A Role for Prospective Longitudinal Investigations in the Study of Traumatic Stress and Disasters

John B. Reid (1990)

Originally appeared in the Journal of Applied Social Psychology, 1990, 20(20), 1695-1703. Note that this online version may have minor differences from the published version.
Oregon Social Learning Center, Eugene, Oregon
Journal of Applied Social Psychology, 1990, 20(20), 1695-1703
Author Notes: The preparation of this manuscript was supported by Grants No. MH 37940 and MH 17126 from the Center for Studies of Antisocial and Violent Behavior, NIMH, U.S. PHS, and Grant No. MH 38730 from the Child and Adolescent Disorders Research Branch, NIMH, U.S. PHS. The author gratefully acknowledges the advice and critical input of Dr. Susan Solomon and Dr. Jack Mazur. Correspondence should be addressed to John B. Reid, Oregon Social Learning Center, 207 E. 5th, Suite 202, Eugene, Oregon, 97401.


Much of our knowledge of the short- and long-term effects of traumatic stress and disasters has been developed through the study of victims and control subjects after the event has happened. A number of problems inherent in such a post-hoc research strategy make it difficult to isolate the specific contributions of the severe stressor itself to subsequent adjustment. Common problems include the difficulty in specifying baseline adjustment, necessary reliance on retrospective and sometimes monomethod data, difficulties in recruiting representative samples of victims, and convincing control groups. Many of these problems could be circumvented by using ongoing longitudinal and multimeasure studies of psychosocial development. Such a strategy would allow study of the effects of severe and traumatic events as they occur to subjects during their participation in such studies. Systematic and multiple measures of baseline adjustment of victims and nonvictims would be available, as well as preplanned assessments for postevent adjustment. Problems involved in implementing such collaborative research ventures are reviewed and solutions proposed.

The purpose of this paper is to discuss the use of prospective longitudinal research strategies and the exploitation of ongoing longitudinal investigations for the study of direct, indirect, immediate, and long-term effects of severely stressful events. Previous studies of such events have typically assessed groups who have been exposed to some common and severe stressor of interest. Victims are recruited and examined at some point after the event, often comparing their adjustment against epidemiological or normative data, or against a control group also selected after the event. A powerful feature of such investigations, and their results, has been the face validity of the stressful evens studied, and those events have often been quite dramatic, including exposure to battle in WWII, Korea, and Vietnam, the holocaust, floods, and to technological disasters such as toxic waste exposure and the meltdown at Three-Mile Island.

One major problem with past research is that it is usually difficult or impossible to accurately and systematically ascertain the pre-event adjustment of the victims. Given that most severe stressors cannot be predicted very well by victims, survivors, or the scientists who study them, it is not surprising that baseline data of scientific quality are seldom available for disaster samples. The evaluation of pre-event adjustment must often be made after the event. Because available pre-event data on survivors may differ markedly (e.g., relevant archival data or informed others who can make objective observations regarding the adjustment), the investigator is often left with the alternative of using self-report or psychiatric interview data to assess pre-event adjustment, exposure to and experience of the event, and post-event adjustment. Although it is possible and common to use individual psychiatric interviews and questionnaires to make post hoc assessments, their validity is still somewhat problematic, particularly when one-time diagnostic interviews are used without corroborating information of good quality (Achenbach & Edelbrock, 1984; Reid, Baldwin, Patterson, & Dishion, 1988; Spiker & Ehler, 1984), and it is not unreasonable to think that such data might face even further challenges to their validity by the effects of the event itself. Since clear estimates of pre-event adjustment are indispensible if the actual effects of the disaster are to be determined, this is a very serious issue.

Not only is interpretation of retrospective data generally problematic (e.g., Bower, 1980), but using the same informant for measures of each major variable makes it extremely difficult to determine to what extent observed relations among the variables are due to shared method variance, and how much is due to predicted theoretical relationships. In the case of a person's recollection of past adjustment after a devastating experience, it is possible that the recollections of past problems are biased to fit the person's current adjustment problems, as shown in Vaillant's (1983) classic longitudinal study of alcoholism. A related problem concerns ethical or practical considerations. Survivors of a disaster or trauma are not always excited about the chance to participate in research forcing them to deal with a personal tragedy, its sequelae, and the way things used to be. Not only have they likely discussed these issues extensively with their family and various friends, investigators, bureaucrats, Red Cross workers, medical helpers, and counselors (which in itself probably affects the integrity of subsequent interview and report data), but the victims most seriously affected may be the least anxious to cooperate. Refusal rates can vary greatly across studies. As an extreme example, Jackson and Mukergee (1974) studied reactions to earthquakes in two cities using the same recruitment strategy. In San Francisco they experienced a refusal rate of nearly 80%, in Los Angeles only 10% to 15%. One can only get so much information from victims, and choices probably have to be made about which variables to assess and which to forego. These and other factors seriously limit the amount and veracity of pre-event data that can be collected after the fact.

