Post a (200 word APA Format) an analysis of implications of vicarious trauma, burnout, and compassion fatigue for counselors and first responders. Be specific and provide examples

The Effects of Vicarious Exposure to the Recent Massacre at Virginia Tech

Carolyn R. Fallahi and Sally A. Lesik Central Connecticut State University

The authors examined whether exposure to the April 2007 Virginia Tech school shootings would increase symptoms of acute stress in students at another university who were not personally involved, but who followed the case vicariously through news media. The authors ran a series of regression analyses using multinomial logit (MNL) models, a methodology which can be used for dealing with categorical outcome variables. The authors found that as TV viewing of the Virginia Tech case increased, the probability that a student would respond with moderate or acute stress symptoms also increased. The authors were able to describe the magnitude of the relationship between vicarious exposure to the Virginia Tech case and acute stress symptomatology.

Keywords: Virginia Tech tragedy, vicarious exposure, multinomial logit models, acute stress disorder, posttraumatic stress disorder

On April 16, 2007, Cho Seung-Hui, a 23-year- old student, murdered 32 people and wounded 25 others before committing suicide at Virginia Poly- technic Institute (Virginia Tech) in Blacksburg, Virginia. It was the deadliest campus shooting in the history of the United States. The media cov- erage of this incident was extensive with daily images of the shooter and victims available on news stations throughout the country for weeks following the incident. We wondered about the effects on students at another university who were not personally involved with the Virginia Tech shootings, but who followed the case vicariously through news media.

Responses to Other Tragedies

The destruction of the World Trade Center is probably the most “imaged disaster in human history” (Mason, 2004, p. 1). Symptoms were not contained to New York, people living

throughout the United States experienced symp- toms of stress as a direct result of 9/11 (Schuster et al., 2001; Stein et al., 2004). A study of children who experienced indirect exposure to 9/11 through the media showed increased levels of worry and posttraumatic stress at levels com- parable to those in children experiencing the disasters directly (Lengua, Long, Smith, & Meltzoff, 2005). This research confirms Pfeffer- baum, Gurwitch, et al. (2000) and Pfefferbaum, Seale, et al. (2000) who found that children experienced symptoms of posttraumatic stress disorder (PTSD) 2 years after the Oklahoma City bombing if they lived within 100 miles of the bombing and had a connection to someone who either died or was hurt in the bombing. Spang (1999) found few symptoms of distress among adults living 900 miles away from Okla- homa City as compared to the Oklahoma City residents 6 months after the Oklahoma City bombing. Stein et al. (2004) found that 2–3 months following the 9/11 terrorist attacks, many adults continued to show stress-related symptoms as a direct result of the attacks. Fol- lowing the explosion of the Shuttle Challenger in 1986, Terr, Block, Michel, Shi, Reinhardt, and Metayer (1999) studied the responses of children living in the east and west coast both 5–7 weeks and 14 months after the explosion. They found that children who watched the Chal- lenger explode and cared more about the teacher on the Challenger, demonstrated more symp- toms of PTSD initially.

Carolyn R. Fallahi and Sally A. Lesik, Department of Psychology and Department of Mathematical Sciences, Central Connecticut State University.

We thank Lisa Leishman, Melissa Cotter, and Sara R. Fallahi, who coded much of the data for this study; Sally K. Laden, who provided editorial assistance; and Dr. Bradley Waite for his comments on earlier drafts of this article.

Correspondence concerning this article should be ad- dressed to Carolyn R. Fallahi, Central Connecticut State University, Department of Psychology, 208 Marcus White Hall, 1615 Stanley Street, New Britain, CT 06050-4010. E-mail: [email protected]

Psychological Trauma: Theory, Research, Practice, and Policy © 2009 American Psychological Association 2009, Vol. 1, No. 3, 220 –230 1942-9681/09/$12.00 DOI: 10.1037/a0015052

220

On April 20, 1999, Littleton, Colorado expe- rienced the deadliest act of school violence re- corded in history prior to the Virginia Tech case. Two students killed 12 students and one teacher while wounding 21 others before com- mitting suicide. The images of this tragedy were aired continuously (Addington, 2003). Colum- bine students throughout the country reported fear of victimization at school (Brooks, Schiraldi, & Ziedenberg, 2000). Addington (2003) found students responded to Columbine with an increase in fear at school, albeit a small increase. This brings up an important question: does exposure to a disaster vicariously increase psychiatric symptomatology in samples of peo- ple not directly exposed to the tragedy?

Media Violence Research

During the last half century, the negative short-term and long-term effects of media vio- lence have been well documented. Exposure to media violence has been associated with the formation of aggressive scripts in memory, hos- tile attributional biases, and aggressive beliefs (Huesmann, Moise-Titus, Podolski, & Eron, 2003). There is also a link between a heavy diet of media violence and later aggression (Bush- man & Anderson, 2001; Paik & Comstock, 1994). Further, the effects of media violence include the augmentation of negative mood states (Caprara, Renzi, Amolini, D’Imperio & Travaglia, 1984), including aggressive emo- tions (Anderson et al., 2003). Viewing indirect aggression on TV early in life has been shown to predict actual indirect aggression in real life (Huesmann et al., 2003; Coyne & Archer, 2005).

