Harnessing social support to empower individuals with spinal cord injury to increase physical activity levels: Observations from feasibility studies

Journal of Rehabilitation and Assistive Technologies Engineering · 2025-07-01

Introduction: We report findings from two feasibility studies with persons with SCI that aimed to leverage social support to increase PA: a two-week study with nine participants and a four-week study with six participants. Method: We recruited a total of 17 participants across two phases (10 participants in Phase 1 and seven participants in Phase 2). In dyads, participants used a smartphone-smartwatch application that we iteratively developed based on participant feedback. The application delivered just-in-time support based on measured PA and encouraged reciprocal self-disclosure to increase closeness within dyads. Results: Participants found the delivery of messages during detected PA to be motivational. Some liked automatically sharing tracked PA with others and did not have privacy concerns about doing so, and most preferred exchanging real-time messages more than context- or activity-triggered messages. Participants also expressed that feeling connected to their partner increased motivation to engage in PA. Conclusion: Participants expressed that they liked being able to connect with individuals with shared life experience to exchange encouraging messages. There are, however, challenges that need to be addressed before a large-scale deployment of this technology, including user concerns about automatically detected activity.

Understanding Longitudinal Ecological Momentary Assessment Completion: Results From 12 Months of Burst Sampling in the TIME Study

JMIR mhealth and uhealth · 2025-10-22 · 2 citations

BACKGROUND: Ecological momentary assessment (EMA) is a valuable method for capturing real-time data on behaviors and experiences in naturalistic settings. However, maintaining participant engagement in longitudinal (ie, multiburst) EMA studies remains challenging, particularly when collecting intensive data over extended periods. Understanding factors affecting completion rates is essential for designing more effective EMA protocols and interpreting results accurately. OBJECTIVE: This study investigated factors influencing EMA completion rates in a 12-month intensive longitudinal study among young adults in the United States, examining both time-varying factors and stable individual characteristics. METHODS: Young adults (N=246, ages 18-29 years) participated in the Temporal Influences on Movement and Exercise (TIME) study, responding to smartphone-based EMA prompts during biweekly measurement bursts (4-day periods of intensive sampling), with continuous passive data collection via smartwatches. Each burst included signal-contingent prompts delivered approximately once per hour during waking hours, resulting in an average of 12.1 (SD 1.3) prompts per day. Multilevel logistic regression models examined the effects of time-varying temporal factors (time of day, day of week, season, and time in study), contextual factors (phone screen status, phone usage, and location), behavioral factors (sleep duration, physical activity levels, and travel status), and psychological factors (momentary affect and stress) on prompt completion. Models also included time-invariant demographic characteristics (sex, race, ethnicity, education, and employment) and tested interactions between time in study and other predictors. RESULTS: Mean completion rate was 77% (SD 13%). Hispanic participants showed lower odds of completion compared to non-Hispanic participants (odds ratio [OR] 0.79, 95% CI 0.63-0.99; P=.04) and employed participants were less likely to complete prompts than unemployed participants (OR 0.75, 95% CI 0.61-0.92; P<.01). Having the phone screen on at prompt delivery substantially increased completion odds (OR 3.39, 95% CI 2.81-4.09; P<.001), while being away from home reduced completion likelihood, with particularly low odds when at sports facilities (OR 0.58, 95% CI 0.47-0.74; P<.001) or restaurants and shops (OR 0.61, 95% CI 0.51-0.72; P<.001). Short sleep duration the previous night (OR 0.92, 95% CI 0.87-0.99; P=.02) and traveling status (OR 0.78, 95% CI 0.75-0.82; P<.001) were associated with lower completion odds. Higher momentary stress levels predicted lower completion of subsequent prompts (OR 0.85, 95% CI 0.78-0.93; P<.001). Completion odds declined over the 12-month study (OR 0.95, 95% CI 0.94-0.96; P<.001), with significant interactions between time in study and various predictors, indicating changing patterns of engagement over time. CONCLUSIONS: Findings highlight the dynamic nature of EMA engagement in longitudinal multiburst studies and underscore the importance of considering time-varying and time-invariant factors in study design and analysis. This study provides valuable insights for researchers designing intensive longitudinal studies in behavioral science and digital health. Potential strategies for optimizing EMA protocols could include tailoring prompt schedules to individual contexts and developing adaptive sampling techniques.

