___________________

¹ PL 112-260 Dignified Burial and Other Veterans’ Benefits Improvement Act of 2012 § 201 (a)(1)(B).

or soil, while important considerations for the overall deployed environment, were considered to be outside of the committee’s Statement of Task.

The initial assessment committee’s Statement of Task focused on conducting an analysis of data collected in the early months of the registry’s operation and offering observations and recommendations on its operation and the best use of the data; however, the reassessment committee’s Statement of Task is broader with less emphasis on the analysis of registry data. The initial assessment committee also thoroughly reviewed the registry’s self-assessment questionnaire, and the reassessment committee concurs with the previous committee’s findings, conclusions, and recommendations related to the questionnaire. The reassessment committee limited its comments primarily to the changes made to the questionnaire since the 2017 assessment (see Chapter 3).

The reassessment committee looked to related exposure registries, epidemiologic cohorts, and military data sources for best practices and for “effective and prudent” approaches to learning about health conditions likely to result from exposure to airborne hazards and open burn pits. The committee limited its discussion to active registries, cohorts, or data sources that share some characteristic, such as target population or exposure, with the AH&OBP Registry. Some of the data sources examined as alternative means or mechanisms to the registry—i.e., the Individual Longitudinal Exposure Record (ILER) and the Millennium Cohort Study—were specified in the Statement of Task. Other sources, such as the congressionally directed World Trade Center Health Registry and VA’s Gulf War Registry, were chosen because they rely on self-reported health and exposure information and had exposures to similar airborne hazards (e.g., exhaust, jet fuel, particulate matter). These other sources are described in detail in Chapter 4 and are also referenced in discussions throughout the report.

The primary purpose of the AH&OBP Registry is to “ascertain and monitor the health effects of the exposure of members of the Armed Forces to toxic airborne chemicals and fumes caused by open burn pits”;² however, registries can serve multiple functions which can vary widely from promoting informal communication and outreach to complex, rigorous, long-term medical registries for specific diseases. Therefore, the reassessment committee considered whether the AH&OBP Registry might be of use for any or all of the following purposes or functions that were proposed by law or in publications and written materials, presentations, and open discussions with VA representatives:

• Research to ascertain health effects associated with exposures to deployment-related airborne hazards and hypothesis generation about exposure–response relationships (etiologic research) (Chapter 5);
• Population health surveillance to monitor the health of veterans exposed to burn pit emissions and other airborne hazards while deployed (Chapter 6);
• Health care access and use to improve clinical care (Chapter 7);
• Informing VA policies and processes, including claims, and improving VA programs to help veterans and service members with deployment exposure concerns (Chapter 8); and
• Communications and outreach from VA to veterans who had exposure to airborne hazards during deployment, to health care providers, and to other stakeholders, and communications from veterans to VA (Chapter 9).

The committee’s conclusions and recommendations center on the ability of the registry in its current form to perform each stated purpose bulleted above and, when applicable, on ways to improve or augment it.

The Statement of Task asked the reassessment committee to consider the point at which this or any registry that is based on self-reported information would complete its intended purpose and to consider what an end state for the registry would look like. It also asked the committee to consider “what alternative form of surveillance and identification of research candidates would need to be identified if the AH&OBP Registry was closed.” Any policy decision to maintain, expand, or close the AH&OBP Registry will necessarily be based on a range of considerations, including congressional interest and action, VA’s priorities, costs and budget, competing priorities, and value to veterans and stakeholders. The relative weights of those considerations and priorities are not necessarily equal. Because of these competing priorities, which may change over time, the committee does not make a recommendation about the specific end state of the registry.

___________________

² PL 112-260 Dignified Burial and Other Veterans’ Benefits Improvement Act of 2012 § 201 (a)(1)(B).

___________________

³ Agendas and recordings of the information gathering sessions are available at https://www.nationalacademies.org/our-work/reassessment-of-the-department-of-veterans-affairs-airborne-hazards-and-open-burn-pit-registry.

___________________

⁴ Almost half of the Department of Defense’s medical research funding is administered by the Congressionally Directed Medical Research Programs (CDMRP) (Viergever and Hendriks, 2016), located within the U.S. Army Medical Research and Materiel Command. In 2022, CDMRP listed 30 health conditions or areas of medical research for which Congress has appropriated funding (CDMRP, 2022a), with a focus on the improved prevention, diagnosis, and treatment of diseases, injuries, or conditions that affect service members and their families, and the general public. Among the military-focused programs are the Toxic Exposures Research Program, the Joint Warfighter Medical Research Program, and the Gulf War Illness Research Program. All CDMRP research programs follow the same general multistep process for soliciting, reviewing, and making funding decisions for applications. The Toxic Exposures Research Program seeks to fund research that can advance the state of the science and improve patient care related to many military exposures, including exposures to airborne hazards and burn pits (CDMRP, 2022b).

CONSIDERATIONS OF EXPOSURE REGISTRY DATA

Well-designed epidemiologic studies are one of the best means of measuring the strength of associations between exposures and health outcomes, but large epidemiologic cohort studies are expensive and may take years to design, conduct, and assess. Therefore, epidemiologists may use other methods to conduct observational or experimental research studies. Epidemiologic study designs may identify and follow cohorts prospectively or retrospectively. A registry may be used to define a cohort if most of a group of individuals who meet some specified inclusion criteria are enrolled and are followed over time.

Registries are structured systems for collecting and maintaining data on a group of people characterized by a specific disease, condition, exposure, or event as a means to facilitate research, monitor health, or provide information to participants (Gliklich et al., 2014). A registry may be as simple as a patient list, or it may be developed by extensive outreach and require that participants complete lengthy self-assessment questionnaires and provide biologic samples or other physical health data. A registry may be established to serve one or more predetermined scientific, clinical, or policy purposes, such as addressing a public concern (Gliklich et al., 2014), or to help evaluate potential health outcomes resulting from an exposure, especially if the exposure is uncommon or its potential effects are not well characterized (Antao et al., 2015). Such registries are most useful when designed for a specific goal or research question that directs how and what data are to be collected. When designed appropriately, registry-based studies can be used to facilitate research, monitor health, inform policy, or to provide information to participants, especially when cost or time constraints do not allow for more robust epidemiologic studies to be conducted.

