Psychological Reactions to Terrorist Attacks: Findings From the National Study of Americans' Reactions to September 11

Context The terrorist attacks of September 11, 2001, represent an unprecedented exposure to trauma in the United States.

Objectives To assess psychological symptom levels in the United States following the events of September 11 and to examine the association between postattack symptoms and a variety of indices of exposure to the events.

Design Web-based epidemiological survey of a nationally representative cross-sectional sample using the Posttraumatic Stress Disorder (PTSD) Checklist and the Brief Symptom Inventory, administered 1 to 2 months following the attacks.

Setting and Participants Sample of 2273 adults, including oversamples of the New York, NY, and Washington, DC, metropolitan areas.

Main Outcome Measures Self-reports of the symptoms of PTSD and of clinically significant nonspecific psychological distress; adult reports of symptoms of distress among children living in their households.

Results The prevalence of probable PTSD was significantly higher in the New York City metropolitan area (11.2%) than in Washington, DC (2.7%), other major metropolitan areas (3.6%), and the rest of the country (4.0%). A broader measure of clinically significant psychological distress suggests that overall distress levels across the country, however, were within expected ranges for a general community sample. In multivariate models, sex, age, direct exposure to the attacks, and the amount of time spent viewing TV coverage of the attacks on September 11 and the few days afterward were associated with PTSD symptom levels; sex, the number of hours of television coverage viewed, and an index of the content of that coverage were associated with the broader distress measure. More than 60% of adults in New York City households with children reported that 1 or more children were upset by the attacks.

Conclusions One to 2 months following the events of September 11, probable PTSD was associated with direct exposure to the terrorist attacks among adults, and the prevalence in the New York City metropolitan area was substantially higher than elsewhere in the country. However, overall distress levels in the country were within normal ranges. Further research should document the course of symptoms and recovery among adults following exposure to the events of September 11 and further specify the types and severity of distress in children.

The coordinated attacks on the World Trade Center (WTC), the Pentagon, and commercial civilian aircraft on the morning of September 11, 2001, were the most deadly terrorist acts ever to occur in the United States. For those directly exposed, the attacks undoubtedly meet the "traumatic event" criterion for posttraumatic stress disorder (PTSD), as set forth in the current edition of the Diagnostic and Statistical Manual of Mental Disorders.¹ Considerable research indicates that individuals directly exposed to a traumatic event are at increased risk for PTSD, for other psychiatric disorders, and for somatic symptoms and physical illnesses.² Additionally, a review of research on the mental health consequences of natural and man-made disasters found that events that involve intentional violence are more likely to be associated with symptoms of severe psychological distress, including PTSD, than are disasters not characterized by human malfeasance.³

Because there have been few such incidents in the United States, there have been few studies of their impact. Studies of the bombing of the Murrah Federal Building in Oklahoma City, Okla, are the most relevant. North et al⁴ found that one third of a sample of 182 persons directly exposed to the blast reported patterns of symptoms that met criteria for PTSD 6 months later. Smith et al⁵ conducted random telephone interviews in Oklahoma City and Indianapolis, Ind (a comparison site), 3 to 4 months after the bombing and found that 43% of those living in Oklahoma City reported 4 or more stress symptoms compared with 11% of those living in Indianapolis.

Initial empirical information on the psychological effects of the September 11 attacks became available soon after the attacks. Based on a random-digit-dialing survey of 560 US adults conducted 3 to 5 days after September 11, Schuster et al⁶ reported that 44% of their national sample of Americans were bothered "quite a bit" or "extremely" by at least 1 of 5 selected symptoms from the PTSD Checklist. Results varied by sex, race/ethnicity, and distance from the WTC, and 35% of the adults surveyed said that their children had 1 or more stress symptoms. Galea et al⁷ studied the prevalence of symptoms consistent with PTSD and depression among 1008 adults living south of 110th Street in Manhattan using random-digit-dialing techniques with telephone interviews conducted 5 to 9 weeks after the WTC attacks. Findings indicated that 7.5% of the adults living south of 110th Street reported symptoms consistent with current PTSD, and 9.7% reported symptoms consistent with current major depression. Additionally, those living closest to the WTC site were nearly 3 times as likely to have PTSD as those living farther away.

