Physical Health Conditions Among a Population-Based Cohort of Vietnam-Era Women Veterans: Agreement Between Self-Report and Medical Records

Abstract

Background:

Little is known about medical morbidity among women Vietnam-era veterans, or the long-term physical health problems associated with their service. This study assessed agreement comparing data on physical health conditions from self-report and medical records from a population-based cohort of women Vietnam-era Veterans from the Health of Vietnam Era Women's Study (HealthViEWS).

Materials and Methods:

Women Vietnam-era veterans (n = 4219) self-completed a survey and interview on common medical conditions. A subsample (n = 900) were contacted to provide permission to obtain medical records from as many as three of their providers. Medical record reviews were conducted using a standardized checklist. Agreement and kappa (agreement beyond chance) were calculated for physical health condition groups.

Results:

Of the 900, 449 had medical records returned, and of those, 412 had complete surveys/interviews. The most commonly reported conditions based on self-report or medical record review included hypertension, hyperlipidemia, or arthritis. Kappa scores between self-reported conditions and medical record documentation were 0.75–0.91 for hypertension, diabetes, most cancers, and neurological conditions, but lower (k = 0.29–0.55) for cardiovascular diseases, musculoskeletal, and gastrointestinal conditions. Generally, agreement did not significantly vary by different sociodemographic groups.

Conclusions:

There was relatively high agreement for physical health conditions when self-report was compared with medical record review. As more women are increasingly represented in the military and more veterans in general seek care outside the Veterans Health Administration, accurate measurement of physical health conditions among population-based samples is crucial.

Introduction

Women are increasingly serving in the United States Armed Forces.¹ The role of women in the military has been longstanding, including many who served during the Vietnam War era, although not studied to the same extent as male veterans. As Vietnam era women enter an age range where medical need typically escalates, it is important to be able to accurately characterize their burden of illness, both to inform their care, and to anticipate potential needs of the rapidly expanding population of women veterans in the generation that follows them.² Indeed, it is estimated that by 2040, up to 16% of veterans will be women.¹

However, little is known about the prevalence and course of physical health conditions among women veterans, especially those serving during the Vietnam War era.³ Women serving during the Vietnam War era may have been at increased risk of physical health conditions, including cancers, endocrinological, or neurological disorders, compared with those serving in the United States, potentially from environmental or other exposures.⁴

Previous research involving the health status of women veterans has primarily focused on mental health conditions, particularly post-traumatic stress disorder (PTSD).^5,6 Recent data on physical health conditions have been limited and mainly ascertained from mortality records.⁷ Moreover, physical health diagnoses can be difficult to ascertain among women veterans, as many do not rely on a single provider such as the VA. Population-based studies that include individuals beyond treated patient groups are essential in estimating the prevalence of physical health conditions among women veterans to understand the long-term impact of wartime exposures.

Current studies regarding women's physical health have primarily involved patients from integrated healthcare systems (e.g., Veterans Health Administration [VA] system), because of easy access to comprehensive medical record data. A substantial proportion of women veterans receive care outside VA and other integrated healthcare systems, and hence, are often not included in these studies.^8,9 Moreover, medical diagnoses from providers outside the VA are currently not available through the VA's electronic medical record system. In-person clinical exams are expensive and often unfeasible to implement in population-based samples, and administrative (claims) data from health plans may not be consistently available, and could potentially miss underdiagnosed conditions.^10,11 Without access to consistent medical information, measurement of physical health status can be incomplete. It is crucial to understand the strengths and limitations of different methods for assessing medical conditions, especially methods that can characterize the health of populations that are not sampled from clinical settings.

Ascertainment of self-reported diagnoses from population-based samples is a potentially viable option for conducting population health studies of physical health conditions and their impacts over time. While good agreement was found for self-reported physical health conditions in a previous study of male veterans when compared with clinical records (e.g., kappa scores for diabetes = 0.82; hypertension = 0.76; cancer = 0.70),¹² there have been no current studies assessing agreement between self-report and medical record data in a population-based sample among women veterans.

The goal of this study is to assess the agreement between self-report and medical records (VA and non-VA) for an array of physical health conditions among women veterans who served during the Vietnam era. We also assessed demographic factors associated with agreement between self-report and medical records to determine whether agreement is consistent across subgroups.

Materials and Methods

This study is based on the VA Cooperative Studies Program (CSP) #579, the Health of Vietnam Era Women's Study (HealthViEWS), which was designed to characterize the health status of women Vietnam-era veterans. Further details regarding HealthViEWS are described elsewhere (ClinicalTrials.gov identifier—NCT00958061).⁶ In brief, women veterans who served in Vietnam, near Vietnam (e.g., in military hospitals or bases in Guam, Philippines, Japan, Korea, or Thailand), and in the United States during the Vietnam Era (July 4, 1965 to March 28, 1973) (n = 4219, 48.3% overall survey response rate) were identified mainly based on existing VA rosters and military personnel record data.^13,14 Study respondents completed a mail survey and telephone interview that included self-reported physical health conditions, as well as demographic factors.

