Hysterectomy Prevalence by Hispanic Ethnicity: Evidence From a National Survey

Abstract

Objectives. We investigated hysterectomy prevalence among Hispanic women.

Methods. We obtained data from 4684 Hispanic women and 20 604 non-Hispanic White women from the 1998–1999 National Health Interview Survey. We calculated nationally representative odds ratios of previous hysterectomy, controlling for confounders.

Results. Compared with non-Hispanic White women, the odds ratio for hysterectomy was 0.36 (95% confidence interval [CI] = 0.30, 0.44) for Hispanic women with no high school diploma, 0.57 (95% CI = 0.44, 0.74) for high school graduates, and 0.67 (95% CI = 0.42, 0.87) for college attenders. Country of origin had little influence on hysterectomy prevalence. Hysterectomy was positively associated with acculturation.

Conclusions. Hispanic women undergo fewer hysterectomies than do non-Hispanic White women. The reasons for this, as well as information on ethnicity-specific appropriateness of hysterectomy, should be explored.

Approximately 633 000 hysterectomies were performed in the United States in 2000.1 As the most frequent nonobstetric surgical procedure among women in the United States,2 hysterectomy results in more than 1 in 3 US women undergoing major surgery by age 60.3 The US hysterectomy rate is among the highest in the developed world.3–5

Many factors seem to influence hysterectomy within the United States. Biological factors such as cancer and the presence of uterine leiomyoma are strong predictors of the procedure.5–7 Body mass index8 and parity7–9 also are strongly associated with hysterectomy. Hysterectomy prevalence has also been shown to differ significantly along a number of sociodemographic lines, including region of residence,2,3 education level,7–11 income,9,11,12 and occupation.9 This pattern suggests that a number of factors influencing hysterectomy prevalence are unrelated to medical necessity.

Race is another sociodemographic variable that has been investigated as a risk factor for hysterectomy, although no consistent associations have been found. Some studies present no significant differences between Black women and White women.7,13 Others indicate a greater rate of premenopausal hysterectomy among Blacks, owing largely to the higher prevalence of uterine leiomyoma among Black women.14

Despite a recent qualitative study of hysterectomy among several ethnic groups, including Hispanic women,15,16 little is known about either the prevalence of hysterectomy among Hispanics or the factors associated with hysterectomy in this subpopulation. A study using national data from 1988 found nearly identical hysterectomy rates for Blacks, Whites, and Hispanics.11 Nationally representative survey data obtained from women younger than 45 years revealed that 3.6% of Hispanic women had had a hysterectomy, compared with 5.2% of non-Hispanic White women.17 An ecological study of Los Angeles County found that zip code areas with higher proportions of Latino residents had higher rates of hysterectomy.18 However, no other published US reports have included an analysis by Hispanic ethnicity.

Overall use of medical care by Hispanics is lower than the national average.19 In general, recent immigrants have been found to make fewer medical care visits than native-born Americans.20 Hispanic persons tend to lack a usual source of medical care, even after controlling for health care coverage.19,21 Language barriers and the concepts of fatalismo (that everything is predestined) and familismo (that important decisions are made with an eye toward how the outcome of the decision will affect family members) have been postulated to affect hysterectomy among certain Hispanics in the United States.22 However, studies of Hispanic women’s use of general health services,19 as well as of mammography and Papanicolaou tests,23–25 have found little difference by acculturation after control for other socioeconomic variables.

The purpose of our article is to reduce the information gap regarding hysterectomy prevalence among Hispanic women. This gap is at least partially a result of the paucity of accurate data on ethnicity collected in surveys. National surveillance of hysterectomy conducted by the Centers for Disease Control and Prevention uses as its source of data the National Hospital Discharge Survey,2 in which ethnicity information is left blank approximately 70% of the time (R. Pokras, personal communication, February 15, 2001). Other studies have not included enough Hispanic women in their samples to evaluate this ethnic group separately. Given the rapidly expanding size of the US Hispanic population,26 it is important for health planners to know the prevalence of hysterectomy in this group. Not only are potentially large numbers of surgeries and associated costs involved; some women may not be receiving necessary medical treatment that could prevent hysterectomy.

METHODS

Data

Data for our analysis came from the 1998 and 1999 National Health Interview Survey (NHIS), a multipurpose health survey gathering representative information on the health and health care utilization of the civilian noninstitutionalized US population.27,28 Interviews have been conducted continuously since the survey began in 1957. Beginning in 1995, a new sampling design was implemented, which oversampled both Hispanic and Black households so that researchers using NHIS data could derive precise estimates for these subpopulations.