Vicarious exposure in terms of the number of hours of media viewing of 9/11 was studied by Blanchard et al. (2004). They found that vicar- ious exposure, as measured by the number of hours of media viewing of 9/11 events, pre- dicted acute stress disorder scores in two sam- ples of college students who were not geograph- ically positioned near New York City. Pfeffer- baum, Gurwitch, et al. (2000) also found that media exposure and indirect interpersonal ex- posure, as defined by having a friend who knew someone hurt or killed in the bombing, were significant predictors of symptomatology. Gil- Rivas, Holman, and Silver (2004) found that adolescents who were indirectly exposed

through media coverage to the events of 9/11 experienced mild to moderate acute stress symptoms, especially when there was signifi- cant conflict with their parents. Propper, Stick- gold, Keeley, and Christman (2007) found that there was a strong relationship between media exposure for 9/11 and changes in dream features following the attack. Specifically, they found that with every hour of TV viewing, this re- sulted in a 5% to 6% increase in September 11th dream references. Talking about the events with friends and relatives was not significantly re- lated to specific dream references. The authors concluded that the media may have a “deterious impact on the emotional well-being of U.S. citizens in the aftermath of September 11” (p. 340). Pfefferbaum, Gurwitch, et al. (2000) found that following the Oklahoma City bomb- ing, children who lost a friend reported signif- icantly more symptoms of PTSD than those who lost an acquaintance. However, when look- ing at children within the community, there was not a significant difference between those per- sonally involved in the bombing, for example, knew someone that was hurt or killed, and those children not personally involved. The authors speculate that this may be due to the fact that children were watching only bomb-related pro- gramming that aired for months after the explo- sion. They speculated that high levels of media coverage contributed to the symptoms of the children within the community.

Rosen, Quyen, Cavella, Finney, and Lee (2005) found that exposure to 9/11 images did not change the average symptoms of distress in a group of chronic PTSD patients. However, they reported an increase in their perceptions of stress and the authors speculate that the 9/11 events may have caused the patients to misat- tribute fluctuations in chronic symptoms to the recent terrorist attacks.

The News Media

News coverage can be a vehicle through which individuals can experience indirect vic- timization (Warr, 1994). The news media can play an important role in defining ‘what is a public tragedy’ (Balk, 2004). The news media can present a distorted image of the risks that individuals face in response to tragedies like Columbine’s school shootings (Brooks, Schiraldi, & Zeidenberg, 2000). Public trage-

221EFFECTS OF VICARIOUS EXPOSURE TO VIRGINIA TECH

dies affect our views on life and shatter our basic assumptions about life (Balk, 2004). Chiricos, Padgett, and Gertz (2006), as well as Chiricos, Eschholz, and Gertz (2000) found a relationship between watching TV news cover- age of traumatic events and an increase in fear.

We were interested in studying the subjective reactions of students at a large state university in the northeast following the shooting at Vir- ginia Tech. We hypothesized that there would be a significant relationship between vicarious exposure through the news media to the Vir- ginia Tech case and acute stress symptoms.

Method

Student Participants

We recruited 145 female and 167 male par- ticipants from undergraduate and graduate psy- chology courses. Students who were enrolled in the introductory psychology and life span de- velopment courses fulfilled course research re- quirements by participating in psychology stud- ies. Students signed up for any of a number of different studies, and all received extra credit for participating.

Demographics

The participants primarily consisted of fresh- man, sophomores, and juniors (96.4%), with a mean age of 19.56 years (SD � 3.72), and a median grade point average of 3.00, who iden- tified themselves as single (96.7%). This was a primarily Caucasian sample (82.1%), with 9.2% self-identifying as African American/ Black, 5.1% as Hispanic or Latino, 1.5% as Asian, and 0.9% as Other.

Survey

Participants were asked to estimate the num- ber of hours spent viewing news coverage of the shootings at Virginia Tech that included both TV and internet viewing of this case. They participated approximately 3 weeks after the incident occurred and provided an estimate of hours spent viewing news coverage since the event. In addition, they were asked to rate their own symptoms of depression, anxiety, and stress-related symptoms, (scale of 1 [not at all] to 5 [very much so]). Symptoms were converted

to a categorical scale by rating 1 (no symptoms), 2–3 (moderate symptoms), and 4 –5 (acute symptoms). This conversion was done for two reasons. First, by combining outcome catego- ries, we were able to obtain more efficient esti- mates by combining indistinguishable catego- ries. Second, these conversions aligned with the severity of the given symptom.

The survey measures included self-ratings of the following symptoms of acute stress disorder as taken from the symptom list presented in the Diagnostic and Statistical Manual IV-TR (American Psychiatric Association, 2000):

(1) Intrusive Thoughts: Experiencing thoughts associated with the Virginia Tech case.

(2) Sleep Disturbance: Experiencing sleep disturbance, for example, trouble falling asleep, trouble staying awake at night, sleeping longer than usual.

(3) Appetite Disturbance: Experiencing ei- ther an increase or decrease in appetite.

(4) Nightmares: Experiencing nightmares about the Virginia Tech case.

(5) Fear: Increasing feelings of fear that something like the Virginia Tech case could either happen again somewhere else or at this university.

(6) Stomach Upset: Experiencing gastroin- testinal distress, for example, upset stomach, butterflies in your stomach, and so forth

(7) Depressive Symptoms: Experiencing a sad or down mood.

(8) Symptoms of Suicide: Experiencing an increase in suicidal ideation as a direct result of the Virginia Tech case.

(9) Disorganization: Feeling disorganized, confused, and “in a daze.”

