Abstract
Mass shooting episodes have increased over recent decades and received substantial media coverage. Despite the potentially widespread and increasing mental health impact of mass shootings, no efforts to our knowledge have been made to review the empirical literature on this topic. We identified 49 peer-reviewed articles, comprised of 27 independent samples in the aftermath of 15 mass shooting incidents. Based on our review, we concluded that mass shootings are associated with a variety of adverse psychological outcomes in survivors and members of affected communities. Less is known about the psychological effects of mass shootings on indirectly exposed populations; however, there is evidence that such events lead to at least short-term increases in fears and declines in perceived safety. A variety of risk factors for adverse psychological outcomes have been identified, including demographic and pre-incident characteristics (e.g., female gender and pre-incident psychological symptoms), event exposure (e.g., greater proximity to the attack and acquaintance with the deceased), and fewer psychosocial resources (e.g., emotion regulation difficulties and lower social support). Further research that draws on pre-incident and longitudinal data will yield important insights into the processes that exacerbate or sustain post-incident psychological symptoms over time and provide important information for crisis preparedness and post-incident mental health interventions.
Keywords
Over the past few decades, mass shooting episodes—defined as events involving one or more persons attempting to kill multiple people, and at least one unrelated person, in an area occupied by multiple unrelated persons (Blair & Martaindale, 2013)—have received substantial media coverage and captured public attention. Recent investigations have suggested that mass shooting episodes are becoming more frequent. For example, a report by researchers at the Texas State University identified 84 episodes in the United States between 2000 and 2010 and noted a trend toward increasing frequency of time (Blair & Martaindale, 2013). An investigation by the Mother Jones news organization, using a slightly different definition of mass shootings (events that were “senseless, random, or at least public in nature”), identified 70 episodes occurring between 1982 and 2012 and noted a recent surge in these events, with nearly half occurring since 2006.
In addition to the increasing frequency of mass shootings episodes, key events over the past decade have had unusually large numbers of fatalities and injuries—for example, 33 fatalities and 23 injuries in the 2007 Virginia Tech massacre, 12 fatalities and 58 injuries in the 2012 Aurora, Colorado theater shooting, and 28 fatalities and 2 injuries in the Sandy Hook elementary school shooting in Newtown, Connecticut (Follman, Aronsen, & Pan, 2014). In an epidemiological review of school shootings, Shultz, Cohen, Muschert, and Flores de Apodaca, (2013) noted that just three events (Columbine, Sandy Hook, and Virginia Tech) accounted for over half of the fatalities in the 215 incidents between 1990 and 2012.
Research indicates that exposure to assaultive violence, or learning that a close friend or loved one has faced such exposure, is associated with an increased incidence of a range of negative mental health outcomes, among them posttraumatic stress disorder (PTSD) and major depression (MD; e.g., Breslau et al., 1996; Lowe, Blachman-Forshay, & Koenen, 2015). It is therefore likely that mass shootings exert a psychological toll on their direct victims and members of the communities in which they took place. Moreover, media coverage of mass shootings and their aftermath reaches far beyond affected communities to the entire nation and beyond. As shown in the aftermath of the September 11 terrorist attacks (9/11), such indirect exposure can have mental health consequences (Henricksen, Bolton, & Sareen, 2010). For example, in the National Epidemiologic Survey of Alcohol and Related Conditions, indirect exposure to 9/11 through the media was associated with increased risk for mood, anxiety and substance use disorders, and PTSD, relative to no reported 9/11 exposure (Henricksen et al., 2010).
Despite the potentially widespread and increasing mental health impact of mass shooting episodes, no efforts to our knowledge have been made to synthesize the extant literature on this topic. In this article, we therefore aim to conduct a review of empirical investigations on the mental health consequences of mass shootings. We provide an overview of this body of research, including the prevalence and predictors of various mental health outcomes. Based on this review, we make recommendations for future research and post-incident interventions.
Method
We conducted a literature search in PsycInfo and PubMed databases, using both general terms (e.g., shooting, tragedy), and names of specific events (e.g., Columbine, Sandy Hook) compiled from comprehensive lists of mass shootings (Follman, Aronsen, Pan, & Caldwell, 2013; Henriques, 2013; Shultz et al., 2013). In our review, we used the aforementioned definition of mass shootings: events involving one or more persons attempting to kill multiple people, and at least one unrelated person, in an area occupied by multiple unrelated persons (Blair & Martaindale, 2013). We included events that took place in any country in the world. We limited our search to articles in peer-reviewed, English language journals that included quantitative indices of post-event mental health, including symptoms of psychiatric disorders (e.g., PTSD and MD) and general symptoms that cut across disorders (e.g., psychological distress and fear). We therefore excluded qualitative studies, and quantitative studies that focused solely on other outcomes (e.g., perceptions of social solidarity and coping strategies). In addition to studies identified through database searches, reference lists of articles on this topic were reviewed to identify additional studies.
