The association between Digital Biomarkers and Suicidality: a Scoping Review and Bibliometric Analysis

Predicting suicide behaviour remains a major global health issue, with current predictive models showing limited success. Digital phenotyping methods allow for passive monitoring of behavioural and physiological biomarkers, offering potential to predict suicidality in real-time and in daily life. The current study combines a scoping review and a bibliometric analysis to map existing empirical research on digital biomarkers for the prediction of suicidality. It assesses the associations and predictive validity of digital biomarkers for suicidality and identifies key research trends, contributors, and themes. PubMed, Web of Science, and PsycINFO were searched for empirical studies reporting associations between digital biomarkers and suicidality. The search yielded 1120 records, with 33 records meeting eligibility criteria for inclusion. Most studies feature small, predominantly female, white, clinical, and western samples, often with mood or anxiety disorders. Sleep and extralinguistic features (e.g., speech disfluency) showed the strongest associations with suicidality. In total, 80% of sleep biomarkers and 87.5% of extralinguistic feature biomarkers were significantly associated with…. Combined active and passive sensing methods achieved the highest predictive accuracy (AUC = 0.84), far surpassing models based solely on passive data (AUC = 0.56). The bibliometric analysis revealed exponential growth in publication output after 2018 and a Western-centric authorship pattern. To conclude, behavioural digital biomarkers such as sleep disturbances and speech patterns capture core clinical features of suicidality and currently show the strongest potential to predict suicidality. Combining passive sensor data with active selfreports substantially improves predictive accuracy, as passive data alone have not yet demonstrated sufficient validity for suicide prediction. Investigation of longitudinal relationships has received limited attention, with relatively few studies investigating temporal associations between digital biomarkers and suicidality. The evidence is constrained by western, educated, industrialised, rich and democratic (WIERD) sampling biases, limiting generalisability.