A stratified random sample of 900 women (described in further detail in Table 1 footnotes) was selected from the 4219 for medical record review and abstraction. Initial mailings to obtain permission for release of medical records were conducted by Westat, a contractor based in Rockville, Maryland. The medical record review was conducted by VA Ann Arbor Healthcare System study staff. All aspects of the study were approved by the VA Central Institutional Review Board, and informed consent was obtained from all participants.

Table 1.

Characteristics of Full Sample of Respondents and Medical Record Review Subsample

Area of service	Full sample (n = 4219), n (%)	Medical record sample (n = 412), n (%)	Chi-square	p
Vietnam	1956 (46.4)	228 (55.3)	20.0	<0.001
Near Vietnam	657 (15.5)	36 (8.7)
In United States	1606 (38.1)	148 (35.9)
Age ≥65 years^a	2249 (53.3)	227 (55.1)	0.5	0.46
Currently married (yes)	2130 (50.5)	198 (48.1)	0.9	0.32
Race (non-white)	241 (5.7)	26 (6.3)	0.3	0.59
Nurse occupation	2982 (70.8)	311 (75.5)	4.4	0.03
Hazardous alcohol use	207 (4.9)	26 (6.3)	1.7	0.18
Smoked at least 100 cigarettes	2168 (51.4)	239 (58.0)	7.2	0.007
Five or more current self-reported physical health conditions	654 (15.5)	115 (27.9)	48.5	<0.0001
PTSD diagnosis	718 (17.0)	76 (18.5)	0.6	0.44

This table reports the frequency of each demographic factor based on self-report from the survey, comparing the frequency of those in the full survey sample with those from the medical record subsample. The subsample included: (1) all women reporting any of the following neurological conditions: ALS, multiple sclerosis, Parkinson's disease, or brain cancer; (2) all women reporting gynecological or breast cancer, but not reporting any neurological conditions; (3) a random sample of women reporting diabetes, but not reporting any neurological conditions or gynecological/breast cancers; and (4) a random sample of women reporting hypertension, but not reporting neurological conditions, gynecological/breast cancers, or diabetes. These physical health conditions were chosen because they represent gender-specific conditions of particular interest to military and veteran cohorts.

Mean age ± standard deviation (SD), in years = 67.0 ± 5.3, range: 59–88 years.

Hazardous alcohol use was defined by the Alcohol Use Disorders Identification test question on having six or more drinks on a single occasion.

The number of medical health conditions (five or more) was based on a total count of self-reported physical health conditions the woman received treatment for in the past year.

PTSD diagnosis was assessed based on the CIDI telephone interview.

CIDI, Composite International Diagnostic Interview; PTSD, post-traumatic stress disorder.

Study population: medical record cohort

The sample of women (n = 900) were initially selected for the medical record review based on positive responses to the mail survey's physical health condition questionnaire, which included 21 conditions based on the Medicare Health Outcomes Study condition assessment.¹² For each condition listed, women were asked (1) whether they had ever been told by a doctor or other health professional that they had the specific physical health condition, and (2) whether they had received treatment in the past year for each condition.

A subsample of women reporting “yes” to ever being told they had a select group of physical health conditions was included in the medical record review (Table 1).¹⁵ Medical records were requested directly from participants' principal care providers (up to three per participant which could include specialists), as evidence suggests that most women veterans primarily use health services outside the VA.⁸ The women were sent up to three mailings, each of which included an introductory letter, information sheet, three Health Insurance Portability and Accountability Act (HIPAA) release of information forms (for study staff to send to up to three separate providers—primary care and specialists), and a self-addressed prepaid envelope. Respondents agreeing to participate were instructed to return the release of information forms using the prepaid envelope to study investigators, who then mailed the completed release of information forms to the identified providers. Providers were mailed a packet that requested copies of medical records for treatment that had occurred within the past 12 months to be mailed back using a prepaid envelope.

