For a time during the middle of the 20th century, hysterectomy was the most common major surgical procedure among women in the United States.1 However, by the 1990s, reports of large variations in the rates of hysterectomy across the country gave rise to controversy regarding the inappropriate use of hysterectomy.2 Managed care plans were seen to have an “exceptional opportunity” to manage the care of patients needing hysterectomy.3 New technical and clinical innovations provided improved medical or minimally invasive alternatives to hysterectomy. For example, between 1990 and 1997, the incidence of endometrial ablation for benign uterine conditions increased steadily in six states, reaching levels of 5–10 per 10,000 women per year, while the incidence of hysterectomy simultaneously declined.4 Concurrently, there was a widespread introduction of laparoscopically assisted vaginal hysterectomy (LAVH) for benign uterine conditions throughout the United States5,6 after the initial description of the technique in 1989.7
In this analysis, we studied how the safety of inpatient hysterectomy changed across a large population during these dramatic shifts in hysterectomy technique and use. We also analyzed complications associated with inpatient LAVH during its widespread introduction in California, looking for evidence of a population-wide “learning curve.”
MATERIALS AND METHODS
The State of California has legislatively mandated all nonfederal licensed hospitals to submit coded hospital discharge data to the Office of Statewide Health Planning and Development, and since 1990 such records have included an encrypted record linkage number that allows identification of all hospitalizations for each patient in the California Patient Discharge Database.
We analyzed data from 649,758 women admitted to nonfederal hospitals in California during the period 1991–2004 who underwent hysterectomy for nonmalignant and nonobstetric indications. We identified hysterectomy procedures using codes from the International Classification of Diseases, 9th Revision, Clinical Modification (ICD-9-CM). To identify LAVH, for cases occurring in 1997 and thereafter, we used the ICD-9-CM code for LAVH (68.51). Before the introduction of this code in 1997, we used a previously reported convention8 to identify LAVH cases, that is, when both laparoscopy (54.21) and vaginal hysterectomy (68.59) were coded in the same admission. Other hysterectomy types were identified as follows: vaginal hysterectomy without laparoscopy, total abdominal hysterectomy (TAH) (68.4), subtotal hysterectomy (68.3). The code for total laparoscopic hysterectomy (68.41) was introduced after 2004, and there were no such cases identified within our database. The code for laparoscopic-assisted supracervical hysterectomy (68.31) was introduced in 2003, and these cases were included in the subtotal hysterectomy category unless otherwise noted. The principal diagnosis was used to classify the primary indication for hysterectomy as follows: fibroids (218.0–218.2, 218.9), endometriosis or pelvic pain (617.0–617.6, 617.8, 617.9, 456.5, 568.0, 625.0, 625.1, 625.2, 625.3, 625.5, 789.0, 789.3, 302.76, 789.00, 789.03, 789.04, 789.06, 789.09), prolapse (618.0–618.6, 618.8, 618.9), bleeding (626.2–627.1, 233.2, 621.0, 621.2, 621.3, 621.4), benign adnexal mass (220, 215.6, 221.0, 221.8, 620.0,-620.2, 620.5, 620.8, 620.9, 752.11), infection (039.2, 014.00, 014.03, 014.05, 014.80, 016.66, 016.70, 016.75, 567.2, 567.8, 567.9, 614.0–614.5, 614.7–615.1, 615.9, 098.17, 098.19, 098.86), noninvasive cervical disease (219.0, 233.1, 622.0–622.2, 622.4, 622.6, 795.0), and “other” indications (all other cases in which primary or secondary diagnosis failed to fall into one of the above categories). When the principal diagnosis was one of the bleeding codes but a fibroid code was a secondary diagnosis, the primary diagnosis was reassigned to the fibroid category. Concurrent surgeries were identified as follows: lysis of adhesions (54.5), appendectomy (47.1), oophorectomy/salpingectomy (65.0, 65.3–65.6, 66.4–66.6), vaginal repair (70.1–70.9), and urinary incontinence procedures (59.3–59.7).
Comorbidities were defined using previously described methods9 with secondary diagnosis codes as follows: obesity (278), chronic anemia (280, 282), cardiovascular (401, 402, 411–414, 416, 417, 420–428), pulmonary (491,492,493), gastrointestinal (555, 556, 571), renal (585), and peritoneal adhesive disease (568.0, 614.6).
