|
|
| |||||||||||||||||||||||||||||||||
Background Since the Persian Gulf War ended in 1991, many veterans of that conflict have reported diverse, unexplained symptoms. To evaluate the health of Gulf War veterans, we studied their postwar hospitalization experience and compared it with that of other military personnel serving at the same time who did not go to the Persian Gulf.
Methods Using a retrospective cohort approach and data from Department of Defense hospitals, we studied hospitalizations of 547,076 veterans of the Gulf War who were serving in the Army, Navy, Marine Corps, and Air Force and 618,335 other veterans from the same era who did not serve in the Persian Gulf. Using multivariate logistic-regression models, we analyzed risk factors for hospitalization both overall and in 14 broad diagnostic categories during three periods from August 1991 through September 1993 (a total of 45 comparisons).
Results After the war, the overall odds ratio for hospitalization of the Gulf War veterans was not higher than that of the other veterans, even after adjustment for selection effects related to deployment. In 16 of the 42 comparisons involving specific diagnoses, the risk of hospitalization among Gulf War veterans differed significantly from that among other veterans. Among these 16 comparisons, Gulf War veterans were at higher risk in 5: neoplasms (largely benign) during 1991, diseases of the genitourinary system during 1991, diseases of the blood and blood-forming organs (mostly forms of anemia) during 1992, and mental disorders during both 1992 and 1993. The differences were not consistent over time and could be accounted for by deferred care, postwar pregnancies, and postwar stress.
Conclusions During the two years after the Persian Gulf War, there was no excess of unexplained hospitalization among Americans who remained on active duty after serving in that conflict.
Using the computerized hospitalization records of the Department of Defense, we conducted a study to determine whether Gulf War veterans were at increased risk for hospitalization after the war, as compared with veterans from the same era who did not go to the Persian Gulf.
Methods
Sources of Data
The data on service in the Gulf War, demographic variables, military separations, and hospitalizations that were used in this study were obtained from the Defense Manpower Data Center, Monterey Bay, California. Data on Gulf War service were compiled from Army, Navy (including Marine Corps), and Air Force records of unit-deployment locations and pay for exposure to hostile fire. Demographic data were obtained from routine data files on U.S. military personnel.
Study Population
On January 21, 1991, President George Bush issued an executive order28 designating much of the land mass of the Arabian Peninsula and the seas surrounding it as a combat zone. Military personnel deployed to serve in the Gulf War for one or more days between August 8, 1990, and July 31, 1991, were considered Gulf War veterans. A total of 579,931 of the 696,562 U.S. Gulf War veterans (83.3 percent) served on regular, active duty in the Army, Navy (including the Marine Corps), or Air Force. Since such personnel are rarely hospitalized in facilities other than those operated by the Department of Defense and since data on their hospitalizations were readily available from that department, we examined the hospitalization experience of these veterans. The hospitalizations of U.S. reserve and National Guard forces could not be studied with our data.
Using a random-number procedure, the Defense Manpower Data Center selected a comparison group of 700,000 military personnel on active duty who were listed on the Department of Defense rosters as of September 30, 1990, but did not go to the Persian Gulf War theater before July 31, 1991. This group included approximately half of all such personnel. The number of subjects selected from each service branch was proportional to the number sent to the Gulf War. The date of September 30, 1990, was selected from the Defense Manpower Data Center's quarterly files showing overall military strength because it was the first such file compiled after the start of the war.
The Defense Manpower Data Center provided personal identifiers for the Gulf War veterans and the other veterans in addition to data on their sex, age, race or ethnic group, marital status, branch of service, occupation, rank, pay grade, and total number of months of active military service. The sample was divided into quartiles according to age. Rank was categorized in three levels: enlisted, warrant officer, and officer. Pay grade and length of service were used to determine each veteran's base salary (unadjusted for bonuses and housing allowances), which was categorized in five levels. Occupation (classified by 226 codes) was categorized in 10 major groups as defined in the Occupational Conversion Manual of the Department of Defense.29
Periods of Study
Four periods were selected for study: October 1, 1988, through July 31, 1990 (the "period before the war"); August 1, 1991, to December 31, 1991 ("1991"); January 1 to December 31, 1992 ("1992"); and January 1 to September 30, 1993 ("1993"). The "period before the war" began with the first available triservice (Army, Navy, and Air Force) hospitalization data. Hospitalizations during the Gulf War period itself (from August 1990 through July 1991) were not specifically evaluated, because access to care for Gulf War veterans differed markedly from that for other veterans. In addition, most illnesses said to be related to the Gulf War were reported after July 1991. Because of the high rate of attrition from military service (43.6 percent by October 1993) and the subsequent loss of follow-up, hospitalizations occurring after September 30, 1993, were not studied.
