5
Gulf War Veterans Health Study

The previous chapter presented the committee's recommendation for the development of a research portfolio to guide studies of the health of Gulf War veterans. Within this portfolio, a prospective cohort study is seen as an essential foundation for the broader array of research. We have called this the Gulf War Veterans Health Study (GWVHS).

As of March 1998, the Department of Veterans Affairs (VA) was projecting cumulative expenditures of $115.2 million for research related to Gulf War health issues from FY 1994 through FY 1998 (VA, 1998a). The types and foci of research topics encompassed by these federally funded projects are broad and include studies at all levels of the research portfolio developed by the committee.

The contributions of the GWVHS and associated research portfolio to these myriad federal efforts are twofold. First, although several studies have focused on Gulf War veterans' health, there has been no population-based study that is both representative of the entire U.S. Gulf War veteran population and that measures current health status, compares it to groups of other veterans and the general public, and provides for the longitudinal tracking through time of changes in health status. To understand the extent to which service in the Gulf War may have affected the health of individuals deployed to this conflict, one must implement such a prospective cohort study.

Second, despite the numerous and high-quality research efforts undertaken at all levels of investigation, no mechanism has been in place that allows comparisons of health to be made between the individual study populations on important and key factors. The research portfolio designed by the committee and described in Chapter 4 would accomplish this.

To understand the contributions of the Gulf War, as well as other conflicts, to the long-term health of the men and women who served, the committee concluded that implementing a prospective cohort study that provides for comparisons with



The National Academies | 500 Fifth St. N.W. | Washington, D.C. 20001
Copyright © National Academy of Sciences. All rights reserved.
Terms of Use and Privacy Statement



Below are the first 10 and last 10 pages of uncorrected machine-read text (when available) of this chapter, followed by the top 30 algorithmically extracted key phrases from the chapter as a whole.
Intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text on the opening pages of each chapter. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

Do not use for reproduction, copying, pasting, or reading; exclusively for search engines.

OCR for page 63
5 Gulf War Veterans Health Study The previous chapter presented the committee's recommendation for the development of a research portfolio to guide studies of the health of Gulf War veterans. Within this portfolio, a prospective cohort study is seen as an essential foundation for the broader array of research. We have called this the Gulf War Veterans Health Study (GWVHS). As of March 1998, the Department of Veterans Affairs (VA) was projecting cumulative expenditures of $115.2 million for research related to Gulf War health issues from FY 1994 through FY 1998 (VA, 1998a). The types and foci of research topics encompassed by these federally funded projects are broad and include studies at all levels of the research portfolio developed by the committee. The contributions of the GWVHS and associated research portfolio to these myriad federal efforts are twofold. First, although several studies have focused on Gulf War veterans' health, there has been no population-based study that is both representative of the entire U.S. Gulf War veteran population and that measures current health status, compares it to groups of other veterans and the general public, and provides for the longitudinal tracking through time of changes in health status. To understand the extent to which service in the Gulf War may have affected the health of individuals deployed to this conflict, one must implement such a prospective cohort study. Second, despite the numerous and high-quality research efforts undertaken at all levels of investigation, no mechanism has been in place that allows comparisons of health to be made between the individual study populations on important and key factors. The research portfolio designed by the committee and described in Chapter 4 would accomplish this. To understand the contributions of the Gulf War, as well as other conflicts, to the long-term health of the men and women who served, the committee concluded that implementing a prospective cohort study that provides for comparisons with

OCR for page 63
other populations will be necessary. Additionally, collecting a common set of data on key factors in all newly funded efforts is important to enable analysts to make essential comparisons and to conduct key analyses of research results. For these reasons, the committee believes that implementation of the prospective cohort study—the GWVHS—and the accompanying research portfolio merits the time, effort, and cost required. This chapter describes the key features of this study. It does not, however, provide detailed design specifics, because the committee believes that the principal investigator who will actually implement the GWVHS should develop those aspects. Rather, the committee has provided a template for the GWVHS that addresses the questions the study is intended to answer, its general design, sampling and scheduling issues, considerations for the selection of survey modes and instruments, a proposed pilot study, cost information, ethical considerations, and oversight responsibilities. Study Questions and Design The specific design of a study is dictated by the research questions to be addressed and the target population to be studied. As discussed in Chapter 1, a variety of questions are being asked about the health of Gulf War veterans by the veterans themselves and by other interested parties, including Congress, the VA, and the DoD. The committee believes that among these many questions, of fundamental importance is ascertaining how many Gulf War veterans are suffering from health problems that affect their ability to function; whether the nature of those problems and the frequency of their occurrence in the veteran population are consistent with the experience of the general public or other groups of veterans; and whether the health of Gulf War veterans is getting better, staying the same, or deteriorating with the passage of time. Because these fundamental questions address both the health of Gulf War veterans at specific points in time and changes in their health over time, the committee recommends that a prospective cohort study of the population of Gulf War veterans be conducted. Such a study should include appropriate comparison groups. Additionally, the committee recommends that the prospective cohort study investigate the following four questions: 1.   How healthy are Gulf War veterans? 2.   In what ways does the health of Gulf War veterans change over time? 3.   Now and in the future, how does the health of Gulf War veterans compare with that of: the general population; persons in the military at the time of the Gulf War but not deployed;