Medical record abstraction

The medical record review was conducted by a trained nurse abstractor, who was blind to the woman's self-reported physical health diagnoses. The medical record abstractor determined the presence or absence of physical health conditions within the past year from the survey date using a standard paper checklist for each condition. The abstractor reviewed key sources of medical information to determine diagnoses from the medical records, including discharge summaries, problem lists, medications prescribed, and outpatient progress notes within the past year. Among women who listed non-VA providers, paper copies of medical records were requested given that access to electronic health record information across multiple healthcare providers or health plans was not feasible in this study. For participating women who listed VA providers, medical records were abstracted using the same standardized paper checklist by electronically viewing the VA's computerized patient record system. The chart components reviewed and the time frame examined were the same for VA and non-VA medical records. To assess inter-rater reliability, a VA physician independently reviewed a randomly selected 5% of the first 100 medical records examined and was kept blind as to the abstractor's findings. The physician independently reviewed the presence or absence of each condition listed in the medical record abstraction. The physician's responses were compared with the medical record abstractor's responses for this subsample of medical records, and the mean kappa between the abstractor and physician assessment across all conditions was 0.83, considered strong agreement.¹⁶

Covariates

Additional sociodemographic and health variables were determined using existing military service records and the HealthViEWS mailed survey and phone interview. These variables were used to compare respondents to the medical record review with those responding to the overall survey and phone interview, and assess correlates of agreement between self-report and medical record review. We chose to compare agreement across demographic subgroups to determine whether some patient subgroups are more or less vulnerable to missed diagnoses in the medical record, or whether they were more or less likely to self-report certain conditions. Date of birth, service location (Vietnam, near Vietnam, United States), and military occupation (nurse vs. non-nurse) were obtained from available military service record data available in the HealthViEWS.⁶

Survey variables included the aforementioned physical health condition questionnaire (indication of receipt of treatment for any of 29 physical health conditions within the past year), as well as race, marital status, whether medical records were obtained from VA or non-VA providers, and health behaviors (hazardous alcohol use, smoking). Hazardous alcohol use was defined based on the Alcohol Use Disorders Identification Test question on whether the woman self-reported having six or more drinks on a single occasion within the past 2 weeks.¹⁷ Smoking status was based on the question whether the woman self-reported smoking at least 100 cigarettes in her lifetime or not.¹⁸

Phone interview variables included diagnosis of PTSD or major depressive disorder based on the Composite International Diagnostic Interview survey,¹⁹ as these diagnoses are thought to influence prevalence and level of agreement between self-report and medical records.^20,21

Analyses

All analyses were conducted using SAS version 9.4 (Cary, NC). Bivariate analyses (chi-square) were used to compare sociodemographic and health characteristics between the full cohort (n = 4219) and the medical record review subcohort (n = 412). Key independent variables were categorized as follows: age (≥65 years, <65 years), race (white, non-white), nursing military occupation (yes, no), marital status (currently married, not married), smoked at least 100 cigarettes in lifetime (yes, no), had six or more drinks on a single occasion (yes, no), PTSD or depression diagnosis (yes, no), and number of self-reported physical health conditions reported receiving treatment for within the past year (5 or more vs. <5). The number of self-reported conditions were further categorized given the skewness of the data (average number of conditions was 3.3, standard deviation = 2.5, and range of 0–13).

For analyses comparing self-report with medical records, physical health conditions were categorized by the following major categories: hypertension, endocrine (diabetes, hyperlipidemia), other cardiovascular/cerebrovascular conditions (angina pectoris/coronary artery disease, congestive heart failure, myocardial infarction/heart attack, stroke/transient ischemic attack, other valvular/arrhythmia), pulmonary (emphysema, asthma/chronic obstructive pulmonary disease, sleep apnea), gastrointestinal (GI) (hepatitis, irritable bowel syndrome), neurological (multiple sclerosis, Parkinson's disease, amyotrophic lateral sclerosis [ALS], or Lou Gehrig's disease), musculoskeletal (arthritis, osteoporosis, sciatica, fibromyalgia), breast and gynecological cancers (ovarian, uterine, cervical), and other cancers (lung, colon).

Overall agreement and Cohen's kappa (agreement beyond chance) were calculated for each physical health condition category. The percent overall agreement was calculated as the number of women in the sample who agreed on presence/absence of diagnosis (across the survey and the chart review), divided by the number in the sample, times 100.

Multivariable logistic regression analyses (reporting adjusted odds ratios [ORs] and 95% confidence intervals [CIs]) were conducted to identify correlates associated with the odds of agreement between self-report and medical record data for each physical health condition category, controlling for the aforementioned demographic, military, and health variables.

Results

Of the 900 women contacted for medical record reviews, 449 had available medical records for review (Fig. 1). Of the 449, 412 women (mean age in years = 67.0 ± 5.3, range: 59–88 years of age) had responses to all three data sources (survey, telephone, medical record) and were included in the analyses. Among the medical record subsample, 281 (68.2%) were extracted solely from non-VA paper medical records.

FIG. 1.