Although many registries are established to collect quantitative data, in reality they more often produce qualitative information. For example, registries may serve as a sponsor’s response to public concerns and provide

a forum for receiving information from the public or stakeholders. In this way, registries can be used to facilitate communication with and outreach to specific populations, including providing updates on scientific and medical developments, new programs, or new policies relevant to the participants (see Chapter 9). Several registries have been established by government agencies to be responsive to concerned constituents, but that should not be the primary motivation for establishing a registry. Rather, a registry should be based on “the potential for harm from exposure, the potential benefits of a registry, and public concerns” (Antao et al., 2015, p. 1544).

As noted in the 2017 initial assessment, there are inherent limitations in using registries to draw inferences on the strength of an association between an exposure and a health outcome. Registries generally lack the design features, such as sampling, that address biases and increase the representativeness of the cohort. Motivating factors for participation and retention in a registry vary but may include the apparent relevance, importance, and credibility of the registry or its sponsoring organization (Gliklich et al., 2014). Participants may be recruited through active or passive means (or a combination), and participant data may also be collected using active (e.g., clinical examination) or passive (e.g., administrative records) means, or a combination of the two. Registries may collect data on an ongoing basis or at a single point in time, whereas cohort studies typically collect data at specified times (Gliklich et al., 2014). Data based on a single point in time or that are infrequently updated lead to concerns about validity and generalizability. The inclusion criteria are another registry consideration; registries with broader eligibility or that are more easily accessible may include many individuals who are not representative of the eligible population. Other considerations when using registries to draw scientific conclusions about associations and inferences are summarized in Table 2-1. The committee kept these factors in mind as it directed updated analyses and reviewed the AH&OBP Registry data for each of its intended purposes.

TABLE 2-1 Limitations of Registries and Their Potential Effects

SOURCE: Modified from Table 2-3 in NASEM, 2017.

___________________

⁵ A deployment—defined as any movement from a service member’s home station to somewhere outside of the continental United States and its territories (VA, 2015) until the service member returns to his or her home station—may include several segments. For example, an individual whose military unit is based in the United States and who is deployed to Kuwait for several days or weeks, then enters Iraq for several months, and returns to Kuwait before returning to the United States would have three deployment segments that would be eligible to include in the AH&OBP Registry. Deployment in total or as individual segments is not restricted to combat areas or operations; individuals or units can be deployed to assist with humanitarian aid, evacuation of U.S. citizens, peacekeeping missions, or for increased security.

⁶ Paradata include the time that a respondent spent on each question or section, time to complete the entire questionnaire, time of day it was completed, and number of stops and starts. This information is important for assessing questionnaire quality and data accuracy.

⁷ The committee is aware that at least in the first few months of registry operation waivers were granted to allow individuals who did not meet the eligibility requirements based on deployment locations or timeframes to participate.

comparisons with other participants or with the eligible population. Data elements describing characteristics of participants but that were not collected by the registry questionnaire, such as fact of death and separation status, were also specifically requested.

The committee notes that deployment information is necessary to conduct an assessment of exposure information in the registry and should include:

• Number of eligible deployments,
• Country of deployment,
• Primary base location for each deployment as available,
• Month and year of deployment start and end (dates of deployment were not provided in the initial assessment and are necessary to examine the effects of time on potential recall bias, latency, and other issues related to the quality of data for longitudinal analyses), and
• Number of days of each deployment.

VA provided a SAS dataset of the AH&OBP Registry data that was extracted June 3, 2021, a data dictionary, and an annotated questionnaire. The registry dataset contained responses to the questionnaire and several variables imported from other sources or calculated (e.g., gender, age, service branch; see Box 2-3), but the dataset did not include data from partially complete questionnaires beyond the last section of the questionnaire viewed. The data dictionary provided the following for each variable: the variable name, question wording, value labels, and notes. The annotated questionnaire included variable names and skip patterns.⁸

In order to provide the most updated information possible on registry participants, VA provided the committee with a second registry dataset extracted February 1, 2022. Along with the updated AH&OBP Registry data files, VA also provided paradata and clarifications regarding the source of some demographic characteristic variables that were imported from other VA databases into the AH&OBP Registry data files. The AH&OBP Registry data files did not contain information or responses on partial completers (those who did not click “submit”); in those cases the only information was the last questionnaire section viewed. By not having information on partial completers, the committee was not able to conduct analyses that might explain nonresponses in the questionnaire or examine differences between partial completers and full completers (participants) or between partial completers and the eligible population.

The data required cleaning before they could be analyzed. For example, several questions that are part of skip patterns did not have the skip patterns coded in the data. These skip patterns had to be added, but the existence of a skip pattern was not always indicated in the data dictionary or in the questionnaire. Creating a value in the data for valid skips is necessary to conduct analyses, especially of item nonresponse. Some inconsistencies were observed among the annotated questionnaire, the data dictionary, and the data. For example, the response options or values for a given variable in the dataset were sometimes different than those indicated in the annotated questionnaire

___________________

⁸ Skip patterns are a question or series of questions that are associated with a conditional response. For example, if a respondent answers that they have “never smoked,” questions pertaining to duration or frequency of smoking would not be applicable to them and would be skipped.

or the data dictionary. At the time of its deliberations, the committee was aware of several VA contractors and researchers that were also working with AH&OBP Registry data and notes that it is possible that if different cleaning methods or assumptions were applied, it may affect consistency across results.

The February 1, 2022, extract analysis file contained 300,472 registry respondents with a status of “participant.” After cleaning the file, 20,691 individuals (6.9%) had missing questionnaire completion date data, and 1,136 individuals (0.4%) were identified who had no eligible or validated deployment segments; these respondents were subsequently excluded from use in analyses of registry participants. The final dataset used for analysis includes 278,645 participants and 1,190,888 deployment segments. However, not all participants from the first dataset extract from June 3, 2021, remained in the final dataset. Of the 243,458 June 2021 extract participants, 81 had their status changed to a nonparticipant status (66 to eligible, 3 to review for eligibility, and 12 to missing [i.e., they were not in the February 2022 extract file]). The committee notes that additional extracts of these data files would have similar issues where participants do not match exactly between datasets, making it difficult to replicate prior analyses. Moreover, each new extract will need cleaning and cannot immediately be substituted into programs that were used for a prior extract, making the process less than user-friendly.

