Response Latency and Latency Variability: Valuable Meta-Data from Digital Assessments

Luke Scheuer, BA

Beth Israel Deaconess Medical Center – Clinical Research Intern

Scientific Abstract


            As suicide rates continue to increase not only in the United States, but across the globe, finding effective techniques to assess an individual’s mental status is increasingly important – early interventions are key to preventing suicide attempts. While smartphone-based assessments, such as administration of the Patient Health Questionnaire (PHQ-9), can provide valuable insight into a patient’s mental status, the meta-data captured by the device can also help. In particular, response latency to specific questions and response latency variability across an entire survey, both measures which cannot be captured by traditional pen-and-paper tests, could shed light into a patient’s mental status – previous research has shown that scores on the PHQ-9 self-harm question (item 9) is correlated with latency in patients with schizophrenia, but not for healthy controls.  


             Approximately 300 otherwise healthy participants who had previously scored highly on the Perceived Stress Scale (PSS) were administered a PHQ-9 survey through our lab’s open source smartphone app, mindLAMP, as part of an ongoing study begun in 2021. Over the course of the study, participants were told to take the PHQ-9 survey up to two times over the course of four weeks. A total of 405 surveys were completed across all participants.


            We found that scores on item 9 were negatively correlated with total survey duration, and thus also average item latency (Spearman’s ρ = -.16, p <.001). Item 9 scores were also negatively correlated with reaction time variability (Spearman’s ρ = -.13, p =.003). Item 9 scores were slightly negatively correlated with latency on item 9, but this correlation was not significant (Spearman’s ρ = -.06, p =.17). Upon filtering for item 9 scores greater than 1, selected participants showed a positive correlation between item 9 score and latency as well as item 9 score and latency variability, but this correlation was also not significant (Latency Spearman’s ρ = .06, p = .717, Variability Spearman’s ρ = .08, p = .616).


            We found no significant correlation between PHQ-9 item 9 latency and item 9 score for healthy participants, and negative correlation between latency variability and item 9 score. For participants who scored highly on item 9, we found a slight positive correlation between both latency and score and latency variability and score, but were unable to conclude it was significant, possibly in part due to the much smaller size of the subgroup, only 10% of our total sample. Given that previous research has found positive correlation between item 9 latency and score in patients with schizophrenia, our findings indicate that both latency and latency variability have potential value as a precision measure of risk for self-harm.

research Areas


Luke S. Scheuer, BA, Jenny Melcher, BA, John Torous, MD, MBI

Principal Investigator

John Torous, MD, MBI