Participant characteristics.

Purpose
To evaluate the accuracy of a wrist-based heart rate (HR) monitor at different exercise intensities across different skin tones.
Methods
Using a cross-sectional design, we compared HR measures from the wrist-based photoplethysmography Fitbit Charge 5 to the Polar H10 chest strap at rest and during the YMCA Protocol using a recumbent cycle ergometer. Participant were grouped into three skin tone categories: light (Fitzpatrick Scale Skin Types 1+2), medium (Types 3+4), and darker skin tone (Types 5+6). HR measures using the Polar chest strap during the exercise test were categorized as <40%, 40–60%, or >60% HR reserve (HRR). Absolute error in beats per minute (bpm) between the two devices was calculated for each measure. A linear mixed effects model was used to assess interaction effects between skin tone and exercise intensity, with participants as the random effect. Bland-Altman plots were used for visual analyses.
Results
Twenty-five participants [mean (SD): 25.8 (1.9) years old; 64% female] were included with 495 observations of simultaneous Fitbit and Polar HR recordings collected during exercise. During exercise, we observed a statistically significant interaction effect between skin tone and exercise intensity. Compared with light skin tone at <40% HRR, mean error was greater for medium skin tone at >60% HRR [mean error (95%CI): 11.8 (5.6–17.9) bpm, p<0.001] and darker skin tone at 40–60% HRR [7.6 (1.7–13.5) bpm, p = 0.011] and >60% HRR [11.7 (5.3–18.0) bpm, p<0.001].
Conclusion
HR measurement error using a wrist-based device was greater with increasing exercise intensity for people with darker skin tones.