Magee, C.A., Kang, H.K., Mahan, C.M., and Murphy,
F.
Veterans Health Administration, Environmental Epidemiology
Service Washington, D.C.
In a national health survey of Gulf War-era veterans designed
to compare prevalences of selected health outcomes for stratified random
sample of all Gulf veterans and of non-deployed contemporary military
controls, 20,931 respondents reported presence during their 'past 12 months'
of medical conditions, symptomatology, psychologic status, hospitalizations,
doctor/clinic visits, and pregnancy outcomes. To assess false-positive
reporting and differential reporting bias between two study groups, a
random sample of 2,000 respondents from each group was selected for medical
record review of any office visits or hospitalizations (up to 4 of each).
Half of the selected veterans had one or more encounters (mean 1.1 apiece);
91% of these veterans gave permission for records release. Over 80% of
requested records had already been obtained by 3/98. From each record,
presenting reasons/diagnoses were abstracted. Veterans' lay terminology
and often incomplete 'reasons' (e.g., only organ/ "prostate"
or only "physical therapy"), necessitates employing a subjective
scale to rank relative concordance of veteran's reason with physician's
symptoms/diagnoses (none/0, partial/1, or good/2 agreement). For veterans
having more than one acquired medical record, a weighted score will be
calculated. Chi-square tests on categorical rank, Kappa tests (using cut-off
value for-/+ agreement), as well as means & t-tests on veteran scores,
can evaluate differential bias.
With 70% of veterans reporting ever achieving pregnancy,
the 4,000 records-sample was also used to solicit records/vital certificates
for: all birth defects & all non-viable outcomes (miscarriages, stillbirths,
'other' (e.g. hydatidiform mole), & 1st-year infant deaths). Since
analysis of liveborn outcomes will focus on the span of 1/86 - '96 (5-yr
pre vs. 5-yr post -war experience), validation encompassed all incident
birth defects, as well as all 'normal' births (birth certificates Parts
I, II) to detect false negative reporting.
This extensive medical records review of randomly selected
veteran respondents will help in assessing validity of self-reported health
data, including clinic visits and reasons, overnight hospitalizations
and diagnoses, pregnancy outcomes, and birth defect and 12-month infant
mortality status among liveborn offspring. Assessment will allow detection
of both false positive and false negative reporting for given outcomes.
Keywords: Health Outcome Validation; Reproductive Outcome
Validation; Differential Reporting Bias |