Recent attention to the high rate of hysterectomies in the United States has led to recommendations to assist clinical decision making. Of the 590,000 hysterectomies done annually, approximately 90% are in response to nonmalignant, nonemergency conditions,1 presumably to improve the woman's quality of life. Unfortunately, most quality-of-life studies combined subjects according to surgical procedures rather than analyzing them by their presenting complaints.2–8 Failure to report on differences in quality of life among various gynecologic symptoms that indicate the need for hysterectomy might obscure important differences in impairment. There is general agreement that mood or psychiatric disorder, sexual function, bleeding, pain, fatigue, perception of health, and interference of symptoms with daily activities are important aspects of quality of life in women with gynecologic conditions that might lead to hysterectomy.2–8 However, studies often measure only one of those areas2,5 or assess multiple domains but do not compare subject samples.3,4,6–8 Despite those limitations, studies showed that many women have high levels of impairment before hysterectomy and others do not. For example, the Maine Women's Health Study6,7 found that about half of the women studied reported fatigue in the month before hysterectomy, whereas more than a third seldom or never reported fatigue. That wide range of impairment supports grouping women in a more refined way.
An analysis according to clinical symptoms might explain the wide variation in function among women who eventually have hysterectomies. Although the Maine Women's Health Study7 grouped women into four categories according to their primary preoperative conditions (leiomyomata, abnormal bleeding, chronic pelvic pain, and other), no statistical comparisons between diagnostic categories were conducted. The comparisons reported were within each group, comparisons of pre- to postsurgery differences in quality of life, and between groups of women treated surgically or medically.7 Mathias et al4 grouped women according to cause of pelvic pain and found that women with pelvic pain due to gynecologic conditions were significantly more impaired than women with pelvic pain due to nongynecologic conditions, but the data were not analyzed at a condition-specific level.
The present study was conducted as part of the Women's Health and Hysterectomy Project, a quality improvement–demonstration project evaluating clinical practice recommendations for hysterectomies in capitated medical groups (in which a fixed payment was made for each woman's treatment during a month or year). As part of the baseline assessment, women who had had hysterectomies in the previous 12 months were interviewed by telephone. The primary goals of that component of the study were to develop and validate statistically a set of scales to measure quality of life in women who had hysterectomies, to measure the degree of impairment experienced by women before hysterectomy across nine common gynecologic conditions, and to examine differences in quality of life among conditions categorized by their primary symptoms (pain, bleeding, pelvic discomfort, or no symptoms).
Materials and Methods
Nine capitated medical groups in Southern California agreed to participate in the study. From the member lists of those organizations, we identified 1089 women age 25 years and older who had had hysterectomies between August 1, 1993, and July 31, 1995. We excluded 310 women because the procedure codes for their surgeries did not meet our criteria (ie, International Classification of Diseases codes 68.3 to 68.8. International Classification of Diseases codes index hospital records by diseases and procedures); surgery was done urgently or for a previously diagnosed cancer; or women were non–English speaking, cognitively impaired, or dead. We obtained informed consent from the remaining 779 women. Five hundred thirty-nine (69%) agreed to be interviewed, 538 completed interviews, and 500 (64%) were included in the study, on the basis of availability of medical records for abstraction. Those 500 women completed telephone interviews (done by a trained nurse using a prepared script) within 12 months (median 9) of their hysterectomies. For this report, 18 women were excluded because their conditions could not be categorized into one of the nine study conditions. The final study sample was 482 women. The study was approved by the RAND Institutional Review Board.