Veterans Affairs/Department of Defense Identity Repository

The VA/DoD Identity Repository (VADIR) contains the demographic and military characteristic data for AH&OBP Registry participants (and the eligible population). To understand how AH&OBP Registry participants differ from the eligible population, comparisons of demographic and military characteristics between participants and eligible nonparticipants are needed. This information addresses the representativeness of the AH&OBP Registry data, which is particularly important for etiologic research. Furthermore, it can be used to inform VA’s recruitment efforts and identify areas where these efforts could be improved, such as to groups that are underrepresented (e.g., a particular service branch or unit component). Previously, this information was made available by the Contingency Tracking System (CTS)⁹ extract provided by DMDC and the Gulf War Oil Well Fire Smoke Registry File. The following data were requested:

• For all eligible service members: sex, month and year of birth, race, ethnicity, education, marital status, service branch, component, number of eligible deployment segments, separation status, separation date, fact of death, and date of death.
• For each eligible deployment: rank, duty occupational specialty code, start month and year of deployment, end month and year of deployment, length of deployment, and country/countries of deployment.

Other data sources or roster files available to VA may be suitable sources of demographic and deployment information for individuals eligible for the AH&OBP Registry, for example, the United States Veterans Eligibility Trends and Statistics (USVETS) file. Because the CTS was not updated after 2015 and neither USVETS nor VADIR was made available to the committee’s subcontractor, a table shell of demographic and military characteristics was provided to VA to complete for the post-9/11 eligible population. VA used VADIR to complete this table (Table 2-4), and it included gender, age, race/ethnicity, service branch, unit component, number of deployment tours, and deployment country (limited to Afghanistan, Iraq, Kuwait, and other) for 2,368,542 post-9/11 veterans separated from the military as of November 30, 2021.

Gulf War Oil Well Fire Smoke Registry

This registry contains records of all service members who served in Operation Desert Storm and Operation Desert Shield while the oil-well fires were burning to allow for comparisons with eligible nonparticipants of that

___________________

⁹ In the initial assessment by NASEM (2017), that committee used the CTS, which is a subset of DMDC data, for service personnel who were physically located in the Operation Enduring Freedom and Operation Iraqi Freedom areas of operations or who were specifically identified as directly supporting those missions outside of the designated combat zone, such as aircrew or support personnel. The CTS contains demographic and military characteristics on respondents that is not collected by the registry. The committee was told that the CTS has not been updated since December 2015.

era. These data were provided to the National Academies for the initial assessment of the AH&OBP Registry. The request included the following variables:

• For all eligible service members: sex, month and year of birth, race, ethnicity, education, marital status, service branch, component, number of eligible deployment segments, separation status, separation date, fact of death, and date of death.
• For each eligible deployment: rank, duty occupational specialty code, start month and year of deployment, end month and year of deployment, length of deployment, and country/countries of deployment.

No data file from the Gulf War Oil Well Fire Smoke Registry was provided. However, because the demographic and military information from it is unlikely to have changed substantially since the initial assessment, this information was taken from Table 4-4 of the initial assessment report. It is presented later in this chapter (Table 2-4) where AH&OBP Registry participants are compared to the eligible population by service era. One limitation of reusing the Gulf War Oil Well Fire Smoke Registry file information as presented in the initial assessment is that the inclusion and exclusion factors cannot be revised.¹⁰ As such, the 545,383 individuals who constituted the eligible population for 1990–1991 Gulf War veterans as included in the Gulf War Oil Well Fire Smoke Registry in 2015 were used as the eligible comparison for the current update.¹¹

VA Electronic Health Records

To examine self-reported and diagnosed health conditions, VA electronic health record data were requested for both the entire eligible population and the participants of the AH&OBP Registry who use VA health care. On December 2, 2021, VA provided a data file of 243,331 registry participants (using registry participants from the June 3, 2021, extract) with binary (yes or no) annual VA health care use flags from 1992 to 2021 (no months or days provided). VA did not provide data for 127 registry participants, so the analyses of health care use were restricted to 243,331 registry participants with eligible deployments and VA use data. On April 1, 2022, VA provided a data file of 23,249,138 VA health care system outpatient ICD-10 (International Classification of Diseases, 10th Revision) diagnosis codes and dates (October 2015 through February 1, 2022) corresponding to 199,735 AH&OBP Registry participants. Data for ICD-9 codes were not provided, and therefore diagnoses prior to October, 2015 were not available.

Assuring Quality Control of Analyses

The committee’s subcontractor, Westat, used a variety of systems and procedures to ensure that the committee’s analyses were performed accurately and that the results were high quality. Such processes included a review of the analysis plan and any modifications made to it based on what data were received, checks of raw data files, and a multistep review process of results, as described below.

The first step in ensuring the integrity of the data files received from VA—and therefore the accuracy of the committee’s data analysis—was to generate frequencies of each item in the data files. Data were examined for logic errors, range errors, and other types of errors. Several questions that were part of skip patterns did not

___________________

¹⁰ For example, in the initial assessment, a total of 5,621 1990–1991 Gulf War veterans completed the AH&OBP Registry questionnaire; of those, 496 were found to be ineligible and were excluded based on data from the Gulf War Oil Well Fire Smoke Registry file. The 801 participants who were eligible according to the Gulf War Oil Well Fire Smoke Registry file but were ineligible based on questionnaire deployment data were included in the analysis because the DoD data were considered to be the gold standard for determining eligibility when manual verification was not an option (NASEM, 2017).

¹¹ The total number of 1990–1991 Gulf War veterans who served in Southwest Asia as part of Operation Desert Storm and Operation Desert Shield is 696,530 as reported by VA (U.S. Senate, 1998). The Gulf War Oil Well Fire Smoke Registry includes only those individuals who were present in Southwest Asia while the oil-well fires were burning (n = 545,383). The initial assessment committee determined that using the Gulf War Oil Well Fire Smoke Registry would more accurately reflect 1990–1991 Gulf War veterans who were eligible to participate in the Airborne Hazards and Open Burn Pit Registry for comparison purposes.

have the skip patterns coded in the SAS data files, making it impossible to distinguish valid skips from missing data; Westat coded the skip patterns in the data. All corrections, updates, or additions to the data were recorded and documented in the SAS analysis programs. Full documentation was maintained on all programs used in the process of building the analysis file.