The survey consists of both an annual “core” set of questions and additional sets of questions that are asked either periodically or on an as-needed basis. In the 1998 survey’s Sample Adult Prevention supplement, information on preventive health behaviors was obtained from a subset of approximately 32 000 adults (aged 18 years and older). This section included a question regarding whether a woman had ever had a hysterectomy. The response rate for this supplement was 73%.27 In 1999, the question regarding hysterectomy was included in the Sample Adult Core questionnaire. This section of the NHIS was given to approximately 31 000 adults. The response rate was 70%.28

The main independent variable of interest, Hispanic ethnicity, was obtained by asking the respondent whether any of a list of Hispanic groups (e.g., Chicano, Puerto Rican) represented the person’s national origin or ancestry. Any positive response, even if the respondent did not know or refused to provide a specific subgroup, was used to define the respondent as Hispanic. We also analyzed specific type of Hispanic origin in this investigation. The groups were designated “Puerto Rican” (n = 558), “Mexican/Mexican American/Chicano” (n = 1963), “Cuban/Cuban American” (n = 626), and “Other Latin American” (n = 804). Respondents who were classified as Hispanic but who could not be assigned to a specific group owing to lack of additional information (n = 733) were omitted from analyses of Hispanic subgroup.

Two data items were used to measure acculturation among Hispanic women. Information was collected on whether the survey was conducted in English, Spanish, or some other language. Hispanic women interviewed in either Spanish or in both Spanish and English were classified as being less acculturated than Hispanic women interviewed in English only. Information was also collected on both place of birth and length of time spent living in the United States for those not born there. Hispanic women were categorized into 3 groups: born on foreign soil and living in the United States for less than 10 years, born on foreign soil and living in the United States for 10 or more years, and born in the United States.

Several characteristics previously associated with hysterectomy were analyzed as potential confounders. These covariates consisted of both socioeconomic variables and variables measuring health characteristics and health care access. The sociodemographic characteristics included in the analysis were age at interview (25–44 years, 45–64 years, 65 or more years), level of education (no high school diploma, high school diploma or General Education Development diploma, at least some college), marital status (ever married, never married), family income relative to the US poverty line (< 200%, ≥ 200%), and region of current residence (Northeast, Midwest, South, West). Health characteristics included weight for height (body mass index [BMI; weight in kilograms by height in meters squared] < 25.0, BMI ≥ 25.0), whether the respondent had a usual source of medical care (yes, no), and self-reported health status (excellent/very good/good, fair/poor). Because having a usual source of care is so important for many types of service utilization,29,30 and because this factor is lower among Hispanic populations than among non-Hispanic Whites,19 it was also included in this analysis. Finally, self-reported health was included, because a large majority of gynecologic surgery is performed for nonmalignant conditions to improve quality of life.31

Analysis Data Set

Of the 26 685 non-Hispanic White women and Hispanic women aged 25 years and older who completed the 1998 to 1999 interview, the 636 women (2.4%) who did not answer the hysterectomy question were excluded. Of the remaining women, 761 (2.9%) had to be eliminated because of missing values for 1 or more of the confounders. An additional 107 Hispanic women (2.3%) were dropped only from the analysis by level of acculturation, owing to missing information on survey language or birthplace.

Analytic Methods

We compared the proportions of non-Hispanic White women and Hispanic women in each potential covariate category. Using the direct method, we adjusted these proportions to the age structure of the full analysis data set. We then constructed 3-way contingency tables, along with stratified odds ratios (ORs), to assess the strength of the association between ethnicity and hysterectomy status, controlling individually for each of the covariates. We considered any variable to be a potential effect modifier that showed substantial differences in the strength of association between ethnicity and hysterectomy by its different levels. We then incorporated all variables of interest in logistic models, retaining all potential interactions. We dropped interaction terms from the model that were not statistically significant. We also dropped covariates that were not significantly associated with hysterectomy and that did not appreciably change the odds ratio between ethnicity and hysterectomy.

We conducted additional analyses by dividing the Hispanic women by level of acculturation. This set of analyses took the final model from the first set of analyses as its starting point. We modified the model to increase the precision of the estimates by eliminating factors with small sample sizes and those that made little difference in the overall estimate.

The NHIS sample is designed to be representative of the US population. We analyzed these data with the sample weights so that the findings could be reported in terms of the US population. All analyses were conducted with SUDAAN software (Version 7.0; Research Triangle Institute, Research Triangle, NC) to produce accurate estimates of sampling error.