Interviews consisted of 44 questions covering demographics, gynecologic and medical histories, symptoms 3 months before surgery, quality of life, and participation in decisions on treatment of identified gynecologic conditions. The majority of questions were based on two sources: an interview developed by Kjerulff et al for the Maryland Women's Health Study (personal communication) and the Medical Outcomes Study Short Form-36.9
We developed the scales to satisfy traditional psychometric criteria.10 To create a set of scales applicable to all women, not only the employed, scores on three items relevant only to employed women were imputed (a statistical technique used to estimate missing values) for women who were not employed, using linear regression equations based on the functional scores for three other domains: home, outside, and social-recreational activities. After that, we included all items that assessed symptom severity, change in symptom severity, and quality of life in a principal-components factor analysis using a varimax rotation to determine the natural structure of the data. After a series of iterations, eight factors emerged. These eight factors were examined using the multitrait method, which tested for item convergent and discriminant validity.11 In accordance with recommended statistical criteria, one scale was dropped and two were collapsed into one scale, resulting in six final scales.11
The final scales consisted of scales of role impairment, impairment in sexual function, mood impairment, impairment due to bleeding, change in symptom severity, and general health perception. Table 1 shows the intercorrelations among the six scales, illustrating an appropriately moderate level of relationship among them. We used Cronbach's alpha, which measures the internal consistency among items in a scale, to judge the reliability of our quality-of-life scales. A score of .70 or greater generally is regarded as an acceptable level of reliability. The role-impairment scale consisted of 12 items that measured the number of days women were unable to perform their usual activities around the house or outside the house, social or recreational activities, or work because of pain or discomfort (specifically pain, cramps, or discomfort in the abdominal or pelvic area), bleeding, or fatigue. Cronbach's alpha for that scale is .93. The scale of impairment of sexual function consisted of five items that assessed the degree to which pain or discomfort, bleeding, and fatigue interfered with usual sexual relations and the degree to which enjoyment and interest in sexual activity were compromised in the 3 months before hysterectomy (alpha = .90). Mood was assessed with the five mood items that measured depression and anxiety from the Mental Health Index of the Medical Outcomes Study Short Form-36 (alpha = .87).9 The impairment-due-to-bleeding scale consisted of two items assessing the degree to which bleeding was a problem 3 months before hysterectomy and the degree to which using napkins and tampons was a problem (alpha = .91). The scale of change in symptom severity included three items assessing changes in pain or discomfort, bleeding, and fatigue over the previous 3 months (alpha = .69). The scale of general health perception was a one-item scale adopted from the General Health Index of the Medical Outcomes Study Short Form-36,9 which asked women to give global ratings of their health from excellent to poor. All scales except the scale of role impairment, which was measured in days per month, were converted to 0 to 100–point scales, where 0 represented no impairment and 100 severe impairment.
We reviewed the literature regarding the conditions for which hysterectomy was used, its efficacy, and risks by conducting a MEDLINE search for the period 1970 through April 1994.12 From that review and discussions with clinicians, we developed a list of 2332 potential indications, or specific clinical scenarios, in which nonemergency, nononcologic hysterectomy might be considered. The indications then were grouped into nine clinical conditions: endometriosis, pelvic pain (pelvic pain and adhesion, dysmenorrhea, and chronic noncyclic pelvic pain), leiomyomata with bleeding, leiomyomata without bleeding, leiomyomata in postmenopausal women, abnormal uterine bleeding, pelvic relaxation (with or without incontinence or pain), cervical dysplasia, and unilateral adnexal mass. Women were assigned to diagnostic categories on the basis of clinical data in their medical records. Using a standardized form, trained abstractors examined inpatient and outpatient medical records, operative reports, laboratory results, and pathology reports for each case. When a woman could be assigned to more than one diagnostic category (eg, when a woman had pelvic pain and cervical dysplasia), we defined the primary condition as the one for which hysterectomy would be most appropriate, based on an expert panel's rating of appropriateness for the indication that corresponded to the woman's condition, menopausal status, degree of impairment from symptoms, diagnostic work-up, and medical and surgical treatment.12
We then placed women from those nine clinical categories into four symptom-based groups: pain, bleeding, pelvic discomfort, and asymptomatic groups. Women who presented with endometriosis and pelvic pain were placed in the primary pain group. Women with abnormal uterine bleeding or leiomyomata with bleeding were placed in the primary bleeding group. Women with pelvic relaxation, leiomyomata without bleeding, and postmenopausal leiomyomata were placed in the primary pelvic discomfort group because the symptoms most commonly reported by those women were pelvic discomfort and related urinary problems. Women with unilateral adnexal mass and cervical dysplasia were categorized as primarily asymptomatic.