The coding processes consisted of two main steps. First, we coded for study characteristics. We recorded the mass shooting event that was the focus of each study, including the year, location, context, demographic characteristics of the perpetrator, and numbers of injuries and fatalities during the incident, and whether the study focused on more than one event. We noted instances in which data from the same sample were used across different analyses to determine how many independent samples were there across the studies. In addition, we noted whether the sample was affected (defined as direct victims of the shooting or members of the community in which the shooting took place) or remote (defined as consisting primarily of members outside of the affected community, e.g., national samples or students at a university in a difficult region), basic characteristics of the sample (e.g., whether participants were students or emergency personnel), sample sizes, and timing of assessments.
Second, we coded the prevalence estimates and predictors of mental health outcomes in each study. We recorded the mental health outcomes assessed, the measures and diagnostic classification systems used, and the prevalence of psychiatric disorders. We then listed significant predictors of mental health outcomes and, in doing so, specified the mental outcome included in the analysis (e.g., diagnosis or symptom severity score, change in symptoms over time) and timing of assessment. Finally, we noted results that went beyond prevalence estimates and predictors, such as those focused on mechanisms leading to mental health outcomes.
Both authors formulated the article selection process, inclusion criteria, and coding scheme. The first author conducted the literature search, coded selected articles, and checked and rechecked coding for accuracy, and the second author oversaw the review process and provided regular feedback.
Study Characteristics
We identified a total of 49 studies on 15 different mass shooting incidents that took place from 1984 to 2008. Three of the studies focused on two different events combined. Table 1 provides basic information on each event (e.g., location, number of injuries, and fatalities) in chronological order. Thirteen events took place in the United States, and two took place in Finland. The majority of events (n = 9) were in a secondary school or university context, whereas the remainder took place in other locations (e.g., local businesses). In total, there were 27 independent samples (three from studies focusing on two different events combined). Of the 27 independent samples, 22 were classified as affected samples and 5 as remote samples (two affected and one remote from the studies focusing on two different events combined).
Summary of Mass Shooting Incidents and Characteristics of Peer-Reviewed Studies.
Note. Affected samples included participants who lived in the communities in which the incident occurred; participants did not have to be directly exposed to the event. Remote samples included members of other communities and nationally representative samples. Table does not include three studies that each focused on two different events (two affected samples, one remote sample).
Post-Shooting Mental Health Outcomes
Table 2 denotes the mental health outcomes included in each study, as well as the measures used to assess them and the classification system used to determine prevalence estimates if this information was available.
Summary of the Findings From Peer-Reviewed Studies on the Psychological Effects of Mass Shooting Incidents.
Note. Under each event, numbering is used to denote when studies used data from the same sample, with studies presented in chronological order by publication date. For example, for the 1984 elementary school shooting in Los Angeles, CA, two studies from the same sample are listed and labeled accordingly: (1a) Pynoos et al. (1987) and (1b) Nader, Pynoos, Fairbanks, and Frederick (1990). When there were multiple samples under each event, the samples were ordered chronologically by publication date of the first study within each sample. The table lists bivariate associations only if multivariate results were unavailable; if prevalence was reported in multiple publications for the same sample, the figure that included a higher # of participants is included. AA/AD = alcohol abuse or dependence; ASD = acute stress disorder; ASPD = antisocial personality disorder; AUD = alcohol use disorder; BAI = Beck Anxiety Inventory; BDI = Beck Depression Inventory; CDC = Centers for Disease Control; CES-D = 10-item Center for Epidemiologic Studies Short Depression Scale; DASS-21 = Depression and Anxiety Stress Scales-21 item version; DA/DD = drug abuse or dependence; DEQ = Distressing Events Questionnaire; DIS = diagnostic interview schedule; DTS = Davidson Trauma Scale; DUD = drug use disorder; FDAS = Four Dimensional Anxiety Scale; GAD = generalized anxiety disorder; GHQ = General Health Questionnaire; IES = Impact of Events Scale; MD = major depression; NIU = Northern Illinois University; PD = panic disorder; PDI = psychiatric diagnostic interview; PoMS = Profile of Mood States; PTGI = Posttraumatic Growth Inventory; PSS-SR = PTSD Symptom Scale-Self Report; PTSD = posttraumatic stress disorder; PTSD-RI =; Posttraumatic Stress-Reaction Index; PTSS = posttraumatic stress symptoms/severity; SASRQ = Stanford Acute Stress Reaction Questionnaire; STAI = State Trait Anxiety Inventory; TSQ = Trauma Screening Questionnaire; VT = Virginia Tech; WAI = Weinberger Adjustment Inventory; WHO-QOL = World Health Organization Quality of Life Scale–Brief.
aSame sample description as the preceding study.
