Mental health service utilization following a campus mass shooting: The role of pre-shooting emotion dysregulation and posttraumatic cognitions

Abstract

Objective:

This study investigated pre-shooting emotion dysregulation and posttraumatic cognitions as predictors of mental health service utilization (MHU; therapy/medication) following a campus mass shooting among undergraduate women, while controlling for time, age, and post-shooting posttraumatic stress (PTS) and depressive symptoms.

Method:

Undergraduate women (N = 483, M_age = 19.23, SD = 2.39) were engaged in a study when a mass shooting occurred on Northern Illinois University’s campus. A separate, longitudinal study was then implemented to monitor post-shooting adjustment among these same women. The present study examined predictors of MHU using data from the pre-shooting assessment and the following post-shooting timepoints: nine months (T1; n = 416); 14 months (T2; n = 416); 20 months (T3; n = 417); 26 months (T4; n = 405); and 33 months (T5; n = 397).

Results:

Multilevel models showed pre-shooting emotion dysregulation and post-shooting PTS and depressive symptoms positively predicted increased likelihood of MHU while controlling for covariates. Posttraumatic cognitions initially predicted increased therapy utilization, but this relationship became nonsignificant after accounting for pre-shooting emotion dysregulation. Pre-shooting emotion dysregulation also weakened the positive relationship between depressive symptoms and therapy utilization and strengthened the positive relationship between age and therapy utilization.

Conclusions:

Pre-shooting emotion dysregulation and post-shooting mental health symptoms were the most robust predictors of increased MHU following a mass shooting. Findings suggest women exposed to a mass shooting engage in treatment when needed, but preexisting emotion dysregulation may serve as a barrier for those who go on to develop depression.

Introduction

Mass shootings, as defined as episodes of gun violence resulting in at least four victim deaths (Smith & Hughes, 2016), have seen an increase in the past 50 years (Duwe, 2017). Although most individuals demonstrate resilience following a mass shooting (Orcutt et al., 2014), some go on to develop depression and posttraumatic stress disorder (PTSD) (Lowe & Galea, 2017). This subgroup of individuals may be vulnerable to chronic dysfunction warranting clinical services, as PTSD is often comorbid with other medical and psychological conditions, may endure if left untreated, and confers a heightened risk for suicide that is compounded by comorbid depression (Foa et al., 2006; Panagioti et al., 2012). However, those experiencing psychopathology following a mass shooting report low levels of mental health service utilization (MHU) (Boykin & Orcutt, 2019; Miquelon et al., 2014), potentially signaling an unmet mental health need among this population. Still, the reasons for this low treatment engagement are unclear. Given that comparable rates of MHU are seen following disasters (e.g., Hurricane Katrina, September 11 terrorist attacks; Smith et al., 2017), this literature may elucidate who utilizes treatment in the wake of a mass shooting and which factors stand in the way.

Much of the research on MHU comes from the disaster literature and has largely relied upon Andersen’s (1995) behavioral model as a framework to understand three classes of factors that affect health care utilization: predisposing factors (e.g., age, gender), enabling/disabling factors (e.g., stigma, financial constraints), and perceived or evaluated need (e.g., emotional distress) (Andersen, 1995; Rodriguez & Kohn, 2008). Although the relevancy of these factors varies and appears dependent on the sample, need has been identified as the strongest predictor of MHU (Andersen, 1995; Smith et al., 2016). For example, across the MHU post-disaster and post-trauma literature, variables such as severity of distress, psychopathological symptoms (e.g., depression, PTSD), and exposure are consistent predictors of increased MHU (e.g., Gavrilovic et al., 2005; Smith et al., 2017; Kantor et al., 2017). However, within the context of mass shootings, the sparse data that are available paint a murkier picture.

Mental Health Service Utilization Following Mass Shootings

To the authors’ knowledge, there have been only four studies investigating predictors of MHU in the wake of a mass shooting (i.e., Boykin & Orcutt, 2019; Felix et al., 2020; Miquelon et al., 2014; and Schwarz & Kowalski, 1992). Ultimately, findings provide mixed support for the otherwise intuitive idea that need promotes MHU. Following a shooting in an elementary school, Schwarz and Kowalski (1992) found that while the majority of individuals utilized an intervention offered in the first two weeks post-shooting, those who reported more peritraumatic fear for their life and worse PTSD dropped out of later free mental health screenings. In contrast, Miquelon and colleagues (2014) found the prevalence of different mental health disorders (including, but not limited to, PTSD) predicted more MHU in the 18 months following the shooting at Dawson College, Montreal. More recently, Boykin and Orcutt (2019) found that individuals with probable PTSD six months after the shooting at Northern Illinois University (NIU) utilized treatment at a significantly higher rate than individuals at minimal to no risk for PTSD. Although this seemingly suggests that need may have driven MHU, Boykin and Orcutt did not find that mental health symptoms reliably predicted MHU, as was the case in the Miquelon et al. study. Similarly, mental health symptoms did not prospectively predict MHU following an episode of mass violence (Felix et al., 2020).

