Expert Conversations: Using Technology to Diagnose HIV-Associated Cognitive Impairment
In this podcast, Raeanne Moore, PhD, and Maulika Kohli talk about their recent article, which investigated using technology to capture digital phenotyping data in HIV-associated neurocognitive disorders (HAND).
Additional Resource:
- Kohli M, Moore DJ, Moore RC. Using health technology to capture digital phenotyping data in HIV-associated neurocognitive disorders. AIDS. Published online October 8, 2020. https://doi.org/10.1097/qad.0000000000002726
Raeanne Moore, PhD, is an Associate Professor of Psychiatry at the University of California San Diego. Trained as a clinical neuropsychologist, her research focuses on utilizing digital health technologies to improve assessment of cognitive functioning. She is also the cofounder of KeyWise AI, a developer of AI-enabled digital detection of mental health symptoms through smartphone keyboard interactions.
Maulika Kohli, BA, is a graduate student at the San Diego State University/University of California San Diego Joint Doctoral Program in Clinical Psychology. Maulika is supported by the National Institute on Alcohol Abuse and Alcoholism of the National Institutes of Health under Award Number T32AA013525.
TRANSCRIPT:
Amanda Balbi: Hello everyone, and welcome to another installment of Podcasts360—your go-to resource for medical news and clinical updates. I’m your moderator, Amanda Balbi with Consultant360 Specialty Network.
An article was published recently about using technology to capture digital phenotyping data in HIV-associated neurocognitive disorders (or HAND). My guests today are the lead authors of the article:
Raeanne Moore: My name is Dr Raeanne Moore. I'm an associate professor in the Department of Psychiatry at the University of California, San Diego. And as disclosures, I’m the cofounder of KeyWise AI, and I am a consultant for NeuroUX.
Maulika Kohli: And I'm Malika Kohli. I am a graduate student at the San Diego State University and University of California, San Diego, Joint Doctoral Program in clinical psychology.
Amanda Balbi: Thank you both for joining me today. To start, can you tell us more about how this article came about?
Maulika Kohli: So, life expectancy of people living with HIV receiving antiretroviral therapy is steadily increasing and with that comes a higher risk of age-related problems.
We're particularly interested in the potential compounding effects of both HIV and aging on the brain. We know that older people with HIV are at a higher risk of HIV-associated neurocognitive disorders, or HAND, which describes neurocognitive impairments associated with HIV infection.
Even people with HIV are also likely to be at an increased risk of age-related neurodegenerative diseases like Alzheimer disease and its precursor agnostic mild cognitive impairment, because of both the effects of HIV and aging on the brain, as well as a higher prevalence of disease risk factors like chronic inflammation and hypertension that really complicates their clinical presentation.
Where neurodegenerative conditions are commonly associated with progressive cognitive and functional decline, HAND is usually nonprogressive. So, cognitive and functional abilities may become impaired, but then remain stable rather than progressively declining like they would in a neurodegenerative condition.
But aside from the progressive vs nonprogressive trajectories, both have really similar comment or presentations. So, for example, they're both associated with impaired memory and impaired executive functions like planning and organizing.
And really, because of this overlap in neurocognitive profiles, identifying preclinical factors that may distinguish among those with HAND vs those on Alzheimer disease trajectory is fairly challenging, but considering the negative health outcomes associated with cognitive dysfunction in people with HIV, like unemployment, medication nonadherence, and depressed mood, as well as the multisystem impact of aging, it's really important to focus work on improving our understanding and treatment of neurocognitive outcomes in this population of aging people with HIV.
More than that, trying to find differentiate HAND from neurodegenerative disease pathology is also really important to understand the likelihood of cognitive impairment progression and for providing, then, effective and targeted treatments.
Raeanne Moore: Yeah, and as Maulika was mentioning, existing tools for measuring cognitive impairment among persons with HIV currently really lacks sensitivity and their cost- and resource-intensive.
Quite more often than not, existing methods require a person to go to a clinic or hospital. They have to spend the day doing long cumbersome neuropsychological testing. They also quite often have to get invasive medical procedures such as MRI, lumbar punctures, especially if there's concerns for an urgent degenerative process such as Alzheimer disease.
And by and large, people really are not getting the care they need. Furthermore, early detection is very difficult because initial changes in cognition are very subtle, and people often do not present with clinical symptoms until the disease has significantly progressed.
We're really in need of novel multimodal measurement tools that can fuse subjective data with objective data, fuse episodic data collection with continuous data collection. In the real world, as people are going about their everyday life, to paint a more holistic picture of a person with HIV is cognitive health.
