# Digital phenotyping for assessment and prediction of interoception, chronic stress, and self-regulation in adults: a scoping review

## TL;DR

This review explores how digital devices can assess and predict interoception, chronic stress, and self-regulation in adults, highlighting the potential and current limitations of digital phenotyping.

## Contribution

The study provides a comprehensive overview of recent digital phenotyping applications for mental health domains, emphasizing gaps and future directions.

## Key findings

- Most studies focused on chronic stress using wearable devices and heart rate variability.
- Integration of smartphone sensing and long-term monitoring remains limited.
- Machine learning models showed modest accuracy in classifying stress or self-regulation.

## Abstract

Digital phenotyping, the real-time quantification of human phenotype in situ via digital devices, offers opportunities to understand how behavior change interventions influence brain and mental health. Interoception, chronic stress, and self-regulation are key domains, benefiting from real-world, continuous assessment beyond what traditional methods can provide.

The aim of this scoping review was to map and synthesize the literature of the last five years on the use of digital phenotyping to measure or predict interoception, chronic stress, and self-regulation in adults. We focused on the types of devices and sensors employed, the psychological domains targeted, the nature of the data collected, feature extraction, data processing methods, and technological platforms utilized.

Following Joanna Briggs Institute methodology and PRISMA-ScR guidelines, we systematically searched PubMed, Web of Science, and Scopus, complemented with Google Scholar. Eligibility criteria included studies published since 2018, using smartphones or commercial wearables to assess or predict interoception, chronic stress, or self-regulation in adults.

From 850 retrieved records, 18 studies met inclusion criteria. Of these, 11 addressed chronic stress or stress reactivity, five self-regulation, and two interoception. Thirteen studies used wearable devices, three used smartphones, and two combined both approaches. Ecological momentary assessment (EMA) via smartphones was applied in eight studies. Heart rate variability (HRV) was the most common physiological measure (n = 14), followed by electrodermal activity and heart rate (n = 4 each). Nine studies analyzed behavioral data, including smartphone use, sleep, and activity. Six studies applied machine learning models, though only three reported classification accuracy (56.8%–79%). Eight used statistical methods to link features with stress or interoception, while four examined self-regulation using predefined features without identifying new biomarkers.

This review highlights that the field is still in its early stages, with most work focused on chronic stress and predominantly reliant on wearable devices. Integration of smartphone sensing and long-term monitoring remains limited, and analytical performance is modest. Nevertheless, the ubiquity of smartphones and wearables positions digital phenotyping as a promising, scalable approach for assessing brain and mental health in daily life. Future research should emphasize multimodal, longer-term data collection, innovative analytic methods, and transparent reporting.