John P. Pollak

It's Time to Eat! Using Mobile Games to Promote Healthy Eating

Its never been more important to teach youth the importance of healthy eating habits. Time to Eat, a mobile-phone-based game, motivates children to practice healthy eating habits by letting them care for a virtual pet. Players send the pet photos of the food they consume throughout the day; the foods healthiness determines the games outcome. An examination of the games design provides insight into the potential of deploying health games on mobile phones.

PAM: a photographic affect meter for frequent, in situ measurement of affect

The assessment of emotion, or affect, is critical for anyone trying to understand human behavior. But there is a problem: affect as a state is frequently changing and difficult to recall and express, yet in research, we typically only assess it via a single questionnaire at the end of a study. This work presents PAM, the Photographic Affect Meter, a novel tool for measuring affect in which users select from a wide variety of photos the one which best suits their current mood. Our findings indicate that PAM-which takes seconds to complete and is designed to run on modern mobile phones and mobile computing devices-demonstrates strong construct validity across two studies and is very well suited for frequent sampling in context. This work provides a tool to researchers in need of frequent assessment of affect and guidance to others interested in developing similar measurement tools.

Metamodels for Transdisciplinary Analysis of Wildlife Population Dynamics

Wildlife population models have been criticized for their narrow disciplinary perspective when analyzing complexity in coupled biological – physical – human systems. We describe a “metamodel” approach to species risk assessment when diverse threats act at different spatiotemporal scales, interact in non-linear ways, and are addressed by distinct disciplines. A metamodel links discrete, individual models that depict components of a complex system, governing the flow of information among models and the sequence of simulated events. Each model simulates processes specific to its disciplinary realm while being informed of changes in other metamodel components by accessing common descriptors of the system, populations, and individuals. Interactions among models are revealed as emergent properties of the system. We introduce a new metamodel platform, both to further explain key elements of the metamodel approach and as an example that we hope will facilitate the development of other platforms for implementing metamodels in population biology, species risk assessments, and conservation planning. We present two examples – one exploring the interactions of dispersal in metapopulations and the spread of infectious disease, the other examining predator-prey dynamics – to illustrate how metamodels can reveal complex processes and unexpected patterns when population dynamics are linked to additional extrinsic factors. Metamodels provide a flexible, extensible method for expanding population viability analyses beyond models of isolated population demographics into more complete representations of the external and intrinsic threats that must be understood and managed for species conservation.

Yum-Me: A Personalized Nutrient-Based Meal Recommender System

Nutrient-based meal recommendations have the potential to help individuals prevent or manage conditions such as diabetes and obesity. However, learning people’s food preferences and making recommendations that simultaneously appeal to their palate and satisfy nutritional expectations are challenging. Existing approaches either only learn high-level preferences or require a prolonged learning period. We propose Yum-me, a personalized nutrient-based meal recommender system designed to meet individuals’ nutritional expectations, dietary restrictions, and fine-grained food preferences. Yum-me enables a simple and accurate food preference profiling procedure via a visual quiz-based user interface and projects the learned profile into the domain of nutritionally appropriate food options to find ones that will appeal to the user. We present the design and implementation of Yum-me and further describe and evaluate two innovative contributions. The first contriution is an open source state-of-the-art food image analysis model, named FoodDist. We demonstrate FoodDist’s superior performance through careful benchmarking and discuss its applicability across a wide array of dietary applications. The second contribution is a novel online learning framework that learns food preference from itemwise and pairwise image comparisons. We evaluate the framework in a field study of 227 anonymous users and demonstrate that it outperforms other baselines by a significant margin. We further conducted an end-to-end validation of the feasibility and effectiveness of Yum-me through a 60-person user study, in which Yum-me improves the recommendation acceptance rate by 42.63%.

Leveraging Multi-Modal Sensing for Mobile Health: A Case Review in Chronic Pain

Active and passive mobile sensing has garnered much attention in recent years. In this paper, we focus on chronic pain measurement and management as a case application to exemplify the state of the art. We present a consolidated discussion on the leveraging of various sensing modalities along with modular server-side and on-device architectures required for this task. Modalities included are: activity monitoring from accelerometry and location sensing, audio analysis of speech, image processing for facial expressions as well as modern methods for effective patient self-reporting. We review examples that deliver actionable information to clinicians and patients while addressing privacy, usability, and computational constraints. We also discuss open challenges in the higher level inferencing of patient state and effective feedback with potential directions to address them. The methods and challenges presented here are also generalizable and relevant to a broad range of other applications in mobile sensing.