OCR for page 63
persons in the military at the time of the Gulf War who were deployed to nonconflict areas; and persons in the military deployed to other conflicts, such as Bosnia or Somalia? 4.   What individual and environmental characteristics are associated with observed differences in health between Gulf War veterans and comparison groups? Further, the committee recommends that this prospective study serve as the foundation for a portfolio of needed research activities. (The research portfolio was discussed in Chapter 4.) As noted earlier, the prospective cohort study is designated by the committee as the Gulf War Veterans Health Study (GWVHS). Appendix B provides a more detailed discussion of the design, methodological and implementation issues. Although the specific decisions regarding data to be collected and modes of administration to be used are left to those actually implementing the GWVHS, the committee recommends that the prospective cohort study incorporate the following features: multiple cohorts, one for each group of interest; multistage sampling with initial cluster sampling followed by stratified random sampling within clusters; random and representative selection of participants within clusters with hypothesis-driven oversampling of specific population subgroups; and multiple modes of interviewing, including telephone and in-person interviewing. Use of a Panel Design In selecting the prospective cohort design the committee evaluated War veteran population. Based on these considerations, the committee various alternatives and considered several features of the Gulf concluded that either a permanent panel design or repeated panel surveys without temporal overlap is the preferred choice. As reflected in the questions the GWVHS is designed to answer, there is interest in both the level of health of Gulf War veterans and changes in their health over time. Panel studies with temporal overlap (rotating panel studies) are the commonly used approach when both levels and change are of interest. Use of this design calls for periodic selection of a new panel from each of the cohorts of interest because the cohorts are continuously gaining and losing members. That is, refreshing the study with a new sample that accurately represents the changed population is usually necessary. The committee concluded, however, that a study design calling for repeated selection of new panels would not be necessary because the Gulf War veteran

OCR for page 63
population is closed, that is, the population will not gain new members. Membership in this population was determined by participation in the Gulf War, specifically service in the Gulf War theater of operations between August 2, 1990, and June 13, 1991. Because the population of Gulf War Veterans is a closed population, a representative sample can be selected at one point in time; insofar as the sample is successfully followed, it will continue to be representative of the population of Gulf War veterans. The committee also concluded that the ability to locate Gulf War veterans and members of the comparison groups of veterans is likely to deteriorate over time. Although DoD maintains records with locating information for all individuals who have served in the armed forces, including Gulf War veterans, the accuracy of address information for those who have left military service becomes less reliable as the years pass. The VA's Benefit Identification and Record Locator Subsystem (BIRLS) is another source of locating information for those who are receiving benefits from VA. As this report is being written, however, 8 years have elapsed since the Gulf War. Given the anticipated deterioration of the address information in DoD and VA records, the committee anticipates that a substantial tracing and tracking effort will be necessary to locate, recruit, and follow a representative sample of Gulf War veterans in the prospective cohort study it recommends. Because the accuracy of information for selecting the initial sample will become less reliable over time, it is important to select a representative sample of Gulf War veterans as soon as possible. To minimize the cost of sample selection, it is desirable to avoid conducting costly recruitment on a periodic basis. The quality of the study data should be reviewed periodically to determine the extent to which the validity of inference based on the panels is compromised by cumulative attrition or other factors. A cohort study that is well designed and implemented holds the attrition rates over time to the lowest possible level, 5% or less for each wave. If the quality of the panel for each cohort of interest is judged satisfactory, the study would continue following the same panels. It is important, if at all possible, to continue following the same panel throughout the study to allow data collected at baseline to be used to identify predictors and correlates of health status changes at subsequent points in time. If attrition leads to a serious degradation of the sample, it may be necessary to select a new panel. If the representativeness of the sample is somewhat impaired, it may be advisable to implement a hybrid design that continues to follow random subsamples of the initial panels, and also draws new panels for each cohort to make up for the discontinued portion of the initial panels. Comparison Groups Key comparison groups are included in the design to provide a basis for identifying unique characteristics of and changes in the health of Gulf War veterans. Comparisons with civilians provide a basis for ascertaining whether levels

