This study was conducted from July 2023 to December 2023 and was approved by the University of Florida Institutional Review Board as exempt under Protocol #ET00042583. Participants provided informed consent electronically prior to study enrollment, in accordance with institutional review board exempt protocol guidelines. Data were deidentified to safeguard participant information. Participants were not compensated for their participation. All procedures were performed in accordance with the ethical standards of the institutional and/or national research committee and with the 1964 Helsinki Declaration and its later amendments or comparable ethical standards.

The study explored the feasibility of long-term wearable sleep tracking and observed exploratory, descriptive trends regarding its potential visual correspondence with balance performance using different wearable technologies (Apple Watch, OURA ring, and Fitbit). Participants were assigned different wearable devices to assess device-specific feasibility and sleep-tracking variability. Two students (1 undergraduate and 1 graduate student) volunteered for this study and were recruited through convenience sampling. The participants were selected based on their willingness to complete a daily sleep diary, regular use of a wearable device, and availability for repeated balance assessments over the study period. Participant 1 wore an Apple Watch Series 5 and an OURA ring (Figure 1A). Participant 2 wore a Fitbit Charge 5 and an OURA ring for four months (Figure 1B). The devices were worn according to the manufacturer’s instructions. Both participants documented time in bed (TIB) and sleep onset time using the National Sleep Foundation Sleep Diary [33]. Both participants were trained to use the diary until they felt confident enough to log their daily sleep. Wearable sleep data were also cross-referenced with self-reported TIB via sleep diary [33]. Sleep diaries are widely recognized as a standard subjective tool for assessing sleep patterns over time. Previous validation studies have shown that self-reported TST from sleep diaries demonstrates reasonable agreement with objective measures such as polysomnography and actigraphy, particularly among healthy adults [4,21]. Balance was assessed monthly with sensory organization tests (SOTs) on the Bertec Balance Advantage system (Figure 2) [34]. Balance assessments were conducted between 9 AM and 11 AM to minimize circadian variability and were standardized across sessions. The participant was instructed to complete up to 18 trials, divided into 3 trials for each of 6 sensory scenarios (Table 1). Balance was assessed monthly (4 times over 4 months). The system generates equilibrium scores to quantify the evaluation sway. Equilibrium scores range from 0% to 100%, with 100% indicating perfect stability and 0% indicating fall [35,36]. A list of sensory scores is calculated by the software, including somatosensory (SOM), visual (VIS), vestibular (VEST), and preference (PREF) scores, as well as composite scores, to provide an understanding of which system is impaired [37]. This case study was reported in accordance with the CARE (Case Report) guidelines (Checklist 1).

The sleep activities and patterns were collected from the Fitbit Charge 5, OURA ring, and Apple Watch Series 5. Participants wore the devices from bedtime until morning, except during showers. We followed the manufacturers’ manuals for data extraction. Fitbit data were downloaded through the Fitbit web interface, OURA data were obtained via the OURA web portal, and Apple Watch data were exported using the “Health Export CSV” app. The extracted sleep variables included daily sleep start time (HH:MM:SS), TIB (minutes), TST (minutes), sleep efficiency (SE, %), wake after sleep onset (WASO, minutes), and different sleep stages duration (light sleep, rapid eye movement [REM] sleep, deep sleep). The Sensory Organization Test data for both participants were downloaded using the Bertec Balance Advantage Dynamic System software, resulting in a printable document for the 4 time points.

The collected sleep and balance data were analyzed using descriptive statistical methods and visualizations. Descriptive statistics were used to summarize sleep data collected from wearable devices and the sleep diary over the 4 months to explore agreement between the 2 sources. Parameters such as mean, SD, and range were calculated for each sleep measure. Intraclass correlation coefficients (ICCs) were calculated to assess agreement among wearable devices for various sleep parameters. The ICC values were interpreted as follows: excellent (0.75‐1.00), good (0.60‐0.74), moderate (0.40‐0.59), and poor (<0.40). For all Bland-Altman and ICC analyses, only complete paired observations between each wearable device and the sleep diary were included. Data points with missing values in either the wearable device recording or the sleep diary entry for a given day were excluded from agreement calculations. Following statistical best practices [38], we did not report P values and effect sizes. The feasibility of using wearables for long-term sleep monitoring was evaluated by calculating the proportion of missing data for each device and the sleep diary. Mean, SD, and range were calculated for each balance measure. We used data visualization techniques to see the relationships between TST and different sensory scores over time. Balance data visualized in figures were plotted as difference scores (participant’s raw score minus demographic-matched reference values) to visually track deviations from normative stability over time. No inferential statistics were applied due to the exploratory nature of the study. All analyses and data visualization were performed using R Studio 4.2.2 [39].

