Materials and Methods
Participants and procedures
Potential participants were identified from the electronic records of 2 Midwestern VA medical centers. Female veterans 52 years of age and younger enrolled at one of the centers or associated community-based outreach clinics in the 5 years previous to, or during, the interview dates of July 2005 to August 2008 were recruited to participate in a computer-assisted telephone interview.
VA enrollment could have been initiated to receive health care, complete a disability claim, enroll in a registry, or respond to veteran outreach. Potential subjects were recruited via mail and telephone protocols, which have been described in detail elsewhere. Women who returned their signed consent forms and met inclusion criteria completed a computer-assisted telephone interview that included questions related to demographics, lifetime sexual assault, gynecological diagnoses and procedures, general and mental health history, health risk behaviors, and care seeking.
The majority of interviews were completed in a single phone call (89%) and lasted an average of 1 hour 16 minutes. Participants who completed the interview were reimbursed $30.00 (US dollars). The study was approved by Institutional Review Boards of the University of Iowa and the Iowa City VA Medical Center.
For the purpose of the funded grant and primary goal of the larger study (to determine premenopausal-aged female veterans’ competing risk factors, including lifetime sexual assault, associated with cervical cytological abnormalities), subjects were excluded from participation if they were older than 52 years, were aware of in utero diethylstilbestrol exposure, or were currently receiving immunosuppressants. This reduced the involvement of those at unusually high risk of cervical dysplasia or genital malignancy.
Sixty-nine percent of the identified sample (1670 of 2414) were located and invited to participate, 63% of whom agreed to participate. Women who refused participation did not differ significantly from those who participated with regard to age, self-report of very good or excellent health, number of gynecological visits in the last year, or ever having had an abnormal Papanicolaou smear. Ninety-five percent of women who agreed to participate completed their interviews; however, 15 were missing data for items related to hysterectomy or sexual assault and were excluded from all analyses, giving a sample size for analysis of 989.
Measures
The primary outcome for this study was hysterectomy, which was operationalized as removal of the uterus, with or without oophorectomy or salpingectomy. Women who reported a hysterectomy were asked follow-up questions to identify reasons for the surgery. Possibilities within the computer-assisted interview program included the following: excessive bleeding, chronic pelvic pain, abnormal Papanicolaou smear, prolapsed uterus, cancer (ovarian, cervical, or uterine), noncancerous fibroids, benign tumor, “tired of pelvic examinations,” childbirth complications, sexual assault damage, and some other reason. Women who reported having a hysterectomy for some other reason were asked to specify the reason as an open-ended response.
Sexual assault (SA) was assessed using the definition adopted by the American Medical Association (1995) and the American College of Obstetricians and Gynecologists (1997), which includes any sexual act that occurred without a woman’s consent involving the use or threat of force or against the woman’s wishes and includes attempted or completed sexual penetration of the vagina, mouth, or rectum by penis, fingers, or object.
Sexual assaults in which an attempt was made but penetration did not occur were deemed attempted SA; completed SA were assaults in which penetration did occur. This definition was read to participants, and women were asked whether they had experienced any such act. Women reporting 1 or more SA were asked to disclose the following: the number of incidents, whether these occurred during childhood (< 18 years old), in adulthood before military service (age 18 years to military entry), during military service, or in adulthood after military service, and the age at which the first assault, first military assault, and most recent assault occurred.
For the analyses presented in this paper, lifetime sexual assault with vaginal penetration (LSA-V) is defined as completed penetration of the vagina by penis, fingers, or objects. Lifetime nonvaginal sexual assault is defined as completed sexual penetration of the mouth or rectum without completed penetration of the vagina.
History of PTSD was assessed by asking subjects whether they had ever been diagnosed with this disorder by a clinician. This PTSD diagnosis history was operationalized as a dichotomous variable for analysis. Participants were also asked for the age at which they received their first PTSD diagnosis.
Gynecological symptoms examined in this study included the following: history of gynecological pain, gynecological bleeding, and pelvic inflammatory disease (PID). History of gynecological pain was operationalized as 1 or more visits to a doctor for gynecological issues during at least 1 time period (the 5 years prior to their military service, during military service, or in the first year after military service) and experience of gynecological pain, which could include pelvic, crotch, or vaginal pain, including pain during periods, but not including pregnancy, labor, or delivery pain.
History of gynecological bleeding was operationalized as 1 or more visits to a doctor for gynecological issues during at least one of the time periods listed above and experience of abnormal gynecological bleeding, which could include heavy bleeding during your period, bleeding in between periods, or postmenopausal bleeding.
