Suicide Mortality Among Veterans Health Administration Care Recipients With Suicide Risk Record Flags

Veterans constituted 8.0% of the 2018 U.S. adult population and 13.8% of suicide deaths (1). The 2018 age- and sex-adjusted suicide rate among veterans was 1.5 times greater than the rate among nonveterans (1). Nearly one-third of these veterans were recent Veterans Health Administration (VHA) care recipients and had a higher 2018 age- and sex-adjusted suicide rate (28.6 per 100,000) than other veterans (27.3 per 100,000) (1). Suicide prevention is the top clinical priority of the U.S. Department of Veterans Affairs (VA).

VHA seeks to identify and support veterans at high suicide risk, including through use of predictive modeling based on the electronic health record (EHR) (2, 3), EHR alerts to reduce opioid and benzodiazepine prescribing among veterans with high-risk conditions (4), and patient record flags. In 2008, local VHA facilities began implementing a high-risk flag (HRF) for suicide in patient records to alert providers of patients at acute high (5) suicide risk (6, 7). Indicators of acute high risk include a current verified report of a suicide attempt; identification of current serious suicidal ideation; and presence of warning signs, such as threatening self-harm (6). Active HRFs appear for all providers each time the veteran’s EHR is opened.

HRF activation occurs after the facility’s suicide prevention coordinator (SPC) reviews referrals from providers because of clinically evaluated acute high suicide risk or recent suicidal behavior (6). The veteran is placed on a facility high-risk list, and the SPC contacts the veteran’s primary and mental health care providers to ensure that the veteran’s care plan has been reviewed considering current suicide risk, including ongoing suicidality monitoring. These expectations apply for initial HRF activations and reactivations. The HRF is reevaluated every 90 days to ensure prompt inactivation if acute high-risk status is resolved or to determine whether HRF continuation is needed. HRF reviews involve clinical consultation, suicide risk assessment, and safety and treatment plan review. HRFs are inactivated after an acute high risk has been resolved, but veterans may still experience chronic suicide risk. VHA does not require that HRF inactivation occur within a specific time frame.

To date, information is limited regarding VHA HRF recipients and their suicide-related outcomes. Among VHA users with substance use disorder diagnoses in 2012, predictors of HRF activation included age <35 years, a depressive disorder diagnosis, and a documented suicide attempt (8). Among veterans with substance use disorder diagnoses in 2012, homelessness was related to higher probability of HRF activation and subsequent care (9). HRFs may not be evenly distributed across the VHA user population; therefore, assessing demographic differences is important. To date, the clinical context immediately preceding HRF activation has not been assessed. We expected that HRF activation often occurs after a suicide attempt or inpatient mental health hospitalization for suicidal ideation.

To inform VA suicide prevention, it is important to assess suicide death following HRF actions and, for context, to provide information regarding suicide mortality rates among VHA patients without HRFs. Assessment of demographic and clinical contexts of HRF receipt and of suicide mortality after flag activation and inactivation may inform local suicide prevention efforts and national policies. In this retrospective study, we considered demographic and clinical characteristics associated with HRF receipt and assessed suicide mortality rates after HRF activations and inactivations.

Methods

Population

To examine trends in HRF utilization and suicide risk over time, we identified annual cohorts of active VHA users in the 2014–2016 period from VHA administrative data sources. Cohorts were defined as all veterans with recent use, meaning they received inpatient or outpatient care in the year or the previous year and were alive as of January 1 of the year (e.g., the 2014 cohort received VHA care in 2013 or 2014 and was alive as of January 1, 2014); this approach has been used in previous VHA studies (10, 11). Analyses excluded veterans whose location of last VHA use or death occurred outside of the 50 U.S. states or the District of Columbia. Excluded patients (N=289,729) accounted for 1.1% of the combined cohort.

Demographic characteristics included age as of January 1 of their reference year (2014–2016; categorized as ages 18–34, 35–54, 55–74, and 75–115 years, with values <18 or >115 set to missing), sex, and race-ethnicity (mutually exclusive categories: White, Black, Hispanic, other, and unknown). Small numbers of Asian or Pacific Islander (API), American Indian or Alaska Native (AIAN), and multiracial veterans precluded separate categories but were combined in the “other” race-ethnicity category to avoid removal from analyses. Analyses were conducted as part of the VHA Office of Mental Health and Suicide Prevention operations and quality improvement work and did not require institutional review board approval.