Taken together, it is not clear from the extant literature that need is always sufficient to explain who goes on to utilize services following exposure to a mass shooting. Instead, these divergent findings suggest it may be necessary to consider how MHU may be inhibited by other constructs associated with need besides psychopathology, such as different attitudes or beliefs (Smith et al., 2017). For instance, Miquelon and colleagues (2014) found that, among individuals who endorsed symptoms consistent with at least one mental health diagnosis, the most commonly cited reason for not seeking services was that they wanted to “figure things out on their own,” suggesting the acknowledgment of needing services yet a reluctance to do so. Further, Schwarz and Kowalski (1992) posited that the participants in their study who dropped out did so to avoid the distress of trauma reminders despite possibly still having a need for services. This set of results points to how the decision to utilize mental health services following a shooting might be influenced by thoughts about treatment, symptoms, and recovery. Consistent with this idea, researchers have speculated that MHU may be hindered by posttraumatic cognitions (Smith et al., 2017), a construct yet to be studied in this context.

Posttraumatic Cognitions

Posttraumatic cognitions refer to negative thoughts and beliefs about the world, the self, and others that might develop following a traumatic event (e.g., “People can’t be trusted,” “I have no future;” Foa et al., 1999; Janoff-Bulman, 1992). These cognitions may be relevant to post-shooting MHU because they play a central role in many theories of PTSD development and maintenance for their interference with the natural recovery process (Ehlers & Clark, 2000; Rauch & Foa, 2006). Emotion Processing Theory suggests that posttraumatic cognitions directly increase avoidance of trauma-related stimuli (e.g., “If I think about the event, I will not be able to handle it”), thus preventing the corrective learning required to fully process the trauma memory (Rauch & Foa, 2006). Cognitive Theory of PTSD suggests that posttraumatic cognitions lead to interpretations of the environment, symptoms, etc. as being related to current threat, producing a fear response regardless of whether real threat is truly present (Ehlers & Clark, 2000). Despite these theoretical differences, both models consider posttraumatic cognitions to ultimately promote rigid, maladaptive avoidance of trauma cues that maintains PTSD over time. Accordingly, posttraumatic cognitions may also discourage MHU after a shooting since these services entail confronting trauma-related content that is being avoided. This motivation to avoid trauma reminders and hence treatment may be even stronger for individuals who historically have difficulty regulating their emotions, suggesting another possible barrier to MHU.

Emotion Dysregulation

Emotion regulation is the ability to remain aware of and accept emotions, inhibit impulsivity, act in goal-directed behavior despite negative emotions, and apply emotion regulation strategies in a flexible, contextual manner (Gratz & Roemer, 2004). Problems in these areas (i.e., emotion dysregulation) may serve as a barrier to MHU following a mass shooting, as individuals with difficulty expressing their emotions are disinclined to seek treatment for fear that their self-disclosures be met with a negative response (Ciarrochi & Deane, 2001; Komiya et al., 2000; Vogel et al., 2008). However, because emotion dysregulation contributes to the development and maintenance of PTSD (Seligowski et al., 2015), the relationship between post-shooting emotion dysregulation and MHU is difficult to disentangle from associated mental health symptoms that may emerge after shooting exposure. Pre-shooting data are necessary to clarify how regulatory tendencies predict post-shooting MHU. For instance, pre-existing emotion dysregulation may inhibit post-shooting MHU when coupled with posttraumatic cognitions—without a foundational ability to regulate emotions, negative appraisals that emerge after a shooting may be more difficult to challenge, re-evaluate, or manage (e.g., McLean et al., 2019). Consequently, posttraumatic cognitions may take hold and deter individuals from utilizing treatment despite an increase in emotional distress (e.g., “I can’t deal with even the slightest upset”). Understanding these dynamics may help inform ways to promote more engagement in mental health services following mass violence. Therefore, using the same sample as Boykin and Orcutt (2019), the current study investigated pre-shooting emotion dysregulation and posttraumatic cognitions as prospective predictors of MHU following the mass shooting at NIU.

Current Study

There are several important differences between this study and Boykin and Orcutt’s (2019). First, this study examined different predictors of MHU and used different covariates, an important step toward identifying which variables are reliably associated with post-shooting MHU. Second, this study utilized more post-shooting time points, increasing reliability and statistical power (Willett, 1989). Third, this study analyzed the data using multilevel modeling (MLM), whereas Boykin and Orcutt utilized regression analyses. MLM is a more appropriate statistical approach for handling longitudinal data because repeated-measures collected at each time point (i.e., level-1) are nested within participants (i.e., level-2). Such clustered, hierarchical data are interdependent, with responses at different time points inherently tied to individual characteristics (Nezlek, 2012). Although the study by Boykin and Orcutt provided a valuable contribution to the MHU literature, their analyses assumed observations were independent of each other, thus not accounting for the nested data structure.

The current study tested the following hypotheses while controlling for time, age, depressive and posttraumatic stress symptoms (PTSS): (1) greater posttraumatic cognitions will negatively predict the likelihood of MHU; (2) greater pre-shooting emotion dysregulation will negatively predict the likelihood of MHU; and (3) the relationship between posttraumatic cognitions and MHU will be strengthened by higher levels of pre-shooting emotion dysregulation. Said differently, it is expected that pre-shooting emotion dysregulation, and the development of posttraumatic cognitions, will both function as barriers to post-shooting MHU.

