(PS9-C73) Identifying Suicide Attempts After Hospital Discharge Using Daily Self-report and EHR Data

Identifying suicide attempts after hospital discharge using daily self-report and EHR data

            Suicide is a leading cause of death, and suicide prevention is a public health priority. Studies had shown that 50% of suicide decedents accessed health care the month before their death (Ribeiro et 2017), and 40% visited the Emergency Department within the year (Nock et al. 2022, Ahmedani et al. 2014). These results show that the months following healthcare visits for suicidal thoughts and behaviors are critical high-risk windows for suicide. Two methods that can be used to assess the presence of suicide attempts (SAs) during these periods include self-report through phone surveys and Electronic Health Records (EHR) review. EHR captures suicide attempts that result in hospitalization or are reported in clinical notes. Prior work has shown that patients are more likely to disclose sensitive information to a device than their clinicians (Nock et al., 2022), so daily surveys may capture additional information. Hence combining both self-report and EHR might yield superior results for describing reports of suicide attempts in patients during high-risk windows.

            The current study identified the incidence and frequency of SAs among adults and adolescents after hospital discharge using EHR and daily surveys. We parse how many SAs are ascertained using each data stream. Adults presenting to a psychiatric emergency room and adolescents treated in a psychiatric inpatient unit were recruited to participate in an intensive longitudinal study of suicidal thoughts and behaviors. Of the participants, n=240 adults and n=262 adolescents completed daily self-report surveys indicating whether they made a suicide attempt that day. In addition, health records were available for n=243 adult and n=308 adolescent participants for the same 6-month period. Skilled research personnel carefully extracted and coded SA events from emergency or behavioral health visit records. We also identified the proportion of unique events coded through each data stream versus potential duplicates. Across the sample, 69 SAs were identified (42 in adults and 27 in adolescents), 55% of unique instances of SAs were identified only through EHR (36% among adults and 19% among adolescents), 30% were identified only through daily surveys (14% for adults and 16% for adolescents), and 14% of SAs were identified using both sources (10% for adults and 4% for adolescents).

Overall, adults had nearly twice as many SA instances identified using EHR than daily surveys, whereas adolescents had similar numbers identified through each source. The number of redundant SAs identified through both sources was relatively small in adults and adolescents, which shows the value of having both data sources during this high-risk period. Further analysis explores the rates of SAs from each data source within each 1-month window during the 6 months. This study improves our understanding of the degree of risk for SAs after hospital discharge, supporting the need for interventions and protective measures during this time. It also has implications for future research, suggesting that EHR and daily surveys are valuable and non-redundant methods for ascertaining SAs for an at-risk population at a particularly high-risk time.