OCR for page 63
and trends in the health status of Gulf War veterans reflect the experience of the general population or show differences that are associated with some aspect of military service. Comparisons with those in the military at the time of the Gulf War but not deployed provide a basis for ascertaining whether deployment is associated with differences in levels and trends in health status. Comparisons with a sample of individuals who were in the military and deployed to a "safe" area provide a basis for ascertaining whether war-theater deployment is associated with health consequences or whether levels and trends in health status of Gulf War veterans are associated with general deployment to the Gulf region. Finally, comparisons with a sample of veterans of other conflicts (e.g., Bosnia and Somalia) provide a basis for ascertaining whether levels and trends in the health status of Gulf War veterans were associated with serving in any conflict or with unique aspects of the Gulf War situation. Other comparison groups could be added, as new conflicts arise in the future, to provide an ongoing basis for assessing the health effects of service in those conflicts for U.S. military personnel. For the study recommended by the committee, comparison groups are to be sampled and surveyed using the same design as that used for Gulf War veterans to maximize the comparability of the data obtained from all groups. Because members of the Gulf War veteran population and the comparison groups were not randomly assigned to those groups, the comparison will be subject to potential selection bias problems common to all observational studies: Gulf War veterans might be different from the comparison groups even in the absence of the Gulf War experience. To account for such differences, data are to be collected on potential confounding factors, such as sex, years of schooling, and other correlates of health status. Sampling Sampling Methods Many issues regarding sample selection procedures must be addressed. The most straightforward approach is to use simple random sampling. A disadvantage of this technique, for purposes of the GWVHS, is that it requires a listing of the entire study population, that is, not only Gulf War veterans but also the comparison groups. Another disadvantage is the potential costliness if data collection involves conducting in-person interviews or evaluations on a study population that is widely dispersed across the country. Stratified random sampling involves dividing a total population into groups (e.g., based on age, sex, military branch of service), and selecting a random sample within each group. This approach is frequently used when attempting to ensure that certain groups are adequately represented in the sample populations (e.g., women, American Indians). As with simple random sampling, however,

OCR for page 63
every member of the population must be listed and, in addition, must be categorized using the stratification variables. Cluster sampling is an approach used when lists of an entire population are not available, but when lists of clusters (e.g., schools within a state or school district) are available. A sample of clusters is selected and all members of selected clusters (e.g., all students in a selected school) are included in the study population. Multistage sampling, often referred to as two-stage sampling, involves initial selection of clusters, followed by listing and sampling of members within each cluster to produce the final sample (Henry, 1990). Using this approach, a sample of geographical locales is drawn first. A stratified random sample of individuals is then drawn from each sampled locale. Geographical clustering of the study population within the sampling units reduces the cost of data collection, particularly for some types of data and modes of survey implementation. For example, some of the analyses included in the GWVHS or the portfolio of studies might require the use of contextual data, such as the supply of health care services in the geographical locale in which the respondent resides. Although some contextual data can be obtained with little effort for all geographical locales, some detailed contextual data require direct data collection in the specific locales being studied. Additionally, if face-to-face interviewing is used to administer questionnaires, a geographically clustered sample is more efficient than random distribution of the study population across the country. Similarly, conducting physical examinations on a subsample of the GWVHS participants would be facilitated by geographical clustering of the sample. Over time the level of clustering can be expected to dissipate gradually, due to migration. Some of the participants who migrate out of the original locales from which they were recruited will move to a new locale that is part of or near another cluster sampled, making it relatively easy to continue face-to-face interviews. Those who move away from any cluster included in the study are likely to be costly to follow if face-to-face interviews are required. Nevertheless, the use of geographic clustering is an efficient, cost-effective approach to sampling. The committee believes that such an approach is the most appropriate method for choosing the samples in the GWVHS. Within clusters the samples would be randomly selected and stratified by demographic characteristics such as age, sex, race/ethnicity, and military characteristics such as branch of service and duty status. Sample Size Sufficient sample sizes for each cohort in the study are crucial to ensure adequate statistical power to find differences as well as to reliably identify the lack of differences between groups. The sample size required depends, in part, on the baseline level of the variable of interest in the comparison group and the magnitude of difference expected in the affected group. The committee con-