Analysis of variance conducted for each scale detected significant variance among the four groups for all six scales (P < .05). To detect specific pair-wise differences among the four groups (eg, significant differences between pain and bleeding conditions for the level of role impairment), analyses were conducted using the Tukey test for pair-wise comparisons with the Tukey-Kramer method to correct for unequal sample sizes. Scale means and standard deviations for the original nine clinical subgroups can be obtained from the authors. Two-tailed t tests were used to compare the mood of women in our study with the mood of patients with other medical conditions.
Subjects were primarily white, employed, married women who had at least some college education (Table 2). Most had at least one child and were premenopausal at the time of surgery. As expected, subjects in the primary pain group were generally younger than those in the other three groups, followed by the subjects in the primary bleeding group, whose ages clustered between 35 and 54 years. Women in the primary pain and bleeding groups, nearly all of whom were premenopausal, were more likely to be employed full time, be married, have completed some college, and have fewer children than were women in the primary pelvic discomfort and asymptomatic groups.
Bleeding (25%), pain or discomfort (28%), fatigue (8%), and mood impairment (8%) were the main ways everyday life was affected before hysterectomy. Six percent of women reported their conditions did not affect everyday life. Ninety-four percent of women reported they understood all of the medical and surgical treatment options for their conditions. Eighty-nine percent of women thought their decisions for surgery matched what they wanted (to some extent) and said they made the final decision to have a hysterectomy alone (43%). The most common primary condition in the study population was leiomyomata (61%), followed by pelvic relaxation (11%) and abnormal uterine bleeding (9%). Each of the other conditions accounted for less than 5% of the total sample.
Quality of life varied significantly among the four major groups (Table 3). In general, subjects in the primary pain group had the highest levels of impairment, women in the primary bleeding group had the second highest levels, and the primary pelvic discomfort and asymptomatic groups experienced relatively low levels of impairment. Women with primary pain conditions had the greatest levels of role, sexual, and mood impairment; the poorest perceptions of general health; and the greatest increase in severity of symptoms before hysterectomy compared with women with primary bleeding, pelvic discomfort, or asymptomatic conditions (P < .05 for all comparisons). The primary bleeding group scored significantly higher on the impairment-due-to-bleeding scale than did any of the other groups (P < .05 for all comparisons). Compared with the primary pelvic discomfort and asymptomatic groups, the primary bleeding group also had significantly greater role and sexual impairment and a significantly greater increase in symptom severity before hysterectomy (P < .05 for all comparisons). The asymptomatic group scored as poorly on the scales of mood impairment and perception of general health as the primary bleeding group (P < .05 for both comparisons). The primary pelvic discomfort group was less impaired on every scale than the primary bleeding and pain groups (P < .05 for all comparisons) and showed less mood impairment and better general health perception than the asymptomatic group (P < .05 for both comparisons).
To understand the quality of life of women before hysterectomy in a larger context, we compared the emotional well-being (ie, mood) of our subjects with that of women with other medical and psychiatric conditions (Table 4).13 Each of the four gynecologic groups scored significantly worse on the emotional well-being index than women with hypertension, diabetes, heart disease, and arthritis (P < .05 for all comparisons). Women with primary pain or primary bleeding conditions scored significantly worse than women with chronic lung problems (P < .05) or panic disorder (P < .05). Women with primary pain conditions scored significantly worse than those with depression (P < .05).