Method

Participants

Undergraduate women were participating in a longitudinal study on sexual victimization when a mass shooting occurred at NIU’s campus (for a detailed discussion of study design, see Boykin & Orcutt, 2019). Following the shooting, the study was expanded to monitor post-shooting adjustment among these same women. For the current study, participants were excluded from analysis if they reported not being on campus during the shooting¹, resulting in a final sample of 483 participants at the pre-shooting assessment (M_age = 19.23, SD = 2.39, age range = 18–53, Mdn_age= 18.83). Timing of subsequent post-shooting assessments² were variable, but approximately 90% of each follow-up sample was collected approximately nine months (T1; N = 416), 14 months (T2; N = 416), 20 months (T3; N = 417), 26 months (T4; N = 405), and 33 months (T5; N = 397) after the shooting. Regarding race/ethnicity, 65.6% identified as White (n = 317), 22.2% identified as Black or African-American (n = 107), 2.9% identified as Asian (n = 14), 0.2% identified as Native Hawaiian or other Pacific Islander (n = 1), 7.2% identified as “Other” (n = 35), and 1.7% preferred not to respond (n = 8); 7.2% identified as Hispanic or Latino (n = 35). Most participants reported being freshmen at the pre-shooting assessment (77.4%, n = 374).

Measures

Pre-Shooting Measures

Emotion Dysregulation.

The current study used the 36-item Difficulties in Emotion Regulation Scale (DERS; Gratz & Roemer, 2004) administered prior to the shooting as a measure of pre-shooting emotion dysregulation. Participants reported how often the items applied to them along a five-point scale (1 = Almost never to 5 = Almost always). A total sum score was used as a level-2 predictor of MHU and the slope of posttraumatic cognitions. Previous research supports the scale’s factor structure and its use across racially and demographically diverse samples (Ritschel et al., 2015). In the current study, the DERS total score demonstrated excellent internal consistency: α = .93.

Post-Shooting Measures

Mental Health Service Utilization (MHU).

Two binary variables adapted from Wang et al. (2007) were used as the current study’s outcome of therapy and/or medication utilization in the wake of the shooting across T1–5 (“Since the mass shooting on February 14, 2008, have you received any sort of professional counseling or therapy [have you taken any prescription medicines] for problems with your emotions, nerves, or mental health?”). Due to the nature of the study, pre-shooting data on MHU was not collected. Participants responded either yes (coded as 1) or no (coded as 0). Affirmative responses were fixed to remain constant for all subsequent timepoints³. Rates of MHU across timepoints are available as Supplemental Materials.

Posttraumatic Cognitions Inventory (PTCI; Foa et al., 1999).

The PTCI was used to measure negative appraisals following exposure to the shooting across T1–5. The PTCI is a 36-item questionnaire that assesses negative appraisals along three subscales: negative cognitions about the self, negative cognitions about the world, and self-blame. The self-blame scale (5 items) was omitted from the original study to minimize participant distress about the shooting, thus resulting in 31 items total. Participants indicated how much they agreed with each item along a seven-point scale (1 = Totally disagree to 7 = Totally agree), with higher scores indicating more negative appraisals. A total sum score was used as a level-1 predictor of MHU. Prior research found the two PTCI scales used in this study differentiate between traumatized individuals with and without PTSD (Beck et al., 2004; Foa et al., 1999). In the current study, the PTCI total score demonstrated excellent internal consistency across all time points: α = .95 (T1); α = .95 (T2); α = .94 (T3); α = .94 (T4); α = .94 (T5).

Covariates

Distressing Events Questionnaire (DEQ; Kubany, Leisen, Kaplan, & Kelly, 2000).

The 17-item DEQ was used to assess for PTSS according to the Diagnostic and Statistical Manual of Mental Disorders (4^th ed.; DSM-IV; American Psychiatric Association, 1994) in conjunction with the Traumatic Life Events Questionnaire (TLEQ). Across T1–T5, participants indicated the severity of their PTSS within the past month in relation to an index trauma along a five-point scale (0 = Almost never to 4 = Almost always), with higher scores indicating greater PTSS. The DEQ was administered in relation to the mass shooting, but if participants indicated another event more distressing than the shooting on the TLEQ, a separate DEQ was administered. In the current study, total shooting-related PTSS was used as a level-1 predictor of MHU. The DEQ has previously shown excellent psychometric properties (Kubany, Leisen, Kaplan, Watson et al., 2000) and demonstrated excellent internal consistency across time points in this study: α = .92 (T1); α = .91 (T2); α = .92 (T3); α = .92 (T4); α = .91 (T5).

Depression, Anxiety, and Stress Scale (DASS-21; Lovibond & Lovibond, 1995a).

The DASS-21 depression subscale (seven items) was used to measure depressive symptoms across T1–5. Participants reported depressive symptoms over the past week using a four-point scale (0 = Did not apply to me at all to 3 = Applied to me very much, or most of the time). Scores were doubled to render them equivalent to those computed using its full 42-item counterpart (Lovibond & Lovibond, 1995a) and then used as a level-1 predictor of MHU. Previous research supports the scale’s factor structure and its use in nonclinical samples (Lovibond & Lovibond, 1995b). In this study, the DASS-21 depression subscale demonstrated good internal consistency across time points: α = .89 (T1); α = .89 (T2); α = .88 (T3); α = .88 (T4); α = .89 (T5).

Time.

Time since the shooting was used a level-1 predictor of MHU and was coded to correspond with the number of months elapsed between assessments: T1 = 0; T2 = 5; T3 = 11; T4 = 17; T5 = 24.

Age.

