9
Future Work
Many reports of this kind end with a call for more research. The circumstances in this case, however, justify such a prompting. For despite the indications that a large amount of data has been collected and reported, more has been left undone than has been accomplished. This chapter outlines some of this remaining work in the hope that it will encourage future investigations.
First, some data have not been analyzed. It was noted earlier that POWs examined by the VA under its protocol program before this research began (the so-called volunteers) were reinvited for a second examination. A number of volunteers had a second exam, but because only one examination per subject was analyzed in this report (generally, the last exam), some earlier examination material remains unanalyzed. In particular, for the selected group of twice-examined volunteers, it is possible to compare the results of the first and second examinations; such a comparison should provide interesting information on examination variability. Along the same lines, unbeknownst to the coding contractor, a small number of examinations were submitted for coding a second time. A comparison of the two codings of the same exam would provide useful information on the completeness and accuracy of the coding process.
Other kinds of data are also available. All operations and surgical procedures have been abstracted, coded, and computerized, but have not yet been analyzed. In addition, for each medical condition, all associated medications were listed and keyed, but not coded. Both these procedural and medication data might be expected to shed additional light on selected top-
ics. For example, the Australian findings that showed an increased history of ulcer were confirmed by supplemental data on the use of cimetidine, and a possible increase in the prevalence of arthritis was suggested by increased use of analgesics.
Second, further analyses of the examination data could be undertaken to broaden the preliminary results presented in this report. For example, evidence was presented comparing the examination diagnosis of posttraumatic stress disorder (PTSD) with the Structured Clinical Interview for DSM-III-R (SCID) and the Mississippi scale diagnoses. Data are available for item-by-item analyses of the SCID and Mississippi questionnaires to determine which items are most strongly correlated with the clinical diagnosis of PTSD. Also of interest would be a comparison of lifetime and current PTSD prevalences, given that there was little difference between the two physician-based estimates and a sizable difference for the SCID. Because a diagnosis of PTSD is based on the meeting of four criteria, an examination of changes over time in criterion scores could provide an unparalleled opportunity to study the natural history of PTSD over several decades. In addition, those subjects who met some but not all criteria (''partial PTSD'') could be profitably studied. Along more strictly methodological lines, a sample of the examinations could be reassessed by an outside panel to estimate interexaminer or even interhospital variability in the diagnosis of PTSD.
The psychological data from several sources might also be profitably compared and combined to gain more insight into psychiatric illnesses. A comparison of PTSD and depressive symptoms, for example, would provide important data on the comorbidity, or joint occurrence, of these two diagnoses. For another example, a discriminant analysis might be done, using all the psychiatric data, to compare POWs versus controls or even one POW group with another. Such an analysis could help determine which psychiatric scales or items distinguish subjects with PTSD from those without it by studying these relationships in one random half-sample and validating them in the other. It should also be noted that analyses like those of Chapter 6, which link examination conditions and prison camp treatment, did not include any demographic factors, although such an addition could provide further useful information. Similar analyses (with or without demographic factors) could be done using the SCID and questionnaire-derived measures of psychiatric illness.
Third, perhaps the most potentially fruitful area of secondary analysis (i.e., that would not require the collection of additional data) involves the linkage of previously and currently collected data. For example, the 1984–1985 questionnaire survey included the Center for Epidemiologic Studies depression (CES-D) scale, thus making it possible to compare 1984–1985 CES-D scores with current scores. Because the course of depressive symptoms is known to vary, this kind of test-retest analysis should provide useful
information on the variability of depressive symptoms among POWs. Even the so-called lifetime prevalence data, based on the results of a single exam, could be refined and expanded by supplementing exam findings with earlier data from questionnaires and records to obtain a more complete, accurate measure of lifetime prevalence.
Closely related to these kinds of analyses are longitudinal studies that would link data collected earlier to subsequent outcomes. One example is the use of 1984–1985 data on depressive symptoms, together with demographic data on such factors as age and education, to predict PTSD rates in the current examination. Including some of the 1967 data from the Cornell Medical Index as well as even earlier hospitalization data (although cumulative response rates could be a problem) would permit an analysis linking five decades worth of information—a unique opportunity to explicate the processes by which such diseases progress.
Of course, a study such as this one always suggests new data that might be collected, but the disappointingly low response rates for this exam make new data collection even more important. New studies focused on a narrower clinical area and smaller and perhaps less geographically dispersed groups should have a better chance of achieving high response rates—the nerve conduction studies done by Hong at the Livermore VA Medical Center offer an example. If such efforts were to be mounted using a sample of the current respondents from the MFUA survey, any new clinical findings could be related to the examination data that have already been collected as well as to risk factor data in earlier questionnaires and records. The conditions in Table 6.7 would be obvious candidates for small-scale clinical studies, but others could also be profitably studied. The detailed diagnostic tabulations in Appendix C, which are described in Chapter 8, are provided as reference material to guide such investigations.
Finally, it has been approximately 15 years since the last mortality follow-up was completed. There is considerable speculation about current death rates among POWs but no comprehensive data. Although conditions like peripheral neuropathy and osteoarthritis can only be studied reasonably as morbid conditions, other conditions can be better investigated by using mortality data. An important point in favor of mortality studies—especially given the low response rates confronted in this study—is the completeness of death reporting, customarily 90% or better. The analyses of Chapter 6 suggest that simple comparisons of POWs and controls may not be as powerful a mechanism to identify associations with military captivity as are explicit analyses of the associations between health outcomes and prison camp factors. To date, the mortality analyses of the MFUA cohort have not attempted to associate mortality rates directly with prison camp factors. Such analyses could be undertaken and, with an additional 15–20 years of mortality data, might uncover further unsuspected links between the POW experience and subsequent medical conditions.