Most people check their phones throughout the day—when they wake up, at lunch, at an afternoon coffee, before bed. Just a glance, nothing more than five or 10 minutes of online tinkering or playing a game. It’s a habit so ubiquitous, we do it without thinking.
Which makes it the perfect time to get a snapshot of your brain’s health.
That’s the goal of the Mobile Toolbox project, a research initiative being carried out by teams at Penn State University and Northwestern University. With the help of the Seattle-based nonprofit Sage Bionetworks and the US National Institutes of Health, they’re searching for ways to measure populations’ cognitive health in the least invasive way possible—by capitalizing on the time you’re already spending on your phone.
Right now, the Penn State team is collecting smartphone snapshots on some 300 healthy adults over 70. They’ve created a mobile app that pings users four times a day to play a quick game that tests different aspects of memory. Twice a day, in the morning and in the evening, they check in again—these times, without a ping. Over the course of two weeks, that means they’ll provide researchers with 84 data points about their cognitive health.
This kind of non-invasive cognitive screening has the potential to solve two of the biggest challenges in dementia detection. First, it’s inaccessible: To get a checkup on cognitive health, people have to visit their primary care doctor, or a neurologist if they have a referral. Second and perhaps more importantly, though, a single in-person assessment of someone’s cognition is unlikely to give their health care provider an accurate picture of their overall cognitive health.
Right now, the ongoing Penn State trial provides participants an Android phone with all functions disabled except for the app, called Ambulatory Methods for Measuring Cognitive Change, or M2C2. But one day it could be a secure, platform-agnostic app available to download for anyone who has a smartphone. That could allow researchers to conduct burst measurements—several quick cognitive assessments, given out six times a day, for two weeks straight—to account for normal variability in mental performance.
As scientists have studied dementia, they’ve discovered that its noticeable symptoms—the changes in a person’s ability to remember and process information, and sometimes even their behavior—are the result of complex neurological changes that accumulate over time. The dementia we recognize is actually the result of years of deterioration in the brain. For a while, the brain is able to MacGyver around lost neurons, maintaining normal function, until one day it reaches a tipping point.
That could explain why, after decades of research, drug therapies have failed to prevent the worst symptoms of Alzheimer’s disease and other dementias. By the time people have developed dementia or even its precursor, mild cognitive impairment (MCI), the damage has already been done.
In order to develop an effective treatment for dementia, scientists are going to have to find a way to prevent it before a person notices anything is wrong. A decade ago, this was a daunting task. But in the era of technology that can track aspects of our health ranging from our steps to our sleep, it seems only logical that our smartphones could play a role in monitoring—and perhaps predicting—our cognitive health.
In order to improve eventual treatment options for MCI and dementia, researchers are looking at ways even the tiniest physical, cognitive, and personality changes can indicate the start of bigger neurological shifts.
The idea of diagnosing dementia in a living person is relatively new in the world of neuroscience. It was only in the 1970s that clinicians developed cheap, standardized tests for patients they suspected to have dementia.
The first of these, developed in 1975, is called the mini-mental state evaluation, or MMSE. It’s a 15-minute questionnaire with a scale of 0 to 30, 30 being perfectly healthy. Twenty years later in 1995, Ziad Nasreddine, a neuroscientist living in Canada, created the Montreal Cognitive Assessment, or MoCA, a similarly short doctor’s-office test that doesn’t require as much facility with the English language.
These rather rudimentary screenings have been the standard for confirming cognitive health for over 20 years—making them as obsolete as DVDs. The MMSE asks questions like what year is it, where the taker is, and to count backwards from 100 seven at a time; MoCA asks people to identify animals and complete spatial reasoning tests. If a person can’t answer correctly, they almost certainly have lost some cognitive function. On the MMSE, a score of 27 indicates early stage dementia, and 26 or below indicates a more advanced stage. For MoCA, scores of 25 and below indicate some cognitive impairment, and 17 and below indicate dementia.