RESULTS

Based on the 1998–1999 NHIS data, an estimated 23.3% of all non-Hispanic White women aged 25 years and older had undergone a hysterectomy. In contrast, only 12.2% of all Hispanic women aged 25 years and older had undergone a hysterectomy (Table 1 ▶). The age-adjusted odds ratio for hysterectomy in Hispanic relative to nonHispanic White women was 0.58 (95% confidence interval [CI] = 0.52, 0.65).

TABLE 1.

—Percentage Distribution of Hysterectomy Prevalence and Covariates, by Ethnicity: United States, 1998–1999

	Hispanic (n = 4684)	White (n = 20 607)
Hysterectomy
Yes	12.2*	23.3*
Age, y
25–44	59.8*	43.5*
45–64	27.8	33.7
≥ 65	12.4	22.8
Education
No high school diploma	44.2*	13.5*
High school graduate	23.3	33.0
Some college or higher	32.5	53.4
Marital status
Never married	16.0*	11.0*
Region
Northeast	16.9*	21.3*
Midwest	7.4	28.6
South	34.4	33.5
West	41.3	16.7
Family income
< 200% of federal poverty level	54.8*	23.5*
BMI
Overweight (BMI ≥ 25.0)	60.5*	48.0*
Self-reported health status
Fair/poor	17.1*	11.6*
Usual source of care
No	16.2*	7.3*

Note. Sample size is the number of women respondents included in the 2 categories. However, the proportions do not refer to simple proportions of the sample, but are weighted to represent the US population. BMI = body mass index. The distribution of every factor significantly differed by ethnicity at the P < .01 level.

All of the variables of interest were highly associated with ethnicity (Table 1 ▶). Compared with non-Hispanic White women, Hispanic women in the United States tended to be young, to have fewer years of education, to be more likely to have never been married, and to be more than twice as likely to live in a household with a family income below 200% of the federal poverty level. In addition, a higher proportion of Hispanic women than of nonHispanic White women reported being in fair/poor (vs excellent/very good/good) health, and a lower proportion had a usual source of health care.

The associations between hysterectomy and most of the covariates were similar for both non-Hispanic White women and Hispanic women. There were a few exceptions, however. After adjustment for age, a weak inverse association between years of education and hysterectomy was found among Hispanic women, whereas among non-Hispanic White women there was a strong inverse association. Similarly, a strong association between marital status and hysterectomy was found among non-Hispanic White women, compared with a much weaker association among Hispanic women. The effect of poverty level on hysterectomy also differed by ethnicity. Whereas family income was positively associated with hysterectomy among Hispanic women, it was inversely associated with hysterectomy among non-Hispanic White women.

Based on a review of the preliminary analyses, the initial logistic models included all of the potential covariates, as well as 2-way interaction terms of ethnicity and education, marital status, and family income. Because the interactions with family income and marital status were not significant when all of the other variables were included in the model, they were eliminated.

A significant gradient was observed in the association between ethnicity and hysterectomy by education in the final model (Table 2 ▶). Among women with less than a high school education, the odds ratio for hysterectomy prevalence among Hispanic vs non-Hispanic White women was 0.36. With increasing level of educational attainment, the difference in hysterectomy prevalence by ethnicity (after control for all other factors) became smaller.

TABLE 2.

—Adjusted Odds Ratios (ORs) and 95% Confidence Intervals (CIs) for Hysterectomy Prevalence, by Hispanic Ethnicity and Education: United States, 1998–1999

	Hispanic Women		Non-Hispanic White Women
Education level	OR	95% CI	OR	95% CI
Some college or higher	0.67	(0.42, 0.87)	0.73	(0.65, 0.83)
High school graduate	0.57	(0.44, 0.74)	0.92	(0.81, 1.04)
No high school diploma	0.36	(0.30, 0.44)	1.00a	…

Note. CI = confidence interval; OR = odds ratio.

^aReferent group.

“Hispanic” is an umbrella category containing groups of people from very diverse backgrounds. For that reason, our final model compared each of the survey data–identifiable Hispanic groups with the group non-Hispanic White women. To obtain sufficient power, the education variable was reduced to 2 levels: (1) greater than a high school education or (2) a high school diploma or less. Two different patterns emerged from these analyses (Table 3 ▶). In the models for Puerto Rican and Cuban/Cuban American women, there was no appreciable interaction between ethnicity and education, whereas this interaction was found to be fairly strong in the models for Mexican/Mexican American/Chicano and Other Latin American women. Among Mexican/Mexican American/Chicano women and Other Latin American women at the lower level of education, hysterectomy prevalence was about one third that of non-Hispanic White women; among women in those 2 groups with at least some college education, hysterectomy prevalence was two thirds that of non-Hispanic White women. However, among Cuban/Cuban American and Puerto Rican women, prevalence of hysterectomy was approximately half that of non-Hispanic White women, regardless of education level.