A related problem in this area is the selection and recruitment of control groups. Because of the problems with getting pre-event data on adjustment, it is often hard, if not impossible, to select an appropriate control group on variables other than exposure to the event and demographic variables. The post hoc control group seldom helps in clarifying the specific contribution of the event on subsequent adjustment.

It is possible to find some situations in which at least some historical data are uniformly available on survivors of extremely stressful experiences. One of those is exposure to war. The military does a good job in routinely collecting data on its employees after they enlist, and in keeping current as to their whereabouts after discharge. It is probably no accident that the vast majority of published studies of PTSD have employed veterans as subjects. Even in this situation, such pre-existing file data may not be enough to identify or control for confounding effects of pre-existing adjustment and sampling problems (Robins, 1983). Although the use of specialized populations such as Vietnam veterans, child abuse victims, or Cambodian refugees offers unique advantages for the study of severe stress, those advantages are often mitigated by generalization problems. During the Viet Nam era, the average age of a combat soldier was under 25 years (Kulka et al., l988). Serious child abuse victims are mostly under five years (American Humane Society, 1981), and survivors of atrocities are probably not representative of the original population exposed to them. This is a frustrating situation. The extremely traumatic or stressful events with the most face validity present the researcher with the most challenges to the validity of the research design.

To summarize, the victims or survivors may be unique, limiting generalization of observed relationships. Such events tend to attract the attention of media, government or insurance investigators, and mental health assistance specialists whose interviews and attention may contaminate, condense and smooth out the recollections of subjects before the scientist arrives on the scene. The survivors may be fragile or reluctant to dredge up information about the event and their pre-existing adjustment, forcing the investigators to limit their data collection to variables most salient to the victims, such as their subjective reaction to the experience and their recall of pre-event adjustment. Corroborating data, such as reports by informed others, records, or objective data on pre-event adjustment will not be consistently available for all subjects, forcing a reliance on self-report data for most variables of interest, and influencing the study's focus towards the direct effects of the events on the victims' own mental or physical adjustment. At another level, the use of survivors of a particular disaster or trauma type makes it difficult to compare data across disasters or traumas because of significant sample differences, measures used, and so on.

Even when we consider more common and less clearly defined traumatic experiences such as divorce and its effects on spouses and children, the two major prospective studies in the literature (Hetherington, Cox, & Cox, 1978; Wallerstein & Kelly, 1980), began data collection after the decision to divorce had been made. Although each study included multimethod assessment strategies, making possible fine-grained analyses of the relations among a number of sociological, social interactional and psychological processes relating the event to its sequelae for both involved adults and children, one still cannot evaluate the independent effect of the divorce itself on subsequent adjustment.

Because divorce is becoming a relatively common event, it has been possible to employ longitudinal studies to address this issue. Menaghan (1985), for example, examined the specific effects of divorce on subsequent adjustment, using data from a panel study of Chicago adults who were assessed in 1972 and again in 1976. A subset were divorced between the two assessments, and their adjustment was compared with those whose marriages were stable across the study. Menaghan found that time-1 measures of depression did not predict divorce, but the divorced group was more depressed at time-2. Her analyses provided evidence, then, that the divorce itself precipitated depressed affect, but analyses of other variables measured indicated that the effect was probably mediated through the increase of subsequent economic hardship caused by the divorce. Using a prospective design and measuring multiple variables, it was not only possible to examine the specific contribution of the event to subsequent adjustment, but it was possible to assess the way in which the event functioned to produce the effects.

Another example of the relationship between less clearly defined traumatic experiences and subsequent adjustment is the frequent finding that poverty and extreme disruptions in the economy are related to subsequent increases in coercive and irritable parental discipline, which has been found to be associated with increases in child behavior problems and juvenile delinquency (Loeber & Dishion, 1983). Again, the traditional research model has been to find subjects exposed to the stressor and to assess functioning at some later point in life. Elder and his associates were able to use the data from the Berkeley Growth Study to examine families who began their participation before the Great Depression. In that study, a wide range of individual, family and demographic variables were collected. Elder, Caspi and van Nguyen (1985) were able to recode those data to show that those fathers who experienced precipitous declines in income were indeed at risk to show increases in irritable discipline, but that those most affected were irritable before l929. The increased hardship acted as an amplifier of previous tendencies. In the same area, Laub and Sampson (1988) recoded and reanalyzed data from the classic longitudinal study by Glueck and Glueck (l950). Laub and Sampson demonstrated that the relationship between social and economic stress to subsequent delinquency was almost totally mediated by its disruptive effects on parental supervision and discipline. The advantages of such post hoc analyses of longitudinal studies are that baseline data on a wide variety of variables are often available and not confounded with the experience of subsequent stressful events. Moreover, the full longitudinal samples are often carefully and representatively selected, control subjects are available for comparative analyses, and all subjects have a pre-existing relationship with the research project, making contact and assessment of victims easier after the occurrence of stressful events.