OCR for page 63
cluded that final decisions on necessary sample sizes for the GWVHS should be left to the researchers who will actually implement the study. To calculate the appropriate sample size, the number of subjects per group must be multiplied by the number of groups used for comparison. As many as five possible comparison groups have been identified for the GWVHS. If it is anticipated that 25% of potential respondents are expected not to respond, the sample must be increased by 25% to achieve the required size. If the sample is to be followed prospectively, the researchers must anticipate some attrition and they may need to choose a reserve cohort. The following are examples to illustrate possible sample sizes needed for studies comparing Gulf War veterans with other groups. Sample size calculations are made on the basis of formulas used in EpiInfo Version 6 (1/97). Assume, for example, that one or more poorly defined symptoms were present in about 1% of non-Gulf War veterans and 2% of Gulf War veterans. To establish that this difference is true with 80% power and 5% statistical significance would require a sample of 2,514 veterans in each of the two comparison groups. If a higher level of significance (1%) is desired, a sample of 3,647 veterans in each group is needed. If the condition under investigation occurs less frequently than in 1% of the population of interest and all other conditions remain similar, a larger sample will be needed. For example, assume that posttraumatic stress disorder is found in 0.7% of non-Gulf War veterans. If it is twice as common among Gulf War veterans, a sample of 3,600 from each group is needed. If the significance level is raised to 1%, a sample of 5,233 in each group is needed. If the difference in the frequency of the condition is greater than twofold, the necessary sample size is reduced. In the above example, moving from a twofold difference (0.7 to 1.4%) to a threefold difference (0.7 to 2.1%), sample size for 5% significance drops from 3,600 to 1,243; for 1% it drops from 5,233 to 1,784. Detection of differences in more common conditions requires smaller sample sizes. Assuming major depression occurs in 2.7% of non-Gulf War veterans and 5.4% of Gulf War veterans, a sample size of 908 is needed for each group for a 5% significance level; for significance of 1%, the sample size is 1,317 for each group. The committee emphasizes that these examples provide the number of individuals that are needed at the end of a study. To allow for anticipated attrition over the course of a longitudinal study, much larger numbers of participants will need to be recruited at the start of the study to ensure adequate sample size at the end. If subgroup analysis is to be performed, larger samples are also necessary. For example, if comparisons are to be made not just between Gulf War veterans and non-Gulf War veterans, but also between male and female veterans of each group, the appropriate sample size will be larger than that required when comparing only Gulf War and non-Gulf War veterans overall. If a cluster-based sample design is used, the power analyses must take into account the design effect due to intracluster correlation, which usually increases the sample size required.

OCR for page 63
Scheduling An important parameter in the design of longitudinal cohort or panel studies is the frequency with which the participants are surveyed. The committee has designed this study to include a baseline survey with two follow-up surveys (or waves) over a 10-year study period. Ensuring the timely availability of information has been considered in developing the implementation schedule because the information obtained is of value to the Gulf War veterans only during their lifetimes. The committee concluded that the GWVHS should be designed with more frequent data collection in the early years when the information obtained has a longer "useful life" to the Gulf War veterans and, if the study is continued beyond the first 10 years, with less frequent data collection in later years when the information has a shorter useful life. For the initial study period, the interval between surveys should not exceed 3 years. This interval should allow time for preliminary analysis of results and any survey modifications deemed necessary after a review of survey findings, and it should provide a short enough interval to maximize participant retention. Additionally, each wave of data collection should be conducted throughout the year to avoid the effects of seasonal variation, but there should be no temporal overlap in data collection across waves. Figure 5-1 illustrates the timing suggested for the GWVHS. To improve retention among the study participants an effort should be made to maintain contact with them during the intervals between studies. Such activities are generally referred to as tracking. During the baseline survey, the interviewer collects contact information from the participants that will help locate them for follow-up surveys. The contact information is usually updated during each follow-up survey. Additional tracking mechanisms that can be used between surveys include sending postcards, birthday cards, and newsletters to the participants at regular intervals; requesting postal notification of change of address; and requesting that participants submit change-of-address information to the study. Incentives can be offered to encourage the participants to provide this information. Figure 5-1 Timing of the Gulf War Veterans Health Study.

OCR for page 63
Mode of Survey The committee believes that those who implement the GWVHS must evaluate multiple modes of data collection based on the types of information to be collected. With the revolution in communication and information technologies in recent years, many data collection methods can be considered, including face-to-face interviews, telephone interviews, distribution and return of questionnaires by mail, or combinations of these modalities. Computer-assisted interviewing is commonly used in face-to-face and telephone interviews. It can help guide interviewers through the proper sequence of questions, check for invalid responses, and allow responses to be recorded electronically. Audio-assisted interviewing is sometimes used in the face-to-face modality for sensitive topics. It allows respondents to listen and respond to questions privately without direct interaction with the interviewer or household members who might censor responses. Internet-based interviewing via e-mail or the World Wide Web may become a possibility. Face-to-face interviewing is usually considered the most reliable method, because it allows for the use of auxiliary material such as printed response categories and interviewer prompts that are difficult to use in telephone interviews, and it allows for direct examination of the respondents, such as taking physical measurements. Also, a face-to-face interviewer can ensure that the study participant, rather than a proxy or surrogate, actually answers the questionnaire. However, some evidence suggests that this format may result in more socially desirable responses, such as underreporting of stigmatized behavior, and it is the most costly form of interviewing. Telephone interviewing is less costly than face-to-face interviews, but it is limited to those who have access to a telephone. Mail surveys are usually the least costly, but they also usually result in a higher nonresponse rate. These survey modalities can encourage respondents to provide honest answers to questions on sensitive topics, but both methods also make it easier for a proxy or surrogate to respond. Regardless of the method of survey administration adopted, GWVHS researchers should allow for proxy assessment of the health status of panel members who die during the study period or who cannot respond directly due to physical, cognitive, or psychological impairments. Improving Response Rates The willingness of Gulf War veterans and members of the recommended comparison groups to participate in the GWVHS will be crucial to the study's success. Because of the longitudinal nature of the study, the demands on participants will be even greater than for participants in a one-time cross-sectional study. In various survey-based studies of Gulf War veterans conducted to date, response rates have ranged from a low of 31% in the study of active duty and reserve personnel based in Hawaii and Pennsylvania, conducted by Stretch et al.

