[Also available in BibTeX] [See also: all keywords]
These papers relate to mobile health (mHealth), that is, the use of mobile computing and communications technology in the delivery of healthcare or collection of health information.Papers are listed in reverse-chronological order;
click an entry to pop up the abstract.
For full information and pdf, please click Details link.
Follow updates with RSS.
Methods: Study participants included adults receiving MOUD at a large outpatient treatment program. We predicted NPOU (EMA-based), medication nonadherence (Electronic Health Record [EHR]- and EMA-based), and treatment retention (EHR-based) using context-sensitive EMAs (e.g., stress, pain, social setting). We used recurrent deep learning models with 7-day sliding windows to predict the next-day outcomes, using Area Under the ROC Curve (AUC) for assessment. We employed SHapley additive ExPlanations (SHAP) to understand feature latency and importance.
Results: Participants comprised 62 adults with 14,322 observations. Model performance varied across EMA subtypes and outcomes with AUCs spanning 0.59-0.97. Recent substance use was the best performing predictor for EMA-based NPOU (AUC=0.97) and medication nonadherence (AUC=0.68); life-contextual factors performed best for EHR-based medication nonadherence (AUC=0.89) and retention (AUC=0.80). SHAP revealed varying latencies between predictors and outcomes.
Conclusions: Findings support the effectiveness of EMA and deep learning for forecasting actionable outcomes in persons receiving MOUD. These insights will enable the development of personalized dynamic risk profiles and just-in-time adaptive interventions (JITAIs) to mitigate high-risk OUD outcomes.
Objective: The aim is to examine patient engagement with multiple digital phenotyping methods among patients receiving buprenorphine medication for OUD.
Methods: The study enrolled 65 patients receiving buprenorphine for OUD between June 2020 and January 2021 from 4 addiction medicine programs in an integrated health care delivery system in Northern California. Ecological momentary assessment (EMA), sensor data, and social media data were collected by smartphone, smartwatch, and social media platforms over a 12-week period. Primary engagement outcomes were meeting measures of minimum phone carry (≥8 hours per day) and watch wear (≥18 hours per day) criteria, EMA response rates, social media consent rate, and data sparsity. Descriptive analyses, bivariate, and trend tests were performed.
Results: The participants’ average age was 37 years, 47% of them were female, and 71% of them were White. On average, participants met phone carrying criteria on 94% of study days, met watch wearing criteria on 74% of days, and wore the watch to sleep on 77% of days. The mean EMA response rate was 70%, declining from 83% to 56% from week 1 to week 12. Among participants with social media accounts, 88% of them consented to providing data; of them, 55% of Facebook, 54% of Instagram, and 57% of Twitter participants provided data. The amount of social media data available varied widely across participants. No differences by age, sex, race, or ethnicity were observed for any outcomes.
Conclusions: To our knowledge, this is the first study to capture these 3 digital data sources in this clinical population. Our findings demonstrate that patients receiving buprenorphine treatment for OUD had generally high engagement with multiple digital phenotyping data sources, but this was more limited for the social media data.
International Registered Report Identifier (IRRID): RR2-10.3389/fpsyt.2022.871916
Objective: We sought to develop and validate a smartwatch step-counting app for older adults and evaluate the algorithm in free-living settings over a long period of time.
Methods: We developed and evaluated a step-counting app for older adults on an open-source wrist-worn device (Amulet). The app includes algorithms to infer the level of physical activity and to count steps. We validated the step-counting algorithm in the lab (counting steps from a video recording, n=20) and in free-living conditions—one 2-day field study (n=6) and two 12-week field studies (using the Fitbit as ground truth, n=16). During app system development, we evaluated 4 walking patterns: normal, fast, up and down a staircase, and intermittent speed. For the field studies, we evaluated 5 different cut-off values for the algorithm, using correlation and error rate as the evaluation metrics.
Results: The step-counting algorithm performed well. In the lab study, for normal walking (R2=0.5), there was a stronger correlation between the Amulet steps and the video-validated steps; for all activities, the Amulet’s count was on average 3.2 (2.1%) steps lower (SD 25.9) than the video-validated count. For the 2-day field study, the best parameter settings led to an association between Amulet and Fitbit (R2=0.989) and 3.1% (SD 25.1) steps lower than Fitbit, respectively. For the 12-week field study, the best parameter setting led to an R2 value of 0.669.
