportive structures are there? I know there can be challenges with mental health services in terms of treating dementia patients in terms of whose job is it to treat them? The neurologist, the primary care doctor. I do wonder about how do we get at if it matters and how can we go that using--?

Because something like dementia, even being able to study it, we can study it with the big data, but we can't study social support in a standardized very good way. We have to find proxy measures, or we have to use something like natural language processing and hope it shows up in the clinician's notes.

Alex: Your hypothesis, Amy, that the elevated suicide rates, for example, in the FTD are moderated by social support that this happens in people who are like?

Amy: I don't know. I'd like to be able to study it, but I'm challenged by the data. I can't unless I do something with primary data collection or maybe NLP. Theoretically, I think it probably matters, but I'm struggling on how to study it and EMR data. I noticed I think it was Kim Van Orden who made a comment about some of these. The assessment can actually be the promise so social isolation scale can easily be administered and integrated into EMR. I don't know how come more health systems aren't doing that or thinking about doing that and how could we navigate making that more important? I think we need studies to show it as important, right?

Nancy: [unintelligible 03:47:04] Are you trying to assess suicide risk or ---

[cross talk]

Amy: Actually assess social disconnection.

Nancy: In my talk, I didn't include another portion of the report, which draws the connection between social disconnection and agitation and aggression and dementia and the idea that physical agitation and other forms of agitation really are manifestations of disconnection. In fact, interventions that have been used as you probably well know, simulate a family presence or use of robotic or living animal-assisted therapy, those do tend to reduce agitation and to the extent that they've studied loneliness. It seems to reduce loneliness too, and music therapy as well.

I just feel like if we're going to be measuring disconnection in persons with dementia, especially, moderate or severe depression, it's not going to be language-based, it's going to be based on other observed features of their behavior.

Maria: This is Maria. I was just thinking that based on the broad impact of loneliness on a lot of health outcomes like cardiovascular outcomes, et cetera, that we have an opportunity to team up with our colleagues and other specialties to-- Really, this is about policy change. That working with regulators to increase the likelihood that hospitals, for example, or health systems would adopt these types of measures.

Jovier: Dr. Oquendo. Am I'm muted? I can't even tell anymore. I was going to ask you a question about your talk and your comments around aspects of sleep and other aspects of behavior across species and what do you think then would be likely, I don't know if I want to go that far as the treatment targets, but other things to look at behaviorally that might be mutable for us to consider.

Maria: That's a really good question. I have to say that there are a couple of reviews that actually list out a number of therapeutic interventions, mostly in the cardiovascular space. The last study I mentioned in my talk, which is actually from the sleep center at the University of Pennsylvania, I thought had very compelling data about the impact of something like Zolpidem, which is a sleep medication in terms of reversing the physiological impact of sleep deprivation. One thing that I think is really a terrible compound is oftentimes in the epidemiological data. People link the use of, say benzodiazepines to cognitive decline and early death but it's likely that those are not causal relationships but rather that people that have cognitive decline may have difficulties with isolation and therefore sleep and then reach out to sleep aids. I think that there's a really important question there that we need to tease out.

Jovier: I have a comment, I have to make a plug from a colleague of mine. Janine Simmons at NIA wanted me to remind people that we have a social connectiveness and isolation PAR through Opnet that has one more receipt date in March. Thank you, Janine, and then I'll make a plug for ourselves, we actually have another PAR with NIA on neuropsychiatric symptoms and mechanisms in neurodegenerative disease and dementia. I just threw that out there, want to make a plug, so thank you.

Vanetta: Thanks, Jovier. I have a couple of questions that have come through on the chat. One of them, I guess, would be for any of the speakers. Is there a possibility that the personal, economic, and social cost of dementia or psychiatric pathologies could be a factor in suicidal ideation?

Maria: I think that's a really interesting question and certainly the link between economic distress and suicidal path has been documented for a long time. It wouldn't be terribly surprising and of course, it depends on probably the stage of the dementing process and how well the person can really think in terms of the impact of their economic situation to their wellbeing.

Nancy: I think we've seen that clinically.