Each analysis requested by the committee was first reviewed by the subcontractor to ensure that committee’s intent was clear and that the most appropriate statistical methods were used. Then, a research analyst translated the request into detailed specifications for a SAS programmer. Specifications included instructions for constructing new variables and for stratifying the data to specific subgroups (such as deployment by era) as well as statistical procedures to be used. The specifications were reviewed by senior staff prior to sending them to the programmer. All programs used to run the analyses were fully documented by the programmer.

The committee’s analyses required the construction of new variables, including disease measures, such as combining chronic obstructive pulmonary disease (COPD), emphysema, and chronic bronchitis into a single outcome. For each constructed variable, cross-tabulations of the raw variables and the constructed variables were made and reviewed to ensure that all of the variables were coded correctly in the constructed variable. As an additional check, a small sample of variables were selected and reviewed a second time to ensure that the constructed variables were properly coded based on the raw data.

Subcontractor staff met at least weekly to review output from the analyses for accuracy and to determine any potential revisions that the analysis required. National Academies’ staff and committee members were consulted during programing on an as-needed basis. Changes were communicated to the programmer via revised specifications. Before the final written results of an analysis were delivered to National Academies’ staff and committee, they were reviewed internally, typically by senior staff, for accuracy, completeness, and quality. Subcontractor staff who had performed the analyses attended the data-focused portions of the committee meetings. The results were described and reviewed in detail with the committee and requested changes to them or additional analyses were documented by National Academies’ staff and sent to Westat. This iterative process continued throughout the course of the committee’s deliberations.

Analyses Conducted

The reassessment committee has a different Statement of Task than the initial assessment committee, and as such, it was unnecessary to repeat several initial analyses such as the multivariable analyses. Those analyses that were not repeated are discussed first. Then, the committee presents those analyses that were repeated or updated since the initial assessment, including descriptive statistics of the AH&OBP Registry population. The final section describes new and augmented analyses.

Analyses Not Repeated

The initial assessment committee analyzed health outcomes with latencies of more than 6 months and less than 10 years after exposure. Health outcomes were limited to those associated with the respiratory and the cardiovascular systems because these are the most plausible and well-documented health effects associated with airborne hazards and are emphasized in the registry. The multivariable analysis included the outcomes of asthma; emphysema, chronic bronchitis, or COPD as a composite variable; any functional limitation due to a lung or breathing problem; respiratory symptoms as a composite variable; hypertension; and a composite variable of coronary artery disease, angina pectoris, and myocardial infarction. Although separate questions were asked for some of these conditions, that committee chose to group some of them as composite variables because it was concerned about the ability of participants to distinguish among them. A few outcomes in other organ systems were included in the questionnaire, but the initial assessment committee determined that those questions were too general to be of value in its analyses. The initial assessment committee also noted that the health outcomes were not captured with the specificity that would be necessary to draw conclusions about the presence or absence of specific diagnoses in the registry population.

only applicable for those veterans who are eligible for that health care and who choose to use it (see Chapter 7). Follow-up of active-duty service members, veterans not eligible for VA health care, and veterans who choose to receive care from non-VA providers is not possible using VA health care records. One analysis of VA health care services among 1990–1991 Gulf War veterans that used those services between January 1, 2008, and December 31, 2018, noted that more than 50% of veterans did not use the VA health care system (Dursa et al., 2020). Furthermore, the population of veterans who use VA health care for all of their health care needs differs from the entire veteran population and from those who do not use VA health care (e.g., VA users have higher prevalence of 21 chronic conditions than non-VA users [Dursa et al., 2016]). Therefore, the committee sought to link VA health care use with clinical encounter information for registry participants enrolled in VHA to understand whether their characteristics are connected to their use of VA health care.

Participants

This section describes the demographic and military characteristics of individuals who chose to participate in the AH&OBP Registry. The committee examined how well the participants reflected the larger eligible population and whether the characteristics of late participants were different than early participants. The final AH&OBP Registry data analysis set contained 278,645 participants and a total of 1,190,888 deployment segments. There does not appear to be a minimum number of days for a deployment to be counted as eligible. Participants were separated into mutually exclusive eras based on their eligible deployment segments:

• 1990–1991 Gulf War: deployment that started on or before April 6, 1991;
• Peacetime: deployment between April 7, 1991, and October 6, 2001;
• Operation Enduring Freedom (OEF)/Operation Iraqi Freedom (OIF)/Operation New Dawn (OND): deployment between October 7, 2001, and December 31, 2014; and
• Combined Joint Task Force–Operation Inherent Resolve (OIR)/Operation Freedom’s Sentinel (OFS): deployment ended on or after January 1, 2015.

Deployments with more than one era flag were categorized as “multiple eras.” Among AH&OBP Registry participants, 1.5% served only in the 1990–1991 Gulf War and 76.2% of participants served post-9/11; 21.1% of participants served in multiple eras (Table 2-2). Of those who served in multiple eras, the majority (11.9% of all participants) served in OEF/OIF/OND and OIR/OFS.

Location of deployment was examined in two ways: first by individual and second by deployment segments. An individual may have had multiple deployments (mean = 4.3, median = 3) to several countries or areas in Southwest Asia. Therefore, in Table 2-3 there are 278,645 individual participants in the AH&OBP Registry, but the column totals are more than the number of participants. The countries with the highest percent of ever deployments are Iraq (58.2%), Kuwait (50.7%), and Afghanistan (45.2%), and very few deployments were to eligible bodies of water (i.e., Arabian Sea, Gulf of Aden, Gulf of Oman, Persian Gulf, or Red Sea).

When location by deployment segments was used, most were to Iraq (30.0%), Kuwait (34.3%), and Afghanistan (22.1%) (Table 2-3). As discussed in Chapter 1, deployment segment eligibility is expanding to additional countries, such as Syria, Uzbekistan, and Egypt; however, they are not yet captured in the registry data as of February 1, 2022, and therefore are not included as possible countries in the tables.