Conclusions: Our findings demonstrate the importance of an iterative process in algorithm development before field-based deployment. This work highlights various challenges and insights involved in developing and validating monitoring systems in real-world settings. Nonetheless, our step-counting app for older adults had good performance relative to the ground truth (a commercial Fitbit step counter). Our app could potentially be used to help improve physical activity among older adults.
In this thesis, we present an end-to-end solution for providing information provenance for mHealth data, which begins by securing mHealth data at its source: the mHealth device. To this end, we devise a memory-isolation method that combines compiler-inserted code and Memory Protection Unit (MPU) hardware to protect application code and data on ultra-low-power micro-controllers. Then we address the security of mHealth data outside of the source (e.g., data that has been uploaded to smartphone or remote-server) with our health-data system, Amanuensis, which uses Blockchain and Trusted Execution Environment (TEE) technologies to provide confidential, yet verifiable, data storage and computation for mHealth data. Finally, we look at identity privacy and data freshness issues introduced by the use of blockchain and TEEs. Namely, we present a privacy-preserving solution for blockchain transactions, and a freshness solution for data access-control lists retrieved from the blockchain.
Methods: This paper describes the protocol (including the study design and methodological considerations) from a novel study supported by the National Drug Abuse Treatment Clinical Trials Network at the National Institute on Drug Abuse (NIDA). This study (D-TECT) primarily seeks to evaluate the feasibility of collecting ecological momentary assessment (EMA), smartphone and smartwatch sensor data, and social media data among patients in outpatient MOUD treatment. It secondarily seeks to examine the utility of EMA, digital sensing, and social media data (separately and compared to one another) in predicting MOUD treatment retention, opioid use events, and medication adherence [as captured in electronic health records (EHR) and EMA data]. To our knowledge, this is the first project to include all three sources of digitally derived data (EMA, digital sensing, and social media) in understanding the clinical trajectories of patients in MOUD treatment. These multiple data streams will allow us to understand the relative and combined utility of collecting digital data from these diverse data sources. The inclusion of EHR data allows us to focus on the utility of digital health data in predicting objectively measured clinical outcomes.
Discussion: Results may be useful in elucidating novel relations between digital data sources and OUD treatment outcomes. It may also inform approaches to enhancing outcomes measurement in clinical trials by allowing for the assessment of dynamic interactions between individuals' daily lives and their MOUD treatment response.
Clinical Trial Registration: Identifier: NCT04535583.
To address this problem, in this paper, we investigate the use of vibration, generated by a smartRing, as an out-of-band communication channel to unobtrusively share a secret with a smartThing. This exchanged secret can be used to bootstrap a secure wireless channel over which the smartphone (or another trusted device) and the smartThing can communicate. We present the design, implementation, and evaluation of this system, which we call VibeRing. We describe the hardware and software details of the smartThing and smartRing. Through a user study we demonstrate that it is possible to share a secret with various objects quickly, accurately and securely as compared to several existing techniques. Overall, we successfully exchange a secret between a smartRing and various smartThings, at least 85.9% of the time. We show that VibeRing can perform this exchange at 12.5 bits/second at a bit error rate of less than 2.5%. We also show that VibeRing is robust to the smartThing’s constituent material as well as the holding style. Finally, we demonstrate that a nearby adversary cannot decode or modify the message exchanged between the trusted devices.
We leveraged prior work regarding receptivity to JITAIs and deployed a chatbot-based digital coach - Ally - that provided physical-activity interventions and motivated participants to achieve their step goals. We extended the original Ally app to include two types of machine-learning model that used contextual information about a person to predict when a person is receptive: a static model that was built before the study started and remained constant for all participants and an adaptive model that continuously learned the receptivity of individual participants and updated itself as the study progressed. For comparison, we included a control model that sent intervention messages at random times. The app randomly selected a delivery model for each intervention message. We observed that the machine-learning models led up to a 40% improvement in receptivity as compared to the control model. Further, we evaluated the temporal dynamics of the different models and observed that receptivity to messages from the adaptive model increased over the course of the study.
Although past work on generic phone notifications has found evidence that users are more likely to respond to notifications with content they view as useful, there is no existing research on whether users' intrinsic motivation for the underlying topic of mHealth interventions affects their receptivity. In this work, we explore whether relationships exist between intrinsic motivation and receptivity across topics and within topics for mHealth interventions. To this end, we conducted a study with 20 participants over 3 weeks, where participants received interventions about mental health, COVID-19, physical activity, and diet & nutrition. The interventions were delivered by the chatbot-based iOS app called Elena+, and via the MobileCoach platform.