Vanetta: A comment that was made by the same person was that a lot of the work that's been done, has been done in the US and this is from someone who's coming from an international region and I think it goes back to the conversation earlier this morning about the lack of sort of multicultural and multiethnic samples in some of this research and points to the need to really increase the representation of different groups in this kind of work. There's another question for Alex, going back to what you were saying earlier, have recent computational approaches clarified the chicken and egg problem, corticostriatal, or striatal of the cortical direction of the reward learning in error signals?

Alex: I would be a fool for trying to summarize this literature in a minute but for people to whom this is alien, there is an incoming signal from the associative cortex and the motor cortex into the striatum which is the input layer of the basal ganglia and you can think of these signals as loops that represent actions or motor actions or abstract actions or choices.

They are updated in the striatum by the dopaminergic mesostriatal projection, and there are [unintelligible 03:53:31] opioid mechanisms for learning in the striatum. Then there are three different output pathways through which the updated signals like those learning signals are broadcast into the cortex, they go through the output layer of the basal ganglia into the thalamus and some of these pathways are inhibitory and others are excitatory.

I think what's crucial here is that it's a loop, so the cortex makes a hypothesis or a plan about what to do. It gets a reward prediction error update in the striatum, it looks at it and accordingly updates the value of what it was going to do. There are more complex relationships though because the question is, what do you learn from? There's so much happening in the environment and somebody needs to put you in front of the right stimuli that you will be updating.

For example, if you're at a stoplight, you want to look at the lights and the traffic but not the birds in the trees so in reinforcement learning that is known as a state-space partition problem is basically reducing a complex environment to some simple game where you have a limited set of actions that you're updating and that's of course, done by the prefrontal cortex or the frontal-parietal networks and in that sense the frontal cortex orients the striatum as to what to learn from.

Vanetta: Thank you for that. A couple of other questions, how has the directionality between social disconnection and sleep been established?

Maria: It's been established mostly in animal models. I only presented some fruit fly models and some mouse models, but a lot of different animal studies have been conducted. One of the limitations that I think I alluded to was that many of the models implement the isolation very early on in life which is a little bit different than what we're looking at but they can really see the causality across genetically identical animals in terms of the impact of social isolation and they can even look at the dose effects of social isolation in those types of models.

Vanetta: Great, thank you and I think Yates had a question that he wanted to ask.

Yates: I did. Thank you. It's for you, Nancy and others. It's really about the context of dementia research these days and where their opportunities might lie there and so much of the action and this isn't work that I do particularly but does seem to be in MCI and preclinical states and the idea that the earlier you intervene the better the likelihood of success and seems to me that at least the anecdotal evidence is pretty good that it is in those early and preclinical stages of the disease in which the risk for suicide behavior may be much higher for a lot of reasons that you've touched on and others.

We've got things going on with therapies like [unintelligible 03:57:21] the amyloid antibody infusions, we've got advances in biomarkers. I guess the question is where do you see this potentially going? Are there opportunities that are going to arise where we can actually selectively insert some standardized measures of social connectedness in the patients that are undergoing large longitudinal studies of dementia and interventions from which we might be able to begin to address some of the questions that are raised here?

Nancy: No, I'm not aware. Actually, I should know, I can ask around. I don't know that we're really measuring social connection in the clinical trials work. Obviously, we're doing it in our own observational study but I guess one of the things that's striking me and I'm maybe taking this in a different direction is the idea that there's like a number of people have talked about like the non-impulsive form of suicide and what would be the risk factors for suicide in someone who has Alzheimer's disease and maybe it's not psychiatrically driven and it's not neurologically driven, it's the human reaction to having this disorder.

I guess what someone else was referring to making a choice that they either a choice based on finances or based on lack of social support. I think those are people that I do worry about since there's such a huge burden of neurodegenerative disease, people who for non-biological reasons may and indeed for reasons of not having adequate support or not perceiving themselves to have adequate support would go down the direction of suicide. I know that, for instance, for the idea of study, the Medicare study where people or even in the A4 study, which is a prevention trial, and they've determined amyloid status in cognitively normal people. They've tried to determine whether or not there are adverse psychological reactions to the diagnosis, and more work does need to be done to understand how to support people after the diagnosis and in the early stages of these disorders. I do think that that's the highest risk time.