Cooperative Studies Program (CSP) 579 Consort Diagram for Medical Chart Abstraction Flow diagram of the progress through stages of the project—Of the 900 women contacted for medical record reviews, 449 had medical records sent back from at least one provider. Of the 449, 412 women (mean age = 67.0 ± 5.3) had responses to all three data sources (survey, telephone, and medical record) and were included in the analyses. HIPAA, Health Insurance Portability and Accountability Act.

Compared with the mail/telephone survey sample, women in the medical record review sample (n = 412) included a greater proportion having served in Vietnam, women who were nurses during the Vietnam era, those with major depressive disorder, and those who reported a greater number of conditions treated in the past year (Table 1).

The most commonly reported conditions based on self-report or medical record review included hypertension, hyperlipidemia, or arthritis (Table 2). Overall, with the exception of hyperlipidemia, women's' self-reported prevalence for each condition closely matched or was greater compared with medical record review prevalence estimates (Table 2).

Table 2.

Agreement Between Self-Report and Medical Record Report for Selected Physical Health Conditions (Within the Medical Record Review Subsample, n = 412)

Physical health condition	n (%) with condition, self-report	n (%) with condition, medical record ^a	Sensitivity (95% CI)	Specificity (95% CI)	Overall agreement ^b (%)	Agreement beyond chance: Cohen's kappa (95% CI)
Hypertension	275 (67.9)	286 (69.4)	0.91 (0.88–0.94)	0.85 (0.79–0.92)	89.4	0.75 (0.68–0.82)
Endocrinological conditions	274 (66.8)	297 (72.1)	0.84 (0.80–0.89)	0.78 (0.71–0.86)	82.7	0.59 (0.51–0.68)
Diabetes	132 (32.4)	126 (30.6)	0.94 (0.90–0.98)	0.95 (0.93–0.98)	95.1	0.89 (0.84–0.94)
Hyperlipidemia	245 (60.2)	283 (68.7)	0.77 (0.73–0.82)	0.77 (0.70–0.85)	77.4	0.51 (0.43–0.59)
Other cardiovascular/cerebrovascular conditions	146 (35.7)	109 (26.5)	0.72 (0.63–0.80)	0.77 (0.73–0.82)	75.8	0.44 (0.35–0.53)
Coronary artery disease(CAD)/angina	44 (10.8)	41 (9.9)	0.53 (0.37–0.68)	0.94 (0.91–0.96)	89.7	0.44 (0.30–0.58)
Congestive heart failure	21 (5.2)	21 (5.1)	0.63 (0.41–0.85)	0.98 (0.96–0.99)	96.0	0.58 (0.39–0.77)
Myocardial infarction	25 (6.2)	14 (3.4)	0.79 (0.57–1.00)	0.96 (0.95–0.98)	95.8	0.54 (0.35–0.73)
Stroke/Transient Ischemic Attack	16 (3.9)	16 (3.9)	0.56 (0.32–0.81)	0.98 (0.97–0.99)	96.6	0.54 (0.33–0.76)
Other: valvular/arrhythmia	104 (25.9)	59 (14.3)	0.56 (0.43–0.69)	0.79 (0.75–0.83)	75.9	0.26 (0.16–0.37)
Pulmonary conditions	119 (29.2)	122 (29.6)	0.78 (0.71–0.86)	0.91 (0.88–0.95)	87.5	0.69 (0.62–0.77)
Asthma/chronic obstructive pulmonary disease	82 (20.2)	94 (22.8)	0.69 (0.59–0.78)	0.94 (0.92–0.97)	88.4	0.66 (0.57–0.75)
Sleep apnea	58 (14.4)	48 (11.7)	0.87 (0.78–0.97)	0.95 (0.93–0.97)	94.3	0.75 (0.65–0.85)
Gastroenterological conditions	78 (19.2)	28 (6.8)	0.78 (0.62–0.93)	0.85 (0.81–0.89)	84.5	0.33 (0.22–0.45)
Hepatitis	38 (9.4)	8 (1.9)	1.00 (1.00–1.00)	0.92 (0.89–0.95)	92.3	0.29 (0.12–0.46)
Irritable bowel syndrome	45 (11.3)	22 (5.3)	0.65 (0.44–0.86)	0.92 (0.89–0.94)	90.2	0.36 (0.20–0.51)
Neurological conditions^c	25 (6.1)	25 (6.1)	0.92 (0.81–1.00)	0.99 (0.99–1.00)	99.0	0.91 (0.83–0.99)
Musculoskeletal conditions	309 (75.4)	255 (61.9)	0.88 (0.84–0.92)	0.45 (0.37–0.53)	71.7	0.36 (0.27–0.45)
Arthritis	234 (57.6)	197 (47.8)	0.82 (0.76–0.87)	0.64 (0.58–0.70)	72.4	0.45 (0.37–0.54)
Osteoporosis	122 (30.1)	81 (19.7)	0.77 (0.67–0.86)	0.82 (0.77–0.86)	80.5	0.49 (0.39–0.58)
Sciatica^d	143 (35.7)	26 (6.3)	0.73 (0.56–0.90)	0.67 (0.62–0.72)	67.3	0.13 (0.06–0.20)
Fibromyalgia	22 (5.5)	12 (2.9)	0.75 (0.51–0.99)	0.97 (0.95–0.98)	96.0	0.51 (0.30–0.72)
Cancers, breast and gynecological	130 (31.8)	121 (29.4)	0.96 (0.92–0.99)	0.95 (0.93–0.98)	95.4	0.89 (0.84–0.94)
Breast cancer^a	108 (26.6)	111 (26.9)	0.96 (0.93–0.99)	0.99 (0.99–1.00)	98.8	0.97 (0.94–0.99)
Gynecological cancers, cervical, uterine, or ovarian	28 (6.9)	14 (3.4)	0.86 (0.67–1.00)	0.96 (0.94–0.98)	95.6	0.55 (0.37–0.73)
Other cancers (lung, colon)	11 (2.7)	15 (3.6)	0.67 (0.43–0.91)	0.99 (0.99–1.00)	98.5	0.76 (0.58–0.95)

