Research on the Mental Health Effects of Terrorism

The terrorist attacks of September 11, 2001, established a backdrop against which mental health effects of disasters, especially large-scale intentionally created disasters, assumed central stage in US public health. Methodologically sound data are required to understand the mental health effects of terrorism and must guide all postdisaster mental health activities from clinical interventions to administrative policy. However, conducting methodologically solid epidemiologic investigations of mental health is extraordinarily difficult in the chaotic and complex settings of disasters, particularly those associated with terrorism.^{1 - 4} The study by Schlenger and colleagues⁵ reported in this issue of THE JOURNAL assessed postdisaster mental health in one of the most complex and challenging disaster settings in US history.

To develop credible data, mental health studies must address complicated aspects of research design in the postdisaster setting. The ideal study of psychiatric effects of disasters and terrorism would commence quickly after the event, assess a representative if not universal sample, establish comparison groups that are similar to the affected population in every way except exposure to the identified event, examine specific psychiatric disorders as well as general distress and functioning, and follow the study population longitudinally. The well-designed study must address issues of timing, sampling, measurement, and interpretation of data with special care.

Ideally, before entering the postdisaster area, researchers should wait while survivors tend to vital activities such as burying the dead, securing housing for those with destroyed residences, and managing the early administrative activities of postdisaster life such as communicating with relief agencies and insurance companies. Additionally, because posttraumatic stress disorder (PTSD) cannot be diagnosed for at least 1 month,⁶ research designed to assess PTSD may be best delayed until the diagnosis can be fully assessed.

The sudden and unanticipated nature of disasters and the often chaotic, inconvenient, and limiting characteristics of disaster sites severely challenge research planning and can significantly delay efforts to begin data collection early after the event. In particularly challenging disasters, such as the Oklahoma City bombing and the World Trade Center terrorist attacks, the process of establishing contacts and securing permissions and access to adequate samples of appropriate study populations may delay entry into the field. Delay in initiating data collection limits opportunities to obtain early information needed to understand mental health effects of disasters. If researchers cannot act quickly, important data may be lost forever.^{7 - 12}

Schlenger and colleagues⁵ collected population data on psychological symptoms within 1 to 2 months of the September 11 terrorist attacks using a Web-enabled panel recruited before September 11 for other research investigations. Another recent study by Schuster and colleagues¹³ obtained population data on stress reactions 3 to 5 days after the terrorist attacks.

Although early postdisaster research findings provide a valuable glance at the nation's mental health shortly after the September 11 attacks, these data do not necessarily predict what might be expected over time. It is unclear what will become of the early distress responses and psychiatric symptoms identified in the study by Schlenger et al,⁵ such as how many will develop into chronic psychiatric illness and how many will resolve spontaneously.

Various populations have been the subject of trauma mental health research, including military combat veterans and survivors of unintentional injury, sexual abuse, and natural and intentional disasters. Mental health findings across these groups differ considerably.^{2 ,14} Specific subtypes of disasters, such as natural disasters, may vary in mental health effects compared with other types of disasters, especially terrorism.^{15 - 16} Even within a single disaster setting, distinct subpopulations emerge. Traditionally, these subpopulations include those directly in the path of the disaster, bereaved individuals who lost loved ones in the event, families of disaster survivors, rescue workers, and community members whose lives or livelihood were affected (eg, individuals whose workplaces were obliterated).

The dearth of systematic postdisaster mental health data encourages generalization of findings from other populations. Yet, application of findings from one type of event to another and from one population to another may be inappropriate because of the major differences among the populations involved. For example, following the Oklahoma City bombing, the larger surrounding community was described as adversely affected psychologically,^{17 - 19} and the entire nation has been considered psychologically vulnerable to effects of the September 11 terrorist attacks.^{13 ,20 - 21} However, community and general populations may have markedly different postdisaster mental health responses and needs compared with those who were directly exposed to the disaster event. As a result, these groups should be considered separately in disaster mental health research and intervention.

The study by Schlenger et al⁵ examined a general household population. The recruitment of their panel group prior to September 11 reduced the potential for sampling bias attributable to mental health effects of the terrorist attacks. The response rate for the first stage of random-digit dialing to recruit panel participants was 41%, and the within-sample response rate of those selected for the Web-based survey was 73%, providing considerable potential for sampling bias. The study by Schuster et al¹³ had a similar response rate and a low (but not precisely reported) yield rate to their random-digit dialing recruitment process. The concern about sampling bias in the study by Schlenger et al⁵ is mitigated, however, by the demonstration of close demographic conformity with comparison census data from the general population represented by the sample.

