Digital phenotyping of social functioning and employment in people with schizophrenia: Pilot data from an international sample

Abstract

Aim: Individuals living with schizophrenia experience significant impairments in social functioning. As a major clinical outcome, social functioning requires appropriate measurement tools that can capture its dynamic nature. Digital phenotyping, using smartphone technology to collect high volume ecologically valid data, can potentially capture these facets. We investigated the viability of digital data, such as GPS, Accelerometer, and screen activation as a proxy for common social functioning measurements. Methods: We used an ordinary least squares linear regression approach to compare the performance of digital signals with the performance of past social functioning scale (SFS) scores for predicting current SFS scores and subdomain values in 62 individuals with schizophrenia using smartphone and clinical assessments over the course of a year. The outcome of interest was the current SFS, for which we compared the capacity of the digital data (active and passive), and prior SFS scores to predict SFS scores. Results: Overall, the sub-scale models in order of performance (measured by RMSE score) were: (i) employment, (ii) social engagement, (iii) interpersonal behavior, (iv) recreation, (v) prosocial activities, (vii) performance, and (vii) competence. Digital data were particularly capable of predicting subdomain scores for employment (R2 = 0.746, Mean Squared Error (MSE) = 1.663) and social engagement (R2 = 0.710, MSE = 2.318). Conclusions: Digital phenotyping may have the capacity to operate as a proxy for certain social functioning measures. Future research should expand on this pilot data by focusing on establishing the reliability and validity of digital phenotyping to measure social functioning, and exploring which subdomains of social functioning are best measured digitally.