Participants’ age in years at T1–5 was used as a level-1 predictor of MHU. Participants’ average ages at each post-shooting timepoint were as follows: M_age = 20.32, SD = 2.00 (T1), M_age = 20.81, SD = 2.01 (T2), M_age = 21.26, SD = 1.75 (T3), M_age = 21.70, SD = 1.73 (T4), M_age = 22.24, SD = 1.76 (T5).

Data Analytic Plan

Data were first screened for quality using the Statistical Package for the Social Sciences (SPSS) Version 22, during which we detected two outliers on age. Analyses were performed with and without these data and showed no meaningful differences. Thus, all data were retained. All multilevel analyses were conducted in the statistical program HLM for Windows, version 7.03 (HLM 7). Multilevel generalized linear models were fitted to test study hypotheses using a dichotomous (Bernoulli) outcome and penalized quasi-likelihood estimator. Three models tested predictors of the likelihood of utilizing: (1) any MHU (either therapy and/or medication); (2) therapy only; and (3) medication only. For all models, predictor variables were centered so that intercept terms may be interpreted as adjusting for their differential effects. Using SPSS, z-scores were computed for age, depressive symptoms, PTSS, posttraumatic cognitions, and emotion dysregulation by mean-centering each variable and then dividing by their respective standard deviations; time was grand-mean centered using HLM 7.

First, a null model was fitted to test whether the log-odds of MHU vary across individuals. Next, a series of linear growth models were fitted to test whether posttraumatic cognitions predicted the log-odds of MHU while controlling for time, age, PTS and depressive symptoms. Finally, a pre-shooting emotion dysregulation was added to test whether it predicted the log-odds of MHU, and moderated the relationship between posttraumatic cognitions and MHU, while controlling for time, age, PTS and depressive symptoms.

The equation for the full model is as follows:

$$\begin{gathered} \text{Level}1:{\eta }_{ti}={\pi }_{0i}+{\pi }_{1i}(\text{time})+{\pi }_{2\text{i}}(\text{age})+{\pi }_{3\text{i}}(\text{depression})+{\pi }_{4\text{i}}(\text{PTSS})+{\pi }_{5\text{i}}(\text{cognitions}) \\ \text{Level}2:{\pi }_{0i}={\beta }_{00}+{\beta }_{01}\left(\text{emotion}\text{dysregulation}\right)+{\text{r}}_{0\text{i}} \\ {\pi }_{1\text{i}}={\beta }_{10}+{\beta }_{11}\left(\text{emotion}\text{dysregulation}\right) \\ {\pi }_{2\text{i}}={\beta }_{20}+{\beta }_{21}\left(\text{emotion}\text{dysregulation}\right) \\ {\pi }_{3\text{i}}={\beta }_{30}+{\beta }_{31}\left(\text{emotion}\text{dysregulation}\right) \\ {\pi }_{4\text{i}}={\beta }_{40}+{\beta }_{41}\left(\text{emotion}\text{dysregulation}\right) \\ {\pi }_{5\text{i}}={\beta }_{50}+{\beta }_{51}\left(\text{emotion}\text{dysregulation}\right) \end{gathered}$$

where ${\eta }_{ij}=\text{log}\left(\frac{\phi ij}{1-\phi ij}\right)$, or the predicted log-odds of mental health service utilization, and β₀₀ = the average log-odds of utilizing mental health services across individuals; β₁₀ through β₅₀ = the respective effects of time, age, depressive symptoms, PTSS, and posttraumatic cognitions on the predicted log-odds of utilizing mental health services; β₀₁ = the effect of pre-shooting emotion dysregulation on the predicted log-odds of utilizing mental health services (i.e., main effect); β₁₁ through β₅₁ = the cross-level (i.e., moderating) effect of pre-shooting emotion dysregulation on the relationships between each level-1 variable and the log-odds of utilizing mental health services in individuals; and r_0i = the random error variability between individuals⁴.

Results

See Table 1 for descriptive statistics and bivariate point-biserial, Pearson, and Spearman correlations. Age was the only individual difference variable not associated with the others.

Table 1.

Descriptive Statistics and Bivariate Correlations Among Study Variables

Scale	1	2	3	4	5	6	7	8	9
1. Any Mental Health Utilization	--
2. Therapy Only	.967**	--
3. Medication Only	.590**	.462**	--
4. PTSD Symptoms	.338**	.330**	.206**	--
5. Depressive Symptoms	.293**	.273**	.201**	.518**	--
6. Posttraumatic Cognitions	.268**	.257**	.171**	.742**	.622**	--
7. Age	.025	.037	−.019	.023	−.009	.037	--
8. Pre-Shooting Emotion Dysregulation	.265**	.258**	.226**	.253**	.441**	.350**	−.045	--
9. Time	.105*	.081	.067	.016	−.010	.009	.122*	−.001	--
Min	0	0	0	0	0	31	19.63	36	0
Max	1	1	1	37	30	143.2	48.55	146	24
Mean	0.23	0.21	0.07	5.20	5.45	61.59	21.28	77.47	11.24
SD	0.39	0.38	0.20	6.32	5.35	19.82	1.96	20.98	2.35

