Virtual reality-assessment of social interactions and prognosis in depression

Background Freud proposed that excessive self-blame-related motivations such as self-punishing tendencies play a key role in depression. Most of the supporting evidence, however, is based on cross-sectional studies and questionnaire measures. Methods In this pre-registered (NCT04593537) study, we used a novel Virtual Reality (VR) task to determine whether maladaptive self-blame-related action tendencies prospectively identify a subgroup of depression with poor prognosis when treated as usual over four months in primary care. Ninety-eight patients with depression (Patient Health Questionnaire-9 ≥ 15), screening negatively for bipolar and alcohol/substance use disorders, completed the VR-task at baseline (n = 93 completed follow-up). Results Our pre-registered statistical/machine learning model prospectively predicted a cross-validated 19 % of variance in depressive symptoms. Contrary to our specific predictions, and in accordance with Freud's observations, feeling like punishing oneself emerged as prognostically relevant rather than feeling like hiding or creating a distance from oneself. Using a principal components analysis of all pre-registered continuous measures, a factor most strongly loading on feeling like punishing oneself for other people's wrongdoings (β = 0.23, p = 0.01), a baseline symptom factor (β = 0.30, p = 0.006) and Maudsley Staging Method treatment-resistance scores (β = 0.28, p = 0.009) at baseline predicted higher depressive symptoms after four months. Limitations Patients were not assessed with a diagnostic interview. Conclusions Independently and apart from known clinical variables, feeling like punishing oneself emerged as a distinctly relevant prognostic factor and should therefore be assessed and tackled in personalised care pathways for difficult-to-treat depression.