(10) Alcohol and Drugs: An increase in al- cohol or drug use.

(11) Replaying the Event: Reliving the trauma of the Virginia Tech Case Invol- untarily.

222 FALLAHI AND LESIK

(12) Anger: Experiencing symptoms of an- ger as a direct result of the Virgina Tech case.

(13) Guilt: Experiencing symptoms of guilt as a direct result of the Virginia Tech case.

Predictor Variables

The predictor FEMALE represents the respon- dents’ self-identified gender (FEMALE � 1 for female respondents, and FEMALE � 0 for males).

The predictor AGE Is the respondent’s age in years.

A respondent’s self-identified race/ethnicity was coded as the dummy predictor MINORITY (MINORITY � 1 if the respondent self- identified as nonwhite; and MINORITY � 0 when the respondent self-identified as white).

The predictor HOURS represents the number of hours of TV coverage of the VT shootings.

Response Variables

There were 14 response variables that repre- sented different symptoms. The self-reported severity for each of the symptoms was initially represented on a 5-point Likert scale and were combined as a categorical variable on a scale of no symptoms, moderate symptomatology, and acute symptomatology. The 14 symptoms stud- ied represented symptoms of acute stress disor- der and posttraumatic stress disorder (PTSD) based on the DSM–IV–TR manual (American Psychiatric Association, 2000). Table 1 gives the distribution of the three combined catego- ries based on each of the 14 symptoms.

Statistical Analysis

We ran a series of regression analyses us- ing the multinomial logit (MNL) model. We chose the MNL because it is the preferred method for regression models in which the response variable has more than two catego- ries. Even though the initial Likert scale was ordinal, we chose the MNL because it does not rely on the parallel regression assumption (Long & Freese, 2003). The categories were collapsed to represent experiencing none, moderate, or acute symptoms. The initial ba-

sic MNL model equation used for each symp- tom is presented in Equation 1.

mlogit(Symptom) � �0 � �1FEMALE

� �2AGE � �3MINORITY � �4HOURS � ε.

(1)

More formally, the MNL fodel can be written as a series of binary logits as in Equation 2.

ln(Symptom�m)/b�X � ln P(Symptom � m�X P(Symptom � b�X

� ��m�/bX � �0 �m�/b � �1

�m�/bFEMALE

� �2 �m�/bAGE � �3

�m�/bMINORITY

� �4 �m�/bHOURS. (2)

Parameter b represents the base category, and m ranges from 1 to J, where J represents the number of categorical outcomes, and X de- scribes the vector of control variables. For this study, J � 3 because there are three distinct outcome categories representing the three cate- gories of symptomatology.

Since there can be more than one solution for the estimated coefficients in a MNL analysis, one of the coefficients needs to be set to 0 in order to identify the model. This is category is referred to as the base category, and it does not

Table 1 Distribution of Student Response to the Different Symptoms

Symptom None Moderate Acute Missing

Intrusive thoughts 204 95 12 28 Sleeping 263 40 10 26 Appetite 277 33 3 26 Distraction 241 59 13 26 Nightmares 270 38 5 26 Fear 200 88 25 26 Butterflies in stomach 268 35 10 26 Depression 261 38 14 26 Suicide 305 5 3 26 Disorganization 261 43 8 27 Alcohol/drugs 278 28 5 28 Replaying the event 262 42 7 28 Anger 245 55 11 28 Guilt 274 30 7 28

223EFFECTS OF VICARIOUS EXPOSURE TO VIRGINIA TECH

matter which category is chosen because the predicted probabilities will be the same regard- less of which parameterization (or base cate- gory) is used (Long, 1997). For our study, we set the base category to None, thus the MNL model would fit model Equations 3 and 4 for each of the 14 symptoms.

ln(Symptom � ModerateNone)

� �iX �Moderate)/None (3)

ln(Symptom � AcuteNone) � �iX �Acute)/None (4)

Where �iX ���/None is the ith coefficient for

category k using the category None as the base category. The remaining coefficients are then estimated with respect to this base category.

Tests for Dependent Categories

We used a Wald test to determine whether the outcome categories could be combined for the 14 regression models (Long & Freese, 2003). The advantage to combining outcome categories is to combine categories which are indistinguishable in order to obtain more effi- cient estimates (Long & Freese, 2003). The null hypothesis for such a test is that two categories can be combined (Long, 1997), and thus there is no difference in the separate odds ratios for the combined categories. For all of the 14 regres- sion models, we found that categories 2 and 3 could be collapsed into a single category ( p � .10). We are calling this combined category Moderate since this category represents a mod- erate reaction to the given symptom. We also found that categories 4 and 5 could be collapsed ( p � .30), and called this category Acute since it represents a severe reaction to the given symptom. By keeping the 5-point scale, the pairwise comparisons between different catego- ries would be numerous and difficult to under- stand, while not giving us any new information. Further, the Wald test showed that there was not a significant difference between these two cat- egories and combining them made conceptual sense.

Testing the Effects of Independent Variables

Since there are J � 3 outcome categories, then there are J – 1 � 2 coefficients associated with each independent variable. By using cate- gory None as the base category, there are two coefficients that would be associated with each predictor namely �i

�Moderate)/None, and �i �Acute)/None.