Participant 1 exhibited an average TIB (logged from the diary) of 6.66 (SD 0.48), 6.49 (SD 0.79), 5.92 (SD 1.11), and 5.95 (SD 1.07) hours for the 2 weeks before each balance assessment, respectively. In contrast, participant 2 showed longer average TIB values of 6.94 (SD 2.51), 7.29 (SD 2.24), 7.46 (SD 2.42), and 7.5 (SD 2.25) hours, respectively. The means and SDs of the average TST from wearables for the 2 weeks before each balance assessment date are summarized in Table 2. The 4 monthly balance assessment SOT scores, along with participants’ matched demographic reference scores, are listed in Table 3.

Following the protocol refinement on September 15, 2023, we introduced a structured approach involving weekly data checkups and reminders. Prior to this refinement, missing data rates were substantially higher across devices; however, this intervention successfully improved the completeness and reliability of the collected data, as shown in Tables 4 and 5.

The Bland-Altman plots (Figure 3) assessed agreement between wearable-derived TIB and self-reported sleep diary entries. Each plot depicted the mean bias (solid line) and the 95% limits of agreement (LoA, dashed lines) across all paired observations. Only complete, paired daily measurements between each wearable device and the sleep diary were included; incomplete pairs were treated as missing and excluded from the analysis. For participant 1, the Apple Watch showed the highest agreement with the sleep diary, with a minimal mean bias of 0.6 minutes (95% CI –10 to 8.8 min). The 95% LoA indicated that differences between the 2 methods generally ranged from –1.41 to 1.43 hours. The OURA ring slightly overestimated TIB relative to the sleep diary, with a mean bias of 13.8 minutes (95% CI 3-24 min). The 95% LoA ranged from –1.02 to 1.47 hours, indicating moderate variability. For participant 2, the Fitbit Charge 5 overestimated TIB compared to the sleep diary, showing a mean bias of 25.9 minutes. The 95% LoA for these differences ranged from –1.31 to 2.17 hours. The OURA ring also overestimated TIB, revealing a mean bias of 25.5 minutes compared to the sleep diary.

Overall, the Apple Watch Series 5 showed the strongest agreement with the sleep diary (underestimating TIB by less than 1 minute on average), whereas the Fitbit Charge 5 overestimated TIB by approximately 25.9 minutes. The OURA ring consistently overestimated TIB across both participants (by approximately 13.8 minutes for participant 1 and 25.5 minutes for participant 2). Effect sizes, expressed purely as mean differences and LoA ranges, are emphasized in this interpretation. In accordance with the exploratory nature of this study, no inferential statistics were applied.

Figure 4 shows the visual trends and fluctuations of participants’ TST and start times. There was a notable increase in variability in both sleep start times and TST over the observed period. For participant 1, late August and late September visually coincided with greater fluctuations in sleep patterns. Over the observed period, participant 2’s sleep start times varied widely, ranging from 10 PM to 2 AM, with most occurring around 1 AM. A slight increase in TST was observed around the second balance assessment.

Because participant 1 wore the Apple Watch and participant 2 wore the Fitbit Charge 5, these 2 wrist-worn devices cannot be directly compared within the same individual in this study; all ICC and Bland-Altman agreement calculations are strictly between each individual’s specific wearable and their diary. The results showed excellent agreement on TST, with ICC values ranging from 0.97 to 0.99, indicating that all devices provided consistent measurements for TIB and TST. However, the agreement was moderate to good for REM and light sleep and poor for WASO, deep sleep, and SE. For participant 1, comparisons between the OURA ring and Apple Watch demonstrated excellent agreement for TIB (ICC=0.91; 95% CI 0.49‐0.97) and TST (ICC=0.97; 95% CI 0.95‐0.98), good agreement for REM sleep (ICC=0.76; 95% CI 0.58‐0.86), and moderate agreement for light sleep (ICC=0.54; 95% CI –0.19 to 0.81; Table 6). Poor agreement was observed for WASO, deep sleep, and SE (ICCs<0.30).

For participant 2, comparisons between the OURA Ring and Fitbit Charge 5 showed excellent agreement for TIB (ICC=0.99; 95% CI 0.98‐0.99), TST (ICC=0.99; 95% CI 0.97‐1.00), and light sleep (ICC=0.80; 95% CI 0.58‐0.89). Good agreement was observed for deep sleep (ICC=0.66; 95% CI 0.41‐0.80), while moderate-to-poor agreement was found for REM sleep, WASO, and SE. Overall, TIB and TST showed strong agreement across devices, whereas stage-specific metrics, such as deep sleep, REM, and WASO, showed greater variability, reflecting challenges in wearable sleep stage detection.