Psychiatric Disorders and Prevalence Estimates
PTSD
Posttraumatic stress symptoms (PTSS) were reported in the majority of studies—36 studies from 18 independent samples. Eighteen of these studies (from 14 independent samples) included a prevalence estimate of PTSD. The lowest PTSD prevalence reported was 3% among parents of children exposed to an elementary school shooting 6–14 months post-incident, determined using conservative Diagnostic and Statistical Manual of Mental Disorders, Third Edition, Revised (DSM-III-R) criteria with the PTSD–Reaction Index (PTSD-RI; Schwarz & Kowalski, 1991a). The highest prevalence reported was 91% among children in the same study, using liberal (proposed) DSM-IV criteria and the children’s version of the PTSD-RI (Schwarz & Kowalski, 1991b).
Major depression
The second most commonly assessed psychiatric disorder was MD. MD symptoms were assessed in 16 studies using 10 independent samples, and the prevalence of MD was estimated in 8 studies using 7 independent samples. The lowest prevalence of MD was 4.9% among survivors of the 1991 cafeteria shooting in Killeen, TX, assessed 1 year after the event using the Diagnostic Interview Schedule for DSM-III-R (North, Smith, & Spitznagel, 1997). The highest prevalence was 71%, detected in a combined sample of Virginia Tech and Northern Illinois University (NIU) students 2 weeks after the attacks at their respective campuses, assessed using the Center for Epidemiologic Studies Short Depression Scale (Vicary & Fraley, 2010).
Other psychiatric disorders
In addition to PTSD and MD, prevalence estimates were provided for the following psychiatric disorders: generalized anxiety disorder (GAD; three studies and three independent samples): range: 0.0–0.9%; acute stress disorder (one study): 33%; alcohol-related conditions (e.g., alcohol abuse, alcohol dependence, alcoholism; four studies, four independent samples) range: 0–9%; drug use disorder (one study): range: 0–0.7%; panic disorder (two studies and two independent samples) range: 1–2.4%; adjustment disorder (one study): 9.1%; social phobia (one study): 3%; and antisocial personality disorder (one study): 0–0.8% (Classen, Koopman, Hales, & Spiegel, 1998; Johnson, North, & Smith, 2002; North, Smith, McCool, & McShea, 1989; North et al., 1997; Séguin et al., 2013; Trappler & Friednman, 1996).
Comparing Prevalence Estimates
There are at least four issues to consider when comparing the prevalence estimates of mental health outcomes across the studies in our review. First, variation in sample characteristics that, as discussed in more detail later, could influence post-event mental health, including variation in demographic characteristics and exposure to the incident. Second, there is wide variation in the timing of assessments, spanning from approximately 1 week to 32 months post-incident. As shown in studies with multiple waves (e.g., Nader, Pynoos, Fairbanks, & Frederick, 1990; North et al., 1997), the prevalence of psychiatric disorders tends to decrease over time, limiting the extent to which estimates at varying time points after different events can be compared. Third, different measures and diagnostic criteria were used across the studies, which could certainly affect prevalence rates. The influence of diagnostic criteria on prevalence rates was most clearly demonstrated in the aftermath of the 1988 elementary school shooting, wherein the prevalence of PTSD among child survivors at 6–14 months post-incident ranged from 8% to 91% and among their parents from 3% to 54% using conservative DSM-III-R criteria and liberal (proposed) DSM-IV criteria, respectively (Schwarz & Kowalski, 1991a, 1991b).
Other Mental Health Outcomes
Although most studies in our review reported on symptoms and prevalence estimates of psychiatric disorders, others included mental health outcomes that were not specific to any disorder. Fourteen studies (12 independent samples) utilized inventories or items assessing constructs that cut across disorders, including psychiatric disturbance, stress, grief, mood, emotional reactions, worry, and fear (e.g., McIntyre, Spence, & Lachlan, 2011; Suomalainen, Haravuorti, Berg, Kiviruusu, & Marttunen, 2011; Vuori, Hawdon, Atte, & Räsänen, 2013). Four studies (three independent samples) included positive indices of mental health—emotional and behavioral well-being, quality of life, and posttraumatic growth (Grills-Taquechel, Littleton, & Axsom, 2011; Hawdon & Ryan, 2011, 2012; Turunen, Haravuori, Punamäki, Suomalainen, & Marttunen, 2014). Finally, one study (Orcutt, Bonanno, Hanna, & Miron, 2014) utilized trajectory analysis of PTSS over seven waves to (one pre- and six post-incident) in the context of the NIU campus shooting. The majority of participants (60.9%) were in a minimal impact resilience trajectory, reporting low levels of symptoms at each wave and a small elevation in symptoms at the first post-incident wave only.