Women reporting key conditions (diabetes, hypertension, cardiovascular diseases, neurological conditions, or cancers) were oversampled for the medical record review, so rates of conditions (based upon self-report and/or medical record review) are not expected to reflect population prevalence. Women may have more than one condition within a category, and so the total number of women within one category may be lower than the sum of women having individual conditions within the category.

Overall agreement was calculated as the percentage of observed agreement (number who agreed on presence or absence of diagnosis (across the survey and the chart review) divided by the total sample times 100).

Neurological conditions from this table included multiple sclerosis, Parkinson's disease, and amyotrophic lateral sclerosis.

Sciatica also included self-reported pain or numbness traveling down below the knee.

CI, confidence interval.

Agreement beyond chance was relatively high for all pulmonary (kappa = 0.69) and neurological conditions (kappa = 0.91), and for specific conditions such as hypertension (kappa = 0.75), diabetes (kappa = 0.89), and breast cancer (kappa = 0.97).¹⁶ Lower agreement was found for other cardiovascular conditions (kappa = 0.44; 35.7% vs. 26.5% based on self-report vs. medical record), all GI conditions (kappa = 0.33; 19.2% vs. 6.8% based on self-report vs. medical record), and all musculoskeletal conditions (kappa = 0.36; 75.4% vs. 61.9% based on self-report vs. medical record) (Table 2).

Multivariable analyses revealed few covariates associated with estimated probability of agreement between self-report and medical record review across the condition categories, with some exceptions (Table 3). Women who reported having smoked at least 100 cigarettes in their lifetime had significantly lower probability of agreement for pulmonary conditions (OR = 0.39, 95% CI, 0.19–0.79) and GI conditions (OR = 0.54, 95% CI, 0.29–0.97) relative to women who smoked less. Women reporting five or more self-reported physical health conditions, relative to women reporting fewer conditions, also had significantly lower odds of agreement for GI conditions (OR = 0.54, 95% CI: 0.30–0.97), but significantly greater probability of agreement for endocrinological conditions (OR = 4.25, 95% CI: 1.84–9.80). Women who served in Vietnam versus those serving near Vietnam or in the United States had significantly lower probability of agreement between the two data sources for pulmonary conditions (OR = 0.45, 95% CI: 0.22–0.94). There were no other covariates associated with probability of agreement between self-report and medical records. Agreement did not significantly differ by the source of medical record information (i.e., VA-only vs. non-VA medical records).

Table 3.

Correlates of Agreement Between Self-Report and Medical Record Review by Condition Categories

