Empirical data about the dynamics of near-term risk factors for suicide-related crises is lacking, mainly because it is difficult to identify the unique aspects of the period prior to crises through typical research designs. A variety of home-based and mobile technologies collect data repeatedly from Veterans outside of the clinic setting, and these data may be able to provide novel information about the early warning signs of suicide-related crisis. The VA's Care Coordination Hometelehealth (CCHT) Program employs care management augmented by interactive daily questionnaires delivered on a dedicated in-home messaging device. The purpose of CCHT is to support care for chronic illnesses, including mental illnesses. Self-report ratings of suicidal ideation, mood and other illness domains are gathered on devices. When concatenated with data from the medical record on events such as psychiatric hospitalizations or reports of suicidal behavior, we hypothesized that data from hometelehealth may be able to identify unique patterns among variables collected to predict the emergence of suicide-related crises.
The overarching goal of this HSRD Pilot Study was to conduct a 'proof-of-concept' study of a new statistical predictive approach to identify warning signs of behavioral emergencies, by leveraging intensive longitudinal data from a retrospective study of home telehealth participation among Veterans with psychiatric diagnoses in San Diego. Specifically, we harmonized CCHT and electronic medical record datasets to integrate time-varying predictors (data derived from in-home monitoring) and future outcomes (behavioral emergencies logged in the medical record). We then validated the refined statistical prediction models employing retrospective data from CCHT and the VA medical record.
This 1-year pilot involved retrospective analyses of data from two sources:. 1) Care coordination secure clinical dashboards that store and display data derived from in-home messaging devices and 2) the Computerized Patient Record System (CPRS). Data was obtained from 159 Veterans seen at the VA San Diego who were participants in the CCHT program for a primary psychiatric diagnosis. Data includes all participants who were enrolled in CCHT for psychiatric care during January 2014 to June 2016 and corresponding chart data from that same period. From the in-home messaging device data, we gathered data on self-reported suicidal ideation, mood, pain medication adherence and other available self-report data. From CPRS, we gathered the time, frequency, and nature of hospitalizations, records of suicidal behavior reports, emergency department visits, and clinical ratings on the comprehensive suicide risk assessment. The resulting dataset contains over 30,000 data points. In addition to descriptive analyses, our aim was to apply a statistical prediction technical called functional data analyses to model precursors to suicide-related crises.
Given the large and complex dataset, we are continuing to produce results addressing the Aims. To date, we have found that any instance of suicidal ideation as reported on in-home messaging device (reported by 31% of participants) was associated with more than double the risk of instances of suicidal behavior (Odds ratio: 2.9) and psychiatric hospitalization (Odds ratio: 2.1). However, the risk of suicide-related crises among people with one at-home report (n=17) vs. more than one (n=23) was not different for either suicidal behavior (Chi-square =0.1, p=0.983) or psychiatric hospitalization (Chi-square=0.7, p=0.388). Moreover, we found that discrepancies between in home ratings of pain versus clinician ratings (higher ratings at home) were associated with greater risk of suicidal behavior (rho=0.40, p=0.02).
Our project suggests that any instance of self-report of suicidal ideation recorded on at-home messaging devices may be herald subsequent risk for suicidal behavior or hospitalization, and this risk does not increase monotonically with increasing self reports. We also found that Veterans who experience more pain at home than they report in clinic settings may be at elevated risk for suicidal behavior. Although these findings may not generalize to Veterans who participate in other forms of telemonitoring/telehealth, our study provides indication that data integration of data derived from remote technology with medical records may yield new insights into the precursors to suicidal behavior.
None at this time.
Mental, Cognitive and Behavioral Disorders
TRL - Applied/Translational
Healthcare Algorithms, Information Management, Knowledge Integration, Suicide