OCR for page 63
(1995, 1996), to 97% of those located in a survey of women who served in the U.S. Air Force during the Gulf War, conducted by Pierce (1997). Studies that included non-Gulf War comparison groups have generally found lower response rates in those comparison groups (e.g., Goss Gilroy, 1998; Holmes et al., 1998; Iowa Persian Gulf Study Group, 1997; Unwin et al., 1999). Table 5-1 presents response rate information for several such studies conducted. Maximizing response is an important issue in all survey studies, since almost all such studies fail to obtain participation from some individuals in the sample. Reasons for such nonresponse will vary: some subjects cannot be located or reached, some are too sick to be interviewed, some refuse to be interviewed. If the response rate is low, for example, less than 70%, there is the potential that nonresponse bias may seriously undermine the ability to draw conclusions from the data. Statistical and econometric techniques can reduce impact of nonresponse, but efforts can and should be made to maximize response rates. (A detailed discussion of these issues can be found in Appendix B, section 5 entitled Nonresponse, Attrition, Tracking, and Tracing.) The committee identified two particular areas that should be explored as approaches for improving participation in the GWVHS: veteran participation in organizing and implementing the study, and use of incentives. The "Iowa Study" (Iowa Persian Gulf Study Group, 1997) illustrates the potential benefit of veteran participation. Principal investigators identified their Public Advisory Committee, composed of members from veterans organizations, as a key factor in achieving high participation. This committee provided input from the beginning stages of the study, and assisted in generating participation of the veteran community. This IOM committee believes that veteran participation in organizing and implementing the GWVHS is a key element in facilitating the participation of veterans, not only from the Gulf War cohort, but also from other veteran comparison groups. Another approach to improving response rates is the use of incentives. Research has shown that monetary incentives are effective at increasing response rates; are more effective when prepaid than when promised; may induce greater respondent commitment and, therefore, improve quality of data obtained; are most effective at increasing the response rates of individuals with lower income and less education; when combined with follow-up mailings, are significantly more effective than one, two, or three follow-up mailings without incentives; and added costs are likely to be offset by savings in the costs of follow-up activities (Armstrong, 1975; Berk et al., 1987; Berlin et al., 1992; Cannell and Henson, 1974; Duffer et al., 1996; James and Bostein, 1990; Kulka, 1993; Linsky, 1975; and O'Keefe and Homer, 1987). Because of the importance of achieving and maintaining representative participation in the study, the committee urges the principal investigators implementing the GWVHS to consider the use of monetary incentives.

OCR for page 63
TABLE 5-1 Response Rates in Previous Survey-Based Studies of Gulf War Veterans Response Rate Study Characteristics Author/Citation 97% (of 525 located veterans) Mail survey; two subsequent survey rounds Pierce, 1997 92% at follow-up 1 Study population: women in U.S. Air Force (active duty, reserve, or National Guard) during Gulf War period   87% at follow-up 2     95% (of 620) Questionnaires/assessments completed during weekend training sessions (psychological symptoms) Perconte et al., 1993   Study population: Gulf War veterans from reserve units in western Pennsylvania; includes personnel not deployed or deployed to Europe   78.4% (of 2,949) Study population: Ft. Devens ODS Reunion Survey; Army Active Reserve and National Guard soldiers assessed within 5 days of return to United States Wolfe et al., 1998 Gulf War veterans: 78.3% (of 2,421) Telephone survey Iowa Persian Gulf Study Group, 1997 Non-Gulf War veterans: 73.0% (of 2,465) Study population: Active duty, reserve, National Guard troops serving in or during Gulf War period; Iowa as home of record   74.4% (of 160) at 1 month Self-administered questionnaires completed at training sessions at 1 month, 6 months, and 2 years following return Southwick et al., 1993, 1995 52.5% at 6 months     38.8% at 2 years  