Odds ratio (95% CI), n = 412	Hypertension	Endocrinological conditions	Cardiovascular/cerebrovascular conditions	Pulmonary conditions	Gastroenterological conditions	Musculoskeletal conditions
Vietnam Service (vs. U.S. Service)	0.49 (0.23–1.06)	0.76 (0.42–1.36)	0.70 (0.42–1.19)	0.45 (0.22–0.94)^a	1.16 (0.63–2.13)	1.01 (0.62–1.64)
Near Vietnam Service (vs. U.S. Service)	1.41 (0.29–6.96)	1.75 (0.54–5.63)	0.65 (0.27–1.58)	0.62 (0.18–2.13)	0.91 (0.32–2.54)	0.72 (0.32–1.62)
Age ≥65 years (vs. <65)	0.65 (0.33–1.28)	0.76 (0.44–1.32)	1.37 (0.86–2.21)	1.05 (0.56–1.96)	1.25 (0.71–2.21)	1.19 (0.76–1.86)
White (vs. non-White)	1.67 (0.50–5.59)	1.87 (0.64–5.45)	0.83 (0.30–2.25)	1.71 (0.56–5.22)	0.17 (0.02–1.34)	0.89 (0.35–2.30)
Nurse occupation (vs. non-nurse)	1.58 (0.74–3.38)	1.59 (0.85–2.97)	1.50 (0.87–2.60)	1.36 (0.65–2.84)	1.12 (0.57–2.19)	0.99 (0.58–1.70)
VA medical record (vs. non-VA medical record)	1.16 (0.47–2.87)	2.11 (0.86–5.19)	0.90 (0.48–1.69)	1.68 (0.67–4.23)	0.73 (0.35–1.54)	0.85 (0.47–1.57)
Smoked at least 100 cigarettes (yes vs. no)	1.16 (0.60–2.24)	0.82 (0.47–1.43)	1.14 (0.71–1.83)	0.39 (0.19–0.79)^a	0.54 (0.29–0.97)^a	0.90 (0.57–1.41)
Hazardous alcohol use (yes vs. no)	3.17 (0.40–25.02)	0.61 (0.23–1.61)	2.52 (0.72–8.80)	0.94 (0.30–2.98)	1.62 (0.46–5.73)	0.80 (0.34–1.91)
No. of comorbidities ≥5 (vs. <5)	2.19 (0.94–5.06)	4.25 (1.84–9.80)^a	0.96 (0.57–1.63)	1.19 (0.60–2.37)	0.54 (0.30–0.97)^a	1.53 (0.91–2.58)
PTSD diagnosis (vs. no diagnosis)	0.69 (0.30–1.57)	1.91 (0.80–4.56)	0.87 (0.47–1.60)	0.62 (0.30–1.29)	1.03 (0.49–2.14)	0.89 (0.49–1.62)

95% CI does not contain 1.0.

For each condition category, a separate multivariable logistic regression model was run adjusting for age in 2010 (≥65 vs. <65 years), nurse (vs. non-nurse), white race (vs. non-white), whether medical record reviews came from the VA exclusively (VA medical record vs. no VA medical records available/non-VA provider), smoking status (smoked 100 or more cigarettes in lifetime, yes vs. no), hazardous alcohol use (reported having six or more drinks on a single occasion vs. less than six drinks on a single occasion), total count of self-reported current treatment for physical health conditions (n = 21 total) of 5 or more (vs. less than 5), and PTSD diagnosis from the CIDI (vs. no diagnosis).

VA, Veterans Health Administration.

Discussion

We found relatively high agreement between self-report and medical records for most chronic physical health conditions experienced by Vietnam-era women veterans. Similar trends in agreement between self-report and medical record reviews have been found in previous studies of older populations. Notably, previous research suggests that self-report data were in high agreement with medical record reviews for chronic disorders requiring ongoing management (e.g., diabetes), especially among women with more advanced education after adjusting for demographic factors. Okura et al. found that questionnaire data were in high agreement with medical record reviews for life-threatening, acute-onset diseases, as well as chronic disorders requiring ongoing management (e.g., diabetes and hypertension) especially among better-educated women after adjusting for demographic factors.²²

However, in this study there was a greater disagreement for particular conditions (GI, musculoskeletal, and cardiovascular). Observed lower agreement between self-report and medical conditions was mainly driven by greater tendency for women to self-report conditions more frequently than was found in current medical records procured from their providers. This supports previous research on older women that found low agreement between self-report and medical record reviews for cardiovascular conditions such as heart failure (kappa <0.4) as well as for hypertension, myocardial infarction, and stroke (kappa <0.6).^23,24 Teh et al. found lower agreement in older subjects particularly for cardiovascular conditions such as heart failure (kappa <0.4), as well as for hypertension, myocardial infarction, and stroke (kappa <0.6).²³ Lower agreement was also observed for GI or musculoskeletal conditions. In these situations, medical records may have missed these conditions if women were able to self-treat these maladies (e.g., with over-the-counter medications), if the condition's onset predated our observation window, or if treatment was received by specialists at different facilities and was not integrated into the medical records examined here. In contrast, several of the high agreement conditions (e.g., diabetes, neurological conditions, breast cancer) involved active treatment or therapies by a provider and therefore were more likely to be repeatedly documented in a series of follow-up visits, making them more detectable in the medical record.

Medical records came from all sources of care: some women provided records from VA (an integrated healthcare system), some provided records from non-VA sources only (representing a range of systems, including capitated and fee-for-service), and some provided records from both VA and non-VA providers. Nonetheless, agreement with self-report did not differ by source of medical record (VA-only vs. any non-VA). This finding supports the argument that self-reported information in women veterans could provide necessary health information. Moreover, such information is cost effective and immediately attainable.