The committee then examined other military and demographic characteristics stratified by era of service (Table 2-4). Era of service is limited to 1990–1991 Gulf War and post-9/11 operations because information on the eligible population was not available for the Peacetime era or for specific post-9/11 operations (such as differentiating OFS and OIR from OEF, OIF, and OND). As a percentage of the entire eligible population, by 2015 only 1.0% of eligible 1990–1991 Gulf War era veterans had completed the registry questionnaire, and only 1.7% of eligible

TABLE 2-3 Countries and Areas of Deployment by Participants and Eligible Segments

NOTE: Participants can be deployed to more than one country. Although Syria, Uzbekistan, and Egypt have become eligible deployment locations, they were not reflected in the AH&OBP Registry data as of February 1, 2022.

NOTES: Because of the way this table was generated for the initial assessment report and updated here, some categories are combined, whereas the preference would be to present them separately; these include National Guard and reserves (under Component), Navy and Coast Guard (under Branch), and countries that are not Iraq, Afghanistan, or Kuwait. Unit component was not included in the February 1, 2022, extract, but was included in the June 3, 2021, extract; therefore the unit component is based on the June 2021 extract, which had 18,739 1990–1991 Gulf War era participants and 228,812 post-9/11 era participants. The VA-provided data from VADIR on post-9/11 eligibles used the phrase “deployment tours,” and the metric does not appear comparable to deployment segments as reported on the registry. When asked for clarification on this variable, VA responded that it is from a VADIR variable called “Tour Count” and that it should not collapse shorter deployment segments, which the committee interpreted to mean that it should match, but as can be seen, there is a large discrepancy.

* Percent difference indicates the percentage differences for a characteristic, i.e., respondent percentages minus eligible percentages.

post-9/11 era veterans had initially completed the questionnaire. After nearly 7 years of AH&OBP Registry operations, the completion rate has increased to 3.9% and 11.0% for these eras, respectively. Although the questionnaire completion rate continues to be low, the absolute number of registry participants is large, as shown in Table 2-4.

Registry participants are not representative of the total eligible population or by the era of service based on their demographic and military characteristics. Notably, relative to the eligible 1990–1991 Gulf War population, registry participants who served in this era are more likely to be women and over 50 years of age. They were also more likely to have served in the Army and in the National Guard or reserves. Finally, they had a notably different distribution of eligible deployment segments, with 55.7% having a single deployment compared with 38.4% of the entire eligible population. Comparisons of post-9/11 participants to the eligible population of that era found that registry participants were less likely to be women and 50 years of age or older. They were more likely to be less than 30 years of age and of White race. Post-9/11 registry participants were more likely to have served in the Army

and Air Force and in the National Guard or reserves and more likely to have been deployed to Iraq or Afghanistan as well as to have had more deployment segments. Defining the eligible population is also difficult, as it is ever-growing as more service members are deployed or redeployed and new countries or deployment locations, such as Syria, Egypt, and Uzbekistan, are made eligible for the AH&OBP Registry (see Chapter 3).

Participants by Time of Completion

To examine how the characteristics of registry participants had changed since the initial assessment, participants were stratified into early and late by time of participation, as described previously. The initial assessment included 46,404 registry participants; however, in the committee’s AH&OBP Registry file extracted February 1, 2022, 46,444 registry participants had completed the questionnaire as of July 31, 2015—40 more participants than in the initial assessment. There is no way to verify that all of the 46,404 participants included in the initial assessment data are the exact same as those included in the early strata in the current dataset. Although this discrepancy is small, because access to the dataset used for the initial assessment was destroyed per the data use agreement at the release of that report, it is impossible to pinpoint the reason for this small discrepancy. This also reinforces the earlier observation that extracts of these data files do not match included participants exactly, making it difficult to replicate analyses.

Table 2-5 presents descriptive statistics of demographic and military characteristics for the eligible population (does not include Peacetime era and not all characteristics are available), all the AH&OBP Registry participants in the committee’s dataset, and participants by date of registry participation (i.e., early and late participation). Several differences in these demographic and military characteristics can be seen, especially between early and late participants as shown in the “percent difference” column. These statistics are described in more detail in Chapter 3. With regard to the deployment era, 79.8% of early participants served during OEF/OIF/OND only, 2.2% during the 1990–1991 Gulf War only, and 17.5% during multiple eras. The late participants had fewer 1990–1991 Gulf War–only (1.4%) and OEF/OIF/OND-only (67.3%) deployments and more deployments for Peacetime-only (1.3%), OIR/OFS-only (8.1%), and multiple eras (21.9%). Late participants had fewer eligible deployment segments and shorter cumulative deployment durations than early participants. In general, early and late registry participants vary slightly in demographic and service characteristics, with the most notable variation being in deployments, military branches, and age (Table 2-5).

Verification of Deployment Segments

Eligible individuals who consent to participate in the registry are first asked to verify the dates and locations of eligible deployment segments according to information contained in VADIR. The questionnaire lists the individual’s eligible deployment segment records, and participants can confirm the system data, correct them, or enter additional deployment history information. Individuals can modify the dates of deployment, add missing deployments, and select or enter the bases they served at while deployed. The initial assessment report committee heard from veterans who had participated in the registry that the process of updating deployment information and entering the names of bases and dates was difficult and frustrating. A 2021 demonstration of this questionnaire to the reassessment committee confirmed that this process of entering names of bases and dates remains difficult (VA, 2021b).

The registry records whether the deployment segment information was indicated by the participant as accurate (verified) or the participant changed the presented information (user-entered). Of the total 1,190,888 eligible deployment segments in the committee’s registry data file, 77.4% (n = 921,203) were verified as correct and 22.6% (n = 269,685) were user-entered. To determine why so many deployment segments had inaccurate information for registry participants, the committee conducted several analyses that examined whether certain military- or deployment related characteristics were more likely to be associated with either verified or user-entered deployments. The analyses might indicate whether system-level changes are necessary to more accurately identify or code deployment segments as eligible. Whether a segment was verified or user-entered was strongly related to

NOTE: This eligible population excludes the Peacetime era between the Gulf War and post-9/11 eras.

* Percent distributions for these sections have been merged; the percentage applies to the combined total population distribution.

Because of the way this table was generated for the initial assessment report and updated here, some categories are combined, where the preference would be to present them separately; these include National Guard and reserves (under Component), Navy and Coast Guard (under Branch), and countries that are not Iraq, Afghanistan, or Kuwait. Unit component was not included in the February 1, 2022, extract, but was included in the June 3, 2021, extract; therefore, the unit component is based on the June 2021 extract, which had 46,444 early participants and 232,201 late participants.