There's this new field of scientific study, which we're very involved in, called digital phenotyping, and digital phenotyping is based on a theory that personal devices like fitness trackers, smart phones, smart watches—and when we use these devices in combination with advanced data analytics like machine learning and artificial intelligence, it can really act as a digital proxy for human behavior.
What digital phenotype does is it takes data that's stored in our personal devices, data that is being collected as we go about interacting with our devices on a day-to-day basis, such as how we type on our phones, our speech patterns, and how we're using the phone—so how many times we pick it up throughout the day, how many times we click on Instagram or another social media app or open a web browser, our movement because all these devices have accelerometers built into them.
It takes this data, and we can use this data to diagnose diseases. Currently, there's evidence showing that this data can be used to help diagnose early onset Parkinson disease, as well as mental health disorders and depression.
If you think about it, to me at least, it really makes a lot of intuitive sense because smartphones are the first thing we often look at when we wake up in the morning. That’s what we were just joking about—it's election week, and we've been waking up every hour looking at our smartphones, and they're often the last thing that we touch before we go to bed.
We're just constantly interacting with them. Thinking about how everyone has their own unique fingerprint, we also all have our own unique digital signature. We're really finding more and more that these digital signatures are related to behavioral indicators of health, as well as cognitive ability.
So, I have an ongoing program of research at the University of California, San Diego, HIV-neuro behavioral research program. Through some of our ongoing studies , we're finding that there are tremendous potential for these mobile tools in improving assessment methods, especially of neurocognitive outcomes among persons with HIV. That was really what inspired us to write this article.
Amanda Balbi: What technologies/devices can be used to phenotype HAND?
Maulika Kohli: There are several different approaches on using technology and devices like Raeanne just mentioned, actually—using smartphones or smartwatches or different kinds of trackers that can always be used together to create this digital phenotype.
We often think of engaging with technology as either actively or passively, active being that it requires the user or the person to actively engage with the device in order to provide information that can be used later. An example of active data that can be collected from mobile devices is self-report data using ecological momentary assessments or EMA.
EMA is kind of an innovative approach to measuring real-world outcomes, and it collects data subjectively, so by the user as well as episodically, so throughout the day across days. An example of what an EMA question would be, “How stressed do you feel right now?”
The nice thing from a data-collection standpoint is that it involves repeated sampling of behaviors and experiences in real time, as well as in a person's natural environment, which can be used to assess daily fluctuations in self-reported cognition, as well as factors associated with cognition—things like mood, stress, social support, substance use, and everyday activities.
Along with EMAs, mobile cognitive testing is another example of an active-data collection method that can be used to help phenotype HAND. Mobile cognitive testing is having someone take tests of memory, attention, and some other thinking skills using their mobile. It provides the ability to study the impact of someone's environment on cognitive abilities in the real world and in their natural environment.
As an example of how mobile cognitive assessments and EMAs can be used together, we looked at the possibility that fluctuations in cognition were associated with real-time activities in middle-age and older adults with HIV.
What we found is that doing cognitively stimulating activities, like working, reading, or writing before taking these cognitive tests, was related to better performance on the mobile cognitive tests, whereas doing more passive-leisure activities. like watching TV, related to worse cognitive performance.
What this study really shows is that using these technologies may help in understanding the different lifestyle factors or everyday health factors and even environmental contacts that might influence someone's cognitive health.
Raeanne Moore: In addition to the active data-collection methods that Maulika was just mentioning, the active data-collection methods are more episodic and not continuous. Passive data is another approach that can be used to phenotype physiology and behavior that's linked to HAND.
Passive data includes data that is obtained passively and unobtrusively through mobile devices such as smartphones, universally known as wearables. And it can be used in conjunction or independent of active data collection.
Fitness trackers are what immediately come to mind when I think every good example of a passive data collection tool that most people are familiar with.
They're providing you with continuous, objective measurements of your sleep if you wear it overnight, your physical activity, your heart rate variability, and that they're getting more and more advanced in measuring all kinds of other physiological metrics these days.
To give you a real-life example of some of our ongoing work in the lab, we have a newly funded study funded by the National Institute of Aging in which we're examining whether keystroke dynamic features obtained from a smartphone keyboard—we don't measure what people type, but we measure how you type it.
And we're examining how these keystroke dynamic features are linked to cognitive impairment and risk factors for Alzheimer disease among middle-aged and older Hispanics with and without HIV.