Designing for Engaging Experiences in Mobile Social-Health Support Systems

How do you design for sustained engagement in the context of supporting or encouraging health and well-being? Two studies of a mobile social-health application reveal how different aspects of the sociotechnical design affect user engagement with the system.

An mHealth App for Self-Management of Chronic Lower Back Pain (Limbr): Pilot Study

Background Although mobile health (mHealth) interventions can help improve outcomes among patients with chronic lower back pain (CLBP), many available mHealth apps offer content that is not evidence based. Limbr was designed to enhance self-management of CLBP by packaging self-directed rehabilitation tutorial videos, visual self-report tools, remote health coach support, and activity tracking into a suite of mobile phone apps, including Your Activities of Daily Living, an image-based tool for quantifying pain-related disability. Objective The aim is to (1) describe patient engagement with the Limbr program, (2) describe patient-perceived utility of the Limbr program, and (3) assess the validity of the Your Activities of Daily Living module for quantifying functional status among patients with CLBP. Methods This was a single-arm trial utilizing a convenience sample of 93 adult patients with discogenic back pain who visited a single physiatrist from January 2016 to February 2017. Eligible patients were enrolled in 3-month physical therapy program and received the Limbr mobile phone app suite for iOS or Android. The program included three daily visual self-reports to assess pain, activity level, and medication/coping mechanisms; rehabilitation video tutorials; passive activity-level measurement; and chat-based health coaching. Patient characteristics, patient engagement, and perceived utility were analyzed descriptively. Associations between participant characteristics and program interaction were analyzed using multiple linear regression. Associations between Your Activities of Daily Living and Oswestry Disability Index (ODI) assessments were examined using Pearson correlation and hierarchical linear modeling. Results A total of 93 participants were enrolled; of these, 35 (38%) completed the program (age: mean 46, SD 16 years; female: 22/35, 63%). More than half of completers finished assessments at least every 3 days and 70% (19/27) used the rehabilitation component at least once a week. Among respondents to a Web-based feedback survey, 76% (16/21) found the daily notifications helped them remember to complete their exercises, 81% (17/21) found the system easy to use, and 62% (13/21) rated their overall experience good or excellent. Baseline Your Activities of Daily Living score was a significant predictor of baseline ODI score, with ODI increasing by 0.30 units for every 1-unit increase in Your Activities of Daily Living (P<.001). Similarly, hierarchical linear modeling analysis indicated that Your Activities of Daily Living daily assessment scores were significant predictors of ODI scores over the course of the study (P=.01). Conclusions Engagement among participants who completed the Limbr program was high, and program utility was rated positively by most respondents. Your Activities of Daily Living was significantly associated with ODI scores, supporting the validity of this novel tool. Future studies should assess the effect of Limbr on clinical outcomes, evaluate its use among a wider patient sample, and explore strategies for reducing attrition. Trial Registration ClinicalTrials.gov NCT03040310; https://clinicaltrials.gov/ct2/show/NCT03040310 (Archived by WebCite at http://www.webcitation.org/722mEvAiv)

Panel: Designing Studies for Feasibility Testing, Refinement and Validation of Digital Biomarkers

One of the most critical steps for digital biomarker research is the study design for feasibility testing, refinement, and validation of digital biomarkers. Any small to large-scale research studies/other practical explorations of various digital biomarkers requires the investigators to decide various study design parameters, such as characteristics of subject pool, recruitment technique, type of digital tools, the length of the study, the size of the dataset, modeling techniques and evaluation metric etc. Depending on the research question, these design parameters can significantly affect results and research outcomes. The focus of this panel will bring together researchers from academia, industry and medical sciences to facilitate a lively and stimulating discussion about various digital biomarker related user study design challenges and solutions. In this panel, we will have experts focusing on diverse digital biomarker related challenges including modeling psychological health with social media data, health sensing and intervention design with a mobile phone, and health tracking tool development with smartwatch etc.