Our exploratory analysis found that significant differences in mean intrinsic motivation scores across topics were not associated with differences in mean receptivity metrics across topics. We also found that positive relationships exist between intrinsic motivation measures and receptivity for interventions about a topic.
First, we developed Auracle, a wearable earpiece that can automatically detect eating episodes. Using an off-the-shelf contact microphone placed behind the ear, Auracle captures the sound of a person chewing as it passes through the head. This audio data is then processed by a custom circuit board. We collected data with 14 participants for 32 hours in free-living conditions and achieved accuracy exceeding 92.8% and F1 score exceeding77.5% for eating detection with 1-minute resolution.
Second, we adapted Auracle for measuring children’s eating behavior, and improved the accuracy and robustness of the eating-activity detection algorithms. We used this improved prototype in a laboratory study with a sample of 10 children for 60 total sessions and collected 22.3 hours of data in both meal and snack scenarios. Overall, we achieved 95.5% accuracy and 95.7% F1 score for eating detection with 1-minute resolution.
Third, we developed a computer-vision approach for eating detection in free-living scenarios. Using a miniature head-mounted camera, we collected data with 10 participants for about 55 hours. The camera was fixed under the brim of a cap, pointing to the mouth of the wearer and continuously recording video (but not audio) throughout their normal daily activity. We evaluated performance for eating detection using four different Convolutional Neural Network (CNN) models. The best model achieved 90.9% accuracy and 78.7%F1 score for eating detection with 1-minute resolution. Finally, we validated the feasibility of deploying the 3D CNN model in wearable or mobile platforms when considering computation, memory, and power constraints.
Methods: There were 32 young adults participating in three exercise sessions with the exercise band, after which each completed an adapted version of the Usefulness, Satisfaction, and Ease (USE) questionnaire to characterize the exercise system’s strengths and weaknesses in usability.
Results: Questionnaire data reflected a positive and consistent user experience, with all 20 items receiving mean scores greater than 5.0 on a seven-point Likert scale. There were no specific areas of significant weakness in the device’s user experience.
Conclusions: The positive reception among young adults is a promising indication that the device can be successfully incorporated into exercise interventions and that the system can be further developed and tested for the target population of older adults.
To this end, we conducted a two-month longitudinal study with 10 participants, in which each participant was provided with a study car for their daily driving needs. We delivered two in-vehicle interventions - each aimed at improving affective well-being - and simultaneously recorded the participants' driving behavior. In our analysis, we found that several pre-trip characteristics (like trip length, traffic flow, and vehicle occupancy) and the pre-trip affective state of the participants had significant associations with whether the participants started an intervention or canceled a started intervention. Next, we found that several in-the-moment driving characteristics (like current road type, past average speed, and future brake behavior) showed significant associations with drivers' responsiveness to the intervention. Further, we identified several driving behaviors that "negated" the effectiveness of interventions and highlight the potential of using such "negative" driving characteristics to better inform intervention delivery. Finally, we compared trips with and without intervention and found that both interventions employed in our study did not have a negative effect on driving behavior. Based on our analyses, we provide solid recommendations on how to deliver interventions to maximize responsiveness and effectiveness and minimize the burden on the drivers.
Methods: A 6-month, non-randomized, non-blinded, single-arm study was conducted from October 2018 to May 2020 at a community-based aging center of adults aged ≥65 years with a body mass index (BMI) ≥30 kg/m2. Weekly dietitian visits focusing on behavior therapy and caloric restriction and twice-weekly physical therapist-led group strength, flexibility and balance training classes were delivered using video-conferencing to participants in their homes. Participants used a Fitbit Alta HR for remote monitoring with data feedback provided by the interventionists. An aerobic activity prescription was provided and monitored.
Results: Mean age was 72.9±3.9 years (82% female). Baseline anthropometric measures of weight, BMI, and waist circumference were 97.8±16.3 kg, 36.5±5.2 kg/m2, and 115.5±13.0 cm, respectively. A total of 142 participants were screened (n=27 ineligible), and 53 consented. There were nine dropouts (17%). Overall satisfaction with the trial (4.7+0.6, scale: 1 (low) to 5 (high)) and with Fitbit (4.2+0.9) were high. Fitbit was worn an average of 81.7±19.3% of intervention days. In completers, mean weight loss was 4.6±3.5 kg or 4.7±3.5% (p<0.001). Physical function measures of 30-s sit-to-stand repetitions increased from 13.5±5.7 to 16.7±5.9 (p<0.001), 6-min walk improved by 42.0±77.3 m (p=0.005) but no differences were observed in gait speed or grip strength. Subjective measures of late-life function improved (3.4±4.7 points, p<0.001).