We identified surgical and medical complications using 900-Series ICD-9-CM as follows: surgical complications (unspecified laceration [998.2], gastrointestinal complication [997.4], urinary tract complication , wound complication [998.3, 998.13], foreign body [998.4, 998.7], bleeding complication or transfusion [998.0, 998.12, 99.0 (blood transfusion)], surgical site infection [998.51, 998.59] and all other 998 surgical complication codes) and medical complications (cardiovascular complication [997.1], pulmonary complication [997.3], central nervous system complication [997.0], peripheral vascular complication [997.2], and other medical complication [997.9, 998.8, 998.9, 999, 960.0, 968.0, 968.9, 979.9, 968.2, 995.3]).
Because rates of conversion from laparoscopy to laparotomy are reported to decline as surgeons acquire experience in operative laparoscopy,10 we analyzed the rate of simultaneous codes for laparoscopy and TAH. We presumed that when both codes were present, conversion was likely to have occurred. We defined the “conversion rate” as the number of conversions (both laparoscopy and TAH codes present) divided by the number of LAVH cases plus the number of conversions. We analyzed “reoperation” (within 30 days of hysterectomy) using codes for exploratory laparotomy (54.14), reopening of recent laparotomy site (54.12), or other laparotomy, including drainage of intraperitoneal abscess or hematoma (54.19).
We estimated hysterectomy incidence rates using census data from the Department of Finance, State of California, for female residents of California age 20 and older. In making these estimates, we assumed that the proportion of women in California who had already undergone hysterectomy was stable during these 14 years, thus we made no attempt to correct for prior hysterectomy.
For statistical analysis, we used SAS 9.1.3 (SAS Institute, Inc., Cary, NC). The Cochran-Armitage test was used to test for time trends in the rates of hysterectomy and associated complications. We constructed multiple logistic regression models to evaluate factors contributing to the risk of medical and surgical complications, including factors known in the literature to be associated with complications. Those factors available in the California Patient Discharge Database included demographic factors, indications for surgery based on discharge diagnosis, concurrent surgeries, comorbidities, and year of surgery for each type of hysterectomy.
This research project was approved by the Human Subjects Committee at the University of California, Davis School of Medicine, by the Office of Statewide Health Planning and Development, and by the State of California Committee for the Protection of Human Subjects.
We found that there were significant reductions in both overall inpatient hysterectomy rates (P<.001) and the odds of medical or surgical complications associated with inpatient hysterectomy between 1991 and 2004 (P<.001). The demographic characteristics of women undergoing hysterectomy for benign indications in California between 1991 and 2004 varied by type of hysterectomy (Table 1). Figure 1A displays the estimated incidence of each type of hysterectomy and the total number of inpatient hysterectomies done for benign indications. Figure 1B displays the proportion of each hysterectomy type. The incidence of LAVH increased 12-fold from 0.3 per 10,000 women per year in 1991 to a peak of 3.6 per 10,000 women per year in 1997, then declined to 3.0 per 10,000 women per year in 2004. There was a significant reduction in the rate of TAH (38.8%, P<.001) and a 17-fold increase in the rate of subtotal hysterectomy during the study interval. During this time, the incidence of any type of inpatient hysterectomy for benign conditions decreased 17.6%. Marked reductions in total surgical and medical complications were observed in the crude frequencies of medical and surgical complications between 1991 and 2004 as shown in Figure 2A and B. Changes in the frequency of hospital readmission by hysterectomy type are shown in Figure 2C. The median length of hospital stay decreased for each type of hysterectomy between 1991 and 2004 as follows: LAVH from 3 to 2 days, vaginal hysterectomy from 3 to 2 days, TAH from 4 to 3 days, and subtotal hysterectomy from 4 to 2 days.
After adjusting for age, race, insurance payer, indication for surgery, comorbidities, and concurrent surgeries over all years, the year of inpatient hysterectomy was associated with a reduction in odds of both medical and surgical complications for each type of hysterectomy (Table 2). Statistically significant increases in the odds of urinary tract complications were observed for LAVH, vaginal hysterectomy, and TAH in the 1995–1996 interval but declined thereafter. Gastrointestinal tract complications also peaked in the mid-1990s for LAVH, vaginal hysterectomy, and TAH and then declined. We found statistically significant declines in the odds of bleeding complications for all hysterectomy types except TAH and in rates of hospital readmission within 30 days for vaginal hysterectomy and TAH. The largest reduction in odds of medical and surgical complications during the study interval was associated with subtotal hysterectomy. Removal of cases of laparoscopic-assisted supracervical hysterectomy (329 cases in 2003 and 1,651 cases in 2004) resulted in negligible changes to the data on subtotal hysterectomy presented in Table 2.