Size of the Study Group
Among the 1,279,931 subjects initially identified, we studied only those for whom data were complete and who were on active duty on the first day of each postwar period. The resulting subgroups contained 1,165,411 subjects for 1991; 1,075,430 for 1992; and 839,389 for 1993.
Hospitalization Data
The hospitalization records included up to eight discharge diagnoses, which were coded by members of the hospital staff using the International Classification of Diseases, 9th Revision, Clinical Modification (ICD-9-CM),30 which groups these codes into 17 broad categories. These data were abstracted by using the Social Security numbers of the study subjects. The risks of any hospitalization and of hospitalization for a diagnosis in each of 14 broad ICD-9-CM categories (not including diagnoses involving the reproductive system) are examined in this report. In each period, a subject was counted only once with regard to each diagnostic category (or hospitalization for any cause), regardless of how often he or she was hospitalized for conditions in that category. Although the diagnostic fields were scanned in numerical order for the ICD-9-CM codes of interest, no consideration was given to ascertaining the primary cause of hospitalization.
Data Management and Statistical Analysis
Potential demographic risk factors for hospitalization were evaluated by univariate and multivariate tests,31 with SAS software (version 6.09, SAS Institute, Cary, N.C.). Hospitalization rates were calculated per 10,000 person-years for the four periods studied and were standardized by the direct method, on the basis of the age and sex distribution of the combined groups of Gulf War veterans and other veterans. Ninety-five percent confidence intervals for the rate ratios were calculated with the method described by Rothman.32
A multivariate analysis of risk factors was conducted with unconditional, multivariate logistic regression with coefficients estimated by the maximum-likelihood method. To reduce the potential bias due to attrition, only members of the study populations who remained on active duty for at least half the period studied were included in the multivariate models.
Results
Demographic Variables
Information on the demographic characteristics of the study subjects has been placed on file with the National Auxiliary Publications Service (*). As compared with the other veterans serving in the same era, the Gulf War veterans were disproportionately male (94 percent vs. 88 percent); they were also younger than the other veterans (median age, 25 vs. 27 years). The two groups also differed, but to a lesser extent, with respect to race or ethnic group, marital status, branch of service, rank, salary, and occupation. Accordingly, the hospitalization rates and rate ratios were adjusted for age and sex, and the multiple logistic-regression models were adjusted for all observed demographic differences between the groups.
Potential Bias
We studied the data from the two groups on postwar attrition from active duty and disqualification from service for medical reasons, to identify potential sources of bias. Although the overall age- and sex-adjusted postwar attrition of Gulf War veterans was higher than that of other veterans, the difference was not due to disqualification for medical reasons or mortality (Table 1).
|
Univariate analyses were performed of hospitalization for any cause and the 14 ICD-9-CM categories of hospitalization for each of the three postwar periods (a total of 45 comparisons). Almost every demographic risk factor had a potentially important univariate association (P<0.25)31 with each outcome.
Multivariate Analysis of Hospitalization for Any Cause
An exploratory stepwise logistic-regression analysis (data not shown) confirmed that almost all the covariates were independently associated with each outcome. Hence, each of the following covariates was included in each logistic-regression model: Gulf War service, sex, age group, race or ethnic group, marital status, branch of service, occupation, rank, salary, and length of military service.
Logistic-regression analysis showed that Gulf War veterans were at slightly lower risk of hospitalization for any cause than other veterans two years before the war but that the risk did not differ after the war (data not shown). Since no triservice data were available from the period before October 1988, data from Navy hospitals on hospitalizations in the Navy and Marine Corps (Naval Health Research Center, San Diego, Calif.) were studied as a surrogate measure. Approximately 109,702 of the 482,034 persons in the combined groups of Gulf War veterans and other veterans in the Navy and Marine Corps in 1991 (23 percent) were on active duty in 1980.
To examine more closely the reduced risk of prewar hospitalization for any cause among Gulf War veterans, we recalculated the adjusted odds ratios (i.e., the odds among Gulf War veterans divided by that among other veterans) for three-month periods (from November 1988 through July 1990 and from August 1991 through July 1993) and plotted them over time (Figure 1). In a similar fashion, we conducted a multivariate analysis of the Navy and Marine Corps populations from 1980 through July 1990 and from August 1991 through July 1993. There was no difference between Gulf War veterans and other veterans in the risk of hospitalization from 1980 through 1988. However, from 1988 through the time just before the start of the war in August 1990, Gulf War veterans were at lower risk of hospitalization. Shortly after the war ended, the differences in this risk disappeared.