The studies reported to date have either focused on providing a broad overview of the country's reactions using survey assessments whose relationship to clinical diagnosis is unknown or documenting clinically significant distress among those most directly exposed to the events. In this article, we report findings from the National Study of Americans' Reactions to September 11 (N-SARS), a Web-based descriptive epidemiological study based on a national cross-sectional sample of adults. The N-SARS was designed (1) to estimate the prevalence of symptoms of PTSD and clinically significant, nonspecific psychological distress in the second month after the attacks, both nationwide and in the areas most proximal to the attack sites, using screening instruments whose relationship to clinical diagnosis is well documented; and (2) to examine the association of both direct and indirect exposures to the September 11 events with symptoms of PTSD and of clinically significant psychological distress. We also report on adults' perceptions of the reactions of children in their households.

The N-SARS sample was drawn from the Knowledge Networks Web-enabled panel, a probability-based, standing research panel that is designed to be representative of the US population.⁸ The Web-enabled panel is recruited using stratified random-digit-dialing telephone sampling techniques, which make it possible to reach every US household with a telephone (95% of US households),⁹ including those with no computer or Web access. At the time that the N-SARS sample was selected, a total of 58 582 households had been recruited to participate in the panel and 21 128 households were actively participating in the Web-enabled panel.

For N-SARS, we selected from the active participants in the Web-enabled panel a sample that included every household that was located in the New York, NY (1196 households), and Washington, DC (369 households), metropolitan areas. In addition, we included an oversample of other major metropolitan areas (specifically, Boston, Mass, Philadelphia, Pa, Chicago, Ill, Houston, Tex, and Los Angeles, Calif, for a total of 776 households) that were not attacked so that we could assess the relationship of living in a major metropolitan area with postattack symptoms. We also included a sample that represented the rest of the country (790 households). A total of 3131 persons were selected. Sample sizes for the domains of interest provide 80% power to detect a 5-percentage-point difference in prevalence with the exception of the Washington, DC, metropolitan area, where the 369 households drawn were all that were participating in the panel when the sample was selected.

Beginning October 12, 2001, surveys were sent via e-mail to the N-SARS sample, and the field period continued through November 12. When the field period ended, a total of 2273 usable surveys (73%) had been returned (4 returned surveys were deleted for analysis purposes because of high item-level nonresponse).

Because initiation into and participation in the Knowledge Networks panel is itself a multistage process, nonresponse can occur at any of the stages. The response rate for the first stage (random-digit-dialing recruitment), calculated by standards established by the survey research field under the aegis of the American Association for Public Opinion Research,¹⁰ was 41%. The within-sample response rate (ie, the proportion of those selected who responded) was 73% (sample domain-specific rates ranged from 70%-78%).

To account for the multistage design of the Knowledge Networks Web-enabled panel, the stratification used in selecting the N-SARS sample from the panel, and nonresponse at all stages of the process, we created statistical weights scaled to sum to the US noninstitutionalized adult population (as documented by the US Census Bureau's Current Population Survey⁹) and used them in all analyses. The purpose of these weights is to adjust for any biases that may have been introduced so that estimates based on N-SARS respondents are representative of the sample domains. Table 1 shows the unweighted N-SARS sample sizes for selected sociodemographic characteristics, the weighted N-SARS population estimates for those characteristics, and the estimates for those characteristics based on the US Census Bureau's October 2001 Current Population Survey.⁹

An institutional review board reviewed and approved the N-SARS methods and procedures, and informed consent was obtained from all participants through an online consent form at the beginning of the survey.

The N-SARS survey instrument focused primarily on 2 domains: specific exposures to the events of September 11 and self-reported mental health outcomes that might be related to those exposures.

To assess direct exposures, we queried whether N-SARS participants were at or near either of the major crash sites (the New York City and Washington, DC, areas) on September 11, 2001. We further assessed whether participants who were in either of those areas on September 11 had been in the WTC or surrounding buildings or in the Pentagon; had seen 1 of the crash sites in person on September 11 after the crash and collapse of the buildings; or could see smoke emanating from the WTC or the Pentagon firsthand that day.

We assessed involvement of family members and friends in the events of September 11 by asking the participants whether they knew anyone who had been injured or killed or was still missing, and asking about their relationship to those persons. Additionally, because US troop deployment to Afghanistan was imminent at the time we fielded the survey, we inquired about current military service of sample members and their families for examination as a potential correlate of psychological distress.

We assessed indirect exposure by inquiring about both the amount of time spent watching television (TV) and specific elements of the content viewed. However, viewing of traumatic events via TV is not strictly specified as a DSM-IV exposure criterion for PTSD. We asked participants to estimate the average number of hours they watched TV on September 11 and the first few days afterward, and whether they viewed TV coverage of several graphic events live (as they happened), on tape, or not at all. The list of graphic events included a plane crashing into the WTC, the WTC collapsing, someone jumping or falling from the WTC, someone getting killed or who had died, someone who appeared to be seriously injured, parts of bodies or something else that was grisly or gruesome, and persons running away to escape from the crash sites. We created a TV content index, ranging from 0 to 7, that indicated the number of different kinds of graphic events the person reported seeing on TV.