Conclusions: A technology-based obesity intervention is feasible and acceptable to older adults with obesity and may lead to weight loss and improved physical function.
Methods: A 12-week pilot in 28 older rural adults with obesity (body mass index [BMI] ≥ 30 kg/m2) was conducted at a community aging center. The intervention consisted of individualized, weekly dietitian visits focusing on behavior therapy and caloric restriction with twice weekly physical therapist-led group strengthening training classes in a community-based aging center. All participants were provided a Fitbit Flex 2. An aerobic activity prescription outside the strength training classes was provided.
Results: Mean age was 72.9 ± 5.3 years (82% female). Baseline BMI was 37.1 kg/m2, and waist circumference was 120.0 ± 33.0 cm. Mean weight loss (pre/post) was 4.6 ± 3.2 kg (4.9 ± 3.4%; p < .001). Of the 40 eligible participants, 33 (75%) enrolled, and the completion rate was high (84.8%). Objective measures of physical function improved at follow-up: 6-minute walk test improved: 35.7 ± 41.2 m (p < .001); gait speed improved: 0.10 ± 0.24 m/s (p = .04); and five-times sit-to-stand improved by 2.1 seconds (p < .001). Subjective measures of late-life function improved (5.2 ± 7.1 points, p = .003), as did Patient-Reported Outcome Measurement Information Systems mental and physical health scores (5.0 ± 5.7 and 4.4 ± 5.0, both p < .001). Participants wore their Fitbit 93.9% of all intervention days, and were overall satisfied with the trial (4.5/5.0, 1–5 low–high) and with Fitbit (4.0/5.0).
Conclusions: A multicomponent obesity intervention incorporating a wearable device is feasible and acceptable to older adults with obesity, and potentially holds promise in enhancing health.
This paper takes the first step towards testing reproducibility and validity of methods and machine-learning models for stress detection. To this end, we analyzed data from 90 participants, from four independent controlled studies, using two different types of sensors, with different study protocols and research goals. We started by evaluating the performance of models built using data from one study and tested on data from other studies. Next, we evaluated new methods to improve the performance of stress-detection models and found that our methods led to a consistent increase in performance across all studies, irrespective of the device type, sensor type, or the type of stressor. Finally, we developed and evaluated a clustering approach to determine the stressed/not-stressed classification when applying models on data from different studies, and found that our approach performed better than selecting a threshold based on training data. This paper's thorough exploration of reproducibility in a controlled environment provides a critical foundation for deeper study of such methods, and is a prerequisite for tackling reproducibility in free-living conditions.
Objective: This study aims to develop a mobile app for a novel device through a user-centered design process with both older adults and clinicians while exploring whether data collected through this process can be used in NLP and sentiment analysis.
Methods: Through a user-centered design process, we conducted semistructured interviews during the development of a geriatric-friendly Bluetooth-connected resistance exercise band app. We interviewed patients and clinicians at weeks 0, 5, and 10 of the app development. Each semistructured interview consisted of heuristic evaluations, cognitive walkthroughs, and observations. We used the Bing sentiment library for a sentiment analysis of interview transcripts and then applied NLP-based latent Dirichlet allocation (LDA) topic modeling to identify differences and similarities in patient and clinician participant interviews. Sentiment was defined as the sum of positive and negative words (each word with a +1 or −1 value). To assess utility, we used quantitative assessment questionnaires—System Usability Scale (SUS) and Usefulness, Satisfaction, and Ease of use (USE). Finally, we used multivariate linear models—adjusting for age, sex, subject group (clinician vs patient), and development—to explore the association between sentiment analysis and SUS and USE outcomes.
Results: The mean age of the 22 participants was 68 (SD 14) years, and 17 (77%) were female. The overall mean SUS and USE scores were 66.4 (SD 13.6) and 41.3 (SD 15.2), respectively. Both patients and clinicians provided valuable insights into the needs of older adults when designing and building an app. The mean positive-negative sentiment per sentence was 0.19 (SD 0.21) and 0.47 (SD 0.21) for patient and clinician interviews, respectively. We found a positive association with positive sentiment in an interview and SUS score (ß=1.38; 95% CI 0.37 to 2.39; P=.01). There was no significant association between sentiment and the USE score. The LDA analysis found no overlap between patients and clinicians in the 8 identified topics.
