ABSTRACT
Research projects that have multiple objectives must be designed so that a balance is maintained among the (sometimes conflicting) objectives. Although the single most important objective of the NVVRS was to determine the prevalence of PTSD among Vietnam veterans, from a policy perspective it was also important to determine the relationship between PTSD prevalence and service in Vietnam. The Veterans Administration’s primary mandate is to serve the needs of those veterans who have service-connected disabilities. Consequently, from the standpoint of providing information for the planning of VA programs to meet the needs of veterans with service-connected disabilities, it would not be enough for the study simply to show that the prevalence of PTSD is higher among theater veterans than among the comparison groups. Although that would be an important piece of information, such a finding could be explained in a variety of ways. Therefore, it was important that the study be designed in a way that would permit the scientific evaluation of alternative hypotheses that might be offered to explain any differences that were found in PTSD prevalence among the study’s cohorts.
The basic problem that we had arose from the fact that although it was possible to include comparison groups in the NVVRS, people had not been assigned to those groups at random. As a result, the possibility of confounding existed—that is, the current prevalence of PTSD might differ among the groups because of differences in their experiences or because of differences in their characteristics that resulted in their becoming theater veterans, era veterans, or civilian counterparts. For example, suppose that mental instability were related to military service 74 in such a way that the more unstable a person was, the more likely he or she was to join the military, and once in the military to volunteer for service in Vietnam. If that were the case, the group of theater veterans would have been less stable to begin with, and we would not be surprised if we found them to be less stable now.
Although in principle one can never completely overcome the problem of nonrandom assignment of people to study groups, it is possible to minimize the impact by adjusting statistically for characteristics on which the study groups differ and that are related to the outcome that is being studied. Characteristics on which the groups do not differ, and those that are not related to the outcome being studied, are not of concern, because by definition they cannot confound the analysis.
The major focus of the analyses described in this chapter is the issue of predisposition: Is the prevalence of PTSD higher among theater veterans because of some characteristics that they brought with them to the war? At first blush, if it were to turn out that the current prevalence of PTSD is not higher among theater veterans than among the comparison groups when predisposing characteristics are controlled, then the greater prevalence observed among theater veterans could be attributed to their characteristics, rather than to their wartime experiences. However, the reality is somewhat more complicated. Differences observed today in PTSD prevalence among the study groups might be due to differences in premilitary characteristics; differential military experiences of the groups (service in a war zone, service outside a war zone, and no military service); differential postmilitary experiences, since the study is being conducted 15 years or more after the military service of many veterans had ended; or interactions among these factors. A full analysis of these factors and their interactions would provide detailed answers to some very important scientific questions, but it would be a time-consuming process. However, addressing the important policy questions would not require the complete scientific analysis. The critical policy question is: Do theater veterans have PTSD today because of how they were when they went to Vietnam, and if not, is the current PTSD prevalence related to some aspects of their experiences in Vietnam when their premilitary characteristics are controlled for?
The analyses described in this chapter are aimed at addressing this question. Our analytic strategy was a conservative one: We first examined 75 the relationship of predisposing factors to current prevalence, and then examined the relationship of exposure to war-zone stress with predisposing factors controlled. This strategy is “conservative” in the sense that it attributes to predisposition the variance in current PTSD prevalence that is due to predisposing factors acting alone (i.e., the “main effect” for predisposition) and the variance due to the interaction of predisposing factors and war-zone-stress exposure. As a result, the effect of predisposing factors may be overestimated by this procedure, relative to what might be found from a more fully specified model. Thus, the findings about the role of predisposing factors may be seen as providing an upper bound for the role of predisposition. However, a finding that exposure to war-zone stress was significantly related to current PTSD prevalence even after the effects of predisposition and the interactive effects of predisposition and stress exposure are taken into account, would be strong evidence that the current PTSD among theater veterans is service-connected.