We included a measure specifically focused on the symptoms of PTSD and another broader measure of clinically significant psychological distress.

Although the N-SARS survey focused on adults, we were also concerned about the reactions of children to the events of September 11. Therefore, we asked adult respondents in households that included 1 or more children younger than 18 years whether any children in the household were "upset" by the events. If so, we inquired further about the presence of 3 specific distress symptoms among children perceived to be most upset: difficulty sleeping; being irritable, grouchy, or easily upset; and fear of separation from parents.

To account for the design features of the N-SARS sample, all analyses were conducted using statistical weights with SUDAAN software.²⁰ We used χ² tests to identify significant associations between exposures and outcomes. For multivariable analyses aimed at assessing adjusted associations with the outcomes (ie, associations adjusted for the effects of other variables included in the model), we used SUDAAN's LOGISTIC procedure for binary outcomes (prevalence estimates) and the REGRESS procedure for continuous outcomes (symptom scale scores), and we used the Wald F statistic to identify significant adjusted associations. Additionally, we used model-based (least-squares means) methods described by Korn and Graubard²¹ to estimate the adjusted difference in prevalence rates between the New York City metropolitan area and the rest of the country, taking account of sociodemographic differences of the populations therein.

Table 2 shows the prevalence of probable PTSD related to the September 11 attacks associated with specific types of exposure to those events. Geographic proximity to the WTC crash site was significantly related to the prevalence of probable PTSD. The prevalence of probable PTSD during the second month following the terrorist attacks among persons who were in the New York City metropolitan area that day was 11.2% compared with a national prevalence estimate of 4.3%. Although the prevalence in the Washington, DC, metropolitan area (2.7%) and in other major metropolitan areas (3.6%) was slightly lower than the overall national prevalence, those differences were not statistically significant.

The prevalence of probable PTSD was also significantly associated with the number of hours of TV coverage of the attacks that participants reported watching on September 11 and in the following few days and with the number of different kinds of potentially traumatic events participants reported seeing. The prevalence among those who reported that family, friends, or coworkers were killed or injured in the attacks and among those who reported being in the military or having close family members or loved ones in the military was considerably higher than among those who did not, but neither difference was statistically significant.

Although the prevalence of probable PTSD was significantly higher in the New York City metropolitan area than in the rest of the country, some or all of the difference could be attributable to differences in the sociodemographic characteristics of the populations of those areas. When we controlled for age, sex, race/ethnicity, and education using logistic regression, however, we found that those who were in the New York City metropolitan area on September 11 were 2.9 (95% confidence interval, 1.4-5.8) times more likely to be probable cases of PTSD than those who were elsewhere that day. The model-based estimate of the difference in prevalence between the New York City metropolitan area and the rest of the United States, adjusted for sociodemographic differences in the respective populations using methods described by Korn and Graubard,²¹ is 5.1 (95% confidence interval, 0.5-9.7) percentage points.

In addition, we also modeled the association of a more detailed set of exposures with the full range of PTSD symptoms, using the PTSD symptom scale score as the dependent variable. Because the most direct exposures (eg, having been in 1 of the attacked buildings the day of the attacks) occurred in adequate numbers only in the New York City metropolitan area, we conducted these analyses among the subset of N-SARS participants who were in New York City on September 11. The variables included in the model were age; sex; race/ethnicity; education; having been in the WTC or surrounding buildings at the time of the attacks; having seen the WTC in person on September 11 but after the collapse of the buildings; having been close enough to see the smoke from the WTC site on September 11; having family, friends, or coworkers injured or killed in the WTC; number of hours per day of TV coverage watched; and the TV content index (number of different kinds of graphic WTC events seen on TV). With this full set of variables controlled for, only age, sex, having been in the WTC or surrounding buildings at the time of the attacks, and number of hours of TV coverage watched per day were significantly associated with PTSD symptoms (Table 3).

The N-SARS findings for the broader measure of clinically significant psychological distress (Table 4) indicate that during the second month after the attacks, 11.6% of the US population was experiencing clinically significant distress. Normative data suggest, however, that this level of distress is within the expected range for a general community sample.¹⁷ Although rates of clinically significant distress were somewhat higher in New York City (16.6%), Washington, DC (14.9%), and other major metropolitan areas (12.3%) than in the remainder of the United States (11.1%), these differences were not statistically significant; nor were the associations of having family, friends, or coworkers injured or killed in the attacks; of being in or having close family or loved ones in the military; or of the TV content index statistically significant. The only one of these variables with a significant bivariate association with the broader distress measure is hours of TV coverage watched per day.