Conclusions: Involving patients and clinicians allowed us to design and build an app that is user friendly for older adults while supporting compliance. This is the first analysis using NLP and usability questionnaires in the quantification of user-centered design of technology for older adults.
In the Internet of Things (IoT), everyday objects are equipped with the ability to compute and communicate. These smart things have invaded the lives of everyday people, being constantly carried or worn on our bodies, and entering into our homes, our healthcare, and beyond. This has given rise to wireless networks of smart, connected, always-on, personal things that are constantly around us, and have unfettered access to our most personal data as well as all of the other devices that we own and encounter throughout our day. It should, therefore, come as no surprise that our personal devices and data are frequent targets of ever-present threats. Securing these devices and networks, however, is challenging. In this dissertation, we outline three critical problems in the context of Wireless Personal Area Networks (WPANs) and present our solutions to these problems.
First, I present our Trusted I/O solution (BASTION-SGX) for protecting sensitive user data transferred between wirelessly connected (Bluetooth) devices. This work shows how in-transit data can be protected from privileged threats, such as a compromised OS, on commodity systems. I present insights into the Bluetooth architecture, Intel’s Software Guard Extensions (SGX), and how a Trusted I/O solution can be engineered on commodity devices equipped with SGX.
Second, I present our work on AMULET and how we successfully built a wearable health hub that can run multiple health applications, provide strong security properties, and operate on a single charge for weeks or even months at a time. I present the design and evaluation of our highly efficient event-driven programming model, the design of our low-power operating system, and developer tools for profiling ultra-low-power applications at compile time.
Third, I present a new approach (VIA) that helps devices at the center of WPANs (e.g., smartphones) to verify the authenticity of interactions with other devices. This work builds on past work in anomaly detection techniques and shows how these techniques can be applied to Bluetooth network traffic. Specifically, we show how to create normality models based on fine- and course-grained insights from network traffic, which can be used to verify the authenticity of future interactions.
Purpose: To evaluate the effects of incentives, weekly planning, and daily self-monitoring prompts that were used as intervention components as part of the Ally app.
Methods: We conducted an 8 week optimization trial with n = 274 insurees of a health insurance company in Switzerland. At baseline, participants were randomized to different incentive conditions (cash incentives vs. charity incentives vs. no incentives). Over the course of the study, participants were randomized weekly to different planning conditions (action planning vs. coping planning vs. no planning) and daily to receiving or not receiving a self-monitoring prompt. Primary outcome was the achievement of personalized daily step goals.
Results: Study participants were more active and healthier than the general Swiss population. Daily cash incentives increased step-goal achievement by 8.1%, 95% confidence interval (CI): [2.1, 14.1] and, only in the no-incentive control group, action planning increased step-goal achievement by 5.8%, 95% CI: [1.2, 10.4]. Charity incentives, self-monitoring prompts, and coping planning did not affect physical activity. Engagement with planning interventions and self-monitoring prompts was low and 30% of participants stopped using the app over the course of the study.
Conclusions: Daily cash incentives increased physical activity in the short term. Planning interventions and self-monitoring prompts require revision before they can be included in future versions of the app. Selection effects and engagement can be important challenges for physical-activity apps.
Clinical Trial Information: This study was registered on ClinicalTrials.gov, NCT03384550.
Objective: The objective of this study was to evaluate the automatic recognition and segmentation of nocturnal asthmatic coughs and cough epochs in smartphone-based audio recordings that were collected in the field. We also aimed to distinguish partner coughs from patient coughs in contact-free audio recordings by classifying coughs based on sex.
Methods: We used a convolutional neural network model that we had developed in previous work for automated cough recognition. We further used techniques (such as ensemble learning, minibatch balancing, and thresholding) to address the imbalance in the data set. We evaluated the classifier in a classification task and a segmentation task. The cough-recognition classifier served as the basis for the cough-segmentation classifier from continuous audio recordings. We compared automated cough and cough-epoch counts to human-annotated cough and cough-epoch counts. We employed Gaussian mixture models to build a classifier for cough and cough-epoch signals based on sex.
Results: We recorded audio data from 94 adults with asthma (overall: mean 43 years; SD 16 years; female: 54/94, 57%; male 40/94, 43%). Audio data were recorded by each participant in their everyday environment using a smartphone placed next to their bed; recordings were made over a period of 28 nights. Out of 704,697 sounds, we identified 30,304 sounds as coughs. A total of 26,166 coughs occurred without a 2-second pause between coughs, yielding 8238 cough epochs. The ensemble classifier performed well with a Matthews correlation coefficient of 92% in a pure classification task and achieved comparable cough counts to that of human annotators in the segmentation of coughing. The count difference between automated and human-annotated coughs was a mean –0.1 (95% CI –12.11, 11.91) coughs. The count difference between automated and human-annotated cough epochs was a mean 0.24 (95% CI –3.67, 4.15) cough epochs. The Gaussian mixture model cough epoch–based sex classification performed best yielding an accuracy of 83%.