OCR for page 63
Response Rate Study Characteristics Author/Citation   Study population: returning members of two National Guard reserve units (medical company, military police company) Southwick et al., 1993, 1995 74.2% (of 136) Questionnaires distributed to unit members Sostek et al., 1996   Study population: single National Guard unit   Gulf War veterans: 73.0% (of 4,262) Mail-out, mail-back survey; one follow-up questionnaire mailing Health Study of Canadian Forces Personnel (Goss Gilroy, 1998) Not deployed: 60.3% (of 5,699)       Study population: Canadian Gulf War and Gulf War-era veterans   Gulf War veterans: 70.4% (of 4,246) Mail-out, mail-back survey; two follow-up questionnaire mailings Unwin et al., 1999 Bosnia veterans: 61.9% (of 4,250)     Not deployed: 62.9% (of 4,248) Study population: U.K. Gulf War veterans, Bosnia veterans, Gulf War-era veterans   Phase I: 52.8% (of 30,000) Phase I: mail-out, mailback survey; three follow-up mailings VA National Health Survey of Gulf War-era veterans and their families (study ongoing) Phase II: 69.8% (of 30,000) Phase II: telephone interviews with nonrespondents     Study population: 15,000 U.S. Gulf War veterans, 15,000 Gulf War-era veterans   Index unit: 62% (of 1,083) Questionnaires distributed during unit training weekends; for Unit C, during 10-day period Fukuda et al., 1998 Unit A: 35% (of 1,520)     Unit B: 73% (of 1,141)     Unit C: 70% (of 2,407)          

OCR for page 63
Response Rate Study Characteristics Author/Citation   Study population: Index unit: Air National Guard, Lebanon, Pennsylvania; Unit A: other Pennsylvania Air National Guard, Unit B: Air Force Reserve; Unit C: active Air Force (Florida) Fukuda et al., 1998 Deployed: 57.3% (of 517) Mail-out, mail-back survey Holmes et al., 1998 Nondeployed: 42.2% (of 497)       Study population: activated Air National Guard Unit   41% (of 606) Mail and telephone recruiting; questionnaires completed at one of five sessions Haley, et al., 1997   Study population: 24th Reserve Naval Mobile Construction Battalion   31% (of 16,167) Questionnaires distributed through units Stretch et al., 1995, 1996   Study population: Active duty, National Guard, and reserve personnel from units in Hawaii and Pennsylvania     SOURCE: Compiled from published articles cited. Pilot Study Because the GWVHS recommended by this committee requires a major commitment of resources, a pilot study will be essential to determine its feasibility and its cost. The committee recommends a pilot study be conducted to determine the feasibility and cost of the prospective cohort study. The pilot study should include an assessment of the following points: for each of the five cohorts, identification of the universe from which the sample is to be drawn, especially the Gulf War veteran sample;

OCR for page 63
willingness of members of each cohort to participate in the baseline study; modes of data collection; and use of incentives to maximize response rates. Data Collection Instruments Three main approaches can be followed in identifying instruments to use in measuring selected health domains: use existing, validated instruments in their entirety; modify existing questionnaires (including combining existing questionnaires into larger batteries, which might be construed as a "new" instrument); and construct wholly new instruments. Using Existing Instruments Generally, the committee advises that, wherever possible, proposed surveys and studies should attempt to use existing instruments, where such instruments have been documented (ideally in the peer-reviewed literature or through other authoritative avenues) to be reliable, valid, reasonably responsive, and reasonably practical to implement. There are several reasons for this approach. The first reason is resources. Existing validated instruments need not be put through further extensive psychometric testing. Moreover, existing instruments offer the potential advantage that population norm data are available, providing a basis for valuable cross-sectional or longitudinal comparisons (e.g., with civilians or with populations facing different types of stressful events). The second reason is timing. Existing instruments could be put into the field months, or even years, ahead of questionnaires that are developed de novo. The third consideration is comparability. It may well be that at least some instruments (or parts of instruments) have been fielded already within the Gulf War veteran population; to the extent they can be used in the proposed studies over time, they will provide a better comparison with previous studies than will newly created instruments. Finally, issues of practicality arise; for example, instruments already used in federally contracted studies have passed the clearance requirements of the Office of Management and Budget and thus are easier to describe and justify. Identifying instruments is, of course, a critical exercise. Many books review available instruments in terms of their purpose, conceptual basis, reliability and validity, and other characteristics (e.g., Bowling, 1997; Frank-Stromborg and Olsen, 1997; McDowell and Newell, 1996). Documenting the quality of off-the-shelf instruments that are considered for these purposes will also enhance the perceived and actual credibility of the surveys/studies. GWVHS principal investigators should review published compila-