Assessing the Mental Health Impact of Mass Shootings
In reviewing the studies, we noted how investigators assessed whether participants’ symptoms reflected the impact of the mass shooting, versus ongoing mental health difficulties that might have been present before the incident took place. Certainly, personality disorders, which are conceptualized as enduring and pervasive phenomenon, are unlikely to be the result of a single event. Symptoms that were asked without reference to the mass shooting episode, including MD and GAD symptoms, also might have been present prior to the shooting incident. Although other outcomes were directly in reference to the mass shooting incident, including PTSD and acute stress disorder, they too could be influenced by preexisting psychopathology.
Three methods were used to address this issue. First, three independent samples included both pre- and post-incident data. The significance of pre- to post-incident changes in mental health reached was assessed in one of these samples (Hartnett & Skowronski, 2010). The investigators found that students’ ratings of four negative moods states (depression, tension, fatigue, and confusion) did not significantly differ between pre- and post-incident assessment; however, they reported significantly higher anger at the post-incident assessment. In the two other samples, significance of changes in mental health from pre- to post-incident was not assessed. However, descriptive data suggested short-term increases in depression and PTSS (e.g., Littleton, Axsom, & Grill-Taquechel, 2009; Orcutt et al., 2014).
Second, two studies assessed mental health impacts through comparison of samples who faced different levels of exposure, with results indicating higher symptomatology in directly exposed subsamples. Students at Joleka High School in Finland, where a shooting took place, had significantly higher posttraumatic distress and psychiatric disturbance than students in another city in Finland (Suomalainen et al., 2011). In the aftermath of the 1994 Brooklyn Bridge shooting, levels of PTSS, depression, and anxiety were higher in youth who directly experienced the attack than an age-matched comparison group of youth in the same community who were not directly exposed (Trappler & Friednman, 1996).
Finally, three studies drew on data from ongoing investigations of remote samples to assess the mental health impact of mass shootings, as well as their broader effects on indirectly exposed populations. For example, in the aftermath of the 1999 Columbine massacre, researchers drew on national school-based surveys with assessments spanning from pre- to post-incident in 1999 and found that students’ perceptions of safety at school declined after the attack (Addington, 2003; Brener, Simon, Anderson, Barrios, & Small, 2002). Interestingly, students’ reports of suicidal ideation and plans also significantly decreased after Columbine, perhaps indicative some form of widespread posttraumatic growth (e.g., increased appreciation of life), although this finding was limited to students living in rural and suburban areas (Brener et al., 2002). Two other investigations of remote college student samples have shown increased fears (e.g., of walking alone, of crime) and decreased perceived safety from pre- to post-incident (Kaminski, Koons-Witt, Thompson, & Weiss, 2010; Stretesky & Hogan, 2001).
Predictors of Adverse Mental Health Outcomes
Table 2 also includes significant predictors of adverse mental health outcomes in each of the study. These predictors can be divided roughly into three categories: (1) demographics and pre-incident characteristics, (2) incident exposure, and (3) post-incident functioning and psychosocial resources.
Demographic and Pre-Incident Characteristics
Demographic characteristics have frequently been included as predictors of mental health. Most consistently, female gender has been shown to be a predictor of post-incident psychological adversity, associated with increased odds of PTSD (e.g., North, Smith & Spitznagel, 1994; Suomalainen et al., 2011), higher levels of fear (Fallahi & Lesik, 2009), and lower levels of emotional and behavioral well-being (Hawdon & Ryan, 2012) in affected samples, and higher fear in a remote sample (Vicary & Fraley, 2010). One proposed explanation for this difference is that women are more likely to employ a ruminative coping style, increasing the severity and chronicity of their symptoms (e.g., McIntyre et al., 2011; Palus, Fang, & Prawitz, 2012).
Indicators of socioeconomic disadvantage, although less often included in post-incident studies, have also been consistently associated with poor mental health. For example, in the aftermath of the 1984 shooting at a fast food restaurant in California, higher prevalences of severe PTSD were documented among community members with lower income or who were unemployed, relative to their counterparts (Hough et al., 1990). Lower education was also associated with higher PTSS among survivors of the 1992 St. Louis courthouse shooting (Johnson et al., 2002). Among an adolescent sample, not living with two biological parents was associated with higher psychiatric disturbance (Suomalainen et al., 2011).
Other demographic characteristics have been less consistently associated with post-incident mental health outcomes. For example, younger and older age have each been associated with adverse outcomes in both community and college student samples (Fallahi & Lesik, 2009; Hough et al., 1990; Johnson et al., 2002; Kaminski et al., 2010). Racial/ethnic minority status has also been inconsistently associated with outcomes. Two studies have found non-White race to be associated with more severe post-incident symptoms (Fallahi & Lesik, 2009; Kaminski et al., 2010). In contrast, among female NIU students, African American ethnicity was associated with lower PTSS at 2–4 weeks post-incident, whereas Asian ethnicity was associated with higher PTSS at 8 months post-incident (Littleton, Kumpula, & Orcutt, 2011).