** Percent difference indicates the difference in percentages between early and late participants in each of the characteristics, i.e., early participant percentages minus late participant percentages.

era of service, with pre-2001 deployments much more likely to be user-entered. In particular, of the 40,963 eligible 1990–1991 Gulf War and Peacetime deployment segments, 92.1% were user-entered, whereas 20.2% of all eligible post-9/11 segments (n = 1,149,925) were user-entered (see Table 2-6). When user-entered deployment segments were examined by country of deployment for those participants with a deployment start date before September 11, 2001, all deployment segments to Afghanistan, the Arabian Sea, Djibouti, the Gulf of Aden, the Gulf of Oman, the Persian Gulf, and the Red Sea were user-entered. For post-9/11 deployments, the areas with the highest proportion of user-entered deployments were to the Gulf of Oman (59.0%), the Gulf of Aden (47.0%), Djibouti (36.3%), Oman (32.5%), and the Persian Gulf (30.4%); about 26% of deployment segments to both Iraq and Afghanistan were user-entered.

The data were then examined to determine whether participants who had more eligible deployments were adding those deployments manually (as opposed to verifying existing deployments). Participants were stratified based on having at least one user-entered deployment (n = 128,769) segment or no user-entered (all system-generated) deployment segments (n = 149,876), and the numbers of deployments were then compared (see Figure 2-1). The number of deployments for a single individual (including those with 0 days; see below) ranged from 1 to 148, with a mean of 4.3 deployments and standard deviation 4.0. Only 2% of participants had 16 or more eligible deployment segments.

The average number of deployments is lower among those with at least one user-entered deployment than among those with all verified deployments (3.9 vs 4.7 deployments). Often, participants with many deployments had all their deployments generated by the system. Of those participants with fewer than 20 total deployments, 54% had no user-entered deployments, and for those participants with 20 or more deployments, 59% had no user-entered deployments. Therefore, having many eligible deployments does not appear to be due to individual participants adding these deployments manually, and, moreover, adding deployments manually is not shown to be a primary reason why a participant would have a high number of eligible deployments.

To further examine the issue, a selection of participants with very large numbers of deployments were manually examined. Nearly all of their deployments were system-generated, and many of the corresponding exposures questions asked for each deployment had missing or “Don’t know” responses. In some instances, a participant would leave missing all the exposure questions for system-generated deployments and then manually enter one deployment and answer the exposure questions for only that single deployment. One observed anomaly was a

participant who manually entered over 100 deployment segments, including answers to the deployment exposure questions for all of these deployment segments.

The committee examined participant characteristics that may be driving verification of system-generated versus user-entered deployment segments. Participants were first stratified by date of last deployment (before or after September 11, 2001) before being sub-stratified by none or at least one user-entered deployment (see Table 2-7). Only 6% (n = 17,222) of participants had no deployments after September 11, 2001, and, of these, 83% (n = 14,280) had at least one user-entered deployment. Those 1990–1991 Gulf War participants with at least one user-entered deployment were more likely than those participants of the same era with no user-entered deployments to be 50 years of age or older (at the time of questionnaire completion), to be White race, to have served in the Air Force or Navy, to have served during Peacetime, to have had two or more deployments, and to have had spent a total of 9 months or more deployed. Comparatively, the proportion of post-9/11 participants with at least one user-entered deployment was much less than for the 1990–1991 Gulf War participants.

Of the participants who had at least one deployment after September 11, 2001, 43.8% (n = 114,489) had at least one user-entered deployment. Those post-9/11 participants with at least one user-entered deployment were more likely than those participants with no user-entered deployments to be 40 years of age or older (at the time of questionnaire completion), to have served in the Air Force, to have had their first deployment before 1992 or after 2015, to have served in multiple eras, to have had two to nine deployments, and to have had spent a total of at least 12 months or more deployed.

The duration of deployment segments was examined next. Nearly 40% (n = 461,277; 38.7%) of deployment segments were found to be of less than 30 days duration. To gain a better understanding of very short deployment segments (<30 days), the committee compared them with longer deployments (≥30 days). Three findings stood out regarding very short deployments: they were less likely to have been user-entered (8% vs 28%), they were less likely to have self-reported burn pit exposure (39% vs 72%), and they were more likely to have been to Kuwait (64% vs 17%).

Another difference noted was that the percentage of “Don’t know” responses for question 1.2.D (which asks whether a respondent had been exposed to a burn pit on deployment) for deployments of 0–30 days was more than twice as high as for deployments of 31 days or longer. In addition, nearly 6% of all eligible deployments have the same start date and end date, resulting in 0 days duration, and 4% have a 1-day duration (see Table 2-8). These 0- and 1-day deployments were disproportionately system-generated rather than user-entered (96.5% vs 24.9%), were more likely to be from the post-9/11 era than the 1990–1991 Gulf War era (99.7% vs 96.0%), and were more likely to have had a deployment segment to Kuwait than not (65.2% vs 30.7%). Another difference was that the percentage of “Don’t know” responses for segments of 0–30 days was more than twice as high as for deployment segments of 31 days or longer.

Among participants with a deployment segment of less than 31 days, 99% had at least one additional deployment, and most often these deployment segments lasting less than 31 days were contiguous with other, longer deployments. Some of the very short deployments may represent anomalies in the data, or they may potentially reflect the use of Kuwait for the movement of U.S. troops in and out of the Southwest Asia theater.

Time to Complete Each Questionnaire Section

Although the registry website states that the questionnaire may take up to an hour to complete (VA, 2022), the registry factsheet available on VA’s website states that the questionnaire takes about 40 minutes to complete (VA, 2018). An early analysis of the time required to complete the questionnaire, which did not include the first section of deployment segment verification, found that of the nearly 37,000 participants who had completed the questionnaire, about 75% completed it in 45 minutes or less (Ciminera, 2015b); the median time of completion was 31 minutes. An internal VA analysis of time to complete the questionnaire using a smart phone or tablet found that the average same-day completion time was 61 minutes¹²—double the median time. The longer completion time on a smart phone or tablet may be because it is not formatted for these modes.