TABLE 3.

—Adjusted Odds Ratios (ORs) and 95% Confidence Intervals (CIs) for Hysterectomy Prevalence, by Hispanic Subgroup and Education: United States, 1998–1999

Ethnicity	OR	95% CI
Mexican/Mexican American/Chicano (n = 1963)
No college education	0.36	(0.29, 0.44)
At least some college education	0.66	(0.53, 0.83)
Other Latin American (n = 804)
No college education	0.31	(0.24, 0.45)
At least some college education	0.59	(0.37, 0.93)
Cuban/Cuban American (n = 626)a	0.58	(0.43, 0.71)
Puerto Rican (n = 558)a	0.57	(0.42, 0.78)

Note. CI = confidence interval; OR = odds ratio.

^aEducation was not a significant interaction variable for this group, so separate estimates by education are not presented.

In the analysis of hysterectomy by the proxy measures of acculturation, language of survey and birthplace, bivariate analyses revealed small numbers of Hispanic women in several regions and among the upper levels of education. Region of residence, although a strong predictor of hysterectomy, was not a strong confounder in the association with ethnicity and so was dropped from these models. We also continued to divide education into only 2 levels for this subanalysis, as was done in the investigation by Hispanic subgroup.

Overall, hysterectomy was positively associated with level of acculturation as measured by the combined acculturation variables (Table 4 ▶). Hispanic women born outside the United States who had lived in the United States less than 10 years were least likely to have had a hysterectomy, whereas Hispanic women born in the United States had a hysterectomy prevalence closer to that of non-Hispanic White women. Also, hysterectomy prevalence was lower among US-born Hispanic women who were interviewed in Spanish, compared with US-born women who were interviewed solely in English. Because of the small sample sizes, the education interaction was not included in this model.

TABLE 4.

—Adjusted Odds Ratios (ORs) and 95% Confidence Intervals (CIs) for Hysterectomy Prevalence, by Level of Acculturation and Nativity: United States, 1998–1999

	OR	95% CI
Combined Acculturation Variables
Hispanica: Spanish used in interview
Foreign-born, < 10 years in United States	0.29	(0.17, 0.49)
Foreign-born, ≥ 10 years in United States	0.38	(0.30, 0.48)
US-born	0.47	(0.27, 0.81)
Hispanica: only English used in interview
Foreign-born, < 10 years in United States	0.07	(0.01, 0.35)
Foreign-born, ≥ 10 years in United States	0.60	(0.47, 0.77)
US-born	0.71	(0.60, 0.84)
Separate Analyses by US Nativity
Foreign-born Hispanica
Education: high school or less	0.38	(0.32, 0.46)
Education: at least some college	0.59	(0.43, 0.81)
US-born Hispanica	0.69	(0.59, 0.82)

Note. CI = confidence interval; OR = odds ratio.

^aReferent group is non-Hispanic White (OR = 1.00).

We constructed separate models for 2 Hispanic subgroups, US-born and foreign-born, both of which were compared with nonHispanic White women (Table 4 ▶). Most of the coviariates had very similar associations with hysterectomy in the 2 models. The exception was education, for which a significant interaction with ethnicity was found only in the model comparing foreign-born Hispanic women with non-Hispanic White women. Among foreign-born Hispanic women, hysterectomy was 1.5 times lower among those with high school education or less, compared with non-Hispanic White women, whereas foreign-born Hispanic women with at least some college were 70% less likely than non-Hispanic White women to report having had a hysterectomy after all other factors had been controlled. Among the US-born Hispanic women, hysterectomy prevalence was lower than that among White women, although not as substantially.

Separate models were also constructed for Hispanics who were interviewed in Spanish and those who were interviewed in English. We obtained virtually the same findings as in the birthplace analysis.

DISCUSSION

On the basis of this analysis, we conclude that Hispanic women are much less likely to undergo hysterectomy than are non-Hispanic White women, although the association is weakened with increasing levels of education. Furthermore, the difference in prevalence of hysterectomy by ethnicity decreased with increasing acculturation among Hispanic women. Most of the previously identified covariates of hysterectomy were confirmed in this data set.