The null hypothesis that predictor i does not affect the outcome can be written as:

H0:�i (Moderate)/None � �i

(Acute)/None � 0

Table 2 shows the individual parameter esti- mates and standard errors for all 14 regression analyses. Table 3 gives the results using a like- lihood ratio test that was used to determine which groups of predictors are significant for each of the 14 regression analyses (Long, 1997; Long & Freese, 2003).

Similar to Mickey and Greenland’s (1989) criteria for including variables in a logistic re- gression analysis, we used p values that are less than 0.10 as the initial criteria for testing the effects of the independent variables. Using more traditional significance levels such as p � .05 may fail to identify variables that are sig- nificant for some of the categories but not for others.

Perfect Predictions

The MNL model cannot be used if there are perfect predictors. In other words, if there is no variability in a predictor variable within one of the outcome categories, this produces excessively large standard errors and unstable parameter estimates. Perfect predictors can occur if there are only a few observations in a given outcome category and if one or more of the predictor variables does not vary within this outcome category. This scenario can be seen in the Acute category for the predictor variable MINORITY for the symptoms of Ap- petite, Nightmares, Alcohol and Drugs, Re- playing the Event, and Guilt, and for the pre- dictor variable FEMALE for the category of Suicide as is presented in Table 2. By remov- ing these predictor variables and running the regression analysis along with performing the appropriate likelihood ratio test, the effect of

224 FALLAHI AND LESIK

these variables on the other outcome catego- ries gave results similar to those presented in Table 3.

The IIA Assumption

The MNL relies on the assumption known as the independence of irrelevant alternatives (IIA). Essentially, the IIA assumption stipulates that adding or deleting outcome categories will not affect the odds ratio of the remaining out-

comes (Green, 2003; Maddala, 1983). We ran the Hausman test (Hausmann & McFadden, 1984) for all 14 regression analyses with the three combined categories and there was no evidence to suggest that the IIA assumption has been violated ( p � .80).

Interpretation of the Results

Although the MNL can be considered as a simple extension of the logistic regression

Table 2 Estimated MNL Coefficients and Standard Errors [in Brackets] for the MNL Model Comparing All Possible Combinations for Each of the Three Response Categories (N � 339)

Symptom Categories FEMALE AGE MINORITY HOURS CONS

Intrusive thoughts None Moderate 0.171 [0.268] �0.085 [0.059] 0.420 [0.374] 0.024 [0.025] 0.691 [1.171]

None Acute �0.386 [0.670] �0.080 [0.162] �0.405 [1.054] 0.108�� [0.035] �1.630 [3.214] Moderate Acute �0.558 [0.685] 0.005 [0.167] �0.825 [1.064] 0.084� [0.035] �2.322 [3.319]

Sleeping None Moderate 0.225 [0.357] �0.156 [0.104] �0.037 [0.540] 0.035 [0.029] 0.988 [2.032] None Acute 0.117 [0.766] �0.035 [0.154] 0.642 [0.919] 0.097�� [0.035] �3.500 [3.110] Moderate Acute �0.109 [0.818] 0.121 [0.182] 0.679 [1.021] 0.062 [0.039] �4.487 [3.635]

Appetite None Moderate �0.255 [0.400] �0.037 [0.078] �0.041 [0.581] 0.031 [0.032] �1.400 [1.556] None Acute �0.078 [2.363] �0.178 [0.707] 21.815 [13.870] 0.143� [0.067] �23.533 [N/A] Moderate Acute 0.176 [2.383] �0.141 [0.710] 21.856 [13.939] 0.111 [0.070] �22.133 [N/A]

Distraction None Moderate 0.271 [0.317] �0.090 [0.075] 0.090 [0.464] 0.007 [0.030] 0.170 [1.482] None Acute �0.046 [0.604] 0.004 [0.084] 0.532 [0.738] 0.084�� [0.031] �3.454� [1.744] Moderate Acute �0.318 [0.652] 0.094 [0.110] 0.442 [0.821] 0.077� [0.038] �3.624 [2.224]

Nightmares None Moderate 0.656† [0.375] �0.242� [0.124] �0.161 [0.577] 0.042 [0.026] 2.322 [2.385] None Acute 0.651 [0.933] �0.469 [0.434] �35.479 [N/A] 0.073 [0.071] 4.691 [8.205] Moderate Acute �0.005 [0.983] �0.227 [0.447] �34.319 [N/A] 0.031 [0.074] 2.369 [8.455]

Fear None Moderate 1.200�� [0.291] �0.060 [0.049] �0.655 [0.477] 0.051† [0.028] �0.377 [0.995] None Acute 1.838�� [0.582] �0.145 [0.138] �0.037 [0.717] 0.114�� [0.034] �0.988 [2.719] Moderate Acute 0.639 [0.604] �0.085 [0.140] 0.618 [0.768] 0.063† [0.032] �0.610 [2.776]

Butterflies in stomach None Moderate 0.286 [0.391] �0.085 [0.095] 0.181 [0.550] 0.045 [0.031] �0.735 [1.884]

None Acute �0.152 [0.739] 0.020 [0.094] 0.403 [0.915] 0.111�� [0.035] �4.413� [1.975] Moderate Acute �0.438 [0.807] 0.105 [0.131] 0.221 [1.020] 0.066† [0.039] �3.677 [2.678]

Depression None Moderate 0.070 [0.366] �0.169 [0.110] �0.604 [0.652] 0.021 [0.032] 1.399 [2.143] None Acute �0.418 [0.604] 0.018 [0.075] 0.107 [0.785] 0.073� [0.031] �3.474� [1.560] Moderate Acute �0.487 [0.680] 0.187 [0.131] 0.710 [0.979] 0.052 [0.403] �4.873† [2.606]