The logistic regression analysis also revealed that certain patient and demographic characteristics were independently associated with medical and surgical complication risk (Table 3). Older age, African-American race, and the presence of comorbidities were associated with increased odds of complications. In fact, African-American race was associated with elevated odds of almost every category of medical and surgical complication, as well as risk of readmission, compared with subjects categorized as non-Hispanic white. In general, a primary diagnosis of fibroids was associated with higher odds than other primary diagnoses, with some exceptions. A primary diagnosis of infection was associated with the highest risk of surgical complications and readmission when TAH or subtotal hysterectomy was performed. Not surprisingly, concurrent vaginal repair or urinary incontinence procedures were associated with the highest risk of urinary tract complications. For example, the odds of urinary tract injury associated with concurrent vaginal repair was elevated for LAVH (odds ratio [OR] 1.46, 95% confidence interval [CI] 1.26–1.69), vaginal hysterectomy (OR 1.55, 95% CI 1.44–1.67), and TAH (OR 1.21, 95% CI 1.14–1.29). Increased odds of urinary tract injury also was associated with concurrent urinary incontinence procedure: LAVH (OR 1.24, 95% CI 1.04–1.48), vaginal hysterectomy (OR 1.34, 95% CI 1.25–1.43), and TAH (OR 1.25, 95% CI 1.18–1.32).
Death during the hysterectomy hospitalization was a rare event, and there were no apparent trends in the death rates. For vaginal hysterectomy, the death rate ranged between 0.01% and 0.05%, TAH between 0.06% and 0.12%, and subtotal hysterectomy between 0.03% and 1.08%. The death rate associated with LAVH was 0.05% in 1997; there were no deaths associated with LAVH in the other years of the study interval.
The rates of presumed conversion of intended LAVH to open TAH are displayed in Table 4 and demonstrate a declining rate of conversion throughout the study interval.
To assess the effect of length of hospital stay on the odds of medical and surgical complications, we adjusted the regression model for length of stay. The results did not show any statistically significant reductions in the risk of medical or surgical complications when comparing 1991–1992 with 2003–2004. For example, between 1991–1992 and 2003–2004, the length-of-stay–adjusted odds of any surgical complication associated with LAVH did not significantly decline (OR 1.05, 95% CI 0.90–1.22) compared with the odds when length of stay was not adjusted (OR 0.52, 95% CI 0.45–0.60). We evaluated two clinical events likely to be unaffected by decreasing length of hospital stay—blood transfusion and reoperation. The adjusted odds of blood transfusion comparing the 2003–2004 with the 1991–1992 (reference) intervals were as follows: LAVH 0.37 (OR 0.29, 95% CI 0.49), vaginal hysterectomy 0.60 (OR 0.53, 95% CI 0.68), TAH 1.01 (OR 0.96, 95% CI 1.07), and subtotal hysterectomy 0.58 (OR 0.45, 95% CI 0.76). The adjusted odds of reoperation were as follows: LAVH 0.26 (OR 0.13, 95% CI 0.52), vaginal hysterectomy 1.19 (0.77, 95% CI 1.85), TAH 1.16 (0.95, 95% CI 1.43), and subtotal hysterectomy 0.33 (0.17, 95% CI 0.67).
Although our regression models were adjusted for demographic factors, we examined time trends in these variables because large changes conceivably could confound the odds of complications. We observed a slight shift to older age groups of women undergoing hysterectomy between 1991–1992 (29.4%, age 20–39) and 2003–2004 (19.3%, age 20–39), a slightly higher proportion of women of African-American race (7.9% in 1991–1992 and 9.1% in 2003–2004), and a modest increase in the proportion of women with public insurance (18.3% in 1991–1992 and 22.1% in 2003–2004). Although we were unable to differentiate patients with private compared with HMO insurance, we did observe a slight decrease between 1991–1992 and 2003–2004 in the proportion of patients with “private insurance” (77.2% to 76.2%), the category that would include patients covered by HMO plans. Between 1991–1992 and 2003–2004, the proportion of patients undergoing inpatient hysterectomy who had any comorbidity increased significantly (P<.001) for LAVH, vaginal hysterectomy, and TAH but declined for subtotal hysterectomy. For example, for any hysterectomy type, the percentage of patients with any comorbidity increased from 32.7% to 44.9% between 1991–1992 and 2003–2004. For patients with subtotal hysterectomy, the percentage of comorbidities decreased from 55.4% to 49.8% during the same time period.
We searched for changes in complication code usage as evidence of a changing convention in the assignment of such cases. Between 1991 and 1995, surgical site infection codes (998.51 and 998.59) were not used.
Rates of hysterectomy in the United States peaked in the mid-20th century at approximately 86 per 10,000 women (age 15 and older) per year1 and then declined,11 consistent with our observations. Because we were unable to analyze cases of hysterectomy that may have occurred in the outpatient setting in freestanding surgical centers, our observed rates for inpatient hysterectomy may have underestimated, to some degree, the total rate of hysterectomy in California.