The use of existing longitudinal data sets that began with the collection of a wide range of measures on reasonably representative samples can be increasingly exploited. For example, investigators can study the relations of early adjustment, family, social, and demographic variables, various stressors experienced by segments of the sample after the study is in progress, and subsequent and broad measures of adjustment after the stressful experiences. However, the older existing data sets are limited by the use of dated measures and a heavy reliance on self-report and interview data. Because most longitudinal studies were not designed to assess the effects of disaster or trauma, measures of severe stress are limited. Useful measures routinely collected include demographic information (income, employment, family structure), information on divorce, bereavement, medical histories, and sometimes service utilization. No comprehensive longitudinal studies, to my knowledge, have collected systematic data on a wide range of extremely stressful events at each assessment wave. Although there are often file and interview data available--and some studies spanned major events such as the Great Depression, World War II and Viet Nam--information on events such as natural disasters or serious household accidents will likely be incomplete.

Not only would the area of disaster and trauma research profit from the routine incorporation of data on exposure to extreme stressful events into longitudinal studies, but the longitudinal studies would profit as well. Most longitudinal studies collect demographic information in order to control for or examine the possible effects of stressors such as economic disadvantage on the life course questions of central interest to the given investigation. Similar attention to traumatic stressful events would increase such control. In the case of longitudinal research designed to examine the effects of single stressors such as divorce, child abuse, or bereavement, this routine, broad band, assessment would help to insure that other severe stressors are not augmenting or masking the effects of the stressor of interest. Finally, the use of such assessment in longitudinal studies addressing diverse topics, from different theoretical orientations would probably increase the scientific interest and the number of analyses of the effects of disasters and trauma on the life course. A large number of prospective longitudinal studies of the life course are currently in progress, recently funded, or proposed. There is also an increasing number of prevention studies funded each year employing substantial "risk samples" for a wide variety of problems. Because of methodological requirements for the complex multivariate designs employed by most recent studies, broad, multimeasure assessment and sound sampling strategies are typically employed. Such investigations could serve as invaluable vehicles, not only for the prospective study of common stressors such as divorce, but for classes of more traumatic stressors. Two roadblocks currently exist for utilizing such data sets for the investigation of traumatic stress. The first problem is that the base rates for exposure to any particular traumatic event may be so low that meaningful analyses might lack sufficient statistical power to be informative. Second, the right questions are not included in such studies to permit identification of all clearly traumatic events when they happen to subjects in such studies.

Because of cost issues, multimeasure, -agent, and -setting assessment batteries with multiple assessment points may be most likely found in studies with samples of 500 subjects or less. At our own center, for example, we currently have four longitudinal studies in progress, each assessing approximately 150 children and their families. Each of the projects utilize essentially the same yearly assessment battery, involving over 20 hours of assessment on each subject family. Extensive diagnostic, demographic, questionnaire and interview data, social interaction tasks and naturalistic observations, court and other file data, are collected on the target child, siblings, parents, and sometimes the grandparents, at each assessment. Data from these studies could be aggregated for some analyses, yielding a total sample of about 600 children and about 900 adults. Although this is a fairly substantial sample, it would still be problematic to find enough victims of highly specific disasters to permit powerful analyses.

Even though the probability of experiencing any one type of trauma is quite low in a given sample, the probability of experiencing any one of a set of events considered traumatic in this area of research is not. As reported by Fran Norris (this issue), about 7% of the U.S. population (i.e., 17,000,000 individuals) experience severely traumatic events such as robbery, sexual assault, serious motor vehicle accidents, bereavement, and natural disaster each year. Add to this the number of people experiencing technological disaster, family violence, the more traumatic divorces resulting in precipitous drops in family income or in children's out-of-home placements, and the overall base rate of seriously traumatic events may approach 10% per year or more. Given that average family size is over three, then the base rate of seriously stressful events captured in a longitudinal study of children and their families should be even higher. In the samples of subjects being assessed at our center, we might expect to observe about 60 individual disasters/traumas affecting children, and 90 affecting adults each year. Given that the average duration of our longitudinal studies is at least 5 years, one could expect to observe even greater numbers of target events. If the estimates were reduced by half, one could still conduct powerful analyses of a number of interesting theoretical issues. For example, what variables make people more likely to experience one or more disaster? What are the short- and long-term effects of experiencing a disaster? What are the effects on the victim's mental health? To what extent and in what way does the experience of a disaster affect one's relationship with, and the mental health of, other family members? What factors lead to increased and decreased effects? Do various classes of disasters produce different effects? Do severe disasters/traumas produce effects different from events such as job loss or divorce?