Theoretically, a drop by a couple of points on these tests could identify cases of early cognitive decline. But it could also easily miss them. Everyone has good days and bad days, says Nelson Roque, a cognitive psychologist at Penn State working on the M2C2 team. If a patient comes into a doctor’s office on a “bad” day one year and a “good” day another year, it could look like their cognition is improving—even if it’s actually getting worse on average.
M2C2 is trying to fix that problem. Its tests are more challenging from the beginning: This means that while a person may have a lower score at first when they’re learning how it works, they’ll improve as they practice over time.
It also makes the test more sensitive. Earlier studies have shown that tests that challenge associative memory—the skill you use to learn new names and faces, for example—can distinguish between people who have mild cognitive impairment and people who don’t. In the M2C2 app, the associative memory test uses colored dots: Three, each with a different color, appear on the screen. Later, you’re asked to give the color of a dot in a specific place. “What we’re interested in is if you say the color was yellow when it was red,” Roque says. That would mean that you have a harder time making associative memories.
The app format offers other advantages. It can test other potential risk factors for dementia, like certain personality traits, explains Stacey Scott, a psychologist at Stony Brook University in New York. Scott started as a post-doc at Penn State, working under Martin Sliwinski, who is now one of the leaders of the Mobile Toolbox projects. Back then, she worked on two major studies spearheaded at the Einstein Medical Center in the Bronx. In addition to tracking cognitive changes, her work measures the personality traits neuroticism and extraversion, which could be good indicators of someone’s stress level and social isolation.
Those who have higher levels of neuroticism are thought to be at higher risk of developing dementia, because they’ve likely experienced prolonged stress—a risk factor for dementia. Extraversion, on the other hand, isn’t necessarily protective against dementia, but it could indicate someone’s degree of social engagement and support systems—or lack thereof. In addition to measuring these changes at discrete moments, some of Scott’s work has measured them over time. She hypothesizes that changes in these measures of personality can indicate that someone may be going through early stages of cognitive decline, too.
So in addition to the cognitive assessment, an app could prompts participants to do a quick check-in on their mood, stress levels, and social engagement. If you were to ask an older adult at a single doctors’ visit about their personality, they’d have a lifetime of data about themselves to report back—meaning it’d be harder to notice changes over time. But if you can prompt them to ask about how they feel multiple times a day, you can track how often they actually report feeling stressed, and whether or not that’s changing.
Scott has continued working with the Einstein Medical Center in New York on the so-called Einstein Aging Study, which hopes to see if these kinds of mood-gauging questions can monitor personality changes over time. In her current work, a similar app pings participants to take a cognitive assessment. They can start with a few questions about their day so far: Did they experience anything stressful? Did they get a chance to see someone else that day, or have they been alone the whole time? She also has plans to work with Roque’s team on the M2C2 study. Together, the data they collect over time may be able to link some of these personality changes to cognitive decline.
At the moment, it’s still a new technology. Roque’s team is trying to get a sense of whether or not older adults would be open to taking this kind of test. “We’re learning about the appropriate way to deal with missing data in a cognitive assessment,” he says. Eighty-four assessments over two weeks is a lot; the team expects that some people will miss at least some tests because they’re busy. But that’s the trade-off for meeting people on a platform they routinely use instead of in a clinic: There’s less the scientists or clinician can actually control.
Ideally, though, this kind of assessment could feel as natural and unobtrusive as wearing an Apple Watch or Fitbit. Then, just like you count your steps and heart rate, you could see how your cognition is doing day to day. Some researchers, including Jukka-Pekka Onnela, a biostatistician over at Harvard, believe that this kind of mobile data collection is the key to public health. He and his team developed an open-source platform called Beiwe that can “digitally phenotype” users who download it from the app store—basically, using data from smartphones to see how you’re using them, and where. Currently, Onnela’s lab is licensing the technology for mental health research.
From early detection to better treatment
Early dementia detection is key for two main reasons. At the moment, early detection is the best way people with dementia and their families can begin the messy process of planning for later-stage care. And as treatments become available, it’ll be the key to administering them soon enough to stop dementia from progressing.