HRFs

HRF actions included the following five types: newly activated flags (“activation”), reviewed and continued or transferred ownership between facilities (“continuation”), reviewed and discontinued flag (“inactivation”), subsequent reactivations of an inactive flag (“reactivation”), and flag discontinued because of error (“error”). We focused on risk associated with activation and inactivation.

HRF activations were defined as veterans’ receipt of a new HRF activation in the 2014–2016 period, among veterans with no HRF actions documented in 2013. If an individual had multiple HRF activation records, the first instance was included. HRF inactivation was identified by receipt of a flag inactivation after a new HRF activation in 2014–2016. Veterans were assessed after their initial HRF activation and HRF inactivation, but they could be included in two yearly cohorts if the activation and inactivation occurred in different years.

Nonfatal Suicide Attempts and Mental Health Inpatient Hospitalizations

Nonfatal suicide attempts were indicated in three VHA sources: ICD diagnosis codes in inpatient or outpatient records, Suicide Prevention Applications Network records, and Suicide Behavior Overdose Report records. Nonfatal suicide attempts were identified via ICD-9 code E95, excluding records with indication of sequelae (E95.9), and all ICD-10 codes for nonsequelae attempts (T14.91; T36–T65 or T71, ending in 2A, 2D, 2XA, or 2XD; and X71–X82, excluding codes ending in “S”).

Mental health hospitalization was identified by stays in a psychiatric inpatient ward, indicated by admission, transfer, and discharge records. Hospitalization for suicidal ideation was indicated by mental health hospitalizations with suicidal ideation as the primary diagnosis, per ICD-9 code V62.84 or ICD-10 code R45.851.

Suicide Mortality

Mortality data were drawn from the VA and U.S. Department of Defense (DoD) Mortality Data Repository, which includes date and cause of death for all VHA users, per annual VA-DoD searches of the Centers for Disease Control and Prevention’s (CDC’s) National Death Index (NDI). ICD-10 cause of death codes used to identify suicide mortality were X60–84, Y87.0, and U03.

Data Analysis

We generated descriptive information about HRF use and demographic characteristics for VHA users with and without HRF receipt. For individuals who were included in more than one annual cohort, demographic characteristics were selected from the earliest cohort. For individuals with more than one HRF designation over the period, demographic characteristics associated with the first instance of a given flag type are presented. We examined the clinical context of HRF activation by assessing the prevalence of suicide attempts, inpatient mental health stays, and inpatient stays for suicidal ideation (not mutually exclusive) on the day of HRF activation as well as the 7 and 30 days before.

We calculated suicide rates for veterans receiving HRF activation overall and by year, 2014–2016, for the 1, 2, 6, 9, and 12 months after activation. Risk time for death was calculated from the date of HRF activation until the end of the period of interest or death, whichever came first, regardless of subsequent flag status changes (e.g., inactivation). Rates are presented per 100,000 person-years and were calculated as the number of suicide deaths in the period of interest divided by the sum of risk time, in years, multiplied by 100,000. We did not compare rates across years statistically and therefore did not adjust for intrasubject correlation.

The same method was used to calculate 1-, 2-, 6-, 9-, and 12-month suicide rates for veterans whose last HRF action before death or the end of the period of interest was inactivation. Risk time for suicide death was calculated from the date of HRF inactivation until the end of the period of interest or death, whichever came first. For HRF activation and inactivation, analyses included unadjusted and adjusted Cox proportional hazards regression models. Risk time began on the date of HRF activation, HRF inactivation, or use (for those without HRF actions) through 365 days or until death, whichever came first. Adjusted models included age, sex, and race-ethnicity. Model proportional hazards assumptions were tested. We conducted all analyses in SAS, version 9.4.

Results

Use of HRFs

For each year in the 2014–2016 period, the number of VHA patients who received any HRF action was 25,759, 27,585, and 28,533, respectively. The numbers of VHA patients who received new HRF activations during this period were 16,086 (62.5%), 15,922 (57.7%), and 15,007 (52.6%), respectively. The mean±SD HRF flag duration from activation until either death, inactivation, or the end of 2017 was 185.2±150.8 days. Within the initial 100 days following HRF activation, 53.6% (N=25,178) of flags were continued, 4.2% (N=1,965) of flags were inactivated, 41.3% (N=19,405) of flags received no further action, and 1.0% (N=467) of veterans died before the flag could be reviewed.