Suicide None Moderate �0.913 [1.189] �0.411 [0.447] 2.104� [1.050] �0.022 [0.097] 3.546 [8.445] None Acute �43.266 [N/A] 0.035 [0.385] 2.300† [1.380] 0.108† [0.058] �6.139 [7.894] Moderate Acute �36.354 [N/A] 0.447 [0.584] 0.196 [1.709] 0.129 [0.109] �9.685 [11.479]

Disorganization None Moderate 0.051 [0.360] �0.084 [0.089] �0.341 [0.575] 0.061� [0.025] �0.428 [1.753] None Acute �0.973 [0.886] 0.062 [0.071] 0.760 [0.899] 0.051 [0.044] �4.780�� [1.564] Moderate Acute �1.024 [0.934] 0.146 [0.112] 1.101 [1.028] �0.011 [0.045] �4.367† [2.317]

Alcohol/drugs None Moderate �0.560 [0.433] �0.038 [0.084] �0.399 [0.666] 0.026 [0.030] �1.303 [1.689] None Acute 1.179 [1.166] �0.255 [0.365] �31.707 [N/A] �0.061 [0.171] 0.468 [6.998] Moderate Acute 1.739 [1.229] �0.217 [0.373] �31.308 [N/A] �0.087 [0.173] 1.770 [7.167]

Replaying None Moderate 0.119 [0.363] �0.063 [0.078] �1.852� [0.900] 0.074�� [0.027] �0.798 [1.569] None Acute 0.859 [0.883] �0.203 [0.276] �32.704 [N/A] 0.070 [0.074] �0.270 [5.364] Moderate Acute 0.741 [0.930] �0.139 [0.284] �35.858 [N/A] �0.004 [0.076] 0.527 [5.525]

Anger None Moderate �0.208 [0.332] 0.006 [0.047] 0.173 [0.456] 0.090�� [0.028] �1.956 [0.962] None Acute 0.281 [0.713] 0.019 [0.086] �0.137 [1.027] 0.130�� [0.040] �4.411� [1.825] Moderate Acute 0.489 [0.747] 0.013 [0.093] �0.310 [1.059] 0.040 [0.036] �2.456 [1.977]

Guilt None Moderate �0.234 [0.437] �0.130 [0.123] 0.394 [0.568] 0.027 [0.029] 0.174 [2.405] None Acute 0.233 [0.856] �0.086 [0.219] �35.771 [N/A] 0.117� [0.053] �2.570 [4.378] Moderate Acute 0.467 [0.942] 0.045 [0.248] �40.166 [N/A] 0.090 [0.058] �2.745 [4.932]

† p � .10. � p � .05. �� p � .01.

225EFFECTS OF VICARIOUS EXPOSURE TO VIRGINIA TECH

model, the interpretations are much more diffi- cult to describe because of the large number of comparisons that are involved. Furthermore, by combining some of the outcome categories, this provides analyses that are more efficient (Long & Freese, 2003). Predicted probabilities can be used as a way to quickly and succinctly sum- marize the probability of a given outcome cat- egory given a specific set of values of the pre- dictor variables (Long & Freese, 2003). For the MNL, this corresponds to estimating the prob- ability that is described in Equation 5.

P̂�Symptom � m�X� � exp�X�̂�m�/J�

� i � 1

J

exp�X�̂�i�/J�

(5)

Where m is the desired outcome category, J is the number of categories and X is the vector of covariates.

Figure 1 presents the graphs of the predicted probabilities for all of the regression models that tested out as significant ( p � .10). For example, as TV viewing increases, so too does the probability that a participant will self-rate as experiencing fear. Specifically, after viewing 10 hours of TV coverage of Virginia Tech, partic- ipants had a 9.4% chance of experiencing acute symptoms of fear. After 25 hours of coverage,

this percentage jumped to 30.7%. As expected, after viewing only 5 hours of news coverage, participants had a 66.2% chance of reporting no symptoms. But as their viewing hours increased to 25 hours of news coverage, the percentage decreased to a 34.1% chance of reporting no symptoms.

Odds Ratios

In order to describe the relationships among the outcome categories, odds ratios (or factor change coefficients) can be used (Long, 1997; Long & Freese, 2003). To describe the factor change in the odds of outcome m versus out- come n as the predictor variable HOURS in- creases by is described by equation (6).

m/n�x,HOURS � � m/n�x,HOURS�

� e�HOURS, m/n

(6)

Table 4 provides the odds comparing each of the outcome categories. To interpret the odds ratio, for a standard deviation increase in the amount of hours watched of the VT shootings, the odds of experiencing an acute intrusive thought are 1.8076 times greater as compared to experiencing nothing (holding all other factors fixed). The range of the odds for experiencing an acute symptom as compared to experiencing no symptoms what-so-ever ranged from 1.4769 for Guilt to 3.1983 for symptoms of Appetite.

Discussion

We assessed college students’ responses to the shooting at Virginia Tech in the first few weeks after the event occurred. The value of using MNL allowed us to develop a predictive model that describes more than simple correla- tional relationships. This technique is a pre- ferred method for looking at categorical depen- dent variables, and allows us to make compar- isons between categories resulting with an odds ratio that allows us to describe the magnitude of the differences in self-reported measures of stress between the categories based on the num- ber of hours watched. In addition, this modeling strategy allows the researcher to include any covariates that might be deemed important.