|
Multivariate Analysis of Hospitalizations According to ICD-9-CM Category
We studied cohort data both before and after adjustment for the prewar selection effect to determine the risk of hospitalization due to diagnoses in the 14 ICD-9-CM categories in each of the three time periods — that is, in a total of 42 comparisons. Adjustment for the prewar selection effect did not change the odds ratios associated with Gulf War service, but all models reported here include the prewar-hospitalization covariate. The odds of hospitalization differed between the two cohorts in 16 of the 42 comparisons of diagnostic categories (Figure 2), with Gulf War veterans at greater risk in five models: neoplasms (in 1991), diseases of the genitourinary system (in 1991), diseases of the blood and blood-forming organs (in 1992), and mental disorders (in both 1992 and 1993).
|
The 10 most frequent diagnoses in each of these five models were examined further with rate calculations adjusted for age and sex. These 10 diagnoses accounted for 68 to 100 percent of the diagnoses in their respective categories.
The 10 most frequent discharge diagnoses involving neoplasms in the last five months of 1991 were mostly for benign conditions (Table 2), and most of the 95 percent confidence intervals of the standardized rate ratios showed no significant difference in rates between Gulf War veterans and other veterans. The one exception was the standardized rate ratio and 95 percent confidence interval for testicular cancer, which was slightly elevated over this five-month period. However, this event was rare (29 admissions), and Gulf War veterans were not hospitalized significantly more often with this diagnosis than other veterans in 1992 (standardized rate ratio, 1.39; 95 percent confidence interval, 0.91 to 2.11) or 1993 (standardized rate ratio, 0.89; 95 percent confidence interval, 0.54 to 1.44).
|
Hospitalizations in 1992 for diseases of the blood and blood-forming organs were usually for anemia (*). Because of a postwar baby boom among Gulf War veterans, 860 pregnancy-related hospitalizations were removed from consideration, and the multivariate analysis was performed again. The resulting values showed no increase in risk among Gulf War veterans (the 95 percent confidence interval for the odds ratio included 1), suggesting that the increase in risk originally found was primarily due to anemias associated with pregnancy.
Finally, we examined the 10 most frequent diagnoses of mental disorders in 1992 and 1993. Gulf War veterans were hospitalized significantly more often than other veterans for conditions related to alcohol and drug use and for adjustment reactions (Table 3).
|
The hospitalization data of the Department of Defense capture the experience of large and demographically well-delineated populations of U.S. adults. These groups are diverse in geographical origin and experience, their socioeconomic status is well known, and because of military requirements, virtually all their hospitalizations are included in the data base, so that they constitute an excellent resource for epidemiologic studies.33,34,35,36,37 We compared the hospitalization experiences of two large cohorts, totaling 1,165,411 persons in 1991, who remained on active duty after the Persian Gulf War.
These comparisons yielded few surprises. The increased overall risk of hospitalization among women and health care workers (*) is consistent with the findings of a previous study of hospitalizations in the Navy.38 The increased rates of hospitalization after the war (Table 3) for conditions related to drug and alcohol use and adjustment reactions have been reported in other groups of combat veterans.39,40 Finally, the comparisons of separation rates (Table 1) are consistent with the results of other recent mortality studies4,5,41 that have not shown Gulf War veterans to have a higher overall or disease-related risk of death than other veterans.
The prewar selection effect we found (Figure 1) is understandable, in that the services permit recently hospitalized personnel to remain attached to their operational units while they convalesce, but the limited-duty status of these personnel makes them ineligible for deployment. The data suggest that this selection effect is transient and that the reduction in the risk of hospitalization seen before the war disappeared shortly after the war (Figure 1). After we created a prewar-hospitalization covariate to control for this selection effect in the multivariate models, the odds ratios in the two cohorts remained essentially the same.
As would be expected with numerous comparisons, we found some differences in the risks associated with specific diagnostic categories and rates of specific diagnoses. These differences were not consistent over time and do not suggest an emerging illness associated with Gulf War service. Many of the observed differences between cohorts with regard to rates of diagnoses suggest that medical care for some conditions was deferred until after the war. Certainly, this is true with regard to the diagnosis of redundant prepuce and phimosis (ICD-9-CM code 605), which usually means that elective circumcision was performed. Deferred diagnostic evaluation or surgery probably also accounts for the slight increases in the rates of various benign neoplasms (Table 2) and of hospitalizations for inflammatory disease or infertility in women (Table 3) that occurred immediately after the war. Because no known associations between an exposure and the appearance of a neoplasm have such a short latency period and because women are hospitalized for infertility only after months, if not years, of outpatient medical care, it is difficult to implicate Gulf War service in causing these conditions. It is more likely that Gulf War veterans waited until they were home before undergoing elective hospitalization.