Schlenger et al⁵ also developed rich comparison data from similarly selected samples in New York City, Washington, DC, and other major US cities as well as from PTSD population norms. While differences between New York City and other cities would be expected and might reflect September 11–related mental health effects, the lack of pre–September 11 comparison data in this study leaves some uncertainty about the degree to which these findings can be assumed to be directly related to the terrorist attacks.

Full clinical diagnostic assessment is labor-intensive and resource-consuming. Structured diagnostic instruments administered by nonclinicians may reduce the resource burden, but even this research is expensive to conduct. Less burdensome alternatives include brief self-report questionnaires that measure symptoms and psychological distress, as used by Schlenger et al,⁵ but these instruments do not provide psychiatric diagnoses. Random telephone surveys have become popular for assessing samples representative of large populations but are limited by the amount of time and nature of the questions that can be asked. Web-based surveys such as that used by Schlenger and colleagues⁵ represent a novel application of technology to facilitate data collection, but this method limits the assessment to brief self-report methods. The symptom checklist used by Schlenger et al⁵ to screen for PTSD has been previously demonstrated to approximate PTSD diagnoses made by structured diagnostic interviews. In contrast, the study by Schuster et al¹³ used a very different 5-item stress question list, thereby making it virtually impossible to compare their results with those of Schlenger et al.

Data consisting of simple symptom counts and symptom frequency/severity scales have limited validity. Symptoms do not constitute psychiatric illness. Diagnostic criteria for disorders include not only specified combinations of symptoms, but also requirements for duration of the symptoms and effects on the individual's ability to function.⁶ In particular, for a diagnosis of PTSD, the symptoms must persist for more than 1 month and must cause clinically significant distress or impair the individual's ability to function.⁶ Additionally, the symptoms must be new after the event ("not present before the trauma")^{6 (p468)} to be counted. Most questionnaires do not distinguish new symptoms associated with the event from endemic symptoms such as sleeplessness that affect many people at one time or another. Failure to address these issues in the assessment of PTSD results in inflated prevalence estimates.

Acknowledgment of symptoms does not necessarily indicate psychopathology. Most individuals directly involved in catastrophic events do not develop diagnosable psychiatric illness, but the majority report experiences such as sleep disturbance, loss of concentration, or feeling emotionally upset afterward.^{14 ,22} Rather than labeling such psychological effects as "symptoms," which unnecessarily implies pathology for the experiences of individuals without psychiatric illness, language such as "reactions" or "responses" might better describe the normative, expected response to extraordinarily upsetting events. Schlenger et al⁵ wisely recognize and report specific psychiatric illness separately from general distress and are careful to describe their findings as pertaining to "probable PTSD" or "symptoms of PTSD" rather than representing them as actual PTSD.

While normalization of psychological effects of disasters can be reassuring for the majority of survivors, it is important to recognize the significant distress of those whose emotional upset does not reach proportions that would qualify for a diagnosis of PTSD. The emotional distress that falls clearly below the diagnostic threshold for PTSD (subdiagnostic distress) that is prevalent among individuals exposed to catastrophic events deserves different mental health interventions from the customary psychiatric treatment for the minority who develop a diagnosable disorder.¹⁴ The distinction between psychiatric illness and this subdiagnostic distress is critical for assessing mental health needs and directing interventions. Despite its central policy implications, this issue is one of the most problematic in disaster mental health research.

Research on the mental health of populations following major catastrophic events encounters a logical dilemma of how to categorize experiences such as jumpiness, loss of concentration, and difficulty sleeping reported among individuals outside the directly endangered area. According to the Diagnostic and Statistical Manual of Mental Disorders, Fourth Edition (DSM-IV),⁶ PTSD cannot be diagnosed without establishing sufficient exposure of the individual to a qualifying event. The DSM-IV diagnostic criteria lack specificity about what constitutes sufficient exposure: "the person experienced, witnessed, or was confronted with" the event.^{6 (p467)} The accompanying explanatory text is more clear on how the exposure criterion is to be interpreted: "The essential feature of Posttraumatic Stress Disorder is the development of characteristic symptoms following exposure to an extreme traumatic stressor involving direct personal experience of an event that involves actual or threatened death or serious injury, or other threat to one's physical integrity; or witnessing an event" of a similar nature or learning that a loved one experienced such an event.^{6 (p463)} Thus, directly exposed individuals, witnesses to the event, and loved ones of those directly exposed are potential candidates for a diagnosis of PTSD. No provision is made, however, for classification of indirect witnessing through viewing media images of the event.