HRF Receipt: Associations With Demographic and Clinical Characteristics

Table 1 presents demographic information for veterans who received HRF actions. Veterans receiving HRF actions were more likely to be female, younger, and with “other” race-ethnicity than veterans who did not receive HRF actions.

Among the 47,015 veterans who received new HRF activations in 2014–2016, 3,389 (7.2%) received the HRF on the date of a documented suicide attempt, 13,492 (28.7%) within the following 7 days, and 18,574 (39.5%) within the following 30 days. In total, 2,084 veterans (4.4%) received HRF activation on the date of a VHA mental health inpatient admission, 15,601 (33.2%) within the following 7 days, and 18,852 (40.1%) within the following 30 days. A total of 7,480 (15.9%) veterans had both a documented suicide attempt and a mental health inpatient admission within 30 days before HRF activation. In addition, 10 veterans (0.02%) received HRF activation on the date of a VHA inpatient mental health admission for suicidal ideation, 56 within the following 7 days (0.1%), and 75 within the following 30 days (0.2%).

Suicide Risk After HRF Activation

The suicide rate in the 12 months after HRF activation in 2014–2016 was 682 per 100,000 person-years (Table 2). Rates were highest in the month after HRF activation (2,058 per 100,000 person-years). Rates decreased over the first 6 months yet remained >600 per 100,000 person-years through 12 months post-HRF activation (Figure 1). Examining rates by year of HRF activation, we found that the 12-month suicide rate after activation was 758 per 100,000 person-years in 2014 and 625 and 662 per 100,000 person-years in 2015 and 2016, respectively. For each year, rates were highest in the month following HRF activation and then decreased. VHA patients with HRF activation had increased risk for suicide in both unadjusted (hazard ratio [HR]=21.30, 95% confidence interval [CI]=18.95–23.95) and adjusted (HR=21.00, 95% CI=18.55–23.72) Cox proportional hazards models compared with risk among VHA users without HRF activation (Table 3).

Suicide Risk After HRF Inactivation

Suicide risk in the 12 months following HRF inactivation in 2014–2016, was 408 per 100,000 person-years (Table 4). Rates were highest in the month after HRF inactivation (600 per 100,000 person-years) and remained >400 per 100,000 person-years through 12 months (Figure 1). Examining by year, we noted that the 12-month suicide rate after HRF inactivation was 501 per 100,000 person-years in 2014 and was 368 and 370 per 100,000 person-years in 2015 and 2016, respectively. The overall pattern that rates were highest in the month after HRF inactivation and then decreased over time was consistent across the individual years, although less pronounced for 2016. HRF inactivation was associated with suicide death in unadjusted (HR=12.74, 95% CI=10.89–14.91) and adjusted (HR=12.43, 95% CI= 10.57–14.63) Cox proportional hazards models (Table 3) compared with risk among VHA users without HRF inactivation.

Discussion

This first study of suicide outcomes after HRFs documents high suicide risk following HRF activations and, to a lesser degree, after HRF inactivations. In the year after HRF activation and according to models adjusted for age, sex, and race-ethnicity, veterans were >20 times more likely to die by suicide compared with the veterans in VHA care who did not have HRF activity. The assessed 12-month suicide rate (682 per 100,000 person-years) was comparable to that observed for VHA patients in the top 0.5% risk tier of a comprehensive suicide predictive model (631 per 100,000 person-years) (2).

Current processes for HRF activation identify a VHA patient subpopulation with high suicide risk. Suicide rates were highest in the month after activation yet remained substantially elevated through 1 year. Further work is needed to distinguish risk associated with HRF receipt from the impact of postactivation care enhancements. To assess the impact of HRFs, researchers will need to assess suicide risk after HRFs relative to patients who have similar suicide risk but do not receive an HRF.

Suicide rates were also substantially elevated in the year following HRF inactivation (408 per 100,000 person-years) relative to the general population of veterans in VHA care without HRF activity. It is unsurprising that suicide rates after HRF inactivation were lower than those after HRF activation because HRFs are intended to be inactivated after acute high-risk status is resolved. However, it is noteworthy that suicide risk after HRF inactivations remained elevated. This finding may reflect chronic risk, unresolved acute risk, or both. The findings highlight the need to understand how best to address postinactivation risk. For example, increasing flag duration would ensure longer periods of provider notification regarding elevated risk yet could also result in alert fatigue. Nevertheless, enhanced care should be considered postinactivation.