A Multi-Agent LLM Network for Suggesting and Correcting Human Activity and Posture Annotations

2025-10-12

Accurate human activity recognition (HAR) is critical for health monitoring and behavior-aware systems. Developing reliable HAR models, however, requires large, high-quality labeled datasets that are challenging to collect in free-living settings. Although self-reports offer a practical solution for acquiring activity annotations, they are prone to recall biases, missing data, and human errors. Context-assisted recall can help participants remember their activities more accurately by providing visualizations of multiple data streams, but triangulating this information remains a burdensome and cognitively demanding task. In this work, we adapt GLOSS, a multi-agent LLM system that can triangulate self-reports and passive sensing data to assist participants in activity recall and annotation by suggesting the most likely activities. Our results show that GLOSS provides reasonable activity suggestions that align with human recall (63-75% agreement) and even effectively identifies and corrects common human annotation errors. These findings demonstrate the potential of LLM-powered, human-in-the-loop approaches to improve the quality and scalability of activity annotation in real-world HAR studies.

A Context-Assisted, Semi-Automated Activity Recall Interface Allowing Uncertainty

Proceedings of the ACM on Interactive Mobile Wearable and Ubiquitous Technologies · 2025-12-02

Measuring activities and postures is an important area of research in ubiquitous computing, human-computer interaction, and personal health informatics. One approach that researchers use to collect large amounts of labeled data to develop models for activity recognition and measurement is asking participants to self-report their daily activities. Although participants can typically recall their sequence of daily activities, remembering the precise start and end times of each activity is significantly more challenging. ACAI is a novel, context-assisted Activity Annotation Inttivity Annotation Interface that enables participants to efficiently label their activities by accepting or adjusting system-generated activity suggestions while explicitly expressing uncertainty about temporal boundaries. We evaluated ACAI using two complementary studies: a usability study with 11 participants and a two-week, free-living study with 14 participants. We compared our activity annotation system with the current gold-standard methods for activity recall in health sciences research: 24PAR and its computerized version, ACT24. Our system reduced annotation time and perceived effort while significantly improving data validity and fidelity compared to both standard human-supervised and unsupervised activity recall approaches. We discuss the limitations of our design and implications for developing adaptive, human-in-the-loop activity recognition systems used to collect self-report data on activity.

Exploring person‐centred sleep and rest–activity cycle dynamics over 6 months

Journal of Sleep Research · 2025-02-05

Sleep and circadian characteristics are associated with health outcomes, but are often examined cross-sectionally or using variable-centred analyses. Person-centred longitudinal research is needed to identify combined effects of sleep and circadian characteristics while allowing for change over time. We aimed to classify individuals into sleep-circadian statuses (aim 1), determine whether they transitioned between statuses over time (aim 2), and explore associated covariates and health outcomes (aim 3). Young adults (N = 151) wore smartwatches continuously for 6 months. Sleep (total sleep time, wake after sleep onset) and circadian rest-activity cycle indicators (interdaily stability, intradaily variability, relative amplitude) were derived from acceleration data and aggregated into person-means for months 1, 3, and 6. These values were entered into a latent transition model for aims 1 and 2. Multinomial logistic regressions, ANOVA, and ANCOVA addressed aim 3. Four statuses were extracted (entropy = 0.88): optimal sleepers, restless sleepers, short sleepers, and nappers. 10%-13% of optimal sleepers and 21% of restless sleepers became nappers, 7%-18% of nappers transitioned to other statuses, and 94%-100% of short sleepers remained unchanged. Males were more likely than females to be short versus optimal sleepers (p < 0.001). Restless sleepers had more physical dysfunction than nappers and short sleepers (p = 0.014, 0.022), while short sleepers reported more excessive sleepiness than optimal sleepers and nappers (p = 0.006, 0.060). This study identified four sleep-circadian statuses and found evidence for change over time. Our longitudinal person-centred approach could help inform the development of tailored diagnostic guidelines for sleep and circadian-related disorders that fluctuate within-individuals.