Few studies have investigated whether marital and parent status affect risk for post-incident adversity. Hough and colleagues (1990) found higher prevalences of severe PTSD among widowed, divorced, or separated persons (relative to married and single persons) and among adults with no children at home (relative to those with 1–3 or more children); however, the significance of these trends was not assessed. In contrast, Johnson, North, and Smith (2002) found that being married, relative to single or divorced, widowed, or separated, was associated with higher PTSS.
Only one study to our knowledge has investigated the role of genetic risk variants in predicting post-incident outcomes. Among a subsample of NIU female students who provided DNA samples, variants within the serotonin transporter gene were associated with significantly greater increases in PTSS from pre-incident to 2–4 weeks post-incident (Mercer et al., 2012). An additional finding showing that family history of mental illness predicted lack of recovery in MD at 3 years after the 1991 Killeen, TX, restaurant shooting also suggests a potential genetic contribution to post-incident responses (North, McCutcheon, Spitznagel, & Smith, 2002).
Similarly, the results of several investigations have shown that pre-incident psychological functioning is a strong predictor of post-incident functioning. A retrospective assessment of PTSD was found to predict post-incident PTSD among survivors of the 1991 Texas restaurant shooting (Sewell, 1996). In the aftermath of the same incident, reports of pre-incident psychiatric diagnosis were also associated with increased risk for PTSD among female survivors (North et al., 1994, 1997). Among adolescents exposed to the Joleka High School shooting, previous “mental support” from a nonguardian adult, a proxy for pre-event functioning, was associated with increased risk for psychiatric disturbance (Suomalainen et al., 2011). More recent studies in the aftermath of college shootings have drawn on pre-incident data and found significant associations between pre- and post-incident assessments of mental health (e.g., Grills-Taquechel et al., 2011; Littleton, Kumpula, et al., 2011).
A related set of findings has found that prior trauma exposure increases risk for psychological adversity in the aftermath of shooting incidents. For example, in the study of female Virginia Tech survivors, those who had experienced sexual victimization prior to the event were at increased risk of PTSS and depression 1 year after the shooting (Littleton, Grills-Taquechel, Axsom, Bye, & Buck, 2012). Among the NIU student sample, higher pre-incident trauma exposure was predictive of higher PTSS at two post-incident time points (Littleton, Kumpula, et al., 2011), as well as increased odds of a nonresilient trajectory of PTSS over time (Orcutt et al., 2014).
Incident Exposure
Indices of greater incident exposure, including proximity to an attack, acquaintance with the deceased, and higher scores on exposure inventories (with items assessing, e.g., seeing or hearing the events and physical injuries), have consistently been associated with more severe psychological reactions (e.g., Hawdon & Ryan, 2012; Littleton, Kumpula, et al., 2011; Pynoos et al., 1987). There is some evidence that the impact of milder forms of exposure on mental health decreases over time. For example, in the aftermath of the NIU shooting, moderate, severe, and extreme exposure (relative to no direct exposure) were associated with higher PTSS 2–4 weeks post-incident, whereas only extreme exposure was associated with higher PTSS at 8 months post-incident (Littleton, Kumpula et al., 2011).
In addition, emotional reactions during and after the incident have been found to predict later psychological responses. For example, students’ perceptions that they would be shot or were in danger during the 1988 Illinois elementary school shooting increased risk for PTSD (Schwarz & Kowalski, 1991a). Kumpula, Orcutt, Bardeen, and Varkovitzky (2011) assessed NIU students’ experiences of peritraumatic dissociation—altered awareness and depersonalization or derealization—during the event and found them to be predictive of higher PTSS 2–4 weeks post-incident. Other investigators have drawn on longitudinal data to show that earlier post-incident symptoms are positively associated with symptoms at later time points (e.g., Bardeen, Kumpula, & Orcutt, 2013; Smith, Abeyta, Hughes, & Jones, 2014). For example, acute stress disorder symptoms 1 week after the 1993 San Francisco office building shooting were associated with increased odds for PTSD 7–10 months post-incident. Being offered post-incident crisis support and early post-incident use of mental health services have also served as proxies of adverse initial responses, and have been predictive of higher psychiatric symptoms later on (e.g., Murtonen, Suomalainen, Haravuori, & Marttunen, 2012; North et al., 2002). Interestingly, Murtonen, Suomalainen, Haravuori, and Marttunen (2012) found that students’ perceptions that early crisis support was unhelpful was also associated with more severe symptoms, suggesting that survivors who do not benefit from early interventions might be at particular risk of long-term mental health problems.