Note. Point-biserial (variables 1–3), Pearson (variables 4–8), and Spearman correlations (variable 9). All variables (except emotion dysregulation) are averaged across post-shooting timepoints for each participant. Any Mental Health Utilization = either therapy and/or medication utilization since the mass shooting; Therapy Only = “Since the mass shooting on February 14, 2008, have you received any sort of professional counseling or therapy for problems with your emotions, nerves, or mental health?” (0 = No, 1 = Yes); Medication Only = “Since the mass shooting on February 14, 2008, have you taken any prescription medicines for problems with emotions, nerves, or mental health?” (0 = No, 1 = Yes); PTSD Symptoms = posttraumatic stress symptom severity based on the Distressing Events Questionnaire total score (0 = Almost never to 4 = Almost always); Depressive Symptoms = depression symptom severity based on the Depression Anxiety Stress Scale depression subscale total score (doubled to be comparable to DASS-42) (0 = Did not apply to me at all to 3 = Applied to me very much, or most of the time); Posttraumatic Cognitions = Posttraumatic Cognitions Inventory total score based on negative cognitions about self and world subscales (1 = Totally disagree to 7 = Totally agree); Age = age in years; Pre-Shooting Emotion Dysregulation = pre-shooting Difficulties in Emotion Regulation Scale total score (1 = Almost never to 5 = Almost always); Time = post-shooting assessment (0 = nine months, 5 = 14 months, 11 = 20 months, 17 = 26 months, 24 = 33 months).

^*= p < .05;

^**= p < .001

The null models indicated statistically significant variation among individuals in the likelihood of utilizing mental health services (see Supplemental Materials). Using these results, the intraclass correlation coefficient (ICC) was calculated to assess the degree to which responses within individuals were similar to each other and different from responses in other individuals $\left(ICC=\frac{\tau }{\tau +\frac{{\pi }^2}{3}}\right)$. We then calculated the design effect (DEFF) to determine the extent to which our standard errors were inflated due to clustering: DEFF = 1 + (m – 1)ICC, where m = average cluster size. Both the ICC and DEFF values suggested a notable amount of clustering.

See Table 2 for results from the multilevel model of posttraumatic cognitions on the likelihood of MHU, controlling for time, age, PTS and depressive symptoms. Although age had a nonsignificant effect on MHU, it was retained as a covariate in subsequent models given its relationship with MHU in the literature. As expected, time positively predicted the likelihood that individuals would utilize any type of mental health service⁵. Both PTS and depressive symptoms predicted an increase in the log-odds that individuals would utilize any mental health service, including therapy and medication specifically. Posttraumatic cognitions only predicted whether individuals utilized therapy services specifically. Notably, the direction of this effect was unexpectedly positive, indicating that, at a midpoint in time, each unit increase in posttraumatic cognitions predicted a 0.28 increase in the log-odds that individuals of an average age would utilize therapy, controlling for PTS and depressive symptoms.

Table 2.

Time, Age, Depression, PTSD, and Posttraumatic Cognitions Predicting the Likelihood of Utilizing Mental Health Services

Model Fixed effect	Log Odds	Robust SE	t-ratio	d.f.	p-value	Odds Ratio	Confidence Interval
Any MHU
Intercept	−2.19	0.16	−14.01	443	< .001	0.11	(0.083,0.153)
Time	0.08	0.01	8.82	1524	< .001	1.08	(1.063,1.101)
Age	0.12	0.15	0.80	1524	.427	1.12	(0.841,1.504)
Depression	0.36	0.10	3.53	1524	< .001	1.43	(1.173,1.748)
PTSD	0.40	0.10	3.80	1524	< .001	1.49	(1.211,1.825)
PTCI	0.22	0.11	1.94	1524	.052	1.25	(0.998,1.560)
Therapy
Intercept	−2.41	0.16	−14.73	443	< .001	0.09	(0.065,0.123)
Time	0.09	0.01	9.92	1524	< .001	1.09	(1.072,1.109)
Age	0.16	0.16	0.97	1524	.33	1.17	(0.854,1.599)
Depression	0.32	0.09	3.73	1524	< .001	1.38	(1.166,1.638)
PTSD	0.35	0.11	3.28	1524	.001	1.42	(1.151,1.748)
PTCI	0.28	0.11	2.60	1524	.009	1.33	(1.072,1.646)
Medication
Intercept	−3.28	0.16	−20.27	443	< .001	0.04	(0.027,0.052)
Time	0.04	0.01	2.97	1526	.003	1.04	(1.014,1.068)
Age	−0.09	0.17	−0.52	1526	.603	0.92	(0.660,1.273)
Depression	0.30	0.12	2.50	1526	.013	1.35	(1.066,1.704)
PTSD	0.30	0.10	2.84	1526	.005	1.34	(1.096,1.649)
PTCI	0.07	0.14	0.45	1526	.652	1.07	(0.804,1.418)

Note. Significant effects have been bolded. Robust SE = robust standard errors; any MHU = Any mental health utilization (either therapy and/or medication since the mass shooting); therapy = “Since the mass shooting on February 14, 2008, have you received any sort of professional counseling or therapy for problems with your emotions, nerves, or mental health?” (0 = No, 1 = Yes); medication = “Since the mass shooting on February 14, 2008, have you taken any prescription medicines for problems with emotions, nerves, or mental health?” (0 = No, 1 = Yes); time = T1 (0), T2 (5), T3 (11), T4 (17), T5 (24); age = age in years; depression = depression symptom severity based on the Depression Anxiety Stress Scale depression subscale total score (0 = Not at all to 3 = Most of the time); PTSD = posttraumatic stress symptom severity based on the Distressing Events Questionnaire total score (0 = Almost never to 4 = Almost always); PTCI = Posttraumatic Cognitions Inventory total score based on negative cognitions about self and world subscales (1 = Totally disagree to 7 = Totally agree).