To assess PID, women were asked whether a health professional had ever told them that they had PID. All gynecological symptoms were operationalized as dichotomous for the purposes of analysis.
Care utilization was examined through past medical and gynecological care settings and past accessibility of care. Medical care setting was assessed by asking, “In the past 5 years, did you receive all, some, or none of your medical care at Veteran Affairs medical centers?” For analysis purposes, “none” and “some” were combined into a single group who received non-VA health care.
Gynecological care setting was assessed by asking, “In the past 5 years, did you receive women’s health care (gynecological, reproductive, etc) somewhere other than a VA medical center?” Past accessibility of care was assessed by asking participants how often they had ready access to gynecological care, including birth control, and routine gynecological health care during the 5 years prior to their military service, during their military service, or in the first year after their military service. Participants responding less than “almost always” for at least 1 time period were coded as experiencing a time period with less than “almost always” access to gynecological care.
Standard demographic measures were assessed, including those variables with a known association with hysterectomy risk (see Table 1 ). Because of the small numbers of participants in each racial/ethnic minority category, race was coded as a dichotomous variable, with all participants describing a nonwhite race/ethnicity (including any participants describing a Hispanic race/ethnicity) combined in the racial minority category. Body mass index (BMI) was calculated using self-reported height and weight. Parity was calculated by subtracting self-reported pregnancy terminations, ectopic pregnancies, miscarriages, and current pregnancy from the number of times women reported having been pregnant.
| Characteristics | Entire sample (n = 989) | No hysterectomy (n = 823) | Hysterectomy (n = 166) | P value |
|---|---|---|---|---|
| Sociodemographics | ||||
| Age, mean ± SD | 38 ± 8.8 | 37 ± 8.8 | 44 ± 5.7 | < .001 a |
| White race, n, % | 790 (80) | 659 (80) | 131 (79) | .816 |
| Current orientation, heterosexual, n, % | 921 (94) | 771 (94) | 150 (92) | .240 |
| History of consensual sex with women, n, % | 117 (12) | 97 (12) | 20 (12) | 1.00 |
| Enlisted rank (vs officer), n, % | 938 (95) | 777 (95) | 161 (98) | .194 |
| Education, n, % | .489 | |||
| High school | 150 (15) | 122 (15) | 28 (17) | |
| Some college | 559 (57) | 462 (56) | 97 (58) | |
| College graduate | 280 (28) | 239 (29) | 41 (25) | |
| Household income, median, $ | 35,000 | 35,000 | 40,000 | .141 b |
| Ever married, n, % | 763 (77) | 612 (74) | 151 (91) | < .001 |
| Care utilization | ||||
| Non-VA women’s health care (in past 5 y), n, % | 518 (53) | 444 (54) | 74 (45) | .037 |
| Non-VA medical care (in the past 5 y), n, % | 667 (68) | 568 (69) | 99 (60) | .021 |
| Experienced a time period with less than almost always access to gynecological care, n, % | 513 (53) | 419 (52) | 94 (58) | .162 |
Comparative population data (see Supplemental Table 1 for a side-by-side comparison of the study groups)
To compare the prevalence of hysterectomy from our sample with the general population, we examined the 2008 Behavioral Risk Factor Surveillance System (BRFSS) data set. The BRFSS is an ongoing data collection program administered by the Centers for Disease Control and Prevention’s Behavioral Surveillance Branch to collect uniform, state-specific data on health practices, and risk behaviors that are linked to chronic diseases, injuries, and preventable infectious diseases that affect the adult population.
Data collection, sampling methodology, and limitations of the BRFSS data set are described elsewhere by the Centers for Disease Control and Prevention. In 2008, the BRFSS asked, “Have you had a hysterectomy?” within the core component of the survey, although it did not query age at hysterectomy.
To assess the prevalence of hysterectomy in the general population, we restricted the BRFSS sample to women who reported never serving on active duty in the United States Armed Forces (regular military or National Guard or military reserve unit). Because the prevalence of hysterectomy increases with age, we further restricted the BRFSS sample to match the age range of our study sample (20-52 years) prior to assessing prevalence of hysterectomy. This yielded a sample size of 107,003 respondents; 100,739 of these respondents provided valid responses to the hysterectomy item.
To compare age at hysterectomy for our sample with the general population, we examined the 2008 American College of Surgeons National Surgical Quality Improvement Program data set (American College of Surgeons National Surgical Quality Improvement Program [NSQIP]). The American College of Surgeons NSQIP extracts data from medical records on major surgical procedures performed in inpatient and outpatient settings at participating sites. Data collection, sampling methodology, and limitations of this data set are described by the American College of Surgeons.