Conclusions: Our study showed longitudinal nocturnal cough and cough-epoch recognition from nightly recorded smartphone-based audio from adults with asthma. The model distinguishes partner cough from patient cough in contact-free recordings by identifying cough and cough-epoch signals that correspond to the sex of the patient. This research represents a step towards enabling passive and scalable cough monitoring for adults with asthma.
Methods: We recruited patients from the Dartmouth-Hitchcock Weight & Wellness Center into a single-arm, non- randomised study of a remotely delivered 16-week evidence-based healthy lifestyle programme. Every 4 weeks, partic- ipants completed surveys that included their willingness to pay for services like those experienced in the intervention. A two-item Willingness-to-Pay survey was administered to participants asking about their willingness to trade their face- to-face visits for videoconference visits based on commute and copay.
Results: Overall, those with a travel duration of 31–45 min had a greater willingness to trade in-person visits for telehealth than any other group. Participants who had a travel duration less than 15 min, 16–30 min and 46–60 min experienced a positive trend in willingness to have telehealth visits until Week 8, where there was a general negative trend in willingness to trade in-person visits for virtual. Participants believed that telemedicine was useful and helpful.
Conclusions: In rural areas where patients travel 30–45 min a telemedicine-delivered, intensive weight-loss interven- tion may be a well-received and cost-effective way for both patients and the clinical care team to connect.
We define several metrics to gauge receptivity towards the interventions, and found that (1) several participant-specific characteristics (age, personality, and device type) show significant associations with the overall participant receptivity over the course of the study, and that (2) several contextual factors (day/time, phone battery, phone interaction, physical activity, and location), show significant associations with the participant receptivity, in-the-moment. Further, we explore the relationship between the effectiveness of the intervention and receptivity towards those interventions; based on our analyses, we speculate that being receptive to interventions helped participants achieve physical activity goals, which in turn motivated participants to be more receptive to future interventions. Finally, we build machine-learning models to detect receptivity, with up to a 77% increase in F1 score over a biased random classifier.
Methods: We conducted a convergent, parallel mixed-methods study using semi-structured interviews, focus groups, and self-reported questionnaires, using purposive sampling of 29 older adults, 4 community leaders and 7 clinicians in a rural setting. We developed codes informed by thematic analysis and assessed the quantitative data using descriptive statistics.
Results: All groups expressed that mHealth could improve health behaviors. Older adults were optimistic that mHealth could track health. Participants believed they could improve patient insight into health, motivating change and assuring accountability. Barriers to using technology were described, including infrastructure.
Conclusions: Older rural adults with obesity expressed excitement about the use of mHealth technologies to improve their health, yet barriers to implementation exist.
Objective: The primary objective of this study is to quantify main effects, interactions, and moderators of 3 intervention components of a smartphone-based intervention for physical activity. The secondary objective is the exploration of participants’ states of receptivity, that is, situations in which participants are more likely to react to intervention notifications through collection of smartphone sensor data.
Methods: In 2017, we developed the Assistant to Lift your Level of activitY (Ally), a chatbot-based mobile health intervention for increasing physical activity that utilizes incentives, planning, and self-monitoring prompts to help participants meet personalized step goals. We used a microrandomized trial design to meet the study objectives. Insurees of a large Swiss insurance company were invited to use the Ally app over a 12-day baseline and a 6-week intervention period. Upon enrollment, participants were randomly allocated to either a financial incentive, a charity incentive, or a no incentive condition. Over the course of the intervention period, participants were repeatedly randomized on a daily basis to either receive prompts that support self-monitoring or not and on a weekly basis to receive 1 of 2 planning interventions or no planning. Participants completed a Web-based questionnaire at baseline and postintervention follow-up.
Results: Data collection was completed in January 2018. In total, 274 insurees (mean age 41.73 years; 57.7% [158/274] female) enrolled in the study and installed the Ally app on their smartphones. Main reasons for declining participation were having an incompatible smartphone (37/191, 19.4%) and collection of sensor data (35/191, 18.3%). Step data are available for 227 (82.8%, 227/274) participants, and smartphone sensor data are available for 247 (90.1%, 247/274) participants.