Viewing live television coverage of major traumatic events (including repetitive video replay after the event is over) does not constitute direct personal exposure. Therefore, if one infers that witnessing requires physical presence, simply viewing the event on television would not qualify an individual for a diagnosis of PTSD. Witnessed events are described in DSM-IV-TR as including, but "not limited to, observing the serious injury or unnatural death of another person due to violent assault, accident, war, or disaster or unexpectedly witnessing a dead body or body parts,"^{6 (p464)} yet it does not explicitly specify that one must observe in person. Although DSM-IV does not specifically exclude television exposure as a form of witnessing, it does address physical proximity: "The likelihood of developing this disorder may increase as the intensity of and physical proximity to the stressor increase."^{5 (p464)}

The earliest versions of the Diagnostic Interview Schedule,²³ developed 2 decades ago during the era of the first appearance of PTSD as a diagnosis, established a tradition that specifically excluded experiences of exposure to an event only through television or other media toward making the diagnosis of PTSD. This tradition extended internationally with the application of the same convention in the Composite International Diagnostic Interview,²⁴ indicating a consensus for interpretation of the criteria until more definitive language is provided. The implications of how to rate indirect exposure through media viewing are enormous. Classification of individuals exposed to the events of September 11 through the media as PTSD cases could potentially identify vastly increased numbers in the New York City area and nationally.

The critical distinction between symptoms generated from a DSM-IV qualifying exposure and the reactions that accompany exposure through the media dictates restraint in assignment of symptoms to PTSD or even probable PTSD in relation to the September 11 attacks in settings outside the directly affected areas. In such populations, symptoms and reactions to the September 11 attacks deserve recognition as psychological sequelae, but these responses are distinct from PTSD.

In another article in this issue of THE JOURNAL, Hollifield and colleagues²⁵ evaluate instruments used to measure trauma and health status in refugees and describe similar problems in defining and studying psychological trauma: different phenomena in different populations have been given the same name and compared. Many of the methodological issues in studying distress and psychopathology in refugees coincide with those of studying nondisplaced populations after natural and intentional disasters.

Schlenger and colleagues⁵ are circumspect in interpreting associations in their data. They avoid the classic logical error post hoc, ergo propter hoc (after the fact, therefore because of the fact) in assigning causal directionality to a mere finding of a statistical association. They do not assume causal directionality in the associations they find between nonspecific distress symptom levels and television viewing, in both amount and content of televised September 11 material. Instead of concluding that these associations necessarily represent adverse mental health effects of exposure to this material on television, Schlenger et al consider the possible logical bases for the association, including potential for association of these 2 variables indirectly through other variables as well as the possibility of reversed directionality from psychological distress to television viewing.

The study by Schlenger et al⁵ of US populations within 2 months of the September 11 terrorist attacks breaks new ground in providing early postdisaster data and comparing these findings with those from other carefully selected populations. Desire for early postdisaster data necessarily creates a tradeoff between entering the postdisaster setting quickly and other methodological strengths such as full assessment of psychiatric diagnoses, specifically including those most directly and proximally exposed to the disaster. Future research may surmount limitations faced by these early studies and push the field to develop new ways to apply more complex and sophisticated clinical and epidemiologic research methods in studies of terrorism and other disasters. The challenges of future disaster mental health research include establishing procedures for more rapid access to disaster-affected populations to facilitate early collection of data without sacrificing the integrity of sampling and quality of measurement tools; recognizing the fundamental differences in subpopulations affected by disasters and examining them separately; developing more precise and validated criteria for PTSD; following samples over time to determine the course of postdisaster mental health recovery; conceptualizing and measuring psychiatric illness separately from normative responses to disaster and general distress; being circumspect in interpreting data and avoiding leaps to assign causal directionality to associated variables; and cultivating public courage and political will to devote resources needed to conduct well-designed research studies.