Epidemiology in Military and Veteran Populations: Proceedings of the Second Biennial Conference, March 7, 1990 (1991)

Han K. Kang^*

The Department of Veterans Affairs (VA) maintains three large automated databases that provide the opportunity for studying long-term health effects of military service in Vietnam. These databases were created for administrative and management purposes, not for research purposes. Nevertheless, these resources can be used either as sampling frames for studies of diseases or as tools for health surveillance and vital status follow-up activities.

VA maintains an automated system to identify veterans and their dependents and to keep track of where their claims folders are located. The system includes records for veterans and dependents of veterans who have or are receiving compensation, pension, education, and other VA benefits. This computerized database, created in 1972, replaced the manual Veterans Administration Master Index. There are approximately 40 million records in the database. The BIRLS is an excellent source of vital status information on veterans. VA pays a lump sum death benefit to eligible survivors of deceased veterans. At the present time up to $300 is paid toward a veteran's burial expenses, and an amount not exceeding $150 is paid for a plot or interment allowance if the veteran is not buried in a national cemetery. Until October 1981, survivors of all Vietnam era veterans who were discharged under conditions other than dishonorable were eligible to receive burial benefits. The BIRLS file offers several advantages in ascertaining veteran deaths:

1. It is considered to be relatively complete in its roster of veteran deaths.

   Beebe and Simons (1) showed that up to 98% of WWII veterans whose deaths in 1962 were independently ascertained were known to VA. A more recent study by the National Academy of Sciences (NAS) Medical Follow-up Agency (2) indicated that 88% of Vietnam era veteran deaths that occurred in 1980 were recorded on the VA BIRLS files. The ascertainment rate was somewhat higher (93.4%) for those who served two or more years on active duty.

2. It has a built-in linkage, through the VA claims folder, to the veteran 's death certificate.

   The BIRLS file, as its name suggests, is a records locator system. The location of a veteran's claim folder is indicated in the BIRLS, and the claim folder holds some kind of notification of death, usually a death certificate.

3. The BIRLS file contains data that can facilitate linkage to a veteran 's military records. It contains all or some of the following information: name, Social Security number (SSN), military service number, date of birth, dates of military service, and branch of service.

We have used BIRLS for many of our mortality follow-up studies of Vietnam veterans:

1. Marine Corps Vietnam Veterans Mortality Study

2. Women Vietnam Veterans Mortality Study (3)

3. Army Chemical Corps Vietnam Veterans Mortality Study (4).

We have also used BIRLS to develop a roster of Vietnam era veteran deaths (5-7). The Vietnam Veterans Mortality Study published in 1988 (5) is entirely based on BIRLS information, which was used to select 90,000 potential study subjects. Over 85% of the death certificates for these veterans were obtained from VA claims folders. A state normally charges a fee of $2-$10 per death certificate. It would have cost VA approximately $300,000 to purchase these death certificates, had they not been in VA claim folders.

We are concerned with the potential impact of Public Law 97-35, which became effective on October 1, 1981, on the reporting of veterans ' deaths to VA. The new eligibility requirement of the law may have caused underreporting of veterans' deaths to VA.

In a preliminary attempt to assess the impact of the law, we have compared the total numbers of each of the three types of death benefits awarded for the three fiscal years (FY's) prior to the eligibility changes

(FY's 1979-81) and for the three fiscal years following the changes (FY 's 1982-84). The estimated number of veterans alive in each of the fiscal year periods ranged from 28 to 30 million. The total number of basic burial benefit claims has declined substantially since the passage of the law and has remained depressed despite the growing number of aging veterans. However, the number of claims for other death benefits such as plot interment allowance and burial flags has not changed substantially during the entire 6-year period. Because the basic burial benefit claim is one of several ways by which a veteran's fact of death is reported to VA and because other death benefit programs do not appear to be affected by the law, the overall impact on the reporting of veteran deaths may not be as severe as one may assume.

Nonetheless, we plan to reassess the completeness of death reporting to the BIRLS file for two post-1981 years and to evaluate the demographic and military characteristics of deceased veterans included and not included in the BIRLS file. This study will be a replication of an earlier NAS/VA study of deaths occurring in 1980 among Vietnam era males (2). As before, eight selected states will be approached for magnetic tape copies of their 1982 and 1987 death records for males born during the years 1936 through 1955. Their veteran status will be determined by matching against the registry of the National Personnel Record Center. A total of 4,000 men who are identified as veterans will be matched to BIRLS by “batch processing.” Veterans not matched during this trial will be researched through individual BIRLS inquiries. The data collected will be analyzed to determine whether death reporting rates differ according to such factors as cause of death, Vietnam service, race, rank, and discharge status.