Previous research found that women's function and quality of life are affected by gynecologic conditions, but reliable and valid measures of symptom effect have been lacking. The development of the six quality-of-life scales in the present study was a useful step toward the goal of obtaining such measures, and our results confirmed previous findings of significant and pervasive impairment in women who had hysterectomies for benign gynecologic conditions.3,6,7 Unlike previous investigations, the current study focused on differences in severity and breadth of impairment in relation to presurgical diagnostic profiles and, more generally, on the primary symptoms underlying presurgical conditions. We found that women with endometriosis or pelvic pain had substantially greater impairment than did women with other conditions that led to hysterectomies. A national survey of women with chronic pelvic pain also found severe impairment across an array of quality-of-life indices.4 Despite its severe impact on quality of life, relatively little attention has been given to chronic pelvic pain.4 Most research focused on conditions characterized by abnormal bleeding,8 for which hysterectomy was an effective treatment. Results of our study, along with those of the study by Mathias et al,4 underscore the strong need to extend research on pelvic pain conditions, particularly endometriosis.
Women with leiomyomata and bleeding or abnormal uterine bleeding also are impaired, although less so than women with pain conditions. Women in the pelvic discomfort group with nonbleeding leiomyomata or pelvic relaxation had the highest quality of life of any group, even higher than that of women with primarily asymptomatic conditions. In cases in which quality of life is high and the likelihood of disease-related fatality is low, woman and physician assessments of quality of life might lead to less invasive and costly treatments. The number of hysterectomies done in the United States has come into question, so the relationship between women's preferences and physician recommendations has gained importance.14 Professional uncertainty about appropriate indications for hysterectomy might be an important cause of the variation in rates.1 With a standardized quality-of-life instrument for clinical settings, physicians could compare candidates for hysterectomies and more accurately gauge effects of women's symptoms on quality of life. Quality-of-life measures also might ease patient-physician discussions about alternative treatments, providing a framework to help women consider the costs and benefits of various options. For example, in a recent study of decision making,8 many women with abnormal uterine bleeding were quite satisfied with the reduction of symptoms from medication. Presurgical quality-of-life assessment might help women and physicians judge effectiveness of treatment.
Although the results of the present study encourage continued development and implementation of standardized quality-of-life measures, results must be considered cautiously because of the composition of the sample and the retrospective nature of the data. Compared with national statistics, a larger proportion of women in our sample had indications for hysterectomies related to leiomyomata (61%) and fewer had endometriosis (4%) or chronic pelvic pain (4%).1 Those differences might result from classification criteria: We assigned the indication for which hysterectomy was rated most appropriate as the primary indication, which might have put larger numbers of women in certain categories than other classification schemes.
Our study addressed the quality of life of women during the 3 months before hysterectomy, measured retrospectively through subject interview. Subject reports of preoperative symptoms obtained retrospectively sometimes differ from contemporary reports at the individual level, although there is evidence that similar results are obtained at the group level,15 especially when scale scores are used.16 Self-reporting of number of work days missed has been found to be reliable and accurate when numbers have been validated by administrative records,17 but the recall interval used here exceeds those previously studied. Therefore, these retrospective reports of quality of life are less reliable than contemporaneous reports would be, and they might have been influenced by postsurgical outcomes. For example, a subject with chronic pain might have an especially poor outcome that could lead to a negatively biased perception of preoperative quality of life. However, the large differences found among groups in our study were consistent with the findings of the Maine Women's Health study7 and the study by Mathias et al,4 supporting results that reflect real differences in presurgical status rather than group differences in forgetting or recall bias. Replication with prospective data is critical in establishing reliable estimates of levels of impairment that can be used for clinical decision making.
Our results pertain to those women who had hysterectomies; women with the same gynecologic conditions who have alternative treatments might have different quality of life before and after treatment. Generalization also might have been limited by lack of more stringent validation of preoperative diagnoses. Prospectively collected data on women with gynecologic conditions clearly diagnosed in compliance with established criteria are needed to examine fully the usefulness of the measures in this study for a broader population. Such studies also need to measure and control for demographic differences and nongynecologic factors that might influence perceptions of health-related quality of life.18
Measuring quality of life in women with gynecologic conditions is critical to understanding the full impact of these conditions. Disease states are important only insofar as they cause decreased longevity or impaired quality of life.19 Measures of quality of life have been developed for many conditions, including some that impose a smaller disease burden on those affected than the gynecologic conditions we studied. Sound measures of quality of life that are specific to gynecologic conditions should make possible better assessments of appropriateness and outcomes of treatments.