To examine the demographics of patients receiving hysterectomy, we included all Current Procedural Terminology codes for hysterectomy and then restricted the sample by age (< 53 years) to match the age range in which our sample could have received their hysterectomies. This yielded a sample size of 3645 patients.
To examine the reason for hysterectomy within the NSQIP sample, the International Classification of Diseases , ninth revision, Clinical Modification codes corresponding to the postoperative diagnosis were collapsed into 14 categories, including the following: (1) abdominal/pelvic mass; (2) abdominal pain; (3) hemorrhage/anemia; (4) benign neoplasm; (5) carcinoma; (6) inflammation; (7) endometrial hyperplasia; (8) prolapse; (9) pelvic pain; (10) cervical dysplasia; (11) endometriosis; (12) abnormal uterine bleeding; (13) fibroids; and (14) cancer. Summary statistics regarding race/ethnicity, BMI, smoking history, age at hysterectomy, and postoperative diagnosis were assessed for this sample.
Data analysis
Preliminary analyses examined self-reported hysterectomy among the following 4 categories of SA: 1) no SA, 2) attempted SA, 3) completed lifetime SA nonvaginal, and 4) at least 1 completed LSA-V. Hysterectomy rates were not significantly different between groups with attempted SA and nonvaginal completed SA, and neither of these groups was different from the group with no SA ( Figure ); therefore, for the purposes of bivariate and multivariate analyses, these 3 groups were combined to create a comparison no LSA-V group.
We examined bivariate relationships between hysterectomy and key demographic variables including race, age, education, marital status, and insurance coverage (see Table 1 ). In addition, we investigated bivariate relationships between hysterectomy, LSA-V, and PTSD and key clinical variables including BMI, parity, smoking history, PID, and self-reported history of gynecological bleeding and gynecological pain.
Bivariate analyses were conducted using χ 2 tests of independence for all variables except income, parity, and number of assaults (completed vaginal and nonvaginal assaults), which were assessed with the Wilcoxon-Mann-Whitney test, and age and BMI, which were assessed with Student t tests. Because of the skewed nature of the data on number of sexual assaults, we used the minimum P value approach to determine the cut point for examining associations between the number of completed assaults and hysterectomy.
Odds ratios were adjusted for age (see Table 2 ). Logistic regression was used to examine the independent contribution of these factors to hysterectomy (see Table 3 ). Initial regression models considered all hypothesized individual factors; those factors found to be statistically associated with hysterectomy were grouped into risk factor categories. Variables significantly associated with hysterectomy at a bivariate level, or clinically relevant, were grouped into 4 categories of theoretically related factors including demographics, health characteristics, gynecological health history, and trauma (LSA-V and PTSD). Logistic regression for each group of related variables was then performed. Variables found to be associated with the outcome at alpha < 0.05 within each group were retained for the final integrated model. All analyses were conducted in SPSS 22.0 (SPSS Inc, Chicago, IL).
| Characteristics | Entire sample (n = 989) | No hysterectomy (n = 823) | Hysterectomy (n = 166) | Age-adjusted AOR (95% CI) a |
|---|---|---|---|---|
| BMI, mean ± SD | 28.2 ± 6.1 | 28.0 ± 6.0 | 29.4 ± 6.2 | 1.01 (0.98–1.04) |
| Parity, mean ± SD | 1.3 ± 1.3 | 1.2 ± 1.2 | 1.7 ± 1.3 | 1.12 (0.98–1.28) |
| Smoking, n, % | 574 (58) | 460 (56) | 115 (69) | 1.42 (0.97–2.07) |
| History of gynecological pain, n, % | 499 (51) | 372 (45) | 127 (77) | 3.75 (2.51–5.59) b |
| History of gynecological bleeding, n, % | 376 (38) | 268 (33) | 108 (65) | 3.65 (2.5–5.28) b |
| Pelvic inflammatory disease, n, % | 84 (9) | 51 (6) | 33 (20) | 2.95 (1.77–4.90) b |
| LSA-V, n, % | 475 (48) | 367 (45) | 108 (65) | 1.85 (1.28–2.66) c |
| History of PTSD, n, % | 228 (23) | 173 (21) | 55 (33) | 1.83 (1.24–2.70) c |
a Adjusted odds ratio from regression model including age
| Models | Demographics model (n = 981) a | Health characteristics model (n = 984) b | Gynecological symptoms model (n = 986) c | Trauma model (n = 984) d | Integrated model (n = 976) e |
|---|---|---|---|---|---|
| Demographic factors | |||||
| Age | 1.13 [1.10–1.16] f | 1.13 (1.10–1.16) f | 1.13 (1.09–1.16) f | 1.13 (1.10–1.16) f | 1.12 [1.09–1.16] f |
| Ever married | 1.94 [1.06–3.56] g | 1.90 (1.03–3.50) g | |||
| Parity | 1.06 [0.92–1.22] | ||||
| Health characteristics | |||||
| BMI | 1.01 (0.99–1.04) | ||||