How events are perceived and remembered have also been found to predict mental health outcomes. For example, after the 1992 St. Louis courthouse shooting, survivors’ perception that the incident had caused them a great deal of harm, that it was very upsetting, and that they had not recovered were each associated with higher PTSS (Johnson et al., 2002). A small study of school personnel in the aftermath of the 1988 Illinois elementary school shooting assessed participants’ changes in reports of emotional, life threat, and sensory experiences during the attack, and found that those whose reports became more intense over time (enlargement of recall) or did not become less intense over time (lack of diminishment of recall) tended to have more severe symptoms (Schwarz, Kowalski, & McNally, 1993).
Other investigations have focused on indirect exposure to events in affected and remote samples. For example, Joleka High School students who reported higher media exposure were at increased risk of post-incident psychiatric disturbance (Haravuori, Suomalainen, Berg, Kiviruusu, & Marttunen, 2011). In a remote college student sample, greater exposure to news media after the Virginia Tech shooting was associated with significantly higher psychiatric symptoms (Fallahi, Austad, Fallon, & Leishman, 2009). Significantly higher symptoms were also found among students who reported more time discussing the incident with family and friends, indicating that informal conversations may serve as an additional form of indirect exposure (Fallahi et al., 2009). Associations between indirect exposure and mental health could be due in part to self-selection and reverse causation, such that participants with more severe symptoms might be more likely to seek out media exposure and initiate conversations on the event. Further research employing longitudinal and experimental designs could address these considerations.
Post-Incident Functioning and Psychosocial Resources
As noted previously, early post-incident mental health responses have been found to prospectively predict later post-incident mental health responses (e.g., Bardeen et al., 2013). In a similar vein, different classes of psychiatric symptoms have been positively associated in cross-sectional assessments. For example, among female survivors of the 1991 Texas restaurant shooting, a diagnosis of any other psychiatric disorder was associated with increased odds of PTSD (North et al., 1994); among the full sample, there was also significant concordance between post-incident MD and PTSD. Studies have also documented associations between post-incident fears and PTSS in affected and remote samples (Fallahi & Lesik, 2009; Schwarz & Kowalski, 1991a).
Research has further suggested interrelations between mental and physical health problems. Higher prevalences of both mild and severe PTSD were observed in affected community members with fair or poor physical health, versus those with good or excellent physical health, in the aftermath of the 1984 California fast-food restaurant shooting. Among children directly exposed to the 1988 Illinois elementary school shooting, increased physical symptoms and visits to the school nurse were associated with increased risk of PTSD (Schwarz & Kowalski, 1991a). In the absence of longitudinal research, the direction of the relationship between post-incident mental and physical health remains unclear. Additional considerations are whether physical health symptoms are manifestations of poor mental health or whether preexisting physical health symptoms or pre- to post-incident changes in physical health account for significant associations.
In terms of psychosocial resources, research has focused on personality characteristics, beliefs and attitudes, coping styles, and social relationships as predictors of mental health outcomes. Personality characteristics that have been associated with adverse outcomes include guilt and resentment, insecurity, and anxiety sensitivity (Schwarz & Kowalski, 1992a; Stephenson, Valentiner, Kumpula, & Orcutt, 2009). Beliefs that events are random and uncontrollable, and punitive attitudes toward crime have also been associated with adverse outcomes, whereas greater self-efficacy, sense of meaning, spirituality and perceived benevolence of others have shown to be protective factors (Littleton et al., 2012; Smith et al., 2014; Vuori, Hawdon, Atte, & Räsänen, 2013).
Coping styles have been differentially associated with outcomes. Forms of coping that involve taking action, cognitive processing of the incident, and acceptance have been associated with lower levels of symptoms (e.g., North, Spitznagel, & Smith, 2001; Sewell, 1996), whereas ruminative and avoidant coping styles have been found to increase risk (e.g., Littleton et al., 2012). To some extent, means of coping in the aftermath of an incident could represent more pervasive difficulties and ways of approaching one’s experiences. In this vein, studies drawing on pre-incident data have found those emotion regulation difficulties and experiential avoidance, or the tendency to disengage from difficult emotions, sensations, thoughts, and memories, to be prospective predictors of post-incident symptoms (e.g., Bardeen et al., 2013; Kumpula, Orcutt, Bardeen, & Varkovitzky, 2011).
Similarly, indicators of fewer social resources (e.g., lower perceived social support and lower social solidarity) have been consistently associated with adverse post-incident outcomes (e.g., Hawdon & Ryan, 2012; Littleton et al., 2012; Suomalainen et al., 2011), and these differences could be driven in part by stable personality characteristics. For example, nonsecure attachment styles were significantly associated with higher PTSS and lower posttraumatic growth in relationships with others among students after a college campus shooting in Finland (Turunen, Haravuori, Punamäki, Suomalainen, & Marttunen, 2014).
