Intentional injury and the risk of subsequent hospitalization for attempted suicide

Abstract

Trauma patients are at increased risk of suicidal behavior. The association between assault injury and subsequent suicidal behavior relative to unintentional injury remains under-studied. This study hypothesized that trauma patients with assault injuries would demonstrate greater risk of subsequent suicide attempt hospitalization compared to patients with unintentional injury. Trauma patients hospitalized in Washington State were identified via administrative records. Proportional hazard analysis was conducted to test differences in risk of suicide attempt hospitalization up to 5 years after the initial trauma hospitalization, and time to onset of first suicide attempt hospitalization by subgroup. Approximately 2% (n=1264) of trauma inpatients were subsequently hospitalized for attempted suicide, and 0.3% died by suicide (n=177) during the follow-up period. Relative to patients with unintentional injuries, those with assault-related injuries (aHR=1.38, 95% CI: 1.02 to 1.86), and self-inflicted injuries (aHR=8.22, 95% CI: 7.24 to 9.33) demonstrated greater risk of suicide attempt hospitalization after discharge. These findings demonstrate a greater risk of suicidal behavior among trauma patients with assault injuries relative to patients with unintentional injuries. This suggests the importance of intentional cause of injury as a risk factor for suicidal behavior to be considered in assessment of suicidality and discharge planning for trauma patients.

Nearly 45,000 people per year in the United States die by suicide (Drapeau & McIntosh, 2016) and more than one million people per year attempt suicide in the United States (Piscopo & Lipari, 2016). Previous scholarship has identified myriad risk and protective factors for suicidal behavior,(O’Connor & Nock, 2014) including traumatic injury (March et al., 2014; Stein et al., 2010). In a large study of medical inpatients in Manitoba, March et al. (2014) found that patients with traumatic injuries were more than four times as likely to attempt suicide in the future compared to uninjured controls from the general population. As less than six cases of assault injuries with subsequent suicidality were observed in that study, the investigators were unable to assess assault injury as a unique predictor of suicidal behavior (March et al., 2014). Although the association of self-inflicted injury and unintentional injury with future risk of suicidal behavior is well characterized (March et al., 2014; Franklin et al., 2017), the risk of suicidal behavior conferred by assault injury remains relatively understudied. Nonetheless, the association between assault injury and subsequent suicidal behavior is theoretically and clinically important. The Interpersonal Psychological Theory of Suicidal Behavior (Joiner, 2005; Chu et al., 2017) posits that the capacity to enact lethal self-harm is acquired over time via habituation to physical pain and psychological provocation, including experiences of victimization (Joiner et al., 2007). Violent victimization and associated injury may be a particularly salient mechanism of acquired capability for suicide, and may represent a tangible indicator of increased suicide risk within the context of clinical assessment and discharge planning for providers who treat trauma patients.

We used data from a statewide retrospective cohort study (Rowhani-Rahbar et al., 2015) to evaluate the association between assault injury and future risk of hospitalization for attempted suicide, as compared to unintentional and self-inflicted causes of injury. We hypothesized that assault injury patients would demonstrate greater risk of subsequent hospitalization for attempted suicide relative to unintentional injury patients, and that self-inflicted injury patients would experience the highest risk of subsequent suicide attempt hospitalization.

Methods

The parent study for this analysis included records of all trauma patients hospitalized in Washington State in 2006 and 2007 (Rowhani-Rahbar et al., 2015); the first hospitalization for each patient during this time frame was identified as their index hospitalization via the Washington State Comprehensive Hospital Abstract Reporting System (CHARS). The present analysis categorized all trauma patients identified as having sustained (1) unintentional, (2) assault, or (3) self-inflicted injuries using ICD-9 and E codes as described previously (Rowhani-Rahbar et al., 2015).

Outcomes for this study were assessed from the day of discharge until 31 December 2011, and included death by suicide and hospitalization for attempted suicide. Study outcomes were identified using the criteria recommended and validated by Simon and colleagues (2007; 2013), and included any inpatient encounter with ICD-9 codes: (1) E950–E959, (2) E980–E988, or (3) Diagnosis of suicidal ideation (V62.84) in combination with 960–989 or 870–897. Due to the low rate of death by suicide observed in this sample, fatality data are offered descriptively without conducting inferential statistics.

Risk of subsequent hospitalization for attempted suicide by cause of injury subgroups was calculated via a proportional hazards model that assessed the time to onset of first hospitalization for attempted suicide following discharge from the index hospitalization. The model was adjusted for age, gender, and insurance payer as an indicator of socio-economic status. The cumulative risk of medically-treated suicide attempts for each subgroup over time was also plotted for visual comparisons. Study procedures were approved by the Human Subjects Division of the Washington State Department of Health.