For participant 1 (Figures 5 and 6), there was an exploratory visual trend toward more stable composite balance difference scores from July to October, particularly during periods of earlier sleep start times and increased TIB. This descriptive trend suggests that more consistent sleep patterns visually coincide with more stable somatosensory processing scores. Similarly, the VIS and VEST scores showed gradual stabilization over time. Participant 2 displayed a similar pattern (Figures 7 and 8), visually displaying more consistent balance scores alongside more regular sleep patterns. Visual fluctuations in the VIS score were observed alongside variations in sleep onset times and TST. While all scores remained within typical ranges, these observations suggest that consistent sleep patterns may pattern together with steadier balance performance in these 2 individuals. It is critical to note that the observed gradual improvements in composite and sensory scores over the 4 monthly SOT assessments, particularly for participant 2, may heavily reflect familiarization and practice effects rather than true physiological adaptations to sleep patterns.

The primary objective of this study was to explore the feasibility of long-term wearable sleep tracking and observe descriptive trends regarding its potential visual correspondence with balance performance. By comparing sleep data from different wearables to a sleep diary and visually exploring fluctuations between sleep parameters and balance measures, this research provided insights into the application of wearables for long-term sleep tracking and health monitoring. Specifically, our findings demonstrated that longitudinal wearable compliance is feasible when supported by structured weekly check-ins and that consumer devices show excellent agreement with one another for TST but vary in their concordance with self-reported TIB. Additionally, exploratory visual analyses suggested that variations in sleep consistency might coincide with fluctuations in postural stability over time. We caution against interpreting these observed trends as causal or generalizable and instead propose this as an illustrative example of how wearables may be used to explore sleep-balance interactions in longitudinal, real-world settings.

Despite its strengths, the study has several limitations that must be acknowledged. The most notable limitation is the small sample size, which includes only 2 participants. This limits the generalizability of the findings, as individual differences may impact the results. Future studies should include a more extensive and diverse sample, accounting for various demographic factors such as age, gender, and health status, to allow a more comprehensive understanding.

Another limitation is the potential learning effect associated with repeated use of the Bertec system. Participants may improve performance over time due to familiarity with the testing procedures rather than actual balance changes [36,46,47]. In a study of this size, these practice effects could potentially explain the observed longitudinal improvements in balance as much as, or more than, sleep-related changes. This could confound results, making it challenging to isolate the impact of sleep on balance. Although adjustment for learning effects was not feasible in this study design due to the small sample size, future studies should consider implementing washout periods, randomizing testing conditions, or using statistical methods to account for practice effects when conducting repeated balance assessments.

Additionally, several important potential confounders, including daily physical activity levels, caffeine and alcohol consumption, psychological stress, and screen use, were not systematically recorded [11,48-50]. These factors are known to influence both sleep quality and balance, and their omission limits the ability to isolate sleep as an independent contributor to balance variability. Future studies should incorporate concurrent monitoring of these variables to more precisely characterize the complex interactions between sleep and postural control.

While sleep diaries offer practical advantages for longitudinal monitoring, they are subject to recall bias and may not capture brief nighttime awakenings as accurately as objective measures. Furthermore, it is important to note that a sleep diary is a subjective tool and not a clinical gold standard for objective sleep architecture, which limits our ability to make definitive criterion validity claims regarding the wearable devices.

Regarding operational feasibility, the missing data before the protocol refinement on September 15, 2023, posed a challenge to the study’s completeness. However, introducing structured weekly data checkups and reminders reduced missing data, thereby improving the study’s reliability. Nevertheless, formal assessments of device comfort, usability, participant preferences, charging burden, and data synchronization issues were not systematically collected, which limits the ability to evaluate overall real-world feasibility.

Despite these limitations, the study offers a methodological foundation for further research on the role of sleep and sensory integration in balance performance. By highlighting descriptive visual trends regarding the potential impact of sleep patterns on balance, this study underscores the importance of considering sleep as a factor in future large-scale postural control research, particularly in populations at risk for sleep disturbances and balance impairments.

This study demonstrates the feasibility of using consumer-grade wearable devices to monitor sleep and explore its potential visual correspondence with balance performance in a real-world, longitudinal context. The excellent agreement among wearables on key sleep parameters supports their use in longitudinal studies, while observed descriptive trends between sleep consistency and postural stability suggest potential relationships warranting further exploration. These findings have important methodological implications for rehabilitation professionals; for instance, literature suggests that while occupational therapists recognize sleep as an important domain [51], they often lack accessible tools to continuously assess it. Technological advancements in wearable devices offer practical opportunities for therapists to conduct ongoing assessments and tailor personalized care. By leveraging continuous tracking, health care providers can better observe the complex interactions between sleep and mobility over time. Future research must use significantly larger, more diverse sample sizes and account for learning effects from repeated balance tests to validate these exploratory trends and fully evaluate sleep’s role in postural control.