In this study, we were able to show that as TV viewing of the Virginia Tech case in-

Table 3 Chi-Squared Statistics and p-Values for Testing the Effects of Each of the Independent Variables for Each of the Fourteen Symptoms (df � 2)

Symptom

Variable

FEMALE AGE MINORITY HOURS

Intrusive thoughts 0.878 3.022 1.565 9.024�

Sleeping 0.408 3.240 0.465 7.831�

Appetite 0.410 0.339 1.860 7.974�

Distraction 0.763 2.048 0.495 7.190�

Nightmares 3.482 7.474� 1.826 3.052 Fear 25.285�� 3.163 2.145 13.183��

Stomach 0.608 1.171 0.264 10.809��

Depression 0.553 3.500 1.022 5.019†

Suicide 5.459† 1.264 6.066� 4.709†

Disorganization 1.400 1.838 1.098 6.103�

Alcohol/drugs 3.043 1.031 1.304 0.815 Replaying the event 1.073 1.652 8.557� 7.203�

Anger 0.627 0.056 0.178 17.396��

Guilt 0.378 1.728 4.247 4.373

Note. † p � .10. � p � .05. �� p � .01.

226 FALLAHI AND LESIK

creased, so too does the probability that a par- ticipant will self-rate as experiencing acute symptoms of intrusive thoughts, sleep distur- bance, distraction, fear, stomach upset, depres-

sion, disorganization, replaying of the event, and symptoms of anger. The probability of ex- periencing acute symptoms for intrusive thoughts, sleep and appetite disturbance, dis-

0 .2

.4 .6

.8

P re

d ic

te d P

ro b

a b

ili ty

0 10 20 30 40 50 Hours

None Moderate Acute

Fear

Fear None Moderate Acute Hours

0.6621 0.2792 0.0587 5 0.5932 0.3129 0.0939 10 0.5148 0.3398 0.1453 15 0.4293 0.3545 0.2162 20 0.3411 0.3525 0.3065 25 0.2567 0.3319 0.4114 30 0.1826 0.2954 0.5221 35 0.1231 0.2491 0.6278 40 0.0791 0.2005 0.7204 45 0.0490 0.1553 0.7957 50

0 .2

.4 .6

.8

P re

d ic

te d P

ro b

a b

ili ty

0 10 20 30 40 50 Hours

None Moderate Acute

Stomach

Stomach None Moderate Acute Hours

0.8637 0.112 0.0243 5 0.8261 0.1325 0.0414 10 0.7767 0.1540 0.0693 15 0.7122 0.1746 0.1131 20 0.6305 0.1911 0.1783 25 0.5324 0.1995 0.2681 30 0.4237 0.1963 0.3800 35 0.3154 0.1807 0.5038 40 0.2197 0.1556 0.6247 45 0.1441 0.1262 0.7297 50

0 .2

.4 .6

.8

P re

d ic

te d P

ro b

a b

ili ty

0 10 20 30 40 50 Hours

None Moderate Acute

Anger

Anger None Moderate Acute Hours

0.7891 0.1832 0.0277 5 0.6877 0.2650 0.0473 10 0.5635 0.3605 0.0760 15 0.4298 0.4564 0.1137 20 0.3048 0.5372 0.1581 25 0.2023 0.5919 0.2058 30 0.1274 0.6185 0.2541 35 0.0771 0.6213 0.3016 40 0.0453 0.6067 0.3479 45 0.0261 0.5805 0.3933 50

Figure 1. Predicted probabilities for all significant regression models based on number of hours of viewing the VT shooting at the p � .01 level.

227EFFECTS OF VICARIOUS EXPOSURE TO VIRGINIA TECH

traction, fear, stomach disturbance, and anger were less than 9% for TV viewing of 10 hours and from 30% to 62% for 40 hours of exposure to the Virginia Tech case. For suicide, disorga- nization, and replaying, the probability of expe- riencing acute symptoms was less than 3% for 10 hours of TV exposure and from 3.55 to 10.73% for 40 hours of exposure. Further, we were able to show that for each hour watched of

the Virginia Tech shootings, the odds of expe- riencing acute symptoms increased from 1.48 to 3.20 times, depending on the symptom. This study improves over past research in allowing us to predict the probability of experiencing acute symptomatology as the result of exposure to real life violence in the media by going beyond showing a relationship and actually quantifying the magnitude of that relationship.

Table 4 Factor Change in the Odds of Symptoms Based on Number of Hours Watched

Symptom Outcome comparison Parameter estimate (b) Factor change in the odds

Intrusive thoughts Acute-None 0.10440 1.8076��

Acute-Moderate 0.07536 1.5331�

Moderate-None 0.02904 1.1790 Sleeping Acute-None 0.10172 1.8017��

Acute-Moderate 0.06801 1.4824†

Moderate-None 0.03370 1.2154 Appetite Acute-None 0.20086 3.1983��

Acute-Moderate 0.17387 2.7357�

Moderate-None 0.02699 1.1691 Distraction Acute-None 0.09621 1.7452��

Acute-Moderate 0.07560 1.5489�

Moderate-None 0.02062 1.1267 Nightmares Acute-None 0.03491 1.2261

Acute-Moderate �0.00487 0.9720 Moderate-None 0.03978 1.2615†

Fear Acute-None 0.11578 1.9692��

Acute-Moderate 0.07095 1.5148�

Moderate-None 0.04482 1.3000 Butterflies in stomach Acute-None 0.11541 1.9503��

Acute-Moderate 0.07297 1.5255�

Moderate-None 0.04244 1.2785 Depression Acute-None 0.07515 1.5449��

Acute-Moderate 0.06220 1.4334†

Moderate-None 0.01295 1.0778 Suicide Acute-None 0.10801 1.8886†

Acute-Moderate 0.12714 2.1136 Moderate-None �0.01913 0.8935

Disorganization Acute-None 0.05780 1.3979 Acute-Moderate 0.00545 1.0321 Moderate-None 0.05235 1.3544�

Alcohol/drugs Acute-None �0.06305 0.6936 Acute-Moderate �0.08400 0.6142 Moderate-None 0.02095 1.1293

Replaying the event Acute-None 0.02901 1.1833 Acute-Moderate �0.01686 0.9068 Moderate-None 0.04586 1.3049�

Anger Acute-None 0.13470 2.1848��

Acute-Moderate 0.03335 1.2135 Moderate-None 0.10135 1.8004

Guilt Acute-None 0.06720 1.4769†

Acute-Moderate 0.02519 1.1574 Moderate-None 0.04201 1.2760

† p � .10. � p � .05. �� p � .01.

228 FALLAHI AND LESIK

As psychologists working in prevention, it is helpful to know how many hours of TV viewing are associated with the development of symp- tomatology. Further, it is now apparent that clinicians working with clients following a trau- matic event need to incorporate questions about vicarious exposure to the event in their assess- ments. Psychologists and Educators need to be better informed about how to help students cope with viewing high-profile media events. Finally, this research gives more evidence that Psychol- ogists should work harder to advise newscasters on the potential negative responses to violent media.

Research exploring gender differences in fear reactions to mass media has consistently shown that females respond with greater frequency and magnitude to audiovisual images in studies pro- duced from 1987 to 1996 and more recently (Peck, 1999; Valkenburg, Cantor, & Peeters, 2007). Similar results were found in the exam- ination of females responses to the Virginia Tech case as females experienced significantly more symptoms of fear via their self-ratings as compared to males. No gender differences were seen in the other 13 acute stress symptomatol- ogy. This makes more of a case for the detri- mental effects of vicarious exposure to TV. The effects seem to be due to TV watching as op- posed to differences between the male and fe- male participants in our study.

Additionally, past research has shown that age and race are factors that make one vulner- able to the effects of violent media exposure (Tucker et al., 2000; Pulcino et al., 2003). In this sample, we did not have a large distribution of participants from varying age or minority status. Our sample was too biased in that par- ticipants were primarily in late adolescence or emerging adulthood and were primarily Cauca- sian. Future research will need to document the probabilities of experiencing increased preva- lence and magnitude of symptomatology based on vicarious exposure to violent media for other samples.

The generalizability of this study is limited because of the lack of standard measures and the reliance on self-report of the participants. In addition, exposure, as measured by the number of hours watching news reports about this inci- dent, was based on self-report without objective corroboration. Finally, without a pretest mea- sure of symptoms of PTSD, we are limited in

drawing a causal inference that the media ex- posure was the cause of the acute stress symp- toms in our sample. Other factors such as stress surrounding upcoming examinations or other stressful events in the life of a student may in fact have influenced their self-reported mea- sures of stress.

References

Addington, L. A. (2003). Students’ fear after Colum- bine: Findings from a randomized experiment. Journal of Quantitative Criminology, 19, 367–387.

American Psychiatric Association. (2000). Diagnos- tic and statistical manual of mental disorders (4th ed., text revision). Washington, DC: Author.

Anderson, C. A., Berkowitz, L., Donnerstein, E., Huesmann, L., Johnson, J. D., et al. (2003). The influence of media violence on youth. Psycholog- ical Science in the Public Interest, 4, 81–110.

Balk, D. E. (2004). Sharing our shock and grief. Death Studies, 28, 971–985.

Blanchard, E. B., Kuhn, E., Rowell, D. L., Hickling, E. J., Wittrock, D., et al. (2004). Studies of the vicarious traumatization of college students by the September 11th attacks: Effects of proximity, ex- posure, and connectedness. Behaviour Research and Therapy, 42, 191–205.

Brooks, K., Schiraldi, V., & Ziedenberg, J. (2000). School house hype: Two years later. Washington, DC: The Children’s Law Center and the Justice Policy Institute.

Bushman, B., & Anderson, C. (2001). Media vio- lence and the American public. Scientific facts versus media misinformation. American Psychol- ogist, 56, 477– 489.

Caprara, G. B., Renzi, P., Amolini, P., D’Imperio, G., & Travaglia, G. (1984). The eliciting cue value of ag- gressive slides reconsidered in a personological per- spective: The weapons effect and irritability. Euro- pean Journal of Social Psychology, 14, 313–322.

Chiricos, T., Eschholz, S., & Gertz, M. (1997). Crime, news, and rear of crime: Toward an Iden- tification of audience effects. Social Problems, 44, 342–357.

Chiricos, T., Padgett, K., & Gertz, M. (2006). Fear, TV news, and the reality of crime. Criminol- ogy, 38, 755–786.

Coyne, S. M., & Archer, J. (2005). The relationship between indirect and physical aggression on tele- vision and in real life. Social Development, 14, 324 –338.