| Smoking | 1.46 (0.99–2.15) | ||||
| Gynecological health history | |||||
| History of gynecological pain | 2.24 (1.43–3.52) f | 2.16 (1.–3.42) h | |||
| History of gynecological bleeding | 2.42 (1.6–3.66) f | 2.33 (1.54–3.55) f | |||
| Pelvic inflammatory disease | 2.00 (1.17–3.43) g | 1.89 (1.10–3.26) g | |||
| Trauma | |||||
| LSA-V | 1.62 (1.10–2.37) g | 1.32 (0.88–1.98) | |||
| History of PTSD | 1.57 (1.04–2.36) g | 1.41 (0.9–2.18) | |||
a Adjusted OR from regression model including age, ever married, and parity
b Adjusted OR from regression model including age, BMI, and smoking
c Adjusted OR from regression model including age, history of gynecological pain, history of gynecological bleeding, and pelvic inflammatory disease
d Adjusted OR from regression model including age, LSA-V, and history of PTSD
e Adjusted OR from regression model including age, ever married, history of gynecological pain, history of gynecological bleeding, pelvic inflammatory disease, LSA-V, and history of PTSD
We compared hysterectomy rates between our veteran study population and BRFSS data using χ 2 tests; age at hysterectomy was compared between our study population and the NSQIP data using Student t tests.
Results
Study population characteristics (see Table 1 )
Subjects ranged from 20 to 52 years of age, with a mean age of 38 (±8.8) and a median age of 40 years. The majority was white (80%), heterosexual (94%, although 12% reported having had sex with women in their lifetimes), enlisted personnel (95%; as opposed to officer), with at least some college education (85%) and some form of health insurance (84%). Median household income was $35,000. Forty percent of those with health insurance were privately insured (not included in table). Sixty-eight percent of the sample had received medical care outside the VA in the past 5 years, 78% of whom had received women’s gynecological or reproductive health care outside the VA. Fifty-three percent had experienced a time period with less than almost always access to gynecological care.
Almost two thirds of the subjects (62%) reported a sexual assault exposure (attempted or completed sexual assault) during their lifetime, and 48% experienced at least 1 completed sexual assault with vaginal penetration (LSA-V). Among those with completed LSA-V, 61% were first assaulted in childhood, 8% during adulthood prior to military service, 26% during military service, and 5% during adulthood after military service. Among those with LSA-V, there was a mean of 7 (SD, 9.1; median, 2; range, 1-56) completed vaginal and nonvaginal assaults. The minimum P value approach indicated a cut point value of 4 completed vaginal and nonvaginal assaults; 41% of those with LSA-V experienced 4 or more of these assaults.
Sixty percent of the sample reported at least 1 of the following: gynecological pain, gynecological bleeding, or PID. Approximately half of those symptomatic women (55%) acknowledged 2 or more of these gynecological symptoms. Those who had been sexually assaulted were more likely to report gynecological pain, gynecological bleeding, or PID (69% vs 51%, P < .0001; data not shown in tables). These variables were examined to determine whether the data met the assumption of collinearity. Collinearity statistics indicated that multicollinearity was not a concern (history of gynecological pain: tolerance, 0.76, variance inflation factor [VIF], 1.31; history of gynecological bleeding: tolerance, 0.79, VIF, 1.26; PID: tolerance, 0.95, VIF, 1.05; exposure to LSA-V: tolerance, 0.95, VIF, 1.05).
Hysterectomy and associated sociodemographic, clinical, and trauma-related factors
Table 1 compares sociodemographic and care utilization characteristics of female veterans with and without hysterectomy. Female veterans who reported hysterectomy were significantly older and more likely to have ever been married; they were significantly less likely to have sought non-VA general medical and women’s gynecological or reproductive health care in the 5 years prior to the interview (see Supplemental Figure for a graph representing the prevalence of hysterectomy by age range).
Table 2 compares health and trauma-related characteristics of female veterans by hysterectomy status. Although all examined health and trauma-related characteristics were significantly associated with hysterectomy in bivariate analyses, only a history of gynecological bleeding (age-adjusted odds ratio [AOR], 3.65, 95% confidence interval [CI], 2.52–5.28), gynecological pain (AOR, 3.75, 95% CI, 2.51–5.59), pelvic inflammatory disease (AOR, 2.95, 95% CI, 1.77–4.90), LSA-V (AOR, 1.85, 95% CI, 1.28–2.66), and PTSD (AOR, 1.83, 95% CI, 1.24–2.70) remained significant after adjusting for age.
