Multitask Learning Model Outperforms Traditional Machine Learning Models in RCTs, Developers Say

The proposed multitask studying (MTL) mannequin was developed utilizing knowledge from wearable gadgets worn by people in a randomized managed trial (RCT) to foretell outcomes of a melancholy remedy.Researchers developed a novel multitask studying (MTL) mannequin designed particularly for randomized managed trials (RCTs), which analyzes knowledge from each the intervention and management affected person teams in RCTs and eliminating the necessity to develop a separate mannequin for every set of sufferers.According to the researchers, this proposed MTL mannequin outperforms the standard and single-task machine studying fashions in predictive efficiency.These findings have been revealed in the Proceedings of the ACM on Interactive, Model, Wearable and Ubiquitous Technologies, and might be offered on the UbiComp 2022 convention in September.RCTs historically depend on statistical analyses to guage variations between remedy and management teams in research, however typically fail to measure the remedy’s impression at a person stage.According to the authors, individualized predictions might facilitate extra focused and cost-effective remedy.“Integrated behavioral remedy could be costly and time consuming,” Chenyang Lu, PhD, professor on the McKelvey School of Engineering at Washington University in St. Louis and creator on the examine, mentioned in a information launch. “If we will make customized predictions for people on whether or not it’s doubtless a affected person could be attentive to a selected remedy, then sufferers might proceed with remedy provided that the mannequin predicts their situations are doubtless to enhance with remedy however much less doubtless with out remedy.”To develop the proposed MTL mannequin, the authors collected knowledge from wearable gadgets worn by people from an RCT in order to foretell outcomes of a melancholy remedy.In the RCT, 106 members with melancholy wore Fitbit gadgets and acquired psychological testing. About two-thirds additionally acquired behavioral remedy, whereas the opposite third didn’t. Patients have been statistically related at baseline so the authors may higher assess whether or not behavioral remedy would result in improved outcomes based mostly on particular person knowledge.The proposed mannequin built-in medical traits and knowledge collected by the gadgets in a multilayer structure, and enabled a information switch between the intervention and management teams to optimize prediction efficiency for each affected person teams.According to the authors, this proposed MTL mannequin marks a promising growth and necessary step in customized drugs.“The implications of this data-driven method prolong past randomized medical trials to implementation in medical care supply, the place the power to make customized prediction of affected person outcomes relying on the remedy acquired, and to take action early and alongside the remedy course, may meaningfully inform shared-decision making by the affected person and the treating doctor in order to tailor the remedy plan for that affected person,” mentioned Jun Ma, MD, PhD, professor of drugs on the University of Illinois Chicago and creator on the examine.To additional take a look at this proposed MTL mannequin, the authors plan to use the mannequin in an identical RCT on telehealth behavioral interventions utilizing Fitbit wristbands and weight scales in sufferers in a weight reduction intervention examine.ReferenceDai R, Kannampallil T, Zhang J, Lv N, Ma J, Lu C. Multi-task studying for randomized managed trials: a case examine on predicting melancholy with wearable knowledge. Proc. ACM Interact. Mob. Wearable Ubiquitous Technol. Published on-line July 7, 2022. doi:10.1145/3534591

https://www.ajmc.com/view/multitask-learning-model-outperforms-traditional-machine-learning-models-in-rcts-developers-say

Recommended For You