In prior work by myself and my collaborators, we've demonstrated that by applying advanced data analytics, such as machine learning and artificial intelligence, to the continuous monitoring of keystroke features and some of these features are things like the time between key presses and use of autocorrect, backspace usage—those types of metrics that we obtained from how a person is typing just when they're interacting with their phone.
We got to do one on one texting or when they're emailing or when they're posting on social media. Any kind of time they’re interacting with their smartphone. We're finding that these features can elucidate changes in circadian rhythm, as well as intraindividual variability is associated with early neurodegenerative changes.
We do have some evidence demonstrating that keystroke features are associated with traditional neurocognitive domains that we assess in a lab, such as processing speed, executive functions in memory, which Maulika has previously mentioned, which are impaired in HAND.
It's pretty exciting, and I look forward to finishing and getting the results out there.
Keystroke data is just one example of a promising metrics that can potentially be used as a standalone tool commonly used in combination with other digital metrics to phenotype HAND.
And currently, the field of digital pnenotyping is in the early stages. But we do see there's some immediate applications for digital phenotyping that can be applied to a research setting. People could use ubiquitous technologies that are in the hands of most people, especially smartphones, as low-cost, low-burden, risk-detection tools.
Ultimately, we're not there yet. But I think ultimately ideas could be integrated into clinical care and provide opportunities for early intervention and reducing health care costs.
Amanda Balbi: Neurocognitive disorders are serious complications associated with HIV infection. How might using technology improve patient care?
Maulika Kohli: That's a really important question. So using both active and positive technology is kind of like we've just described—to frequently monitor cognition, as well as everyday factors that might influence cognition may help us increase and improve our ability to detect change in cognition, mood, behavior, all of which might be indicative of early neurodegenerative disease pathology or HAND.
The real benefit of frequent monitoring is that we're able to get a more accurate picture of someone's cognitive health compared to just like a snapshot every year or a couple years when someone might go to a primary care appointment.
Also, with earlier indications are diagnosis of these conditions comes earlier interventions, too, like earlier initiation of medications, modifying lifestyle factors that might impact cognition—that could be like getting more physical activity, improving sleep habits, or eating healthier. And then it can also help with early family planning. All of these might improve patient health outcomes.
This kind of cognitive monitoring might be especially helpful for someone—to give a more real-life example—who knows that they have a genetic marker that increases their risk of having a neurodegenerative disease like Alzheimer disease.
Or maybe they have a family history of dementia, and they're worried as they're getting older, maybe starting to notice some subtle changes in memory that they are worried about disease onset. They're looking to self-monitor for early risk factors.
This allows the ability for patients to become more proactive members in monitoring their subtle changes in cognition using self-tracking of cognition and associated symptoms.
Kind of like Raeanne was mentioning, this is similar to how people track things like their personal fitness, which could have an impact on guiding clinical care, treatment planning, and life planning.
Amanda Balbi: What is the overall key take-home message for health care providers today?
Raeanne Moore: Ultimately, digital phenotyping I really truly believe has the potential to help clinicians detect diseases such as HAND and Alzheimer disease sooner and earlier along in the disease course.
The visual biomarker tools have the potential to aid with early identification in at-risk individuals, as well as improve diagnosis accuracy and tracking the symptoms disease progression over time can help provide better quantification of the treatment response.
I can also see it having any utility for accelerating clinical trials for drug discovery. Furthermore, this tool can reduce barriers to care and, just as importantly, reduce health disparities.
We're currently in this transition from telehealth, which telehealth has really been accelerated by the pandemic that we’re currently in—but telehealth is still episodic care. So we're going through a transition of episodic care to continuous remote care. It’s a really exciting time. And I can see a future where health care is one that incorporates continuous remote monitoring to measure objective cognitive change, as well as provide actionable insight to the patient and their health care providers.
I just like to leave you with kind of this mental image. Imagine like a future where you have a patient, and they walk into their doctor's office. They're getting our vitals from the nurse. She or he is taking their temperature, their height, their weight, their blood pressure, their cell phone, their Apple Watch data. Then the doctor can use all of these metrics to really assess the patient's overall health.
Amanda Balbi: Thank you again for speaking with me today.
Raeanne Moore: Yeah, thank you so much for having us. We’re really excited about this work we're doing. And as I was mentioning, it's really an emerging field, but we're currently in the transition time that the field has been accelerated. The technology is available. It’s available in people's hands already, and I think it can really make a big impact on improving self-monitoring, self-tracking of cognitive health, as well as interactions between patients and their health care providers.
Maulika Kohli: Yeah, thank you so much for having us. It was a pleasure to talk about our recent publication and like Raeanne said, it's a really exciting time. So, thank you for having us to talk about it.