Using Wear Time for the Analysis of Consumer-Grade Wearables’ Data: Case Study Using Fitbit Data
Bibliographic record
Abstract
Background: Consumer-grade wearables allow researchers to capture a representative picture of human behavior in the real world over extended periods. However, maintaining users' engagement remains a challenge and can lead to a decrease in compliance (eg, wear time in the context of wearable sensors) over time (eg, "wearables' abandonment"). Objective: In this work, we analyzed datasets from diverse populations (eg, caregivers for various health issues, college students, and pediatric oncology patients) to quantify the impact that wear time requirements can have on study results. We found evidence that emphasizes the need to account for participants' wear time in the analysis of consumer-grade wearables data. In Aim 1, we demonstrate the sensitivity of parameter estimates to different data processing methods with respect to wear time. In Aim 2, we demonstrate that not all research questions necessitate the same wear time requirements; some parameter estimates are not sensitive to wear time. Methods: We analyzed 3 Fitbit datasets comprising 6 different clinical and healthy population samples. For Aim 1, we analyzed the sensitivity of average daily step count and average daily heart rate at the population sample and individual levels to different methods of defining "valid" days using wear time. For Aim 2, we evaluated whether some research questions can be answered with data from lower compliance population samples. We explored (1) the estimation of the average daily step count and (2) the estimation of the average heart rate while walking. Results: For Aim 1, we found that the changes in the population sample average daily step count could reach 2000 steps for different methods of analysis and were dependent on the wear time compliance of the sample. As expected, population samples with a low daily wear time (less than 15 hours of wear time per day) showed the most sensitivity to changes in methods of analysis. On the individual level, we observed that around 15% of individuals had a difference in step count higher than 1000 steps for 4 of the 6 population samples analyzed when using different data processing methods. Those individual differences were higher than 3000 steps for close to 5% of individuals across all population samples. Average daily heart rate appeared to be robust to changes in wear time. For Aim 2, we found that, for 5 population samples out of 6, around 11% of individuals had enough data for the estimation of average heart rate while walking but not for the estimation of their average daily step count. Conclusions: We leveraged datasets from diverse populations to demonstrate the direct relationship between parameter estimates from consumer-grade wearable devices and participants' wear time. Our findings highlighted the importance of a thorough analysis of wear time when processing data from consumer-grade wearables to ensure the relevance and reliability of the associated findings.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.014 | 0.058 |
| Meta-epidemiology (narrow) | 0.001 | 0.000 |
| Meta-epidemiology (broad) | 0.001 | 0.001 |
| Bibliometrics | 0.002 | 0.004 |
| Science and technology studies | 0.001 | 0.001 |
| Scholarly communication | 0.002 | 0.002 |
| Open science | 0.001 | 0.001 |
| Research integrity | 0.002 | 0.001 |
| Insufficient payload (model declined to judge) | 0.001 | 0.001 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".