We also modeled the associations of September 11 exposures with clinically significant psychological distress while controlling for selected sociodemographic characteristics, using the BSI global symptom index scale score as the dependent variable. We used the same set of sociodemographic characteristics and exposures that were included in the PTSD symptom model and, again, included only participants who were in the New York City metropolitan area on September 11. With both the sociodemographic characteristics and the detailed exposures controlled, only sex, number of hours of TV coverage watched, and the TV content index were significantly associated with clinically significant psychological distress symptom levels (Table 5).

Finally, adult participants who lived in households with children (n = 729)were asked whether any child in the house was upset by the events of September 11. The proportion indicating that at least 1 child was upset was highest in the New York City metropolitan area (60.7%), followed by other major metropolitan areas (57.3%), Washington, DC (54.9%), and the rest of the United States (48.0%). These differences were not statistically significant.

The mean age of children perceived as most upset was 11 years, and there were no sex differences in terms of which children were perceived to be most upset. Adult reports indicated that 19.8% of the most upset children were having trouble sleeping, 29.9% were described as irritable, grouchy, or easily upset, and 26.5% were described as fearing separation from their parents.

Based on responses of a national sample of adults to a survey conducted in the second month following the terrorist attacks, the prevalence of probable PTSD related to the September 11 attacks was significantly higher in the New York City metropolitan area than in Washington, DC, other major metropolitan areas, or elsewhere in the United States. Given that the population of the New York City metropolitan area exceeds 10 million adults, the 5.1-percentage-point difference in the prevalence of probable PTSD between the New York City metropolitan area and the rest of the United States, adjusted for age, sex, race/ethnicity, and education differences, translates into an estimated 532 240 excess cases of probable PTSD among adults in the New York City metropolitan area following the terrorist attacks.

To provide perspective on the clinical relevance of our findings with respect to probable PTSD, we compared our results with those of other studies that have included both the PCL scale and clinical diagnosis. The mean PCL scale score among N-SARS probable PTSD cases was 58. Blanchard et al¹³ found that survivors of motor vehicle crashes with PTSD diagnosed via structured clinical interviews had a mean score of 60 on this scale and that for sexual assault survivors, the mean score was 55. This suggests that the probable PTSD cases identified in N-SARS have severity in the range that one would expect to find in clinical populations. Together with the prevalence findings, this suggests that the elevated rates of probable PTSD in the New York City metropolitan area represent an important public health problem.

The N-SARS estimate of the prevalence of probable PTSD in the New York City metropolitan area (11.2%) is somewhat higher than the estimate for the portion of Manhattan that lies south of 110th Street (7.5%) reported by Galea et al.⁷ The studies differed in the scope and sociodempgraphic composition of the samples studied and in the ways in which self-reports of PTSD symptoms were translated into "diagnoses," either of which could account in part for the differences in prevalence estimates. Despite these differences in methods, however, the 7.5% PTSD prevalence reported by Galea et al falls within the 95% confidence interval of the N-SARS probable PTSD prevalence rate for the New York City metropolitan area.

The low prevalence of probable PTSD in the Washington, DC, metropolitan area, the other population center that was attacked, is somewhat surprising. Clearly, there are differences in the actual events in the 2 cities that could account in part for the prevalence difference. These include that the Pentagon is more geographically isolated from the city than the WTC towers; that it is a military rather than a civilian target, possibly reducing the perception of personal vulnerability, vulnerability to attack, or identification with the victims; and that the crash into the Pentagon was much less devastating than the crashes into and collapse of the WTC towers, which produced spectacular visual images and an order of magnitude more deaths and injuries. The N-SARS estimates are, however, the only estimates published to date for the Washington, DC, area.

Although there is limited evidence linking indirect exposures to traumatic events via TV to posttraumatic stress symptoms in children and adolescents,²² there is little empirical information about the association in adults. In addition, strict interpretation of current DSM-IV criteria for a diagnosis of PTSD does not specifically include TV viewing as an example of indirect exposure.¹ In a community sample in which there was opportunity for direct or indirect exposure or both, we found a statistically significant association between PTSD symptom levels and the number of hours per day of TV coverage of the attacks that were watched, even after controlling for indices of direct exposure to the attacks, the content of the TV coverage seen, and sociodemographic characteristics. Similar models fit to our measure of nonspecific clinically significant psychological distress symptoms indicated that no direct exposure measures were associated with nonspecific distress symptom levels, but both hours of TV watched and the TV content index were.