OCR for page 63
tions or available databases to ascertain what information is currently available regarding the reliability, validity, and similar properties of existing instruments. Another approach is direct evaluation of potential instruments against established criteria as discussed by Lohr et al. (1996). A third option is for the designers of these surveys/studies (or VA and DoD) to engage the services of evaluators knowledgeable in the assessment of a wide array of existing instruments for the purposes envisioned for this study, for example, the Medical Outcomes Trust. Modifying Existing Instruments Using existing instruments, more or less intact, may blur into modifying existing instruments. This may be desirable, or at least necessary even if not optimal, in several circumstances. For example, combining two or more instruments (e.g., a generic instrument with one or more condition-specific questionnaires) may result in redundant items, and these should be pared such that each item is asked only once. Different instruments are likely to be formatted differently, and achieving a consistent, unique appearance for any "combined" questionnaires to be employed in the GWVHS will necessarily lead to modifications of at least some of the original forms. Such reformatting should be done with care taken not to distort the intent of the original items or response categories. Length of a survey instrument may prove to be a significant consideration, and here the GWVHS principal investigators may be tempted to pick and choose items from different instruments that may appear to suit their purposes (e.g., adding to one instrument specific items from other instruments). This practice should be discouraged for questions that constitute subsets of a larger questionnaire, unless it is essential in view of interview length or time constraints, as it can make irrelevant the psychometric, scoring, and feasibility properties of the "donating" instrument. Use of such subsets of items from existing instruments requires methodological research on the subset to ensure that it is acceptable to use. Creating New Instruments If, in the end, investigators of the proposed studies choose not to use existing instruments at all, then they will face the challenge of constructing a new instrument. By and large, this requires attention to the following types of steps: laying out the conceptual model for the domains to be measured; developing, in an iterative process, successively smaller pools of items to address the domains laid out in the conceptual model; this work can be done through review of existing instruments, review of the literature, and use of focus groups or other cognitive testing procedures;

OCR for page 63
pretesting a draft instrument in the population(s) of interest, with specific attention to reliability (replicability in the test-retest sense), internal consistency (where appropriate), feasibility, and various forms of validity; revising the instrument as needed in response to problems identified in the pretesting stage; and developing (and testing) different versions of these instruments (e.g., both self- and interviewer-administered models). When the measurement model calls for instruments for which pure psychometric properties are not considered relevant, then the developers must undertake equivalent testing activities and document the characteristics of the final instrument in other ways. Care should be taken to format any new instruments in accordance with accepted principles of format and layout (e.g., those of Jenkins and Dillman, 1977). When instruments are to be administered in a computer assisted mode, additional efforts are necessary to develop and test the computer programs that are to be used. Developing and testing data collection instruments can take months and should be led by an experienced instrument development team. Space does not permit detailed description of all the steps of instrument construction, but guidelines for such endeavors can be found in sources such as Fowler (1995). Cost The committee recognizes that the cost of implementing the recommended prospective cohort study is not insignificant. Recent estimates of the cost of other large national panel surveys offer a range of reference points. The Health and Retirement Study, conducted by the Institute for Social Research of the University of Michigan for the National Institute on Aging, has an estimated cost of $13.4 million for a 2-year cycle. The household component of the Medical Expenditure Panel Survey of the Agency for Health Care Policy and Research has a budget of $13.8 million for six rounds of interviews with a single panel over a 2.5-year study period. The cost of the Panel Study of Income Dynamics, also conducted by the Institute for Social Research and funded by the National Science Foundation and other federal agencies, is estimated at $2 to $3 million per year. The budget for the Survey of Income and Program Participation, conducted by the U.S. Census Bureau, is $31 million for FY 1999. These cost estimates for ongoing survey programs do not include the initial costs to select the sample and develop the survey instruments and procedures. Ethical Considerations As with any survey or experiment involving human participants, ethical considerations must be addressed in implementing the GWVHS. For the most

OCR for page 63
part, these are governed by the requirements and regulations of the principal investigators' institutional review board. Certain issues are prominent, however, because of the features of the research portfolio this committee is proposing, because of considerations of compensation and benefits raised by the study population, and because of potential work-related impact for those remaining in the military. The results of the prospective cohort study are viewed as providing information that will generate hypotheses for further research at other levels of the portfolio. Such results could, for example, generate the desire to include a subsample of the study population in additional research efforts, such as determining treatment effectiveness for a particular condition or symptom. How and in what ways can the obligations and requirements for confidentiality of individual responses be balanced with the benefits to be derived from identifying subpopulations for future research? How will determinations be made about the validity of requests for information to conduct additional research? Who should have that decision-making authority? Additionally, questions will arise about whether the individuals enrolled in the study are bearing a disproportionate burden of intrusiveness into their lives over an extended period (relative to those never approached or enrolled in the study). The issue of initial study participation is presumed to be dealt with through informed consent, but consent for further study participation also should be explicit. These are examples of some of the important questions that must be addressed prior to implementation of the GWVHS. The following section discusses a mechanism for dealing with these and other important oversight issues relevant to implementing the research portfolio. Independent Advisory Board The long-term research strategy envisioned through full implementation of the research portfolio laid out by this committee will involve many participants: individuals from the armed services and the civilian sector, and agencies and organizations from both the public and private sectors. Moreover, many factors in the coming years may exert substantial influence on the particular questions to be asked and the specific types of projects to be carried out within this research strategy. The structure and the assumptions that underlie the committee's proposed study design are dynamic, not static. This is particularly true if the portfolio is applied to investigations of the health effects of contemporaneous conflicts (e.g., Somalia, Bosnia) and to those of as yet unforeseen conflicts, peacekeeping efforts, or wars as well as to the health problems of Gulf War veterans. Health-related factors likely to be of importance include well-known and predictable phenomena. Aging of the Gulf War population, for example, will bring more of the "traditional" chronic diseases as an overlay on health problems that some veterans now manifest or could experience in the future. At the