These analyses have a number of strengths. The large groups studied, for which there were detailed demographic data to mitigate confounding, offer unusually high statistical power to detect differences in the risk of hospitalization. Because discharge data are recorded thoroughly by hospitals associated with the Department of Defense and because personnel on active duty have little opportunity to be hospitalized outside these hospitals, nearly all the hospitalizations of Gulf War veterans are represented in the data we used. In addition, all veterans are screened medically, with interviews, physical examinations, and laboratory studies, before their separation from military service. Free medical care by the Department of Defense and potentially lifelong medical-disability benefits are strong incentives for veterans to report their maladies thoroughly. Therefore, it is unlikely that any hospitalization occurring before the time of separation from the military would have gone unnoticed.
These analyses also have limitations. Our broad categorization of a person's status as a Gulf War veteran or another type of veteran may mask the detection of illness that has a temporally or geographically specific exposure as the cause. All the logistic-regression models were adjusted for the prewar selection effect, but if this adjustment was inadequate, slight increases in the risk of hospitalization among Gulf War veterans might not be detected. The study data reflect only the hospitalization experience of persons who remained on active duty through September 1993; illnesses associated with exposure during the Gulf War whose latency periods are longer than two years would have been missed. Similarly, illness occurring after a veteran's separation from the military would have been missed. In addition, illnesses not requiring hospitalization are not represented in our data. Finally, the increased attention in the media and awareness by the public concerning morbidity among Gulf War veterans may have influenced rates of postwar hospitalization among such veterans.23 This bias should have increased the likelihood of such hospitalizations, however.
In conclusion, we constructed multivariate logistic-regression models for hospitalization both overall and for conditions assigned to any of 14 broad diagnostic categories in each of three postwar periods. The risk associated with 16 of these 45 comparisons differed between Gulf War veterans and other veterans. In 5 of these 16 cases, the risk of hospitalization was higher among Gulf War veterans, but the increases were inconsistent over time and were probably due to deferred medical care, a postwar baby boom, chance, or mental conditions known to be associated with war. The data suggest that veterans of the Gulf War who remained on active duty were not at increased risk for unexplained hospitalization during the 25 months after the war.
Supported by a work unit (63738DP4464.001-6423) of the Naval Medical Research and Development Command, Bethesda, Md., Department of the Navy, and by the Department of Defense, Health Affairs.
We are indebted to the following for their most helpful assistance and recommendations in conducting these analyses: Gary D. Gackstetter, D.V.M., Ph.D., of the Department of Defense, Health Affairs, Washington, D.C.; Larry M. Dean, Ph.D., D. Stephen Nice, Ph.D., Frank C. Garland, Ph.D., Stephanie F. Brodine, M.D., Christopher G. Blood, M.A., and Debbie C. Brummitt of the Naval Health Research Center, San Diego, Calif.; Rebecca L. Calderon, Ph.D., M.P.H., of the Environmental Protection Agency, Chapel Hill, N.C.; Robert F. DeFraites, M.D., M.P.H., and David N. Cowan, Ph.D., M.P.H., of the Walter Reed Army Institute of Research, Washington, D.C.; the late Samuel M. Wishik, M.D., M.P.H., of the University of California, San Diego; Mr. Mike Dove and Ms. Sue Butler of the Department of Defense Manpower Data Center, Monterey Bay, Calif.; and Terry L. Thomas, Ph.D., of the Uniformed Services University of Health Sciences, Bethesda, Md.
Source Information
From the Clinical Epidemiology Division, Naval Health Research Center, San Diego, Calif. (G.C.G., B.D.C., C.M.A., J.D.K.); the Environmental Epidemiology Service, Department of Veterans Affairs, Washington, D.C. (H.K.K.); the Navy Environmental Health Center, Norfolk, Va. (S.W.B.); the Naval Medical Research Unit No. 2, Jakarta, Indonesia (F.S.W.); and the Department of Family and Preventive Medicine, School of Medicine, University of California San Diego, La Jolla, Calif. (E.B-C.). The views expressed in this article are those of the authors and do not reflect the official policy or position of the Department of the Navy, the Department of Defense, the Department of Veterans Affairs, or the U.S. government.Presented in part at the 123rd annual meeting of the American Public Health Association, San Diego, Calif., October 29–November 2, 1995.
Address reprint requests to Dr. Gray at the Emerging Illness Research Team, P.O. Box 85122, Naval Health Research Center, San Diego, CA 92186-5122.
References
| |||||||||||||||||||||||||||||||||
This article has been cited by other articles:
HOME | SUBSCRIBE | SEARCH | CURRENT ISSUE | PAST ISSUES | COLLECTIONS | PRIVACY | HELP | beta.nejm.org Comments and questions? Please contact us. The New England Journal of Medicine is owned, published, and copyrighted © 2009 Massachusetts Medical Society. All rights reserved. |
Previous
