Validation and Application of Wrist-Worn Wearable Physiological Measures in Psychotherapy

Abstract Purpose The integration of wrist-worn wearables to assess physiological parameters, such as heart rate (HR) within psychotherapy is increasing due to their intuitive handling and their ability to provide real-time data. However, little is known about the reliability and validity of different wearable parameters (e.g., HR, stress rate) and to what extent the degree of emotional activation associated with different types of intervention influences the accuracy of these parameters. Methods 12,330 thirty-second segments from 159 treatment sessions addressing test anxiety conducted with 26 participants were analyzed. Each session contained emotion-focused and cognitive-oriented interventions. The accuracy of the wearable’s (Garmin Vivosmart 4) HR and stress rate (SR) indices was compared to stationary gold-standard measurements (Electrocardiogram, ECG and Electrodermal Activity, EDA) by calculating Intraclass Correlations (ICC). Using multilevel prediction analyses for next-session outcomes, we examined and compared the predictive validity of client stationary and wearable physiological parameters. Results A very good agreement between wearable HR and lab ECG HR, as well as wearable SR and wearable HR was found. The agreement between wearable SR and lab ECG HR was good, but that between wearable SR, HR and lab EDA was poor. Wearable HR and lab ECG HR were more coherent with each other during cognitive-oriented compared to emotion-focused interventions. Regarding the predictive validity, client HR was positively associated with their next-session symptom severity levels, regardless of the measurement method and intervention type. Conclusions The results suggest that physiological parameters capturing activating and regulatory information may offer valuable insights into psychotherapeutic change processes. In this context, Garmin’s wearable HR may be a valid measure to investigate specific research questions when stationary measurement is not feasible. Nevertheless, existing limitations of wearable parameters are discussed.