___________________

¹² Personal communication, Michael Montopoli, director, Post-9/11 Era Environmental Health Program, VA, September 15, 2016.

TABLE 2-10 Median and Mean Time (in minutes) to Complete the AH&OBP Registry Questionnaire by Number of Deployment Segments

NOTE: Analysis based on 209,715 participants who completed the questionnaire within 360 minutes (6 hours) after starting.

approximately 10% had 10 or more. Table 2-10 shows that the median time to complete was below 40 minutes for those with 1–4 deployment segments, compared with 51 minutes for those with 10 or more deployment segments, a difference of 11 minutes or more. Thus, on average, each additional deployment segment adds 1–3 minutes to the median questionnaire-completion time and 1.1–2.8 minutes to mean questionnaire–completion time. Therefore, given the additional time required for verifying and correcting deployment information, participation in the registry can be a very time-consuming process, which may contribute to not all of the eligible segments being completed or to dropout at that or later stages, which would introduce additional forms of selection or nonresponse bias.

Participation and Completion by Questionnaire Section

The February 1, 2022, data extract did not contain any information on eligible noncompleters, but the committee’s registry data files extracted on June 3, 2021, included 243,458 participants and information on 131,645 eligible noncompleters, which was used for analyses of noncompleters. An eligible noncompleter was defined as an individual who logged in to the registry, completed the consent form, and was found to have at least one eligible deployment segment, but who did not complete the full questionnaire or submit it. Table 2-11 shows the last section of the questionnaire viewed by noncompleters, which was found to be an acceptable proxy for questionnaire section completion. This was confirmed by the committee during the live demonstration of the questionnaire which showed that participants are required to answer all the displayed questions to move onto the next set of questions or section.

Similar to prior VA analyses of noncompleters, the committee found that 42.9% of eligible individuals accepted the consent but did not answer a single question and that 33.8% did not proceed past the first section on deployment history. The detailed review of verified versus user-entered deployment segments presented in the previous section highlights the difficulty of completing the self-assessment questionnaire for some participants, as the first section of the questionnaire requires a great deal of manual review and entry, which is likely a reason for noncompletion.

VA Health Care Users

According to the VA-provided data on 228,696 AH&OBP Registry participants matched to VA health care data between 2001 and 2021, 170,122 (74.4%) of those participants had used VA at least once. Table 2-12 shows the demographic and military characteristics of VA users and non-VA users. Registry participation appears to be

associated with an increased use of VA health care, as discussed in more detail in Chapter 7. Registry participants who used VA health care at least once between 2001 and 2021 were more likely to be over 40 years of age, to be Black or Hispanic, to be retired or separated from the military, to have served in the Army, to have served in the reserves or National Guard, to have served in OEF/OIF/OND, and to have had five or more deployments.

Health Conditions

The AH&OBP Registry questionnaire asks about a wide range of symptoms and health conditions, but the questions and the response options do not have the specificity necessary to allow them to be used to draw conclusions about the presence or absence of specific diagnoses among the registry participants. This section describes the health status of registry participants: overall, by early and late time of participation, and by VA enrollment status.

The health outcomes included in the questionnaire and the proportion of all participants who indicated an affirmative diagnosis stratified by early and late participation are shown in Table 2-13. Three questions that pertain to health conditions ask about frequency (how many days or how often) instead of a yes or no response (1.4.E, 2.2.3.H, 2.2.3.I); for these questions, a response was counted as an affirmative diagnosis if the frequency was greater than 0.

The percentage point difference between early and late participants for health outcomes is relatively small. The greatest differences were for seeking medical care during deployment for wheezing, difficulty breathing, or other symptoms (4.0%); experiencing wheezing, difficulty breathing, or other symptoms during deployment (3.1%), and having neurologic problems in the past 12 months (3.1%). When examining respiratory health conditions specifically (questions 2.2.1.A–H), the differences in percentage points were 1.1% or less. Slightly higher affirmative responses for respiratory conditions among the early participants are observed for asthma; emphysema, chronic bronchitis, COPD; some other lung disease; and constrictive bronchiolitis. Although the difference in percentage points is small, early participants were less likely than late participants to report having hay fever or allergies (-0.8%), hypertension (-0.8%), other heart conditions (-0.9%), stopping breathing during sleep (-2.6%), and having a cancer diagnosis (-0.2%). Given the differences in personal or military characteristics, such as age, number of deployments, cumulative length of deployments (see Table 2-5), it is possible that these characteristics account for the modest differences in health outcomes. Early respondents have lower item nonresponse rates than later respondents on almost every item (data not shown). One possible explanation for this difference is that the early participants had higher levels of engagement in the registry and thus had higher item completion rates.

Table 2-13 shows that 16,072 (5.8%) of participants responded “Yes” to registry question 2.4.A, “Have you ever been told by a doctor or other health professional that you had Cancer or a malignancy (tumor) of any kind?” Of those, 15,170 (5.4%) of registry participants went on to specify one (n = 12,729), two (n = 1,904), or three (n = 537) types of cancer diagnoses. There were 18,148 cancer diagnoses reported among these participants (Table 2-14), the most common being skin cancers and melanoma, followed by prostate cancer.

Exposures

The reassessment committee updated and augmented specific exposure analyses presented in the initial assessment. The analysis of the registry exposure data included a careful inspection of descriptive data and a consideration of qualitative exposure variables. This section describes the exposure experience of registry participants: overall, stratified by early and late time of participation, and by VA enrollment status. It also presents the methods used to crudely assess concordance of self-reported exposures with known time frames and locations where burn pits were operating.

Table 2-15 shows deployment-specific and any military-service-specific exposures collected by the questionnaire for all participants stratified by early and late participation. With the exception of question 1.3.A, the questions

that ask about experiencing these exposures ask about frequency (how many days in a typical month) instead of a yes or no response. Therefore, an exposure was counted as yes if the frequency was greater than 0 days. The early participants reported all exposures more frequently than the late participants. The largest percentage point difference between early and late participants was 7.1, for being close enough to feel the blast of an improvised explosive device (IED) or other explosive device (72.2% of early participants vs 65.1% of late participants).