With pre-shooting emotion dysregulation added into the model, posttraumatic cognitions no longer predicted MHU, and depressive symptoms no longer predicted medication utilization (see Table 3). Pre-shooting emotion dysregulation positively predicted an increase in the likelihood of utilizing any services, and therapy and medication separately, controlling for all other variables. Pre-shooting emotion dysregulation also moderated the relationship between depressive symptoms and therapy utilization, decreasing the log-odds by −0.17, while controlling for all other variables. Pre-shooting emotion dysregulation also strengthened the positive relationship between age and therapy utilization, predicting a 0.38 increase in log-odds.

Table 3.

Pre-Shooting Emotion Dysregulation Moderating the Likelihood of Utilizing Mental Health Services

Model Fixed Effect Moderator	Log Odds	Robust SE	t-ratio	d.f.	p-value	Odds Ratio	Confidence Interval
Any MHU
Intercept	−2.19	0.16	−13.61	442	< .001	0.11	(0.082,0.154)
DERS	0.51	0.16	3.22	442	.001	1.66	(1.219,2.265)
Time Slope	0.08	0.01	8.57	1519	< .001	1.08	(1.064,1.103)
DERS	0.01	0.01	0.61	1519	.540	1.01	(0.989,1.022)
Age Slope	0.36	0.19	1.93	1519	.054	1.43	(0.994,2.070)
DERS	0.33	0.16	2.06	1519	.039	1.39	(1.016,1.907)
Depression Slope	0.34	0.12	2.89	1519	.004	1.41	(1.116,1.776)
DERS	−0.07	0.08	−0.91	1519	.363	0.93	(0.801,1.085)
PTSD Slope	0.40	0.11	3.69	1519	< .001	1.50	(1.208,1.857)
DERS	−0.02	0.09	−0.25	1519	.805	0.98	(0.822,1.164)
PTCI Slope	0.17	0.12	1.42	1519	.156	1.19	(0.937,1.502)
DERS	−0.01	0.11	−0.06	1519	.952	0.99	(0.806,1.224)
Therapy
Intercept	−2.41	0.17	−14.13	442	< .001	0.09	(0.064,0.125)
DERS	0.54	0.17	3.21	442	.001	1.72	(1.234,2.399)
Time Slope	0.09	0.01	9.70	1519	< .001	1.09	(1.074,1.114)
DERS	−0.00	0.01	−0.14	1519	.888	1.00	(0.982,1.016)
Age Slope	0.44	0.19	2.31	1519	.021	1.56	(1.068,2.266)
DERS	0.38	0.17	2.29	1519	.022	1.46	(1.056,2.019)
Depression Slope	0.35	0.10	3.47	1519	< .001	1.43	(1.166,1.743)
DERS	−0.17	0.08	−2.10	1519	.036	0.85	(0.726,0.989)
PTSD Slope	0.35	0.11	3.13	1519	.002	1.41	(1.138,1.757)
DERS	0.01	0.10	0.145	1519	.885	1.01	(0.838,1.228)
PTCI Slope	0.21	0.12	1.80	1519	.071	1.23	(0.982,1.553)
DERS	0.01	0.10	0.08	1519	.935	1.01	(0.825,1.232)
Medication
Intercept	−3.42	0.17	−20.16	442	< .001	0.03	(0.024,0.046)
DERS	0.61	0.17	3.62	442	< .001	1.83	(1.319,2.549)
Time Slope	0.04	0.01	2.79	1521	.005	1.04	(1.012,1.073)
DERS	0.00	0.01	0.22	1521	.828	1.00	(0.980,1.026)
Age Slope	−0.22	0.28	−0.76	1521	.445	0.81	(0.462,1.404)
DERS	0.26	0.27	0.96	1521	.336	1.29	(0.765,2.187)
Depression Slope	0.27	0.14	1.90	1521	.058	1.31	(0.991,1.742)
DERS	−0.05	0.09	−0.56	1521	.574	0.95	(0.789,1.141)
PTSD Slope	0.38	0.12	3.27	1521	.001	1.47	(1.165,1.844)
DERS	−0.17	0.10	−1.77	1521	.076	0.84	(0.696,1.018)
PTCI Slope	−0.11	0.18	−0.64	1521	.522	0.89	(0.634,1.261)
DERS	0.20	0.14	1.45	1521	.015	1.22	(0.931,1.600)

Note. Significant effects have been bolded. Robust SE = robust standard errors; any MHU = Any mental health utilization (either therapy and/or medication since the mass shooting); therapy = “Since the mass shooting on February 14, 2008, have you received any sort of professional counseling or therapy for problems with your emotions, nerves, or mental health?” (0 = No, 1 = Yes); medication = “Since the mass shooting on February 14, 2008, have you taken any prescription medicines for problems with emotions, nerves, or mental health?” (0 = No, 1 = Yes); DERS = pre-shooting Difficulties in Emotion Regulation Scale total score (1 = Almost never to 5 = Almost always); time = T1 (0), T2 (5), T3 (11), T4 (17), T5 (24); age = age in years; depression = depression symptom severity based on the Depression Anxiety Stress Scale depression subscale total score (0 = Not at all to 3 = Most of the time); PTSD = posttraumatic stress symptom severity based on the Distressing Events Questionnaire total score (0 = Almost never to 4 = Almost always); PTCI = posttraumatic cognitions inventory total score based on negative cognitions about self and world subscales (1 = Totally disagree to 7 = Totally agree).