Gil-Rivas, V., Holman. E. A., & Silver, R. C. (2004). Adolescent vulnerability following the September 11th terrorist attacks: A study of parents and their children. Applied Developmental Science, 8, 130 – 142.

229EFFECTS OF VICARIOUS EXPOSURE TO VIRGINIA TECH

Green, W. (2003). Econometric analysis (5th ed.). Upper Saddle River, NJ: Prentice Hall.

Hausmann, J., & McFadden, D. (1984). Specification tests for the multinomial logit model. Economet- rica, 52, 1219 –1240.

Huesmann, L. R., Moise-Titus, J., Podolski, C., & Eron, L. D. (2003). Longitudinal relations between chil- dren’s exposure to TV violence and their aggressive and violent behavior in young adulthood: 1977– 1992. Developmental Psychology, 39, 201–221.

Langua, L. J., Long, A. C., Smith, K. I., & Meltzoff, A. N. (2005). Pre-attack symptomatology and tem- perament as predictors of children’s responses to the September 11 terrorist attacks. Journal of Child Psychology and Psychiatry, 46, 631– 645.

Long, J. (1997). Regression models for categorical and limited dependent variables. Thousand Oaks, CA: Sage.

Long, J., & Freese, J. (2003). Regression models for categorical dependent variables using strata. Col- lege Station, TX: Strata Corporation.

Maddala, G. (1983). Limited dependent and qualita- tive variables in econometrics. New York: Cam- bridge University Press.

Mason, W. (2004). The holes in his head. Retrieved September 11, 2007, from http://www.tnr.com/ doc.mhtml?i� 20040927&s� mason092704

Mickey, J., & Greenland, S. (1989). A study of the impact of confounder-selection Criteria on effect estimation. American Journal of Epidemiology, 129, 125–137.

Paik, H., & Comstock, G. (1994). The effects of television violence on antisocial behavior: A meta- analysis. Communication Research, 21, 516 –546.

Peck, E. Y. (1999). Gender differences in film- induced fear as a function of type of emotion Measure and stimulus content: A meta-analysis and laboratory study. Unpublished doctoral disser- tation, University of Wisconsin, Madison.

Pfefferbaum, B., Gurwitch, R. H., McDonald, N. B., Leftwich, M. J. T., Sconzo, G. M., et al. (2000). Posttraumatic stress among young children after the death of a friend or acquaintance in a terrorist bombing. Psychiatric Services, 51, 386 –388.

Pfefferbaum, B., Seale, T. W., McDonald, N. B., Brandt, E. N., Rainwater, S. M., et al. (2000). Posttraumatic stress two years after the Oklahoma City bombing in youths geographically distant from the explosion. Psychiatry: Interpersonal and Biological Processes, 63, 358 –370.

Propper, R. E., Stickgold, R., Keeley, R., & Christ- man, S. D. (2007). Is television traumatic? Dreams, stress, and media exposure in the after-

math of September 11, 2001. Psychological Sci- ence, 18, 334 –340.

Pulcino, T., Galea, S., Ahern, J., Resnick, H., Foley, M., et al. (2003). Posttraumatic stress in women after the September 11 terrorist attacks in New York City. Journal of Women’s Health, 12, 809 – 820.

Rosen, C., Quyen, T., Cavella, S., Finney, J., & Lee, T. (2005). Chronic PTSD patients’ functioning be- fore and after September 11 attacks. Journal of Traumatic Studies, 18, 781–784.

Schuster, M. A., Stein, B. D., Jaycox, L. H., Collins, R. L., Marshall, G. N., et al. (2001). A national survey of stress reactions after the September 11, 2001, terrorist Attacks. New England Journal of Medicine, 345, 1507–1512.

Spang, G. (1999). Post-disaster stress following the Oklahoma City bombing: An examination of three community groups. Journal of Interpersonal Vio- lence, 14, 168 –183.

Stein, B. D., Elliott, M. N., Jaycox, L. H., Collins, R. L., Berry, S. H., et al. (2004). A national lon- gitudinal study of the psychological consequences of the September 11, 2001 terrorist attacks: reac- tions, impairment, and help-seeking. Psychiatry: Interpersonal & Biological Processes, 67, 105– 117.

Terr, L. C., Block, D. A., Michel, B. A., Shi, H., Reinhardt, J. A., et al. (1999). Children’s symp- toms in the wake of the Challenger: A field study of distant-traumatic effects and an outline of re- lated conditions. American Journal of Psychiatry, 156, 1536 –1544.

Tucker, P., Pfefferbaum, B., Nixon, S. J., Dickson, W. (2000). Predictors of posttraumatic stress symptoms in Oklahoma City: exposure, social sup- port, and peri-traumatic responses. The Journal of Behavioral Health Services & Research, 27, 406 – 416.

Valkenburg, P. M., Cantor, J., & Peeters, A. L. (2007). Fright reactions to television: A child sur- vey. Communication Research, 27, 82–99.

Warr, E. M. (1994). Public perceptions and reactions to violent offending and victimization. In A. J. Reiss Jr. and J. A. Roth (Eds.), Understanding and preventing violence. Volume IV: consequences and control (pp. 1– 66). Washington, DC: The National Academies Press.

Received September 10, 2008 Revision received December 29, 2008

Accepted December 29, 2008 �

230 FALLAHI AND LESIK