# Validation of the Fitbit Charge 5 for the Detection of Heart Rate and Electrodermal Activity

## TL;DR

The Fitbit Charge 5 shows moderate accuracy for heart rate and electrodermal activity but has notable underestimation and overestimation issues compared to a research-grade device.

## Contribution

This study validates the Fitbit Charge 5 against a research-grade device for measuring anxiety-related physiological markers.

## Key findings

- Fitbit Charge 5 showed moderate correlations with the Equivital for heart rate (rs = 0.45–0.58) and EDA (rs = 0.42–0.50).
- Bland–Altman analyses revealed consistent underestimation of HR and overestimation of EDA by the Fitbit.
- Intraclass correlations were moderate for both HR and EDA, indicating some reliability but with accuracy limitations.

## Abstract

Wearable devices are increasingly used to evaluate psychophysiological markers of anxiety for continuous health monitoring. Consumer‐grade wearable devices, such as Fitbits, have the potential for widespread use and dissemination given their affordability and accessibility for both research and clinical settings. However, the validation of consumer‐grade devices against research‐grade devices is required. This study aimed to evaluate and compare the accuracy of the Fitbit Charge 5 against a research‐grade wearable device, the Equivital EQ02, in measuring psychophysiological parameters of anxiety, specifically heart rate (HR) and electrodermal activity (EDA). Fifty‐five undergraduate students (M
age = 19.4, SDage = 1.6, 46% female) wore both Fitbit and Equivital devices whilst completing social stressor and reading tasks. Statistical analyses demonstrated significant moderate correlations between the two devices for heart rate (HR) estimates (rs = 0.45–0.58) and low to moderate correlations for electrodermal activity (EDA) estimates (rs = 0.42–0.50). Intraclass correlations were moderate for both HR (ICCs = 0.53–0.72) and EDA (ICCs = 0.46–0.64) across conditions (ps < 0.05). Furthermore, Bland–Altman analyses revealed that the Fitbit showed a pattern of underestimation of HR (ranging from 24 to 32 bpm) and overestimation of EDA (ranging from −12.92 to 10.29 μS) compared to the Equivital. These findings highlight potential reliability concerns with the Fitbit Charge 5 in measuring physiological data. While the device may have some utility in assessing HR and EDA, it is crucial to approach the interpretation of data from consumer‐grade wearable devices with caution due to potential accuracy limitations.

Our findings provide new insights into the utility and limitations of consumer‐grade wearables for monitoring psychophysiological markers of anxiety. We complement existing research by demonstrating that, while the Fitbit Charge 5 shows moderate correlations with a research‐grade device, its consistent underestimation of heart rate and overestimation of electrodermal activity indicate potential accuracy concerns. These findings emphasize the need for validation studies and cautious interpretation when using consumer‐grade devices in research and clinical applications.