Veterans receiving HRF actions were more likely to be younger, female, and of “other” race-ethnicity compared with other veterans in VHA care. The concentration of HRFs among younger veterans is consistent with findings of higher suicide rates among younger veterans compared with older veterans (1). Proportionally greater receipt of HRF actions by female veterans is consistent with the observation that women are more likely to engage in nonfatal suicidal behavior than are men (12). The “other” race-ethnicity category was composed of API, AIAN, and multiracial veterans. CDC reporting indicates that suicide rates among AIAN individuals are elevated compared with other race-ethnicity groups, whereas rates among API individuals are similar or lower than among individuals in other groups (13). Consistent with the findings in that work and those of previous assessments of demographic characteristics of HRF recipients among veterans with substance use disorder diagnoses (9), we observed that HRF activations were more frequent among veterans classified as “other” race-ethnicity. Further work, particularly with more granular categories, is needed to understand race-ethnicity–related differences in suicidal behavior.

This analysis extends previous work assessing predictors of HRF receipt among veterans with substance use disorder diagnoses (8) by considering clinical contexts of HRF receipt. Approximately 40% of veterans who received HRF activation did so within 30 days of a documented suicide attempt. Approximately 40% did so in the 30 days following inpatient mental health admission. Fewer than 1% received their new HRF activation within 30 days of an inpatient admission for suicidal ideation. However, ICD coding limitations may exist for suicidal ideation; clinician-noted ideation is rarely documented in the EHR with diagnostic codes (14).

After the study period of this work, VHA enhanced HRF practices. In 2017, the VA Office of Inspector General recommended that care after HRF activations and reactivations should include safety plan documentation within 7 days, receipt of at least four mental health visits within 30 days, and case review within 90 days (15). These recommendations have been operationalized in VHA performance monitoring systems. A 2020 VHA notice formalized the HRF inactivation process (16) by providing inactivation guidance, including documented evidence of risk reduction from clinical consultation between the SPC and treatment providers, review of the EHR, and completed suicide prevention safety planning. Furthermore, the VHA notice required that SPCs continue personal contact with veterans monthly for 12 months after HRF inactivation through use of caring communications, an evidence-based intervention for suicide prevention (17). Assessment of suicide risk following HRF inactivation after formalized inactivation processes were implemented is needed.

Implications

The findings of this study indicate high suicide risk among VHA patients who received HRF activations. Furthermore, although reduced after HRF inactivation, suicide risk remained substantially elevated. These results highlight the need for comprehensive suicide risk assessment, protocols for risk reduction, and continued support for veterans after HRF inactivation. VHA enhancements to HRF practices are ongoing; continued assessments will be important to evaluate their impact and to inform ongoing suicide prevention efforts.

Limitations

This study had several limitations. First, our analyses were limited to VHA users in the 2014–2016 period who were followed up for 12 months, because of timing of national HRF implementation and the availability of mortality data. Second, VHA site variation in HRF use complicated our assessment. Further work is needed to enhance the understanding of HRF decision-making processes and organizational factors. Third, although death certificate data from the CDC’s NDI are considered the gold standard of U.S. mortality data (18), misclassification of suicide deaths may occur. Previous work suggests that veteran deaths may be less likely than nonveteran deaths to be misclassified and that inclusion of undetermined deaths has little impact on risk estimates (19). Despite these limitations, this study advances our understanding of the context and risk among veterans with HRFs. Our findings underscore the need to provide ongoing support to veterans who receive HRF inactivation.

Conclusions

Veterans often receive HRFs after documented suicide attempts or mental health inpatient hospitalizations. To the best of our knowledge, this study marks the first assessment of adverse outcomes among patients with clinical suicide risk flags. Suicide risk following HRF activation is increased among VHA patients, particularly in the first month but also throughout the year postactivation, relative to patients who do not receive HRF activation. Although reduced, suicide risk remains high after HRF inactivation, highlighting the importance of ongoing VHA work to ensure support after HRF inactivation. Although this study was not an evaluation of the impact of HRFs, its findings provide a baseline for future assessments of their impact.