The multivariable analyses that examined medical record/self-report consistency as a function of demographic, military, or health correlates largely revealed no reliable associations with the agreement. This suggests that the agreement may be fairly consistent across different subgroups of women veterans serving during the Vietnam era. The few exceptions to this finding included women who were smokers or had served in Vietnam had a lower likelihood of agreement for pulmonary conditions. Reasons for these trends could potentially include underreporting of these conditions in the self-report due to potential social desirability in not self-reporting medical sequelae of smoking such as lung-related conditions, or among women serving in Vietnam, exposure to potentially hazardous airborne elements. Those reporting five or more conditions were more likely to have agreement for endocrinological conditions, such as diabetes, yet potentially poorer agreement for GI conditions, such as hepatitis, although this last relationship appears weak (upper limit of CI = 0.97). This may reflect the potential for some conditions to receive less attention due to other “competing” medical conditions. Prior research has found that an increased burden of physical health conditions can lead to decreased focus on potentially competing medical conditions that are not routinely assessed.²⁵ A better understanding of which groups might have a tendency to under-report can inform efforts to better adjust for trends in medical comorbidities among these subpopulations, and ultimately help better characterize the health of populations that do not have complete medical record information.

To our knowledge, HealthViEWs is the first population-based study to assess agreement among a cohort of older women veterans for a comprehensive list of major physical health conditions. The Vietnam War was one of the most significant American wars of the 20th century with 2.6 million people who served in the country of Vietnam.¹⁵ While a great deal is known about the physical health of the men who served in Vietnam, less is known about the physical health of women veterans who served there. Moreover, while more attention is being paid to mental health conditions experienced by women veterans, less is known about their physical health. While officially not in combat roles at the time of the Vietnam War, many of these women served in a warzone, hence may have been exposed to harmful chemicals or experienced health risks that could lead to lifelong physical health conditions. In addition, little is known regarding the medical risk factors of those serving during war, but not necessarily in an active warzone, such as in military hospitals or bases.

This study is innovative in several ways and has many advantages from previous studies of women Vietnam-era veterans. Notably, it undertakes a comprehensive assessment of physical health conditions based on different data sources among a veteran population beyond current users of VA healthcare. However, there are limitations to this study that warrant consideration. Neither self-report nor medical record review is considered the ultimate gold standard for ascertaining physical health conditions, and cost and logistics precluded further confirmation by in-person clinical exams. As most of the providers were not from the VA, the study relied on paper medical records. The variation in the number of providers that sent paper medical records (up to three) could have altered the overall agreement. Also, providers may not have diagnosed conditions such as fibromyalgia, but rather had documented these in the clinical notes that were not sent as part of the medical record copies. We were also unable to assess documentation of cognition as part of the neurological conditions. Moreover, for some conditions (e.g., hepatitis) we were unable to differentiate specific disease types which have very different courses (e.g., hepatitis A). Finally, the overall good agreement observed might be attributed to the fact that the majority of respondents were nurses, and hence, may have had higher health literacy than non-nurses.

Conclusions

Among women veterans serving during the Vietnam War era, there was relatively high agreement for physical health conditions when self-report was compared with medical record review. Also, agreement did not vary widely by different sociodemographic groups, or between data gathered from VA versus the private sector. This suggests that self-report may be a reliable resource of information on physical health among Vietnam-era women veterans, especially given that most had a nursing background. Nonetheless, inconsistency between self-report and medical records may provide the opportunity to elucidate documentation problems pertaining to potentially serious conditions, notably cardiovascular disease that might be missed in medical records among women Vietnam veterans. Such inconsistency could reflect underdiagnosis or failure to record diagnoses in the medical record, and future studies would do well to deconstruct the sources of these potential inconsistencies.

Footnotes

Acknowledgment

This research was supported by the Department of Veterans Affairs, Veterans Health Administration, Cooperative Studies Program (CSP) #579.

Disclaimer Statement

The views expressed in this article are those of the authors and do not necessarily represent the views of the Department of Veterans Affairs.

Author Disclosure Statement

No competing financial interests exist.

References

Department of Veterans Affairs National Center for Veterans Analysis and Statistics. Projected Veteran Population 2013–2043. Washington, DC: Department of Veterans Affairs, 2014. Available at: www.va.gov/vetdata/docs/quickfacts/Population_slideshow.pdf Accessed June 28, 2016 .

U.S. Department of Veterans Affairs. Evidence Synthesis Program. ESP Report: An Evidence Map of the Women Veterans' Health Research Literature (2008–2015). Available at: https://www.hsrd.research.va.gov/publications/esp/womens-health2.cfm Accessed July 28, 2017 .

