Military Servicemembers and Veterans are at elevated risk for suicide, but rarely self-identify to their leaders or clinicians regarding their experience of suicidal thoughts. We developed an algorithm to identify posts containing suicide-related content on a military-specific social media platform.

Publicly-shared social media posts (n = 8449) from a military-specific social media platform were reviewed and labeled by our team for the presence/absence of suicidal thoughts and behaviors and used to train several machine learning models to identify such posts.

The best performing model was a deep learning (RoBERTa) model that incorporated post text and metadata and detected the presence of suicidal posts with relatively high sensitivity (0.85), specificity (0.96), precision (0.64), F1 score (0.73), and an area under the precision-recall curve of 0.84. Compared to non-suicidal posts, suicidal posts were more likely to contain explicit mentions of suicide, descriptions of risk factors (e.g. depression, PTSD) and help-seeking, and first-person singular pronouns.

Our results demonstrate the feasibility and potential promise of using social media posts to identify at-risk Servicemembers and Veterans. Future work will use this approach to deliver targeted interventions to social media users at risk for suicide.

To assess the effect of different front-of-package labelling (FOPL) schemes on the objective understanding of the nutritional content and intention to purchase products, in Panama.

Single-blinded multi-arm parallel-group randomised controlled trial.

Supermarkets across Panama. Participants were exposed to two-dimensional images of fifteen mock-up products presented at random and balanced orders. Participants assigned to the intervention groups were exposed to mock-ups featuring one FOPL scheme: black octagonal warning labels (OWL), traffic-light labelling (TFL) or guideline daily amounts (GDA). The control group was not exposed to any FOPL scheme.

Adult supermarket shoppers (n 1200). Participants were blinded to group assignment.

A similar number of participants were randomised and analysed in each group: OWL (n 300), TFL (n 300), GDA (n 300) and control (n 300). The odds of choosing to purchase the least harmful or none of the options more often was the highest in the OWL group. Compared with the control group, these odds were two times higher in the OWL group (OR 2·13, 95 % CI 1·60, 2·84) and 57 % higher in the TFL (1·57, 1·40–2·56), with no changes in the GDA (0·97, 0·73–1·29). OWL also resulted in the highest odds for correctly identifying the least harmful option and for correctly identifying a product with excessive amounts of sugars, sodium and/or saturated fats.

OWL performed best in helping shoppers to correctly identify when a product contained excessive amounts of nutrients of concern, to correctly identify the least harmful option and to decide to purchase the least harmful or none of the options, more often.