Conclusions: This study describes the evidence-based development of a JITAI for increasing physical activity. If components prove to be efficacious, they will be included in a revised version of the app that offers scalable promotion of physical activity at low cost.
Trial Registration: ClinicalTrials.gov NCT03384550; https://clinicaltrials.gov/ct2/show/NCT03384550 (Archived by WebCite at http://www.webcitation.org/74IgCiK3d)
International Registered Report Identifier (IRRID): DERR1-10.2196/11540
The app implements an activity-level detection model we developed using a Linear Support Vector Machine (SVM). We trained our model using data from a user study, where subjects performed common physical activities (sit, stand, lay down, walk and run). We obtained accuracies up to 99.2% and 98.5% with 10-fold cross validation and leave-one-subject-out (LOSO) cross-validation respectively. We ran a week-long field study to evaluate the utility, usability and battery life of the ActivityAware system where 5 older adults wore the Amulet as it monitored their activity level. The utility evaluation showed that the app was somewhat useful in achieving the daily physical activity goal. The usability feedback showed that the ActivityAware system has the potential to be used by people for monitoring their activity levels. Our energy-efficiency evaluation revealed a battery life of at least 1 week before needing to recharge. The results are promising, indicating that the app may be used for activity-level monitoring by individuals or researchers for epidemiological studies, and eventually for the development of interventions that could improve the health of older adults.
We present and evaluate a prototype implementation to demonstrate this protocol’s feasibility on low-power wearable devices, and present a case for the system’s ability to meet critical security properties under a specific adversary model and trust assumptions.
We introduce the Amulet Platform for constrained wearable devices, which includes an ultra-low-power hardware architecture and a companion software framework, including a highly efficient event-driven programming model, low-power operating system, and developer tools for profiling ultra-low-power applications at compile time. We present the design and evaluation of our prototype Amulet hardware and software, and show how the framework enables developers to write energy-efficient applications. Our prototype has battery lifetime lasting weeks or even months, depending on the application, and our interactive resource-profiling tool predicts battery lifetime within 6-10% of the measured lifetime.
We built a user-friendly, mobile health-data collection system using wireless medical sensors that interface with an Android application. The data-collection system was designed to support minimally trained, non-clinical health workers to gather data about blood pressure and body weight using off-the-shelf medical sensors. This system comprises a blood-pressure cuff, a weighing scale and a portable point-of-sales printer. With this system, we introduced a new method to record contextual information associated with a blood-pressure reading using a tablet’s touchscreen and accelerometer. This contextual information can be used to verify that a patient’s lower arm remained well-supported and stationary during her blood-pressure measurement. In a preliminary user study, we found that a binary support vector machine classifier could be used to distinguish lower-arm movements from stationary arms with 90% accuracy. Predetermined thresholds for the accelerometer readings suffice to determine whether the tablet, and therefore the arm that rested on it, remained supported. Together, these two methods can allow mHealth applications to guide untrained patients (or health workers) in measuring blood pressure correctly.
Usability is a particularly important design and deployment challenge in remote, rural areas, given the limited resources for technology training and support. We conducted a field study to assess our system’s usability in Kolar town, India, where we logged health worker interactions with the app’s interface using an existing usability toolkit. Researchers analyzed logs from this toolkit to evaluate the app’s user experience and quantify specific usability challenges in the app. We have recorded experiential notes from the field study in this document.
Our recognition method uses bioimpedance, a measurement of how tissue responds when exposed to an electrical current. By collecting bioimpedance samples using a small wearable device we designed, our system can determine that (a)the wearer is indeed the expected person and (b) the device is physically on the wearer’s body. Our recognition method works with 98% balanced-accuracy under a cross-validation of a day’s worth of bioimpedance samples from a cohort of 8 volunteer subjects. We also demonstrate that our system continues to recognize a subset of these subjects even several months later. Finally, we measure the energy requirements of our system as implemented on a Nexus S smart phone and custom-designed module for the Shimmer sensing platform.
We address this problem of balancing disclosure and utility of personal information collected by mobile technologies. We believe subjects can decide how best to share their information if they are aware of the benefits and risks of sharing. We developed ShareBuddy, a privacy-aware architecture that allows recipients to request information and specify the benefits the subjects will receive for sharing each piece of requested information; the architecture displays these benefits and warns subjects about the risks of sharing. We describe the ShareBuddy architecture in this poster.
