Tech Doctor Conducts Data Quality Verification for New Fitbit Air® Model Launch

Tech Doctor Inc. (Head Office: Chuo-ku, Tokyo; Representative Director: Kazunobu Minato; hereinafter referred to as Tech Doctor) has conducted a comparative verification of simultaneous wearing of the new wearable device "Fitbit Air®" released by Google Inc. with existing Fitbit devices (Charge6, Inspire3, Charge5), and an evaluation of data differences accompanying the migration to the Google Health app.

In clinical research utilizing wearable devices, the characteristics of acquired data may change due to device model changes, or updates to sensors, algorithms, and applications. Therefore, continuous data quality evaluation (validation) is an extremely important process for ensuring the reliability of research results in long-term studies or when comparing with past data.

In this verification, high consistency was confirmed for key indicators such as heart rate, resting heart rate, steps, calories burned, and sleep-related metrics between Fitbit Air and existing Fitbit devices. On the other hand, systematic differences (deviations) between models were also confirmed for SpO2 and some physiological indicators, indicating that careful evaluation for each indicator is necessary when changing devices.

Through its track record of supporting over 100 research projects utilizing commercially available wearable devices, Tech Doctor continuously evaluates the impact of changes in devices and platforms on research data. Moving forward, we will continue to promote quality verification and evaluation activities to support the use of reliable data in the development of digital biomarkers and clinical research.

◼️ Background of Verification

In recent years, clinical research using wearable devices has been expanding, particularly among pharmaceutical companies and academia. However, commercially available wearable devices are continuously updated with new models, improvements to sensors and algorithms, and changes to application specifications.

While these changes contribute to improved user experience, they can also affect the continuity and comparability of research data. It is essential to appropriately evaluate the impact of device changes in advance, especially for long-term studies or studies that compare with accumulated past data.

Tech Doctor conducts continuous validation of major devices to ensure data quality when using commercially available wearable devices for research. We have also promptly conducted a comparative evaluation with existing models for the newly released Fitbit Air.

◼️ Verification Overview

<Target Devices>

<Verification Conditions>

To eliminate the influence of differences in dominant hand or arm movement on step count and accelerometer-based indicators, the verification was conducted under the condition of "simultaneously wearing two devices on the same arm."

<Evaluation Metrics>

The following statistics, based on the Bland-Altman method, are used as the primary evaluation metrics for device comparison:

Bias (Systematic Error): Average difference between Fitbit Air and existing devices (overall data bias)

MAE (Mean Absolute Error): Average of the absolute errors for each measurement (variation of individual data)

※ (How to interpret the data): The closer the Bias is to zero and the smaller the MAE, the higher the data consistency and continuity between the two devices (i.e., the less likely it is that device replacement will affect research data).

◼️ Summary of Verification Results

High Consistency Confirmed for Major Indicators with Existing Fitbit

Fitbit Air and existing Fitbit devices (Charge6, Inspire3, Charge5) were simultaneously worn on the same arm, and various physiological indicators such as heart rate, activity levels, and sleep were compared and verified.

As a result, while the difference in minute-by-minute heart rate remained within approximately 1 bpm on average compared to existing models, a tendency for Fitbit Air to show slightly higher readings was confirmed across all comparison devices. Daily resting heart rate also showed nearly equivalent results, with differences within ±1 bpm for all devices. Furthermore, no significant differences were observed in step count and calories burned when compared with Charge6 and Inspire3, confirming high continuity in major activity indicators.

Regarding sleep indicators, key metrics such as total sleep time, REM sleep time, and deep sleep time were generally consistent, and the trend of sleep stage transitions showed similar results to existing models.

No interference between sensors was observed when wearing two devices on the same arm, and no issues affecting the verification were identified.

<Heart Rate>

<Step Count>

<Sleep>

The units for "Air Average" and "Cmp (Comparison Target) Average" are "Time (HH:MM)" only for wake-up and bedtime, and "minutes" for all other indicators. The units for "Bias" and "MAE" are also "minutes (time span)" only for wake-up and bedtime, and "minutes" for all others.

Differences Confirmed for Some Indicators

On the other hand, for some indicators, differences and variations in characteristics were confirmed, likely due to the effects of model changes or algorithm modifications.

For SpO2 (blood oxygen saturation), while no significant bias was observed at the average value level, relatively large variations were confirmed in individual data. Differences were also observed between models for HRV (heart rate variability), respiratory rate, and some sleep indicators.

These results indicate that even when continuity of major indicators is maintained in commercially available wearable devices, some indicators may be affected by changes in models or algorithms.

In clinical research and digital biomarker development, it is important to understand these differences in advance and consider them appropriately in study design and analysis.

<SpO2 (Average per Sleep Session)>

Charge5 has few sessions and is for reference only

No Significant Impact Confirmed from Google Health App Change

In addition, a data comparison was conducted for the migration from the conventional Fitbit app to the Google Health app.

As a result, no significant differences that would be problematic for research use were observed for major indicators, confirming that the impact of the application change is limited.

In recent years, the environment surrounding commercially available wearable devices has been continuously changing, not only in the devices themselves but also in the cloud infrastructure and applications. Tech Doctor continuously evaluates the impact of these changes on research data.

◼️ Tech Doctor's Initiatives

These verification results indicate that even when continuity of major indicators is maintained in commercially available wearable devices, some indicators or specific model pairings may be affected by changes in algorithms or hardware specifications.

In clinical research and digital biomarker development, it is important to understand these differences in advance and consider them appropriately in study design and data analysis.

This verification is not intended to evaluate the superiority or inferiority of specific devices, but rather is part of an ongoing effort to rapidly evaluate changes in commercially available devices and continuously ensure the data quality required for research use.

Moving forward, Tech Doctor will continue to promote the development of quality evaluation and analysis platforms that support the reliability of clinical research and digital biomarker development utilizing wearable devices.

※ Fitbit is a registered trademark of Google LLC.

【 About Tech Doctor 】 Tech Doctor Inc., with its vision of "Towards an era where data improves well-being," develops "digital biomarkers*" that derive health-related insights from everyday sensing data, including wearable devices, and promotes their social implementation. In collaboration with companies in the medical, pharmaceutical, and food industries, as well as research institutions, we aim to realize data-driven AI healthcare.

Representative: Kazunobu Minato Chief Medical Officer: Keisuke Izumi Head Office: Kyobashi Edogrand 4F, 2-2-1 Kyobashi, Chuo-ku, Tokyo Established: June 21, 2019 Business Activities: Development and operation of the digital biomarker development platform "SelfBase," and provision of digital healthcare solutions Website: https://www.technology-doctor.com/

* Digital Biomarker A digital biomarker is an indicator that continuously and objectively evaluates the presence of disease, changes in pathological conditions, or treatment effects based on everyday physiological data obtained from smartphones and wearable devices. While conventional biomarkers are "point data" measured temporarily at medical institutions, digital biomarkers are characterized by their ability to continuously acquire "line data" from daily life. Based on indicators such as exercise, sleep, and heart rate, they are expected to be used for early disease detection, treatment monitoring, and as new endpoints in drug development. Development has been progressing since around 2019 overseas, and attention is also growing in Japan.