1. Carlson KJ, Nichols DH, Schiff I. Indications for hysterectomy. N Engl J Med 1993;328:856–60.
2. Helstrom L, Lundberg PO, Sorbom D, Backstrom T. Sexuality after hysterectomy: A factor analysis of women's sexual lives before and after subtotal hysterectomy. Obstet Gynecol 1993;81:357–62.
3. Clarke A, Black N, Rowe P, Mott S, Howle K. Indications for and outcome of total abdominal hysterectomy for benign disease: A prospective cohort study. Br J Obstet Gynaecol 1995;102:611–20.
4. Mathias SD, Kuppermann M, Liberman RF, Lipschutz RC, Steege JF. Chronic pelvic pain: Prevalence, health-related quality of life, and economic correlates. Obstet Gynecol 1996;87:321–7.
5. Kjerulff KH, Langenberg PW. A comparison of alternative ways of measuring fatigue among patients having hysterectomy. Med Care 1995;33(4 suppl):AS156–63.
6. Carlson KJ, Miller BA, Fowler FJ Jr. The Maine Women's Health Study. I. Outcomes of hysterectomy. Obstet Gynecol 1994;83:556–65.
7. Carlson KJ, Miller BA, Fowler FJ Jr. The Maine Women's Health Study. II. Outcomes of nonsurgical management of leiomyomas, abnormal bleeding, and chronic pelvic pain. Obstet Gynecol 1994; 83:566–72.
8. Geller SE, Bernstein SJ, Harlow SD. The decision-making process for the treatment of abnormal uterine bleeding. J Womens Health 1997;6:559–67.
9. Ware JE Jr, Sherbourne CD. The MOS 36-item short-form health survey (SF-36). I. Conceptual framework and item selection. Med Care 1992;30:473–83.
10. Ware JE Jr. Standards for validating health measures: Definition and content. J Chronic Dis 1987;40:473–80.
11. Stewart AL, Hays RD, Ware JE Jr. The MOS short-form general health survey: Reliability and validity in a patient population. Med Care 1988;26:724–35.
12. Bernstein SJ, Fiske ME, McGlynn EA, Gifford DS. Hysterectomy: A review of the literature on indications, effectiveness, and risks. Publication no. MR-592/2-AHCPR. Santa Monica, California: RAND, 1998.
13. Sherbourne CD, Wells KB, Judd LL. Functioning and well-being of patients with panic disorder. Am J Psychiatry 1996;153:213–8.
14. Bernstein SJ, McGlynn EA, Siu AL, Roth CP, Sherwood MJ, Keesey JW, et al. The appropriateness of hysterectomy: A comparison of care in seven health plans. JAMA 1993;269:2398–402.
15. Emberton M, Challands A, Styles RA, Wightman JA, Black N. Recollected versus contemporary patient reports of pre-operative symptoms in men undergoing transurethral prostatic resection for benign disease. J Clin Epidemiol 1995;48:749–56.
16. Guadagnoli E, Cleary PD. How consistent is patient-reported pre-admission health status when collected during and after hospital stay? Med Care 1995;33:106–12.
17. Revicki DA, Irwin D, Reblando J, Simon GE. The accuracy of self-reported disability days. Med Care 1994;32:401–4.
18. Walling MK, O'Hara MW, Vincent SD. Abuse history and chronic pain in women: A multivariate analysis of abuse and psychological morbidity. Obstet Gynecol 1994;84:200–6.
19. Kaplan RM. An outcomes-based model for directing decisions in women's health care. Clin Obstet Gynecol 1994;37:192–206.