Machine learning model predicts health status

Carnegie Mellon University research has developed a model that can accurately predict how stay-at-home orders such as those placed during the COVID-19 pandemic affect the mental health of people with chronic neurological conditions such as multiple sclerosis.

– Advertising –

Researchers from Carnegie Mellon University, the University of Pittsburgh, and the University of Washington collected data from smartphones and fitness trackers of people with MS before and during the first wave of the pandemic. Plus précisément, ils ont utilisé les données de capteur collectées passivement pour créer des modèles d’apprentissage automatique afin de prédire la dépression, la fatigue, la mauvaise qualité du sommeil et l’aggravation des la symdeptômes précisé de la symdeptôme at home.

Before the pandemic began, an initial research question was whether digital data from smartphones and fitness trackers of people with MS could predict clinical outcomes. By March 2020, as study participants were asked to stay home, their daily behaviors had changed dramatically. The research team realized that the data collected could shed light on the impact of stay-at-home orders on people with MS.

“It gave us an exciting opportunity,” he said. Mayank Joelhead for Intelligent Human Sensing Laboratory (SMASH) at CMU. “If we look at data points before and during the stay-at-home period, can we identify factors that indicate changes in the health of people with MS?”

The team passively collected data for three to six months, collecting information such as the number of calls to participants’ smartphones and the duration of those calls; number of missed calls; and participant location and screen activity data. The team also collected information about heart rate, sleep and step count from their fitness trackers. The search was recently published in Internet Medical Research Journal of Mental Health. Joel, Associate Professor at the College of Computing Department of Software and Social Systems (S3D) and human-computer interaction institute (HCII), in collaboration with Prerna Chikersal, HCII PhD student; Dr. Zongqi Xia, Assistant Professor of Neurology and Director of the Translational and Computational Neuroimmunology Research Program at the University of Pittsburgh; And Anend Day, Professor and Dean of the University of Washington School of Mass Communication.

The work was based on previous studies by the Goel and Dey research groups. In 2020, a team from CMU published research stating that Introduce a machine learning model that can identify depression In students at the end of the semester using data from smartphones and fitness trackers. Participants in the previous study, specifically 138 freshmen at Carnegie Mellon University, were relatively similar to each other compared to the broader population outside of college. The researchers set out to test whether their modeling approach could accurately predict clinically relevant health outcomes in a real-world patient population with greater demographic and clinical diversity, leading them to collaborate with the Chia Disease Research Program in Multiple Sclerosis.

People with MS can have many chronic comorbidities, which allowed the team to test whether their model could predict adverse health effects such as extreme fatigue, poor sleep quality and worsening of MS symptoms as well as depression. Building on this study, the team hopes to advance precision medicine for people with MS by improving early detection of disease progression and implementing targeted digital phenotype-based interventions.

The work can also help inform policy makers tasked with issuing future stay-at-home orders or other similar responses during epidemics or natural disasters. When the initial stay-at-home orders for COVID-19 were issued, there were early concerns about its economic effects, but only a belated appreciation of the mental and physical health consequences for people — especially among vulnerable populations such as those with chronic neurological illnesses. diseases. .

“We were able to monitor people’s behavior change and accurately predict clinical outcomes when they had to stay home for long periods of time,” Joel said. “Now that we have a business model in place, we can assess people at risk of deteriorating mental health or physical health, inform clinical triage decisions, or shape future public health policy.”