Summary: A new multitasking AI model based on data from wearable devices predicts treatment outcomes on an individual basis for people with depression.
Over the past several years, managing an individual’s mental health has become more of a priority with an increased focus on self-care. Depression alone affects more than 300 million people worldwide annually.
Recognizing this, there is great interest in making use of popular wearables to monitor an individual’s mental health by measuring markers such as activity levels, sleep and heart rate.
A team of researchers at Washington University in St. Louis and the University of Illinois at Chicago used data from wearable devices to predict treatment outcomes for depression in individuals who participated in a randomized clinical trial.
They developed a new machine-learning model that analyzes data from two groups of patients — those who were randomly selected to receive treatment and those who did not — rather than developing a separate model for each group.
This unified, multitasking model is a step towards personalized medicine, in which clinicians design a treatment plan specific to each patient’s needs and predict outcomes based on individual data.
The search results are published in ACM facts on interactive, modular, wearable, and ubiquitous technologies It will be presented at the UbiComp 2022 conference in September.
Chenyang Lu, a Fullgraf Professor at the McKelvey School of Engineering, led a team including Ruixuan Dai, who worked in Lu’s lab as a doctoral student and is now a software engineer at Google; Thomas Canampalle, Associate Professor of Anesthesiology and Associate Principal Research Information Officer in the School of Medicine and Associate Professor of Computer Science and Engineering at McKelvey Engineering; and John Ma, MD, PhD, professor of medicine at the University of Illinois Chicago (UIC); and their colleagues developed the model using data from a UIC randomized clinical trial with nearly 100 adults with depression and obesity.
“Integrated behavioral therapy can be expensive and time consuming,” Lu said.
“If we can make personal predictions for individuals about whether a patient is likely to respond to a particular treatment, patients might continue treatment only if the model predicts that their conditions are likely to improve with treatment but less likely without treatment. Such personalized predictions of treatment response would facilitate A more targeted and cost-effective treatment.”
In the trial, patients were given Fitbit bracelets and psychological tests. About two-thirds of the patients received behavioral therapy, and the rest of the patients did not. Patients in both groups were statistically similar at baseline, giving the researchers a level playing field with which to discern whether treatment would produce better outcomes based on individual data.
Clinical trials of behavioral therapies have often included relatively small groups due to the cost and duration of these interventions. The small number of patients created a challenge for the machine learning model, which usually performs better with more data.
However, by combining the data of the two groups, the model could learn from a larger data set, which recorded the differences between those who underwent treatment and those who did not. They found that their multitasking model predicted depression outcomes better than the model that looked at each group individually.
“We devised a multitasking framework that combines an intervention group and a control group in a jointly randomized control trial to train a standardized model to predict subjective outcomes for an individual with and without treatment,” said Day, who has a PhD in computer science. science in 2022.
The model integrates clinical characteristics and wearable data into a multi-layered architecture. This approach avoids dividing study groups into smaller groups for machine learning models and enables dynamic knowledge transfer between groups to improve prediction performance both with and without intervention. “
“The implications of this data-driven approach extend beyond randomized clinical trials to implementation in clinical care delivery, where the ability to predict patient outcomes depending on treatment received, and to do so early and along the course of treatment, can be meaningful. inform joint decision-making by the patient and the treating physician in order to tailor the treatment plan for that patient,” Ma said.
The machine learning approach provides a promising tool for building personalized predictive models based on data collected from randomized controlled trials.
Going forward, the team plans to leverage a machine learning approach in a new randomized controlled trial of behavioral interventions for telehealth using Fitbit bracelets and weight scales among patients in the weight loss intervention study.
About this neurotechnology and depression research news
author: Brandy Jefferson
Contact: Brandi Jefferson – WUSTL
picture: The image is in the public domain
original search: open access.
“Multitasking learning for randomized controlled trials: a case study on depression prediction using wearable dataWritten by Chenyang Lu et al. ACM Actions on Interactive Wearable Technologies Available Everywhere on Mobile Devices
Multitasking learning for randomized controlled trials: a case study on depression prediction using wearable data
A randomized controlled trial (RCT) is used to study the safety and efficacy of new treatments, by comparing patient outcomes for the intervention group with the control group. Traditionally, randomized controlled trials rely on statistical analyzes to assess differences between treatment and control groups.
However, such statistical analyzes are generally not designed to assess the impact of the intervention at the individual level. In this paper, we explore machine learning models combined with RCT for personalized predictions of depression treatment intervention, in which patients were monitored longitudinally using wearable devices.
We formulate individual-level predictions in the intervention and control groups from RCT as a multi-task learning (MTL) problem, and propose a new MTL model designed specifically for randomized controlled trials. Instead of training separate models for the intervention and control groups, the proposed MTL model is trained on both groups, effectively expanding the training data set.
We develop a hierarchical model architecture to aggregate data from different sources and different longitudinal stages of the experiment, allowing the MTL model to exploit commonalities and capture the differences between the two groups. We evaluated the MTL approach in a randomized controlled trial of 106 depressed patients, who were randomly assigned to receive an integrated intervention treatment.
Our proposed MTL model outperforms both the single-task models and the traditional multi-task model in predictive performance, representing a promising step in using data collected in randomized controlled trials to develop predictive models for precision medicine.