Data and algorithms used for the study: An open-source deep learning model for classifying real-world activity intensities in persons with chronic pain using a wrist-worn accelerometer

The code and date were used for the publication of the article: 'An open-source deep learning model for classifying real-world activity intensities in persons with chronic pain using a wrist-worn accelerometer' Summary: In chronic pain (CP) care, patterns of activity and rest are relevant across diagnostic, intervention, and evaluation phases. Objective measurement of these patterns can support individualized care. Wrist-worn tri-axial accelerometers are user-friendly and yield valid and reliable data in the general population. The public availability of a detailed dataset together with a validated deep learning model for classifying activity intensities from raw accelerometer data can promote clinical use and standardize activity intensity measurement in CP research. This study aimed to (1) develop, test, and share a deep learning model that classifies activity intensity from raw accelerometer data, (2) investigate whether individuals with CP need a separate DL model, compared to the general population, and (3) explore its construct validity in real-world settings among individuals with and without CP, using real-world observations as the comparator instrument. Seven strategies were applied to train a Convolutional Neural Network, using three datasets – lab and real-world data from participants with and without CP - comprising 211,864 ten-second windows. Construct validity was evaluated using3,443 one-minute observations in real-world conditions from 5 individuals with and 10 without CP. The model’s classification accuracy was 76.5% in individuals with CP and 80.5% in individuals without CP. Results showed that there is no clear advantage of a CP-specific deep learning model. Real-world construct validity in CP on the individual level was good (agreement 73.6% in CP), and lower than in individuals without CP (agreement 80.2%). constructThe results suggest that the accompanying script and datasets enable accurate processing of raw accelerometer time series to activity intensities in the CP populations included in this study. Further investigation of construct validity and other clinimetric properties is warranted.