An Evaluation of Temporal and Categorical Uncertainty on Timelines: A Case Study in Human Activity Recall Visualizations

2025-11-01

Encoding uncertainty in timelines can provide more precise and informative visualizations (e.g., visual representations of unsure times or locations in event planning timelines). To evaluate the effectiveness of different temporal and categorical uncertainty representations on timelines, we conducted a mixed-methods user study with 81 participants on uncertainty in activity recall timelines (ARTs). We find that participants’ accuracy is better when temporal uncertainty is encoded using transparency instead of dashing, and that a participant’s visual encoding preference does not always align with their performance (e.g., they performed better with a less-preferred visual encoding technique). Additionally, qualitative findings show that existing biases of an individual alter their interpretation of ARTs. A copy of our study materials is available at https://osf.io/98p6m/.

Longitudinal User Engagement with Microinteraction Ecological Momentary Assessment (μEMA)

Proceedings of the ACM on Interactive Mobile Wearable and Ubiquitous Technologies · 2025-09-03 · 1 citations

Microinteraction ecological momentary assessment (μEMA) is a type of EMA that uses single-question prompts on a smartwatch to collect real-world self-reports. Smaller-scale studies show that μEMA yields higher response rates than EMA for up to 4 weeks. In this paper, we evaluated μEMA's longitudinal engagement in a 12-month study. Each participant completed EMA surveys (one smartphone prompt/hour for 96 days in 4-day bursts) and μEMA surveys (four smartwatch prompts/hour for the 270 days). Using data from 177 participants (1.37 million μEMA and 14.9K EMA surveys), we compared engagement across three groups: those who completed 12 months of EMA data collection(Completed), those who voluntarily withdrew after six months of EMA data collection (Withdrew), and those unenrolled by staff after six months of poor EMA response rates (Unenrolled). Compared to EMA, unenrolled participants were 2.25 times, those who withdrew were 1.65 times, and completed participants were 1.53 times more likely to answer μEMA prompts (p &lt; 0.001). Regardless of response rates, |μEMA was perceived as less burdensome than EMA (p &lt; 0.001). These results suggest μEMA is a viable method for intensive longitudinal data collection, particularly for participants who find EMA unsustainable.

The Physical Activity Assessment Using Wearable Sensors (PAAWS) Dataset: Labeled Laboratory and Free-Living Accelerometer Data

Proceedings of the ACM on Interactive Mobile Wearable and Ubiquitous Technologies · 2025-12-02

Poor sleep and sedentary behavior patterns increase the risk of chronic diseases and negatively impact an individual's health and quality of life. Large-scale surveillance studies can unobtrusively measure free-living physical activities, sedentary behaviors, and sleep using wearable sensors; however, many human activity recognition algorithms cannot reliably detect activities in true free-living settings because they are trained on data collected in a controlled, lab setting. We describe the data collection protocol and present the first release of a multimodal, multi-sensor-site dataset (PAAWS R1). The PAAWS R1 release includes ~4 hours of semi-naturalistic activities from 252 individuals and ~7 days of 24-hour, free-living activities from 20 adults. We have annotated waking day activities using video to provide second-by-second, ground-truth labels capturing short, quickly changing bouts of activity with realistic activity transitions. Additionally, we have labeled up to two nights of sleep stages from PSG data collected during some nights of the free-living protocol. The PAAWS dataset enables researchers to directly compare activity recognition algorithms on the same participants' data across multiple collection protocols and days of free-living behaviors, encouraging convergence towards robust algorithms that could aid health research and drive novel mobile computing interventions and applications.