VA Health Care Users

Using the VA data file of registry participants who used VA health care between 2001 and 2021, Table 2-16 shows unadjusted responses to the exposure-related questions. When stratified by registry participants who used VA health care at least once versus those who never used VA health care over the same period, 74.4% used VA health care at least once. Those using VA care were more likely to respond yes than those who did not use VA health care to all questions about exposures. This was also observed for the highest stratum of all military- and deployment-related exposures that asked about days per month of exposures. Some of the largest differences between users and non-users of VA health care were for being close enough to feel a blast from an IED (71.8% vs 58.5%, respectively), exposure to heavy combat smoke for 25–31 days per month (23.7% vs 12.8%, respectively), being in a convoy for 25–31 days per month (31.7% vs 20.3%, respectively), experiencing respiratory symptoms

due to air quality (21.1% vs 13.1%, respectively), and having some burn pit duties on at least one deployment (59.7% vs 47.5%, respectively). Although this table presents a simplistic univariable analysis that may suggest that use of VA health care is associated with high exposure and high symptom burden, without properly adjusting for potential confounders and characterizing the two populations of users and non-users it is difficult to determine the direction of association with certainty.

Self-Reported Exposure Concordance

The committee examined self-reported burn pit exposure to check for concordance of this information against DoD location data using two methods. For both analysis methods, the deployment base was first identified using responses to two questions. First, if participants added a deployment segment in response to question 1.1, they could enter a base name. Second, question 1.2.B asked participants about the base at which they spent most of their time during the deployment segment. Only one base name can be entered for either question. If a participant indicated one of the bases of interest for either of these variables, they were considered to have been at that base during the deployment. The same three burn pit exposure questions were used for both analysis methods: 1.2.D–near a burn pit on this deployment; 1.2.F–duties that included the burn pit on this deployment; and 1.2.G–number of hours per day smoke or fumes entered your worksite or housing on this deployment.

In the first method used to check for concordance, participants were restricted to those with deployment segments to Joint Base Balad, Iraq, because it was documented to have the largest burn pit in the Southwest Asia theater (IOM, 2011). Self-reported burn pit exposure was determined using responses from the three questions from section 1.2. Comparisons of self-reported exposure to burn pits were made before and after burn pit operations were closed at Joint Base Balad. Between July 2007 and October 2009, four incinerators were put into operation at Joint Base Balad, resulting in 100% of solid waste disposal via incineration or off-site recycling (AFHSC et al., 2010). Using base information provided in section 1.2 of the self-assessment questionnaire, participants with deployments at Joint Base Balad were identified and divided into three periods: before incinerator installation (January 2003–June 2007), during incinerator installation (July 2007–October 2009), and after incinerator installation (November 2009–present). Participants with a deployment in two or more of the periods were put into the period during which they were deployed the greatest number of days. Self-reported exposure to burn pits was then compared among deployments in each period (Table 2-17). However, the committee is aware that the implementation of the incinerators did not result in an immediate cessation of the burn pit and that this concordance analysis is quite crude.

For all three of the burn pit exposure questions, the number of “Yes” responses to these exposures for deployments that occurred during and after incinerator installation decreased relative to deployments that occurred before the incinerators were installed. The affirmative response to these three burn pit-specific exposure questions was higher than what would be expected for burn pits no longer in operation, which goes to the point above that the implementation of the incinerators did not result in an immediate cessation of the burn pit. As described in Chapter 1, the information regarding the locations and dates of operation—let alone contents—of burn pits is incomplete and there is no gold standard with which to compare self-report of these exposures.

In the second method used to examine the concordance of self-reported burn pit exposure, deployment segments were used to compare three locations (Joint Base Balad, Camp Taji, and Contingency Operating Base Speicher in Iraq) with known burn pits versus two locations (Camp Arifjan and Camp Buehring, Kuwait) without known burn pits. These locations with and without known burn pits have been used in other epidemiologic studies of health effects of exposure to burn pit emissions which have been summarized in previous National Academies’ reports (Abraham et al., 2014; AFHSC et al., 2010, Smith et al., 2012). A difference between those studies and this analysis is that those analyses limited the location cohorts to those who were deployed within a 3-mile radius of the burn pit, based on DMDC deployment information, whereas the committee was limited to working with the base name alone because information on longitude and latitude for each deployment was not provided. Deployment dates were not restricted for this analysis and generally started in 2003 for these bases, although Camp Arifjan had a small number of deployments in 2002.

Table 2-18 shows the responses to the three deployment exposure questions by location; shaded locations denote where burn pits were known to be operating. While the proportion of deployments with “Yes” responses to “Were you near a burn pit on this deployment?” (1.2.D) is much lower for the two locations (camps Arifjan and Buehring) without burn pits, more than one-third of deployments to these bases still reported burn pit exposure. The proportion responding “Don’t know” to this question for these two bases without burn pits is nearly the same as those reporting “Yes.” The proportion of “Don’t know” responses for locations without documented burn pits is nearly four times higher than for bases with documented burn pits. Moreover, for approximately one-third of deployments to locations without documented burn pits, the respondent indicated having had duties that involved the burn pit. Given that this is a crude analysis for determining concordance of exposure to burn pits, one explanation for the discrepancy may be that there were situations in which a participant responded that his or her primary base location was one that did not have burn pits but actually was at a base with burn pits for at least part of the deployment; because only one base name can be provided, the participant was indeed exposed to burn pits but not at the location where he or she spent the majority of that deployment.

The second concordance method confirms the findings of the first concordance method regarding the difficulty with obtaining quality exposure data on the basis of the questionnaire responses. However, in the absence of a gold-standard exposure-assessment tool, it is difficult to conclude with any reasonable certainty the direction and magnitude of exposure measurement error in the self-reported exposure information.

This chapter described the committee’s approach to responding to its Statement of Task, gathering information to address its charge, considering exposure registries in general, and, finally, offering a descriptive review of the AH&OBP Registry data. Information on registry participants, exposures, and the health conditions will be used in chapters 5–9 to inform the committee’s assessment of the ability of the AH&OBP Registry to perform each of its five stated purposes. Chapter 3: Development and Operations, describes the development and implementation of the registry, including eligibility, the process for participating, and changes that have been made to it since it launched in June 2014. Additional factors of representativeness, participation, and nonresponse are also examined.

REFERENCES

This page intentionally left blank.