OCR for page 63
same time, newly emerging and reemerging diseases, particularly infectious diseases, may complicate the epidemiological picture for veterans. In addition, the number and sophistication of new health technologies—diagnostic, therapeutic, rehabilitative, and preventive—will continue to increase. Understanding of the genetic or molecular basis of disease will also improve, potentially giving rise to approaches to management of illnesses not thought of or considered possible today. To ensure high-quality research throughout the program the committee believes an independent advisory board should be established to set policies for and monitor the progress of the prospective cohort study and research portfolio of the health of Gulf War veterans and veterans of other military conflicts. Such an advisory board could include the following tasks among its functions: provide scientific and public oversight of research on issues related to the health of Gulf War veterans and veterans of other conflicts; establish policies regarding data protection and access, and review and award research grants or contracts; ensure integration of new research findings to advance understanding and treatment of war-related illnesses; and provide advice to federal agencies on research on the health of veterans. The advisory board should be an independent body to ensure its scientific integrity and public perception of validity of research results. It should be composed of acknowledged leaders in fields pertinent to the investigation of veterans' health issues and should include members who represent the veteran population. The benefits of such an advisory board are several. First, it provides for a broad range of expertise to participate in the oversight of this major, and complicated, effort to monitor the health of veterans of military conflict. Second, its agenda can be quite broad and thus encompass more than might be accomplished by any single federal department; it could include relatively immediate concerns such as funding priorities as well as longer-term issues such as how to take account of changing disease epidemiology, new health care technologies, or new populations of veterans of other conflicts. Third, it provides a visible mechanism for public accountability. Finally, such an advisory board can command national attention when it speaks or acts; it is thus in a position to call for direct, immediate, and meaningful action on the conclusions and implications of critical findings. The functions of the advisory board should include a review of the scientific and methodological merit of proposed and ongoing studies in the research portfolio. This review should take into account not only the research activities being supported or carried out within the structure proposed in this report, but also changes in various other programs within the federal government or the private sector. Of particular concern would be modifications to large national surveys (e.g., the National Health Interview Survey, the Medical Expenditures Panel

OCR for page 63
Survey, or similar periodic surveys of the Department of Health and Human Services or other federal agencies) that capture epidemiological and other data relevant to the health and health care utilization of veterans. Should any major changes be judged necessary, they can be set in motion in an orderly, but timely way so that consistency with earlier studies and methods is maintained. Therefore, the committee recommends that an independent advisory board oversee the conduct of the prospective cohort study. The advisory board should be an independent, scientific, and policy-oriented body composed of experts in clinical medicine, epidemiology, health status and health outcomes assessment; veterans' health issues; health services research; social, behavioral, physical, and biomedical sciences; survey research; statistics; national health databases; and health policy, along with members of the public who represent Gulf War veterans. review, in a timely manner, requests for proposals developed by the funding agencies to conduct the prospective cohort study recommended by the committee. evaluate the methodological design of the cohort study. set the minimum requirements for policies on methods for locating and retaining study participants, informed consent, respondent burden, confidentiality and security of data, use of incentives, and responsibility for reporting identified individual or public health threats. evaluate the success of the prospective cohort study at the end of the 10-year study period. submit a report to Congress every 2 years. Summary A prospective cohort study—the Gulf War Veterans Health Study—comparing the health of Gulf War veterans to other veteran and civilian groups provides the opportunity to obtain important information about the current and long-term health effects of military service in conflicts in which the United States engages. Such information will provide the basis for analyzing the extent to which health status is a function of a particular conflict, of participation in

OCR for page 63
conflict in general, or the result of factors common to both military and civilian populations. Special effort must be made to ensure the integrity of the study process and public acceptance of findings because of the many questions and concerns voiced by both members of Congress and the public about the openness and completeness of federal efforts to investigate the health problems of Gulf War veterans. To this end, the committee believes it essential to establish an independent advisory board composed of scientists and members of the public to oversee the GWVHS process and to report findings.