Although VA maintains the BIRLS file for administrative and management purposes, it can serve as an important national source of vital status information for adult males. After all, the 1980 census showed that one in three males above 16 years old was a veteran (8). A detailed characterization of deaths reported and not reported to the BIRLS will be useful to the epidemiologist who plans to tap this national resource.

The PTF is a computerized hospital discharge abstract system of inpatient records, including patients' demographic data, surgical and procedural transactions, and patient movements and diagnosis. One PTF record is prepared for each discharged VA in-patient by the medical records librarian at the discharging station. The VA Data Processing Center in Austin updates the PTF each week. The PTF record contains information on such variables as name, SSN, date of birth (DOB), place of residence, marital status, period of military service, radiation exposure,

self-reported Agent Orange exposure, prisoner of war status, discharge diagnosis, length of stay, surgical and other procedures, and compensation and pension status. Nearly 1.5 million veterans are treated as in-patients in the VA hospitals each year.

The PTF provides an excellent sampling frame for a study of disease. We have conducted three case-control studies based on the veteran patients in the PTF: soft tissue sarcomas (9), non-Hodgkin's lymphoma (10), and Hodgkin's disease (11). To those who want to consider using the PTF for any future study, we would like to share our experience with the data file. Beginning in July 1982, the PTF added a Vietnam Service Indicator or Agent Orange Exposure Indicator to the record. We thought it would be an expedient and practical way to evaluate Vietnam veterans' health problems by comparing discharge diagnoses of Vietnam veterans to those of non-Vietnam veterans. If there is no Vietnam effect, the distribution of various categories of health conditions recorded in the PTF should be similar after adjustment for certain demographic variables between these two patient groups. Since there is a built-in Vietnam service indicator for each patient and his discharge diagnosis is already coded using ICD 9, it would have been a relatively simple process. Before we started this project, we thought we would check the accuracy of the indicator. We randomly selected 1,000 Vietnam era veterans from the PTF for FY 1983. We were able to locate and retrieve military service records for 914 of the veterans from the National Personnel Records Center in St. Louis. We had their military service records abstracted by a contractor and compared Vietnam service against the PTF indicator for such service. The results were more than disappointing. Of 370 veterans who were identified as Vietnam veterans by the PTF, only 196 veterans were determined to have served in Vietnam by the military record. Similarly, of those 544 veterans who did not serve in Vietnam according to the PTF, 159 patients were found to have served in Vietnam. Therefore, the percentage of false positives and false negatives were 31% and 45%, respectively.

Coding for discharge diagnosis was not any better for certain types of cancer. A few years ago we conducted a case-control study of soft tissue sarcoma using veteran patients in the PTF (9). We identified a total of 418 patients with an ICD 171 diagnosis, i.e., malignant neoplasm of connective and other soft tissue, by a complete search of the PTF for Vietnam era veterans who were hospitalized between 1969 and 1983. A pathology report for each ICD 171 case was requested from each treating VA hospital. A review of the 394 pathology reports that we received by an expert pathologist, who during the review had no knowledge of Vietnam service status of any of the patients, yielded the following results. He excluded 151 cases from the study because of miscoding or misclassi-

fication. Nine cases were considered doubtful, leaving only 234 cases that met the classification system for soft tissue sarcoma. That was almost a 40% loss.

We had a similar experience with malignant lymphoma cases among Vietnam era veterans in the PTF (10). In preparation for a case-control study of lymphoma among Vietnam era veterans, we identified the recorded lymphoma cases (ICD 200-202), and as before retrieved the pathology reports and had them reviewed by an expert pathologist. Only one-half (or 450) of the cases coded ICD 200 and 202 were determined to be consistent with the diagnosis of non-Hodgkin's lymphoma. We have not reviewed the PTF for other coded diagnosis. But it was clear to us that you can't rely on ICD coding recorded in the PTF without further verification at least for a study of cancer. The PTF can provide an excellent sampling frame for study of diseases among veterans. But one has to be well aware of the possibility of substantial coding errors.

In order to further evaluate the PTF, we arranged for a pathological review of tissue slides from the cases described as soft tissue sarcomas (ICD 171) in the pathology reports. We obtained the services of the Armed Forces Institute of Pathology (AFIP). The AFIP was provided with sets of slides that VA Medical Centers forwarded to us along with information on the sex and age of the patient and the anatomical site from which the specimen was taken. The AFIP knew neither the Vietnam service status of the patients nor the contents of their VA clinical records and pathological diagnosis. Among the 181 cases the AFIP reviewed, the AFIP disagreed with the VA diagnosis in 10 cases (5.5%).