Results

Patients included approximately equal proportions of males (n=32985, 49.5%) and females (n=33684, 50.5%), and the median patient age was 54 years (IQR: 37 to 78). The majority of patients had public insurance (n=36012, 54.0%), followed by private insurance (n=24852, 37.28%), and self-pay status (n=5805, 8.7%). Approximately 2% (n=1264) of identified trauma inpatients were hospitalized for a medically-serious suicide attempt during the follow-up period, and 0.3% (n=177) died by suicide after discharge from their index hospitalization (see Table 1). Most index hospitalizations were for unintentional injuries (n=58,445; 87.7%), followed by self-inflicted injuries (n=5,622; 8.4%), and assault injuries (n=2,602; 3.9%).

Table 1.

Frequency of hospitalization for attempted suicide and death by suicide among trauma inpatients

	Subsequent self-directed violence
Index Hospitalization	Hospitalized for Attempted Suicide			Died by Suicide
	Females n=754	Males n=519	Total N=1264	Females n=64	Males n=113	Total N=177
Unintentional n=58,445	308(0.5%)	253(0.4%)	561(0.9%)	21(0.04%)	67(0.1%)	88(0.1%)
Assault n=2,602	16(0.6%)	33(1.3%)	49(1.9%)	0(0.0%)	2(0.1%)	2(0.1%)
Self-Inflicted n=5,622	421(7.5%)	233(4.1%)	654(11.6%)	43(0.8%)	44(0.8%)	87(1.5%)

Note. N = 66,669. Row percentages are reported in parentheses.

Consistent with our hypothesis, inpatients with assault injuries were 38% more likely than those with unintentional injuries to have been hospitalized for a suicide attempt during the follow-up period (aHR=1.38, 95% CI: 1.02 to 1.86), and patients with self-inflicted injuries were more than 8 times as likely to have been hospitalized for a suicide attempt (aHR=8.22, 95% CI: 7.24 to 9.33), after adjusting for age, gender, and insurance payer. Relative to the rate of suicide attempt hospitalization in patients with unintentional injuries (234 cases per 100,000 person-years), assault-related injury was associated with approximately 89 additional suicide attempt hospitalizations per 100,000 person-years. The plotted cumulative risk of hospitalization for attempted suicide in Figure 1 demonstrates the hypothesized pattern of results.

Figure 1.

Cumulative risk of hospitalization for attempted suicide during the follow-up period, stratified by cause of injury subgroup. SA = Suicide Attempt.

Discussion

Our findings demonstrate the greater risk of medically serious suicidal behavior among trauma patients with intentional injuries compared to patients with unintentional injuries. Consistent with previous literature, self-inflicted injury was a robust predictor of subsequent hospitalization for attempted suicide (Franklin et al., 2017). Our findings were also consistent with previous cross-sectional and self-report survey findings that have demonstrated strong associations between violent victimization and suicidal behavior, including stronger associations between violent victimization and attempted suicide relative to unintentional forms of trauma exposure (e.g., natural disasters, accidents; Stein et al., 2010).

Our findings should be considered in view of several limitations. First, our outcome only included suicide attempters who were admitted to a hospital that were identified via a statewide registry of hospital admissions. Therefore, it was not possible to analyze the impact of intentional injury on subsequent suicidal ideation or non-medically-treated suicidal behaviors. Given the prevalence of suicidal behaviors that do not lead to inpatient care, and the ubiquity of undisclosed self-directed violence among high risk groups (Geulayov et al., 2018), additional research is needed to evaluate the relative impact of intentional trauma with regard to a broader range of suicide-specific outcomes (e.g., non-suicidal self-injury, suicidal ideation, undisclosed and/or untreated suicide attempts). Additional research that includes more nuanced suicide-related outcomes may also elucidate the extent to which relatively low absolute risk of suicidal behavior observed in this study does not represent the full impact of intentional injury with regard to suicidality in this population. Second, this study only included Washington State hospital records, and thus would not have included patients who emigrated to another State during the study period. Further, our findings may not generalize to trauma patients outside of Washington State or the Pacific Northwest. Finally, due to the low base rate of death by suicide, it was not possible to determine inferential estimates of risk for this outcome among subgroups.

Taken together, our findings demonstrate the relative importance of injury caused by assault as a predictor of suicidal behavior, relative to unintentional causes of injury. These findings suggest the potential utility of including intentional cause of injury as an additional domain within clinical conceptualizations of suicide risk, and assessing the extent to which assault-related injuries may contribute to primary drivers of suicidality (e.g., psychological pain, hopelessness, stress; Jobes, 2016) among patients preparing for discharge following major trauma. In addition, our findings buttress longstanding recommendations to enhance screening of suicidality among all patients admitted to hospital care (Horowitz et al., 2013). Survivors of violence hospitalized for their injuries suffered increased risk of suicidal behavior in this sample, additional research is warranted to explicate the mechanisms underpinning this outcome, and to evaluate the effectiveness of brief interventions to reduce future risk of suicidal behavior in this patient population.