Discussion

This study is the first to examine how pre-shooting shooting emotion dysregulation and posttraumatic cognitions predict MHU among college women following a mass shooting. Ultimately, pre-shooting emotion dysregulation and post-shooting PTS and depressive symptoms were the most robust predictors of increased MHU. Although posttraumatic cognitions initially predicted increased therapy utilization specifically, this effect did not persist after considering pre-shooting emotion dysregulation. Contrary to expectations, pre-shooting emotion dysregulation did not interact with posttraumatic cognitions in predicting MHU, but instead weakened the relationship between depressive symptoms and therapy utilization and strengthened the relationship between age and therapy utilization. Overall, findings suggest that women who need mental health support after a mass shooting utilize services, however, pre-shooting emotion dysregulation hinders such utilization in the presence of depressive symptoms.

Posttraumatic Cognitions

Contrary to hypotheses, posttraumatic cognitions only predicted therapy utilization, but not medication utilization or the combination of the two service types. Notably, the relationship between posttraumatic cognitions and therapy utilization was unexpectedly positive, while controlling for the effects of time, age, PTS and depressive symptoms. These results suggest that the development of negative beliefs after the shooting (e.g., “I can’t deal with even the slightest upset”) may have been a factor driving the decision to utilize psychotherapy, and not a barrier as previously thought. Posttraumatic cognitions may have only predicted increased utilization of therapy, but not medication, because the women in this study may not have considered medication to be a viable solution to beliefs such as “my life has been destroyed by what happened” or “the world is a dangerous place.” This is consistent with evidence that women are inclined to choose psychotherapy over medication to address a trauma memory (Angelo et al., 2008). That said, posttraumatic cognitions were no longer a significant predictor of therapy utilization after the inclusion of pre-shooting emotion dysregulation. This may mean that the ability to cope with the distress and emotions produced by posttraumatic cognitions are a more important facilitator of MHU than the cognitions themselves.

Pre-Shooting Emotion Dysregulation

Pre-shooting emotion dysregulation also predicted an increase in the likelihood of all MHU on its own, controlling for the effects of time, age, PTS and depressive symptoms, and posttraumatic cognitions. Again, this relationship was unexpectedly positive, suggesting that pre-shooting emotion dysregulation functions to promote, rather than inhibit, utilization of mental health services. Although this conflicts with evidence that difficulty relating to and managing emotions deters MHU (Ciarrochi & Deane, 2001; Vogel et al., 2008), this study may have found the opposite because our sample was comprised entirely of college women. Compared to college men, college women report greater openness to emotions and more positive views of counseling (Komiya et al., 2000). As such, it is possible these characteristics facilitated MHU among women in the current study despite a history of emotion dysregulation. Alternatively, reporting more difficulties in emotion regulation prior to the shooting may have reflected a greater level of awareness and understanding of emotions (Nolen-Hoeksema, 2012). If so, these women may have then been more attentive to their emotions following the shooting and thus interpreted later difficulties in managing emotions as a sign they needed help. Regardless, the ability of pre-shooting emotion dysregulation in prospectively predicting MHU presents implications for clinicians working with similarly aged women exposed to shootings and perhaps other acts of mass violence. For instance, women exposed to mass violence may benefit from learning to rely more on certain regulatory strategies (e.g., reappraisal) over others (e.g., rumination) in a way that is flexible and context-dependent (see Nolen-Hoeksema, 2012).

Mental Health Symptoms

In support of the extant literature on MHU, PTS and depressive symptoms predicted more utilization of therapy and medication services separately and when combined (therapy and/or medication). This was true even after controlling for time, age, and posttraumatic cognitions. This finding is in line with the broader literature indicating that those with greater psychopathology, and therefore greater need for MH services, are more likely to utilize services (Smith et al., 2017). Although previous analyses using the same sample did not find any reliable predictors of MHU (i.e., Boykin & Orcutt, 2019), this discrepancy likely reflects differences across studies in statistical modeling, number of timepoints used, and selection of variables.

Interactions between Mental Health Symptoms and Pre-Shooting Emotion Dysregulation

The addition of pre-shooting emotion dysregulation in the models altered the relationships between PTSS, depressive symptoms, and MHU in two ways. First, while PTS and depressive symptoms remained significant positive predictors of both combined MHU and therapy specifically, only PTSS continued to predict medication usage on its own. It is unclear why a unique relationship between PTSS and medication utilization emerged, especially given evidence that psychotherapy is preferred over medication for the treatment of both depressive and anxiety disorders (including PTSD) (McHugh et al., 2013). Future work is needed to understand why PTSS may be uniquely tied to medication utilization above and beyond pre-existing problems with emotion regulation.