I believe, therefore, that the major problem with the PTF is the miscoding by medical clerks rather than misdiagnosis by VA clinicians.

The Agent Orange Registry was initiated by the Department of Veterans Affairs in mid-1978 in response to concerns expressed by Vietnam veterans who were increasingly worried that they may have been exposed to Agent Orange and other herbicides that might be causing a variety of ill effects. It provides the veteran an opportunity to receive a complete health evaluation and answers to questions concerning the current state of knowledge regarding the relationship between herbicide exposure and subsequent health problems. The computerized Agent Orange Registry data include veteran's name, address, some information on military service, and findings at the time of his physical examination.

Because of the self-selected nature of the registry participants, this group of veterans cannot be viewed as being representational of Vietnam

veterans as a whole. There are approximately 200,000 participants in the Registry; that is almost one-tenth of all Vietnam veterans.

Because of the possibility of coding and keying errors in the Agent Orange Registry, we have assessed the overall accuracy of the data in the registry by randomly reexamining original microfilmed medical records of Agent Orange examinations and comparing this information to the data on the code sheet. We then compared the data on the code sheet with the data contained in the computerized registry.

Coding and keying errors do exist in the Registry, especially in the diagnosis and complaint fields. Upon reviewing the records for 200 Agent Orange examinations, it was found that 11% of the diagnoses were miscoded and 2.7% of the diagnoses were miskeyed.

We are conducting two case-control studies using information in the Registry (12,13). Because of the limitations that I described earlier, we are verifying military service data by reviewing the study subjects ' military records and also the diagnosis by reviewing their Agent Orange Registry abstracts. As I stated earlier, the Registry contains the demographic and medical data for approximately 200,000 Vietnam veterans. With appropriate caution and appreciation of its limitations, the Registry can serve as another useful resource for future health studies of Vietnam veterans.

1. Beebe GW Simon AH Ascertainment of mortality in the U.S. veteran population Am J Epidemiol 1969 ; 89 : 636-643

2. National Academy of Sciences Commission on the Life Sciences Medical Follow-up Agency Ascertainment of Mortality in the U.S. Vietnam veteran population A report to contract V101(93)P-937 Washington, D.C. : 1985

3. Thomas TL Kang, HK Mortality among Women Vietnam-era veterans: a preliminary analysis Presented at the 23rd Annual Meeting of the Society for Epidemiologic Research, June 12-15, 1990, Snowbird, Utah

4. Thomas TL Kang, HK Mortality and Morbidity Among Army Chemical Corps Vietnam veterans: A preliminary report AM J Ind Med 1990 ; 18 : in print

5. Breslin P Kang HK Lee Y et al. Proportionate mortality study of US Army and US Marine Corps veterans of the Vietnam war J Occup Med 1988 ; 30 : 412-19

6. Bullman TA Kang HK Watanabe KK Proportionate Mortality among US Army Vietnam veterans who served in military region I Am J Epidemiol 1990 ; 132 : in print

7. Watanabe KK Kang HK Thomas TL Mortality among Vietnam veterans: with methodological considerations J Occup Med 1991 ; in print

8. Veterans Administration Veterans in the United States A statistical portrait from the 1980 census Office of Information Management and Statistics Washington, D.C ; 1984

9. Kang HK Weatherbee L Breslin P et al. Soft tissue sarcomas and military service in Vietnam: A case comparison group analysis of hospital patients J Occup Med 1986 ; 28 : 1215-18

10. Dalager NA Kang HK Burt VL et al. Non-Hodgkin's lymphoma among Vietnam veterans J Occup Med 1991 ; in print

11. Dalager NA Kang HK Burt VL et al. Malignant lymphoma among Vietnam-era veterans Presented at the 9th International Symposium on chlorinated dioxins and related compounds September 17-22, 1989 Toronto, Canada

12. Thomas TL Kang HK Health status of a self-selected group of 104,000 Vietnam veterans Presented at the 22nd Annual Meeting of the Society for Epidemiologic Research June 14-16, 1989 Birmingham, Alabama

13. Bullman T Kang HK A case control analysis of Vietnam veterans from the Agent Orange Registry for posttraumatic stress disorder Presented at the 23rd Annual Meeting of the Society for Epidemiologic Research June 12-15, 1990 Utah