On the other hand, researchers have been informed by Conservation of Resources (COR) theory (Hobfoll, 1989), which suggests that change in psychosocial resources, rather than stability, increases risk for adverse psychological outcomes. Supporting COR theory, Littleton and colleagues found that survivors’ reports of loss of life direction and pride, optimism, and interpersonal resources (e.g., companionship) were associated with higher PTSS and psychological distress (Littleton Axsom, et al., 2009; Littleton, Grills-Taquechel, & Axsom, 2009; Littleton, Kumpula, et al., 2011). Lower levels of resource gain, however, were significantly associated with higher psychological distress, suggesting that only negative changes in resources are associated with adverse outcomes.
Processes Leading to Adverse Mental Health Outcomes
In our review, we noted findings that went beyond documenting prevalence estimates and predictors by exploring the mechanisms contributing to mental health outcomes over time. A total of seven studies (from three independent samples) did so, all conducted within the past decade. In general, two types of mechanisms were explored across these studies. First, studies assessed pathways from pre-incident risk factors to post-incident mental health through post-incident risk factors. One example is a study in the aftermath of the NIU shooting showing that students’ pre-incident experiential avoidance was positively associated with their post-incident reports of peritraumatic dissociative experiences, which in turn were positively associated with PTSS (Kumpula et al., 2011). Another study of student survivors of the Virginia Tech attack found that prior sexual victimization was indirectly associated with PTSS and depression through lower self-worth, benevolence beliefs, and social support (Littleton et al., 2012).
Second, studies explored the ways in which risk factors and mental health influence each other over time through cross-lagged models. For example, a longitudinal study in the aftermath of the Virginia Tech shooting found reciprocal relationships between psychological distress and maladaptive coping from 2 months to 6 months, and 6 months to 1 year post-incident, suggesting that psychological distress could perpetuate itself by undermining coping mechanisms (Littleton, Axsom et al., 2011). This was not the case for the model containing PTSS, wherein the paths from PTSS to maladaptive coping reached statistical significance, but those from maladaptive coping to PTSS did not (Littleton et al., 2011). Another study employed cross-lagged modeling and found significant bidirectional relationships between emotion regulation difficulties and PTSS from pre-incident to 2–4 weeks after the NIU shooting, whereas only the path from emotion regulation difficulties to PTSS was significant from 2-4 weeks to 8 months post-incident (Bardeen et al., 2013). For the model containing general distress, only the path from distress to emotion regulation difficulties was significant from pre-incident to 2—4 weeks post-incident, whereas only the path from emotion regulation difficulties to distress was significant from 2-4 weeks to 8 months post-incident (Bardeen et al., 2013).
Taken together, the few studies in our review that investigated mechanisms contributing to post-incident mental health suggest a complex interplay between risk factors and outcomes that depend in part on the timing of assessment and outcome assessed. Further research exploring these processes will help to establish conceptual models of pathways to post-shooting mental health.
Discussion
The purpose of this review has been to review the extant literature on mental health in the aftermath of mass shootings. The research to date provides evidence that these events can have mental health consequences for victims and members of affected communities, leading to increases in PTSS, depression, and other psychological symptoms. The few studies on remote samples further suggest that these events can have at least short-term psychological effects, for example, increased fears and declines in perceived safety, on persons living far outside of the affected communities. These effects are not distributed equally, however, and research has identified several risk factors for adverse outcomes, including demographic characteristics (e.g., female gender and lower socioeconomic status), higher pre-event trauma exposure and psychological symptoms, greater direct and indirect event exposure, and lack of psychosocial resources (e.g., emotional regulation difficulties, experiential avoidance, and low social support).
Although the extant body of research offers some conclusions about the mental health effects of mass shootings, more research is needed to better understand the mechanisms through which risk and protective factors contribute to longer term outcomes. There is a particular need for studies in the aftermath of high-impact events. For example, we identified no studies of affected samples in two recent events with unusually high numbers of casualties—the 2012 shootings at the Aurora movie theater and Sandy Hook elementary school. Researchers and institutional review boards might be hesitant to conduct studies in the aftermath of such events due to concerns about retraumatizing or otherwise taking advantage of victims. Notably, however, the majority (85%) of NIU students who completed a post-incident experimental study reported that they would participate in the study again (Fergus, Rabenhorst, Orcutt, & Valentiner, 2011). Although a single data source, these results suggest some evidence that study participation might not have adverse effects. Research in the aftermath of other traumatic events suggests that some persons who have experienced trauma may actually derive benefits from research participation (e.g., Griffin, Resick, Waldorp, & Mechanic, 2003).
Further studies that draw on pre-incident data from ongoing investigations would provide greater insight into the role of pre-event functioning and trauma exposure on psychological responses. In a similar vein, studies that include multiple waves of follow-up data would allow for an enhanced understanding of longitudinal patterns of responses and the processes that lead to chronic symptoms. The research could also be enriched by further studies that include positive outcomes (e.g., resilience and posttraumatic growth) and incorporate additional sources of data, including genetic variants and information on community characteristics and resources. In addition to further research on affected samples, more investigations are needed to understand the broader impact of mass shootings on unaffected communities. It was notable that only 5 of the 28 identified samples focused on remote populations; as such, our conclusions regarding the widespread mental health effects of mass shootings are tentative at best.