Second, the effect of emotion dysregulation on the relationship between depressive symptoms and psychotherapy utilization was negative, suggesting greater emotion dysregulation prior to the shooting dampened the positive effect of depressive symptoms on psychotherapy utilization specifically. Said differently, although post-shooting depressive symptoms predicted an increased likelihood of utilizing therapy, this was not true for people who had more difficulties regulating emotions prior to the shooting. The reasons for this interaction are likely varied. Given the uncontrollable and unpredictable nature of a mass shooting, the development of depressive symptoms among women who already had trouble regulating their emotions may have engendered a passive coping response that led to avoidance of therapy—an expression of depressive behavior known as learned helplessness (Vollmayr & Gass, 2013). Conversely, women who were predisposed to regulate their emotions more adaptively may have already been comfortable expressing and experiencing their emotions, and thus more likely to seek out opportunities to do so in a therapeutic context when depressive symptoms developed after the shooting. Another possibility for this interaction could be the tendency for women to engage in rumination (Nolen-Hoeksema, 2012), a repetitive thinking style considered key in maintaining depression (Kennair et al., 2017). Although not explicitly assessed in this study, our measure of pre-shooting emotion dysregulation may have still captured ruminative tendencies (e.g., Ando et al., 2020). This would explain the observed interaction because depressive rumination entails abstract thoughts about a depressed mood’s meaning, causes, and consequences that stymy problem-solving behaviors (e.g., MHU) (Kennair et al., 2017). Communities affected by mass shootings may potentially mitigate this deterring effect by promoting messages that normalize the experience of depression and other mental health difficulties, as strength-based campaigns that emphasize only resilience may alienate those who are struggling (e.g., “BOSTON STRONG;” Smith et al., 2017).

Age

Contrary to expectations, age did not predict any MHU before including pre-shooting emotion dysregulation. However, age and pre-shooting emotion dysregulation interacted to predict an increased likelihood of psychotherapy utilization, suggesting that older women with greater pre-shooting emotion dysregulation were more likely to utilize psychotherapy services. This aligns with the broader MHU literature that shows the effect of age on MHU varies by context (Smith et al., 2017). However, the effect of age in the current study was found only for psychotherapy, not medication use. It is possible that the effect of age is due to developmental factors or changes in resource availability across the lifespan that influence avoidance of emotional expression, such that older individuals are less likely to avoid psychotherapy despite past difficulties with regulating emotion. Furthermore, this effect may be nonlinear considering evidence that middle-aged individuals are more likely to use mental health services than younger or older individuals (Smith et al., 2017). More research is needed to understand the effect of age on MHU, both generally and specifically in the context of mass shootings.

Limitations and Future Directions

The current study is not without its limitations. To start, the unique features of our sample, the trauma, and the location in which it occurred restrict the generalizability of our findings. On the one hand, our sample of women is a strength given the paucity of research on help-seeking among women with PTSD (Smith et al., 2020). Nonetheless, our results may not generalize to men or individuals who identify as male, and our college sample limits the extent to which our results apply to older or community samples. For instance, following disaster, women utilize more MH services than men, and previous MHU work has utilized older samples and those with wider age ranges than ours (Smith et al., 2017). Additionally, younger individuals and women both show a preference for psychotherapy over medication (McHugh et al., 2013). Therefore, our sample’s characteristics may have biased relationships between predictors and MHU. Future researchers may explore whether the links between emotion dysregulation and MHU are moderated by sex, gender identity, or race/ethnicity. It might also be useful to study other moderators of the emotion dysregulation to MHU relationship (e.g., sleep, substance use/self-medication) to inform intervention efforts more precisely.

Regarding the study of a mass shooting, our results may not extend to other kinds of traumas or mass casualty events, such as natural disasters, but it is also worth discussing how findings may vary across different mass shooting events. MHU may have been biased in this study given the increased access to MH services on our campus after the influx of mental health resources in the wake of the shooting⁶. Relatedly, it is not known how differences between evolving guidelines for how to respond to such events could impact MHU. As such, researchers may consider how MHU varies by institutional and community responses, including how MHU may be inadvertently discouraged by strength-based campaigns (Smith et al., 2017).

Finally, although the finding that posttraumatic cognitions enabled greater MHU was unexpected, the a priori hypothesis that they would serve as a barrier to MHU was theoretical (see Smith et al., 2017). In other words, there were no previous data on which to base study predictions, and it seems an argument could be made for either directionality (i.e., negative beliefs either inhibiting or facilitating treatment usage). Future research may find the direction of this effect depends on variables such as sample, trauma type, or severity of exposure.

Conclusions

Among college women exposed to a mass shooting, pre-shooting emotion dysregulation, and post-shooting mental health symptoms, independently predicted an increased likelihood of utilizing mental health services up to 33 months later. These findings converge with the broader literature and offer hope—those with greater need for services are more likely to engage in them, and clinicians may find emotion regulation difficulties to be a useful starting place for treatment. An important caveat is that women with pre-existing difficulties in regulating emotions who went on to develop depressive symptoms after the shooting may have been discouraged from utilizing therapy. Since these women were less likely to enter a clinician’s office, another takeaway from this study is for the communities affected by mass violence: It may be important to raise awareness not only of strength and resilience, but also of struggle and the opportunity for recovery.

Clinical Impact Statement.

This study adds unique data on variables that facilitate and inhibit mental health service utilization among women following a campus mass shooting. Findings suggest that clinicians working with women exposed to mass violence may consider how longstanding problems with emotion regulation are interfering with functioning. Colleges, institutions, and communities affected by acts of mass violence may encourage greater mental health service utilization among those in need by acknowledging and normalizing the difficulty of adjusting afterward.