Goldzweig

, Balekian

, Rolon

, Yano

, Shekelle

. The state of women veterans' health research. Results of a systematic literature review. J Gen Intern Med, 2006; 21 Suppl 3:S82–S92.

Haskell

, Mattocks

, Goulet

, et al. The burden of illness in the first year home: Do male and female VA users differ in health conditions and healthcare utilization. Womens Health Issues, 2011; 21:92–97.

Bean-Mayberry

, Yano

, Washington

, et al. Systematic review of women veterans' health: Update on successes and gaps. Womens Health Issues, 2011; 21(4 Suppl):S84–S97.

Magruder

, Serpi

, Kimerling

, et al. Prevalence of post-traumatic stress disorder in vietnam-era women veterans. JAMA Psychiatry, 2015; 72:1127–1134.

Kang

, Cypel

, Kilbourne

, et al. HealthViEWS: Mortality study of female US Vietnam era veterans, 1965–2010. Am J Epidemiol, 2014; 179:721–730.

Department of Veterans Affairs National Center for Veterans Analysis and Statistics.. American's Women Veterans: Military Service History and VA Benefit Utilization Statistics. Washington, DC: Department of Veterans Affairs, 2011. Available at: www.va.gov/vetdata/docs/specialreports/final_womens_report_3_2_12_v_7.pdf Accessed June 28, 2016 .

Veterans Access, Choice, and Accountability Act of 2014. Pub. L. 113–146. Available at: https://www.congress.gov/bill/113th-congress/house-bill/3230 Accessed June 28, 2016 .

10.

Muggah

, Graves

, Bennett

, Manuel

. Ascertainment of chronic diseases using population health data: A comparison of health administrative data and patient self-report. BMC Public Health, 2013; 13:16.

11.

Yurkovich

, Avina-Zubieta

, Thomas

, Gorenchtein

, Lacaille

. A systematic review identifies valid comorbidity indices derived from administrative health data. J Clin Epidemiol, 2015; 68:3–14.

12.

Miller

, Rogers

, Kazis

, et al. Patients' self-report of diseases in the Medicare Health Outcomes Survey based on comparisons with linked survey and medical data from the Veterans Health Administration. J Ambul Care Manage, 2008; 31:161–177.

13.

Dalager

, Kang

, Thomas

. Cancer mortality patterns among women who served in the military: The Vietnam experience. J Occup Environ Med, 1995; 37:298–305.

14.

Thomas

, Kang

, Dalager

. Mortality among women Vietnam veterans, 1973–1987. Am J Epidemiol, 1991; 134:973–980.

15.

Brooks

, Laditka

. Evidence of greater health care needs among older veterans of the Vietnam War. Mil Med, 2008; 173:715–720.

16.

Landis

, Koch

. The measurement of observer agreement for categorical data. Biometrics, 1977; 33:159–174.

17.

Meneses-Gaya

, Zuardi

, Loureiro

, et al. Is the full version of the AUDIT really necessary? Study of the validity and internal construct of its abbreviated versions. Alcohol Clin Exp Res, 2010; 34:1417–1424.

18.

Tanner

, Egede

, Shamblin

, et al. Attitudes and beliefs toward lung cancer screening among US Veterans. Chest, 2013; 144:1783–1787.

19.

Kimerling

, Serpi

, Weathers

, et al. Diagnostic accuracy of the Composite International Diagnostic Interview (CIDI 3.0) PTSD module among female Vietnam-era veterans. J Traumatic Stress, 2014; 27:160–167.

20.

Beristianos

, Yaffe

, Cohen

, et al. PTSD and risk of incident cardiovascular disease in aging veterans. Am J Geriatr Psychiatry, 2016; 24:192–200.

21.

Kilbourne

, McCarthy

, et al. Recognition of co-occurring medical conditions among patients with serious mental illness. J Nerv Ment Dis, 2006; 194:598–602.

22.

Okura

, Urban

, Mahoney

, Jacobsen

, Rodeheffer

. Agreement between self-report questionnaires and medical record data was substantial for diabetes, hypertension, myocardial infarction and stroke but not for heart failure. J Clin Epidemiol, 2004; 57:1096–1103.

23.

Teh

, Doughty

, et al. Agreement between self-reports and medical records of cardiovascular disease in octogenarians. J Clin Epidemiol, 2013; 66:1135–1143.

24.

Chou

, Scholle

, Weisman

, Bierman

, Correa-de-Araujo

, Mosca

. Gender disparities in the quality of cardiovascular disease care in private managed care plans. Womens Health Issues, 2007; 17:120–130.

25.

Redelmeier

, Tan

, Booth

. The treatment of unrelated disorders in patients with chronic medical diseases. N Engl J Med, 1998; 338:1516–1520.