Amy: I have a question for Nancy. I said, working with cancer patients we- it's just that I've been thinking more about this, I'm like, the dementia patients don't get the same supportive services, neither is arguments for it's not the same as like with cancer, more of a terminal illness, but there is a structural systematic support in place for when one gets a diagnosis of cancer. What has been the barrier obstacle to doing that for patients who get a diagnosis of dementia?

Nancy: I think it's workforce. I think that traditionally people get diagnosed very late for one thing and they're diagnosed either by a neurologist or maybe very late in the game by their primary care physician and there's just a huge undersupply of geriatric psychiatrists. I don't feel people are recognized very early in the disease process. Although I know there are people who are interested in taking more of a palliative approach earlier in the course of many late-life diseases in terms and including cognitive impairment.

I guess just one thing that came up for me in these discussions today was this idea of meaning, and it goes back to Julianne's first talk when she was talking about social relationships in viewing values and meanings to life and I think that when people get a dementia diagnosis they can't make sense of it. It changes the whole meaning and context of their lives. I do feel like supporting people psychologically around the diagnosis could make a really big difference and also their family.

Vanetta: We are starting to run low on time, but I know Marty had a question, we can maybe--

Marty: Oh, it's maybe just a concluding statement. I'm not sure or a really naive question, but I'm really struck that this session has been looking at targets and mostly focusing in on biological and neural cognitive neurological, and Alex's statement of that, in some cases people, they're not perceiving reality like reality might be. I say that's a simple- sorry, Alex- it's very simple, but there's also a lot of circumstances where people are being distant from the rest of society, and that we have ages and we have this perception of people not engaging with older adults and I think it'd be interesting as we move forward thinking about targets of how do you reconcile over two different strains of people's lives are like, and then what's going on neurologically and cognitively? I'm not sure if that's a question or just an observation.

Amy: Well, I just want to say, going back to, what I keep thinking about is- and Maria made a comment about in terms of treatment and thinking through what is- there's something called confounding by indication. Oftentimes, even if you look at psychoactive drugs, or polypharmacy and whatnot, these are often it's more- in EHR data these are like the most severe cases so you often do see an increased risk because these are the sickest individuals, or they have the most mental health issues and whatnot, but I do think that you can still think about it as these are important markers and indicators.

To hone in on that, even if it's not causally related, is it something to monitor and look at and take into consideration, because people are having these issues? I just wanted to add that.

Nancy: I really agree with that and I think that, for instance, for some neuropsychiatrists, but addressing the neuropsychiatric manifestations of these neurodegenerative disorders very early, at least, if you can relieve some of those symptoms, you're not necessarily going to impact the progression but you can reduce the burden and the distress and help just lower the suffering, and certainly bolstering social support in that early phase also would be really helpful to people.

Vanetta: Thank you all for a great discussion. I think it is a little bit past time to turn things over to our wrap up of today.

Marty: Thank you, Vanetta. I think it's the perfect time. It's not past time.

Just very shortly that this has been a great day. We've really had a lot to think about, about mechanisms and targets and I want to remind people that I'm very pleased that people are as we said, in the beginning, are listening to each other and I think we're trying to bring the different strands of where the research is, and what the science is to figure out where the folds are, and what the directions are and it's pretty exciting. We'll be doing more of that tomorrow.

Yates has some things to say I'm sure, but I will just summarize what we're going to do tomorrow is that we'll be reconvening at 10 o'clock and then, beginning with another series of presentations, just like the ones of today, this time talking on the translation of targets, to implementation of interventions and thinking sort of mechanistic thinking that will from there, we will go into- and those will be open to everyone who's on the call now, the public, we will have a talk from Dr. Lazenby of NIMH. Then there will be a closed session of breakout rooms from 1:00 to 1:45, and then that will move into another open session where we will try to synthesize some of the discussions that we've had throughout the two days.

We'll be looking forward to everybody coming back again at 10 o'clock.

Yeates: Thank you all. Have a good evening.

Jovier: Thank you. We’ll see you tomorrow

Marty: Goodnight.

Jovier: Goodnight.