We have been able to use our data sets to ask questions about the direct effects of less traumatic stressors such as divorce, economic disadvantage, and parental drug use on individual family members, and we have recently been able to show that such stressors on parents have substantial indirect and negative effects on the social development of their children to the extent that they disrupt the parents use of sane and consistent discipline practices (cf. Reid and Patterson, 1989). These analyses used multimeasure assessment of each construct, and accounted for at least 30% of the variance of the criterion adjustment measure in each study. Although we have extensive data on our subjects, we are not in a position to study the effects of severe trauma because we simply did not ask the correct questions.

This brings us to the second roadblock to the exploitation of longitudinal studies for the study of disaster and trauma. Given that our projects were not funded specifically for the study of this class of variables, and we already place heavy assessment demands on our subjects, we would need a relatively short assessment device that could be used anually to screen all subjects, and a list of follow-up questions to get more complete descriptive information on those who report experiencing a disaster or trauma. If such an instrument were available, our research group, and several other longitudinal investigators with whom we collaborate, would be willing to seriously consider its incorporation into the existing assessment batteries. The knotty conceptual issues involved, and a brief procedure to screen subjects for recent exposure to severe trauma and disaster, are described by Fran Norris in this issue.

Collaboration among epidemiologists, longitudinal investigators and researchers in the area of trauma and disaster could add significantly to our understanding of the potentially complex relationships of severe stress to a wide array of developmental issues. Such relationships have been suggested by post-hoc studies of trauma and disaster; they could be systematically examined in the context of longitudinal investigations.


American Humane Society. (l980). Annual report: National analysis of official child neglect and abuse reporting. Denver, CO: The American Humane Association.

Achenbach, T. M., & Edelbrock, C. S. (l984). Psychopathology of childhood. Annual Review of Psychology, 35, 227-256.

Bower, G. H. (l980). Emotional mood and memory. Paper presented at the meeting of the American Psychological Association, Montreal, Quebec, August, l980.

Elder, G. H., Caspi, A., & van Nguyen, T. (l985). Resourceful and vulnerable children: Family influences in stressful times. In R. K. Silbereisen, K. Eyferth, & G. Rudinger (Eds.), Development as action in context: Problem behavior and normal youth development (pp. l67-l86). Hillsdale, NJ: Erlbaum.

Glueck, S., & Glueck, E. (l950). Unraveling juvenile delinquency. Cambridge, MA: Harvard University Press.

Hetherington, E. M., Cox, M., & Cox, R. (l978). The aftermath of divorce. In J. H. Stevens, Jr., & M. Matthews (Eds.), Mother-child, father-child relationships (pp. ll0-l55). Washington, DC: National Association for the Education of Young Children.

Jackson, E. L., & Mukergee, T. (l974). Human adjustment to the earthquake hazard of San Francisco, California. In G. F. White (Ed.), Natural hazards: Local, national, global. New York: Oxford University Press.

Kulka, R. A., Schlenger, W. E., Fairbank, J. A., Hough, R. L., Jordan, B. K., Marmar, C. R., & Weiss, D. S. (l988). Contractual report of findings. National Vietnam Veterans Readjustment Study. Research Triangle Park, NC: Research Triangle Institute.

Laub, J. H., & Sampson, R. J. (l988). Unraveling families and delinquency: A reanalysis of the Gluecks' data. Criminology, 26(3), 355-379.

Loeber, R., & Dishion, T. J. (l983). Early predictors of male delinquency. Psychological Bulletin, 94, 68-99.

Menaghan, E. G. (l985). Depressive affect and subsequent divorce. Journal of Family Issues, 6(3), 295-306.

Reid, J. B., Baldwin, D. V., Patterson, G. R., & Dishion, T. J. (l988). Observations in the assessment of childhood disorders. In M. Rutter, A. H. Tuma, & I. S. Lann (Eds.), Assessment and psychopathology in childhood disorders (pp. 156-195). London: Guilford Press.

Reid, J. B., & Patterson, G. R. (l989). The development of antisocial behavior patterns in childhood and adolescence. European Journal of Personality, 3, 107-119.

Robins, L. N. (l983). Some methodological problems and researchissues in the study of the effects of stress on children. In N. Garmezy & M. Rutter (Eds.), Stress, coping, and development in children (pp. 335-346). New York: McGraw Hill.

Spiker, D. G., & Ehler, J. G. (l984). Structured psychiatric interviews for adults. In G. Goldstein & M. Hersen (Eds.), Handbook of psychological assessment (pp. 29l-304). New York: Pergamon.

Vaillant, G. E. (l983). The natural history of alcoholism. Cambridge, MA: Harvard University Press.

Wallerstein, J., & Kelly, J. (l980). Surviving the break-up: How children and parents cope with divorce. New York: Basic Books.