Among those with both a history of PTSD diagnosis and hysterectomy, 51% received their PTSD diagnosis following their hysterectomy, 44% received their PTSD diagnosis before their hysterectomy, and 6% received their PTSD diagnosis and had their hysterectomy within the same year.
Hysterectomy prevalence, ages, and indications (data not shown in tables)
One hundred sixty-six of the 989 veteran subjects reported having had a hysterectomy; this represents a significantly higher rate of hysterectomy compared with the rate found in the general population BRFSS data set (16.8% in our population vs 13.3% in the BRFSS population, P = .002). Of note, the median age was 40 years in our population at the time of the interview and 41 in the BRFSS population at the time of the interview. Interestingly, 2352 women in the age-restricted BRFSS data set self-identified as veteran/military personnel. Consistent with our study findings, this subset of the BRFSS population showed a higher prevalence of hysterectomy, with 17.7% of those 2352 military women reported having had a hysterectomy.
Mean reported age at hysterectomy was significantly lower in our sample of veterans than that seen in the age-restricted NSQIP data set (35 years old [range 17–48 years] in our veteran population vs 43 years old in the NSQIP population, P < .0001). The mean time since hysterectomy for our veteran study population was 10 years (±8), with 41% of hysterectomies occurring in the past 5 years. The majority of the subjects who had a hysterectomy (65%) reported having had their hysterectomy in a public or private non–VA medical center facility.
Subjects typically identified multiple reasons for hysterectomy, with 89% of the sample identifying 2 or more reasons (data not shown in tables). The most frequently identified reasons for hysterectomy included the following: chronic pelvic pain (71%), excessive bleeding (65%), noncancerous fibroids (46%), abnormal Papanicolaou smear (28%), and some other reason (42%). Among reasons cited as some other reason, endometriosis (51%) was the most common, and others included ovarian cysts/problems (18%), birth control (5%), and dyspareunia (4%) (not mutually exclusive).
These indications for hysterectomy among our sample are generally consistent with those commonly cited for hysterectomy within the literature and within the NSQIP data set, in which fibroids were the primary indication in 33% of postoperative diagnoses, abnormal uterine bleeding in 24%, and pelvic pain in 5%.
The large proportion of our respondents selecting some other reason likely was due to the omission of endometriosis from the response set, another common indication (10%) in the NSQIP sample and a common free response in our study. Whereas indications for hysterectomy are difficult to directly compare because they were derived from patient self-report in our study and from medical record extraction in the NSQIP study, it is interesting to note the markedly higher rate of reporting chronic pelvic pain as the reason for hysterectomy in our study population compared with NSQIP (71% vs 5%, respectively).
Logistic regression models
As noted above ( Table 2 ), veterans who had experienced LSA-V were almost twice as likely to have had a hysterectomy as those with no sexual assault exposure (AOR, 1.92, 95% CI, 1.27–2.92). After logistic regression modeling including gynecological pain, bleeding, PID, and PTSD, this association between LSA-V and hysterectomy was no longer significant ( Table 3 ). Results generally remained unchanged in follow-up analyses in which we restricted the sample to an age range matched on hysterectomy history (remaining range 27-51 years), except that the category of ever married was no longer significantly associated with hysterectomy.
Participants reporting LSA-V and hysterectomy had experienced a median of 3.5 completed vaginal and nonvaginal assaults compared with 2.0 vaginal and nonvaginal assaults for those without hysterectomy ( P = NS). Among women with LSA-V, those with 4 or more completed vaginal and nonvaginal assaults were significantly more likely to have had a hysterectomy than those with 3 or fewer completed vaginal and nonvaginal assaults (28% vs 19%, P = .047). Among subjects reporting both hysterectomy and LSA-V, 96% had their hysterectomy after their first assault. The mean interval between first LSA-V and hysterectomy was 22 years (SD, 9.6; median, 22; range, 2-43). There may have been other assaults during the time between this first assault and hysterectomy.
Table 4 provides odds ratios for hysterectomy by time period of the first completed sexual assault with vaginal penetration. Female veterans sexually assaulted in childhood were more likely to report a hysterectomy than those first assaulted in adulthood outside military service. After adjusting for age, female veterans first sexually assaulted during military service were also more likely to report hysterectomy than those first assaulted in adulthood outside military service.
