WASH Study

This project is a smartphone sensor data analytics project that deals with a full set of heterogeneous datasets to be correlated with mood analysis and mental health. The acquired data includes periodically-prompted user survey entries, light sensor output, Bluetooth and WiFi status, GPS data, and multivariate IMU time series e.g. 3-axis accelerometer and 3-axis gyroscope output at simple and complex activity contexts.

The Warfighter Analytics using Smart Phones for Health (WASH) program recruited 25,000 participants across the United States to download and run a newly designed cell phone app named Health & Injury Prediction and Prevention Over Complex Reasoning and Analytic Techniques Integrated on a Cell Phone App (HIPPOCRATIC App) (Android and iOS) on their personal devices and allow specific and partially-prompted data acquisition that ranges from their answers to survey question to sensor data recording at various activity contexts.

The effect of mood and mental on human activity, whether it is smartphone use activity or actual physical activity, has been an active research track for decades. Human activity modeling and recognition may be performed using wearable sensors and/or smartphone sensors, and our study deals with the latter.

This research was inspired by the fact that almost everyone nowadays has a smartphone that they are constantly using, which makes accessing valuable data to analyze and improve people's health an easy-to-achieve task. With the multidisciplinary expertise in our team, we are developing and deploying state-of-the-art data analytics and machine learning pipelines to provide tools for better mental health diagnostics. 

This project aims to make use of our smart devices to capture essential health-related data, analyze it, and provide the community with open-source tools for mental health diagnostics in specific, and human health in general.   

Additional analysis will be done to report compliance and retention rates to describe the quantity of survey data collected and to provide an indication of the quality of the data collection procedures as well as to examine factors that may influence compliance and retention rates. 

This project will set standards in smartphone sensor big data processing and analysis, and it will provide analytical tools for the correlation of human activity with mood and mental health.

In this era where all mankind is suffering from stress, anxiety, and mental instability, the introduction of such tools and analyses for the improvement of mental health is of great value to human beings in general, and more particularly for patients suffering from mental illnesses and their medical caregivers.

Once developed and implemented, the tools and analytical pipelines generated by this project may be embedded on edge-devices (e.g. smartphones, smart watches, microcontroller-based systems) for online mental health assessment, or linked to the healthcare facility's network for advanced EHR, or even on the cloud so that medical providers can monitor and diagnose treatment effectiveness over the web.

The influence of participant behavior and demographic-related factors on the compliance and retention rates of mobile technologies enabled digital health studies remains largely unexplored in the research literature. The compliance study portion of this project fills this gap to provide empirically-based strategies that could support researchers’ choices in designing data collection protocols for future studies.

This project will be carried out as a series of processes: ETL → EDA → QA/QC → ML/AI model engineering → MLOps pipeline integration 

To achieve the desired outcomes, advanced unsupervised analysis and anomaly detection algorithms will be implemented for QA/QC, machine learning (shallow and deep) models will be built, tested and tuned for optimal performance, and operation pipelines will be developed for system automation, optimization, and integration.

The compliance study will examine cohort characteristics and its association with compliance and retention rates. Metrics to assess retention rates may include duration in the study, longitudinal participant engagement (user activity streak), and number of times participants reached out to the research team for tech/payment/survey-related issues (data stored in REDCap). Metrics to assess compliance may include number of surveys completed over the total number of available surveys that could be completed by participants’ unique study time and number of Audio and simple & complex activity context survey completed. 

1. Indicators of retention in remote digital health studies: a cross-study evaluation of 100,000 participants (retention study)
   1. https://github.com/Sage-Bionetworks/digitalHealth_RetentionAnalysis_PublicRelease (code associated with the above study)
   2. https://github.com/Sage-Bionetworks/digitalHealth_RetentionAnalysis_PublicRelease/blob/master/analysis/timeToEvent_TRUE.Right.CensoringApproach.R (code associated with the above study)
2. An Alternative to the Light Touch Digital Health Remote Study: The Stress and Recovery in Frontline COVID-19 Health Care Workers Study (compliance study)

https://healthstudy.tozny.com/help

(Anything that you think is relevant that doesn’t fit under any of the sections above.)

video1086320079.mp4