In this thesis we describe solutions to two of these problems. First, we evaluate the use of bioimpedance for recognizing who is wearing these wireless sensors and show that bioimpedance is a feasible biometric. Second, we investigate the use of accelerometers for verifying whether two of these wireless sensors are on the same person and show that our method is successful as distinguishing between sensors on the same body and on different bodies. We stress that any solution to these problems must be usable, meaning the user should not have to do anything but attach the sensor to their body and have them just work.
These methods solve interesting problems in their own right, but it is the combination of these methods that shows their true power. Combined together they allow a network of wireless sensors to cooperate and determine whom they are sensing even though only one of the wireless sensors might be able to determine this fact. If all the wireless sensors know they are on the same body as each other and one of them knows which person it is on, then they can each exploit the transitive relationship to know that they must all be on that person’s body. We show how these methods can work together in a prototype system. This ability to operate unobtrusively, collecting in situ data and labeling it properly without interrupting the wearer’s activities of daily life, will be vital to the success of these wireless sensors.
In this paper, we describe Plug-n-Trust (PnT), a novel approach to protecting both the confidentiality and integrity of safety-critical medical sensing and data processing on vulnerable mobile phones. With PnT, a plug-in smart card provides a trusted computing environment, keeping data safe even on a compromised mobile phone. By design, PnT is simple to use and deploy, while providing a flexible programming interface amenable to a wide range of applications. We describe our implementation, designed for Java-based smart cards and Android phones, in which we use a split-computation model with a novel path hashing technique to verify proper behavior without exposing confidential data. Our experimental evaluation demonstrates that PnT achieves its security goals while incurring acceptable overhead.
We conducted focus groups to understand the privacy concerns that patients have when they use mHealth devices. We conducted a user study to understand how willing patients are to share their personal health information that was collected using an mHealth device. To the best of our knowledge, ours is the first study that explores users’ privacy concerns by giving them the opportunity to actually share the information collected about them using mHealth devices. We found that patients tend to share more information with third parties than the public and prefer to keep certain information from their family and friends. Finally, based on these discoveries, we propose some guidelines to developing defaults for sharing settings in mHealth systems.
We make three contributions. First, we propose Adapt-lite, a set of two techniques that can be applied to existing wireless protocols to make them energy efficient without compromising their security or privacy properties. The techniques are: adaptive security, which dynamically modifies packet overhead; and MAC striping, which makes forgery difficult even for small-sized MACs. Second, we apply these techniques to an existing wireless protocol, and demonstrate a prototype on a Chronos wrist device. Third, we provide security, privacy, and energy analysis of our techniques.
We make three contributions. First, we propose an mHealth sensing protocol that provides strong security and privacy properties with low energy overhead, suitable for low-power sensors. The protocol uses three novel techniques: adaptive security, to dynamically modify transmission overhead; MAC striping, to make forgery difficult even for small-sized MACs; and an asymmetric resource requirement. Second, we demonstrate a prototype on a Chronos wrist device, and evaluate it experimentally. Third, we provide a security, privacy, and energy analysis of our system.
We provide a method to probabilistically detect this situation. Because accelerometers are relatively cheap and require little power, we imagine that the cellphone and each sensor will have a companion accelerometer embedded with the sensor itself. We extract standard features from these companion accelerometers, and use a pair-wise statistic -- coherence, a measurement of how well two signals are related in the frequency domain -- to determine how well features correlate for different locations on the body. We then use these feature coherences to train a classifier to recognize whether a pair of sensors -- or a sensor and a cellphone -- are on the same body. We evaluate our method over a dataset of several individuals walking around with sensors in various positions on their body and experimentally show that our method is capable of achieving an accuracies over 80%.
This poster describes a simple, flexible, and novel approach to protecting both the confidentiality and integrity medical sensing and data processing on vulnerable mobile phones, using plug-in smart cards---even a phone compromised by malware. We describe our design, implementation, and initial experimental results using real smart cards and Android smartphones.
In this position paper, we propose Mobile-phone based Patient Compliance System (MPCS) that can reduce the time-consuming and error-prone processes of existing self-regulation practice to facilitate self-reporting, non-compliance detection, and compliance reminders. The novelty of this work is to apply social-behavior theories to engineer the MPCS to positively influence patients’ compliance behaviors, including mobile-delivered contextual reminders based on association theory; mobile-triggered questionnaires based on self-perception theory; and mobile-enabled social interactions based on social-construction theory. We discuss the architecture and the research challenges to realize the proposed MPCS.