Additional research would also inform interventions to prevent and treat post-incident mental health problems. The first step in prevention would be to decrease the likelihood that mass shooting events occur in the first place. The rarity of these events precludes the use of statistical modeling to predict their occurrence (Swanson, 2011). Researchers have accordingly advised against the use of risk profiles, which have the potential for false positives bias, and stigmatization (Reddy et al., 2001). Instead, a deductive threat assessment approach in which a team of professionals gathers information about a particular case to assess the likelihood that a violent episode will occur, and formulates a response based on that assessment, have been proposed (e.g., Cornell & Allen, 2011; Reddy et al., 2011). Others have suggested the need for promoting positive school climates in which there is open dialogue between students, teachers, and administrators to reduce the likelihood of school shootings (Mulvey & Cauffman, 2001), and efforts to improve access to and continuity of high-quality mental health services for people with serious mental illness to prevent violent behaviors among this group (Swanson, 2011). Finally, others have noted the high rates of gun ownership and firearm mortalities in the United States compared to those in other developed countries as evidence of the need for a range of efforts to reduce gun access (e.g., Shultz, Cohen, Muschert, & Flores de Apodaca, 2013). An observational study showing declines in firearm-related deaths, suicides and homicides, and a complete absence of mass shooting episodes, in Australia a decade after gun law reforms aiming to limit civilian access to semiautomatic and pump-action shotguns and rifles provides support for such efforts (Chapman, Alpers, Agho, & Jones, 2006).
In addition to preventing mass shootings, it would be important to develop interventions to address mental health problems in their wake. Trained crisis response teams that establish safety, evaluate the psychological needs of victims, connect survivors with a range of services to meet their needs, and evaluate response efforts have been proposed to mitigate the effects of school violence (Crepeau-Hobson, Sievering, Armstrong, & Stonis, 2012). Hobfoll and colleagues (2007) have also identified five empirically supported principles for mental health responses to mass trauma—promoting a sense of safety, calming, a sense of self- and community-efficacy, connectedness, and hope.
The empirical research also lends support for approaches that identify survivors most at risk of adverse outcomes, including women, persons of lower socioeconomic status, those who faced higher levels of exposure, and persons lacking strong social support networks. Furthermore, extant studies suggest interventions that enhance emotion regulation and active coping skills and that encourage engagement with and acceptance of emotions, thoughts, memories, and sensory experiences (e.g., Metz et al., 2013). These skill-building interventions could be part of the standard curriculum and could promote resilience after a range of traumatic events and stressors.
The recommendations for future research, policy, and practice are made with caution, given the limitations of this review. For example, although our efforts to identify studies meeting our inclusion criteria were exhaustive, we did not track such data as the total number of articles identified across all of our searches, and the number that were dropped based on each criterion. In addition, although the first author checked and rechecked coding for accuracy, articles were not double-coded, and a system for establishing the reliability of the article screening and coding process was not developed and applied. Taken together, these limitations suggest a need for a more systematic review, particularly as additional literature on this topic is published. Further reviews could also apply broader inclusion criteria, for example, to provide insight into the influence of direct and indirect exposure to mass shootings on other domains of functioning (e.g., physical health and social functioning). Inclusion of non-English language articles could shed additional light on the influence of mass shootings on mental health cross-culturally. Qualitative studies could also be integrated to provide a richer sense of the patterns and predictors of mental health in the aftermath of shootings. Our review of predictors also focused on factors that account for significant variation in post-shooting mental health within each study, and we did not attend to potential between-study sources of variation, such as differences in samples (e.g., age-group and extent of exposure), procedures (e.g., timing of assessment), and event characteristics (e.g., single vs. multiple shooters, number of injuries and fatalities). Future research could examine within- and between-study variability simultaneously using meta-analytic techniques. Finally, we did not compare the prevalence estimates and predictors of mental health outcomes in the aftermath of mass shootings to those in the aftermath of other traumatic events. Inclusion of mass shootings on trauma inventories in future epidemiological studies would provide insight into this issue.
In summary, the limited research suggests that mass shooting incidents can lead to an array of mental health problems in survivors and members of affected communities. Furthermore, they have been associated with increased fears and decreased perceptions of safety in indirectly exposed populations. A variety of risk and protective factors have been identified, including demographic characteristics, pre-event trauma exposure and functioning, event exposure, and psychosocial resources. Further research that explores the processes contributing to long-term psychological responses will yield important implications for post-incident interventions to reduce mental health impacts.
Footnotes
Declaration of Conflicting Interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The author(s) received no financial support for the research, authorship, and/or publication of this article.