Documentation of these adjusted associations in a community sample of adults raises a number of important questions. The associations could be an indication that exposure via TV contributed to the development of the symptoms, that those who were already distressed by other September 11 exposures watched TV coverage as a coping mechanism, or that psychologically vulnerable persons are more likely to seek out such exposures via TV. Although the N-SARS findings do not speak definitively to the direction of causality, our findings suggest that the N-SARS measures of TV watching—both amount of coverage watched and the specific content—may be better conceived of as correlates of distress (eg, a coping mechanism) than as indices of exposure. This is an issue that requires additional research, however, in designs that support more definitive causal inference.

Although we did not find a statistically significant association between proximity to the attacks and adults' reports of distress among children in their households, the fact that about 61% of adults in New York City and 49.4% of adults in the rest of the United States perceived 1 or more children in their households to be upset by the attacks suggests a need for further study. Studies of children in Oklahoma City after the bombing of the Murrah Federal Building found significant levels of psychological problems related to direct and indirect exposure and to TV viewing.^22-24

Although the children's distress perceived by adults may be self-limiting, biased by the adults' own reactions, or otherwise without clinical significance, further examination of the reactions is clearly indicated. Follow-up should involve direct assessments of children themselves as well as more detailed reports by their parents and teachers.

The national scope, the oversampling in areas where the attacks occurred, and the use of instruments whose relationship to clinical diagnosis is well established are clear strengths of the N-SARS study. The study also has a number of limitations, however. First, the cross-sectional design and lack of preattack measures limit our ability to make causal inferences about the attacks and leave unanswered important questions about the long-term consequences of the attacks. The latter can be addressed via subsequent waves of assessment of the study sample, which we hope to conduct.

Second, the inferential power of the study is limited by several aspects of the sample design. The Knowledge Networks Web-enabled panel from which the sample was drawn makes feasible rapid assessment of community samples following unforeseeable events such as the September 11 attacks. Recruitment and retention of such panels, however, is a logistically complex process that contains multiple potential sources of nonresponse. Since each such source is an opportunity for the introduction of bias, questions arise about the representativeness of the findings. The use of analysis weights that take account of the various components of nonresponse helps to minimize the potential for bias, and the fact that weighted estimates of the distributions of sociodemographic characteristics based on the N-SARS match well with the census estimates of those distributions (Table 1) is reassuring. Additionally, studies of first-stage nonresponse in the Knowledge Networks panel²⁵ have found little detectable bias in study outcomes. Nevertheless, findings based on probability samples with high levels of overall nonresponse should be confirmed by findings based on samples selected by methods that produce more complete response.

Third, the use of screening measures rather than comprehensive clinical assessments to identify probable cases of PTSD and of clinically significant psychological distress increases the likelihood of misclassification with respect to these outcomes. The cross-sectional design does not allow us to rule out the possibility that some N-SARS participants had prior histories of trauma and may have been already experiencing PTSD symptoms at the time of the attacks or that the attacks exacerbated symptoms in previously exposed individuals whose PTSD had been in remission. We did, however, use the "specific stressor" version of the instrument in an effort to minimize these threats to validity. Additionally, some participants who reported elevated PTSD symptoms in N-SARS would be best classified as cases of acute stress disorder. Recent studies, however, have shown that acute stress symptoms are excellent predictors of PTSD, and that approximately 80% of individuals who meet diagnostic criteria for acute stress disorder will later meet PTSD criteria.^26-28 Finally, viewing of traumatic events via TV would not typically be conceptualized as meeting the DSM-IV exposure criteria for PTSD. Therefore, we believe that our PTSD findings must be interpreted with care but nonetheless have acceptable validity for providing estimates of the prevalence of probable PTSD related to the attacks of September 11.

We conclude that the major burden of probable PTSD in the second month following the September 11 terrorist attacks among adults was closely related to direct exposure to the events and that the prevalence in the New York City metropolitan area was much higher than elsewhere in the United States. Nevertheless, although many Americans may have been upset by the attacks of September 11, a broad measure of nonspecific psychological distress indicates that the overall level of clinically significant psychological distress in the second month following the attacks was within normal limits for the country as a whole. Further research should document the course of disorder and recovery among adults with PTSD related to the September 11 attacks and determine whether the distress of children in their households, as reported by adult N-SARS respondents, is clinically significant.