During the study interval (1991–2004), we observed large reductions in the adjusted odds of medical and surgical complications for each type of hysterectomy. The appearance of a reduced risk of inpatient complications during the study interval could have been affected by the following factors:
- A trend toward intrinsically safer hysterectomy because of changes in medical and surgical practice (eg, improved surgical equipment or technique or prophylaxis for infection and venous thromboembolism)
- A selection of patients with more favorable functional status or disease severity, leading to fewer complications among a healthier cohort of patients exposed to hysterectomy
- The artifactual appearance of lower complication risk as a result of progressively shorter hospital stays and reduced opportunity to identify and enter complication codes
- A changing convention in the use of complication codes by hospital abstractors
- Changes in demographic factors (eg, age, race, insurance type) that could have been incompletely adjusted and affected the odds of complications
Our results suggest that factors 2, 4, and 5 listed above are unlikely to explain the observation of reductions in the odds of hysterectomy-associated complications. Likewise, the modest increase in HMO insurance enrollment in California between 2002 and 200412 is unlikely to be a major factor contributing to falling complication rates between 1991 and 2004.
The cause-and-effect relationship between the odds of complications associated with inpatient hysterectomy and the simultaneous decline in the median hospital length of stay is complex. Length of stay may decline because of lower complication rates or as a result of administrative policies designed to reduce hospital use. Shorter length of stay could reduce the opportunity to identify and enter codes for complications, confounding the relationship between complications and length of stay. Although our analysis of trends in odds of complications suggested no significant reduction over time when the regression model was adjusted for length of stay, such adjustment using variables so strongly correlated can be misleading. Hysterectomy cases associated with complications are likely to be less affected by administrative pressures to reduce hospital use, therefore, over time, there appears to be no change in risk if length of stay is adjusted in the analysis of complicated cases. The regression analyses for blood transfusion and reoperation suggest that the observation of falling complication risk, especially for LAVH, vaginal hysterectomy, and subtotal hysterectomy, was not due only to shorter hospital stays. The sustained decline in laparoscopy-to-laparotomy conversion rates and in the reoperation and blood transfusion rates associated with LAVH represent convincing evidence of a population-wide LAVH learning process in California during the study interval. Others13 also have noted the simultaneous decline of TAH rates and increase in subtotal hysterectomy rates. It is likely that a perception that subtotal hysterectomy was associated with a lower risk of urinary incontinence and sexual dysfunction resulted in the rising popularity of subtotal hysterectomy, although these benefits have not been documented in all studies.14
In regard to laparoscopy-to-laparotomy conversion rates, we observed an initial, rapid decline between 1991 and 1996 followed by a slower decline between 1996 and 2004. We believe the initial steep decline represents a learning phenomenon comparable with that described in institution-based studies. We also hypothesize that the conversion rate declined because laparoscopic-assisted hysterectomy became a required element of residency programs in the United States, and newly trained surgeons entered practice with more experience in these procedures than did their predecessors. The method we used to define laparoscopy-to-laparotomy conversion may have resulted in miscoding of some cases of total laparoscopic hysterectomy. Although ICD-9-CM codes for total laparoscopic hysterectomy appeared in 2004, such cases probably became more frequently employed before this date, and coding for such cases may have used codes for both laparoscopy and TAH. Such cases would have been counted in Table 4 as conversions and may have artifactually elevated the observed conversion rate before 2004.
This work is affected by limitations common to similar studies that attempt to use administrative data to analyze postoperative clinical outcomes.15 Two published validation studies suggest that most coded diagnoses within the California Patient Discharge Database may be reasonably predictive of those within the actual medical record.16,17 The use of complication codes provides a uniform method of identifying cases with complications, but the exact nature of the injury, the morbidity involved, and its severity are unknown. In regard to LAVH, we adopted the previously used convention of defining the operation before 1997 as the simultaneous appearance of codes for vaginal hysterectomy and laparoscopy, but we are unable to confirm for such cases how much of each operation was accomplished laparoscopically.
In conjunction with a reduction in rate of inpatient hysterectomy, the odds of hysterectomy-associated complication codes for procedures with benign indications declined substantially in California between 1991 and 2004. Some of this decline may have been affected by shorter lengths of hospitalization or by changes in medical and surgical practice that resulted in intrinsically safer operations. That African-American race was associated with higher odds of complications or readmission suggests an ongoing disparity in the outcomes for African-American women undergoing hysterectomy for benign gynecological conditions in California that deserves further study.