A machine learning approach to determine resilience utilizing wearable device data: analysis of an observational cohort

Objective

To assess whether an individual's degree of psychological resilience can be determined from physiological metrics passively collected from a wearable device.

Materials and Methods

Data were analyzed in this secondary analysis of the Warrior Watch Study dataset, a prospective cohort of healthcare workers enrolled across 7 hospitals in New York City. Subjects wore an Apple Watch for the duration of their participation. Surveys were collected measuring resilience, optimism, and emotional support at baseline.

Results

We evaluated data from 329 subjects (mean age 37.4 years, 37.1% male). Across all testing sets, gradient-boosting machines (GBM) and extreme gradient-boosting models performed best for high- versus low-resilience prediction, stratified on a median Connor-Davidson Resilience Scale-2 score of 6 (interquartile range = 5-7), with an AUC of 0.60. When predicting resilience as a continuous variable, multivariate linear models had a correlation of 0.24 (P = .029) and RMSE of 1.37 in the testing data. A positive psychological construct, comprised of resilience, optimism, and emotional support was also evaluated. The oblique random forest method performed best in estimating high- versus low-composite scores stratified on a median of 32.5, with an AUC of 0.65, a sensitivity of 0.60, and a specificity of 0.70.

Discussion

In a post hoc analysis, machine learning models applied to physiological metrics collected from wearable devices had some predictive ability in identifying resilience states and a positive psychological construct.

Conclusions

These findings support the further assessment of psychological characteristics from passively collected wearable data in dedicated studies.

Evaluation of a machine learning approach utilizing wearable data for prediction of SARS-CoV-2 infection in healthcare workers

Objective

To determine whether a machine learning model can detect SARS-CoV-2 infection from physiological metrics collected from wearable devices.

Materials and Methods

Health care workers from 7 hospitals were enrolled and prospectively followed in a multicenter observational study. Subjects downloaded a custom smart phone app and wore Apple Watches for the duration of the study period. Daily surveys related to symptoms and the diagnosis of Coronavirus Disease 2019 were answered in the app.

Results

We enrolled 407 participants with 49 (12%) having a positive nasal SARS-CoV-2 polymerase chain reaction test during follow-up. We examined 5 machine-learning approaches and found that gradient-boosting machines (GBM) had the most favorable validation performance. Across all testing sets, our GBM model predicted SARS-CoV-2 infection with an average area under the receiver operating characteristic (auROC) = 86.4% (confidence interval [CI] 84-89%). The model was calibrated to value sensitivity over specificity, achieving an average sensitivity of 82% (CI ±∼4%) and specificity of 77% (CI ±∼1%). The most important predictors included parameters describing the circadian heart rate variability mean (MESOR) and peak-timing (acrophase), and age.

Discussion

We show that a tree-based ML algorithm applied to physiological metrics passively collected from a wearable device can identify and predict SARS-CoV-2 infection.

Conclusion

Applying machine learning models to the passively collected physiological metrics from wearable devices may improve SARS-CoV-2 screening methods and infection tracking.