But long before M2C2 can be used to detect dementia, it’ll likely have to be validated with biological indicators that verify something in the brain has changed. At the moment, these biological indicators are also out of date—but researchers are working to improve them.
Positron emission tomography, or PET scans, can pick out the patterns of protein associated with Alzheimer’s. Spinal taps, also known as lumbar punctures, are another way to confirm a diagnosis. But they’re too invasive and expensive to justify using on healthy-seeming people. “They’re very hard to put into practice unless you have the ability and financial resources,” says Greg Jicha, a neuroscientist at the University of Kentucky.
It’s a catch-22: Insurance like Medicare will cover a spinal tap to confirm a diagnosis of Alzheimer’s or another dementia, and some of the costs of PET scans. But most people aren’t diagnosed unless a test like the MMSE or MoCA has already shown they have symptoms of cognitive decline. By that time, it’s too late for therapies to halt the progression of dementia. At the moment, the only people who get screening while they are still cognitively normal are those involved in clinical research—a tiny fraction of the population, Jicha explains.
A better option: Instead of these invasive tests, an early suggestion of cognitive decline could direct people to get an early diagnosis via a simple blood test, a strategy being investigated by research groups around the world.
When our brains are damaged or start to develop the pathologic buildups associated with Alzheimer’s, trace amounts of proteins show up in our blood. For years now, research teams led by Washington University at St. Louis in Missouri, the University of Beijing and the Chinese Academy of Sciences, and National Center for Geriatrics and Gerontology in Japan have all been working to find ways to measure changes in the ratio of two forms of amyloid in the blood that could be indicative of early MCI or Alzheimer’s disease.
So far, no blood test has proven robust enough to find definitive cutoffs for “healthy” and “unhealthy” brains. And amyloid is just one kind of protein, indicative only of Alzheimer’s disease. So the US National Institute on Aging (NIA) is looking at a wider range of biomarkers, including alpha-synuclein, a protein associated with Parkinson’s disease and Lewy Body dementias, and TDP-43, the protein associated with the newly-minted dementia LATE.
A better approach might be to find a blood test that can detect many forms of brain damage. Researchers at Washington University have studied the presence of a protein called neurofilament light chain, which seems to be given off any time that neurons have been damaged. There’s also work on a blood test for extracellular vesicles, little fatty pockets that several types of brain cells may use to swap biological information.
But there likely won’t be a single blood test for the job. “We have to be cognizant that there’s not just one biomarker in blood that defines the disease,” says Eliezer Masliah, the head of neuroscience at the NIA. “It’s a combination of biomarkers and the longitudinal trajectory, and we need to see their behaviors over time.”
This is where smartphone apps like M2C2 could come in handy. Roque acknowledges that a phone buzzing at you six times a day for 14 days is pretty intrusive—so soon, he hopes to identify the minimum cognitive data needed for the app to pick out the earliest stages of MCI.
Of course, smartphones can do more than just give a quiz—they can also track GPS data, and theoretically collect sound and even air quality. If scientists can track these variables, over time they could be used to understand how they impact a person’s risk for dementia. And smartphones may be just the tip of the iceberg.
Theoretically, monitoring any kind of technology we use on a regular basis could be a way to find changes in our behaviors over time. That’s the basis of ongoing work led by Lesley Ross, a psychologist at Penn State who studies healthy aging and function. Her research team is leading a pilot study to see if ability to drive can be used as be a proxy for changes in cognitive health. Right now, they’re working with about 20 healthy adults over 65, using in-car cameras to track how they drive and what else may be happening in the car over time, like passenger chatter or music playing.
It will one day be possible to get an overall picture of someone’s cognitive health at home, with tools they already know. It would require the sacrifice of a certain level of privacy, to be sure. But if early detection could be medically non-invasive, alerting your doctor to cognitive decline before you even notice it might just be worth the trade-off.
This article was written with the support of a journalism fellowship from the Gerontological Society of America, Journalists Network on Generations, and the John A. Hartford Foundation.