Many characteristics, such as low education, low income, and poor health status, were found at a higher prevalence among Hispanic women than among non-Hispanic White women. Therefore, comparing the prevalence of hysterectomy by ethnicity11 does not yield any information regarding how much of the difference is explained by ethnicity alone. The associations between sociodemographic characteristics and hysterectomy that have been found among White women were not found with Hispanic women. If those associations had been the same, one would expect hysterectomy rates among Hispanic women to be even higher than those of non-Hispanic White women.

It has been argued that the use of medical procedures such as hysterectomy by Hispanic women may be influenced by their degree of acculturation. Language barriers and cultural definitions of health and illness affect health behaviors of Hispanics in the United States.22,33 It has also been noted that some Hispanic women report feeling less comfortable being examined by medical professionals of the opposite sex, and more embarrassed during an examination, than do non-Hispanic White women.34 Although the NHIS data do not provide very specific information on acculturation, we did find a striking difference in hysterectomy by level of acculturation as measured by our proxy acculturation variables (country of birth and language of interview).

One of the main limitations of the NHIS data in the analysis of hysterectomy is that the data are obtained cross-sectionally by self-report. Thus, we do not know how long ago a respondent underwent her hysterectomy, nor do we have medical records verifying the procedure. Hysterectomy seems to be fairly accurately self-reported, however.35 The information collected on covariates during the survey may be very different from the values of these items at the time of the hysterectomy. Some variables, such as education, probably are fairly stable over time. Other variables, such as income, health status, and age, may have changed a great deal. There is no way to know how much this limitation of the data affected the results of the analysis. All of the associations between hysterectomy and the covariates were in the same direction as would be expected, at least among nonHispanic White women. Unless the status of these variables changes more often among Hispanic women than among White women, we can assume that the findings are probably robust, although confounding may not have been fully controlled.

The cross-sectional nature of the survey limited our ability to use several potentially important confounders. In this analysis, we found no significant association between household income and hysterectomy, most likely owing to the dynamic nature of this variable. Thus, some confounding by these variables probably remained in the association with ethnicity reported. Such confounding could only be addressed by using a different data source that collects information at the time of hysterectomy.

Some variables considered to be important predictors of hysterectomy—for example, parity and age at first birth7,8—were not available in the NHIS data set. We explored this limitation by conducting a subanalysis of women aged 45 years and older, thereby excluding those women we knew had had a hysterectomy earlier in life, when parity strongly influences hysterectomy prevalence.17 Although this subanalysis only partially controlled for the effect of parity on hysterectomy prevalence, we found almost the exactly same effect of ethnicity among women aged 45 years and older as we did with women aged 25 years and older.

Although the NHIS data are not collected specifically to address the question of hysterectomy prevalence among Hispanic women, their strength stems from the large sample size and oversampling of Hispanic households. Furthermore, because of the nationally representative nature of the survey, the association found in this analysis is not a result of a phenomenon specific to a certain geographic locale or Hispanic subpopulation. When the analyses were conducted by Hispanic subgroup, we still found a significantly lower prevalence of hysterectomy compared with non-Hispanic White women.

In general, this analysis indicates that Hispanic women tend to undergo hysterectomy infrequently, with much smaller differences by covariates than are seen in non-Hispanic White women. This analysis also suggests that the size of the difference in the prevalence of hysterectomy lessens as acculturation increases among Hispanic women. US rates of hysterectomy, which in general is a treatment for conditions that are not life-threatening, are much higher than rates found in most other industrialized countries.4,5 Thus, our results may point to an overuse of the procedure among non-Hispanic White US women, rather than to an underuse of the procedure among Hispanic women. Education and income, in general, tend to be inversely associated with hysterectomy among White women,7–12 possibly indicating that this procedure is used by women for whom other, less invasive forms of medical care are not exploited. It has been speculated by Bunker36 that performing hysterectomies on less educated women may be more expedient than educating them. On the other hand, Hispanic women may be more tolerant of chronic physical discomfort than are non-Hispanic White women, or Hispanic women who are more removed from the health care sector may not be receiving the care they truly need and deserve; the quality of life of women who undergo hysterectomy is reported to improve after surgery.37–39

Because previous research has found a high degree of inappropriate recommendations for hysterectomy,40,41 a next step in this research would be to look at the medical necessity of hysterectomy among Hispanic and non-Hispanic women and assess any differences. This avenue of research could also investigate numbers and types of treatments used before performing hysterectomy within each ethnic/racial group. Additional studies to investigate the specific reasons for lower hysterectomy prevalence among Hispanic women are also warranted.