Goldberg, Jay MD; Holtz, David MD; Hyslop, Terry PhD; Tolosa, Jorge E. MD, MS
For most of the 20th century, the routine use of episiotomy was believed to have multiple benefits for both mother and infant. These benefits were believed to include prevention of pelvic floor damage and its sequelae, including urinary incontinence, poor wound healing, severe lacerations involving the anal sphincter, and fetal intracranial hemorrhage. Thacker and Banta's 1983 review of episiotomy literature from 1860 through 1980, which found few good studies and no evidence of any benefit, sparked further investigation.1 During the past 20 years, a large body of literature has been published, which strongly advocates the selective use of episiotomy.2 Multiple studies during this period were reported demonstrating that the routine use of episiotomy did not protect against pelvic relaxation or fetal intracranial bleeds. Episiotomy actually increased rates of perineal infection, increased blood loss, increased pain during healing, negatively affected body image issues and sexual function, and increased incidence of injuries to the anal sphincter, with subsequent increased risks of incontinence of flatus and fecal material.3
Episiotomy rates widely vary between countries, institutions, and individuals because of differences in attitudes and training. The Argentine episiotomy trial reported an 83% incidence of episiotomies in 1993.4 Henriksen et al in Denmark found a 37% overall rate of episiotomy in 1990.5 The range of episiotomy rates in the United Kingdom's West Berkshire perineal management trial was 14–96% in primiparas in the early 1980s.6 Shiono et al found the mean episiotomy rate in the United States to be 62% in 1990.7 Although there are many cross‐sectional epidemiologic reports of episiotomy rates, few long‐term longitudinal studies of episiotomy rates were found. The National Center for Health Statistics reported episiotomies in 61.1%, 55.6%, 47.2%, and 39.3% of deliveries in 1985, 1990, 1995, and 1998 in the United States, respectively.8 Similarly, Graham and Graham reported a decrease in Canadian episiotomy rates from 66.8% in 1981/1982 to 37.7% in 1993/1994.9
This study was initiated to examine trends in episiotomy rates over the last 20 years to see if practice patterns had been altered by the large body of literature strongly advocating the selective use of episiotomy.
MATERIALS AND METHODS
This study received approval by the Internal Review Board of Thomas Jefferson University Hospital. An electronic audit of the medical procedures database at Thomas Jefferson University Hospital from 1983 to 2000 was completed, and all records on childbirth were extracted based on International Classification of Diseases, 9th Revision, codes. This database also included some limited information on maternal demographics, such as age and ethnicity. Race was categorized as white, black, Asian, Hispanic, or other. Other, as listed in Tables 1 and 2, were mostly patients having no racial designation. Procedure codes were used to determine which births involved episiotomy, spontaneous vaginal delivery (SVD), cesarean, forceps, or vacuum assistance. We eliminated births involving cesarean, so that the remaining data are for vaginal births only (n = 34,048). Births involving both forceps and vacuum (n = 338) were also not included because we believed that this small number of deliveries was very different from other operative deliveries. Rates were determined by computing the total number of episiotomies performed divided by the total number of vaginal births, not including the small number with both vacuum and forceps, in each of the calendar years in the study database.
Univariate analyses of the association of episiotomy status with clinical characteristics was completed based on univariate logistic regression models and generalized estimating equations (GEE) techniques. GEE is used to adjust standard errors for the clustering of multiple childbirths within mothers.10 Multivariable analyses of the association of episiotomy status with clinical characteristics were also completed. The multivariable models were computed using logistic regression models with GEE methods to adjust for clustering of births within mothers. Possible covariates considered in these models were: year of childbirth, race, maternal age, SVD, forceps, large for gestational age, small for gestational age, third‐ or fourth‐degree laceration, and whether the patient had private or Medicaid insurance. Because of the statistical linear relationship between SVD, forceps, and vacuum‐assisted delivery, only two of these can simultaneously be considered in models. Because there was minimal information on insurance status before 1990, all analyses that include insurance status are based on births occurring in 1990 and beyond (n = 18,138).
Because parity and gravidity were not available, we selected the subgroup of mothers having more than one vaginal delivery during the period of the study at Thomas Jefferson University Hospital. Every delivery after the initial delivery within our database for an individual woman was assigned to a multiparous subgroup. SAS statistical software 8.0 (SAS Institute Inc., Cary, NC) was used for data management and descriptive analyses. STATA statistical software 7.0 (Stata Corp., College Station, TX) was used for univariate and multivariable logistic regression GEE models.
We examined 34,048 vaginal births between 1983 and 2000 at Thomas Jefferson University Hospital in Philadelphia. The cumulative percentage of patients classified as white, black, Asian, Hispanic, or other were 40.5%, 49%, 2.5%, 1.2%, and 6.8%, respectively. Episiotomy rates were calculated for all vaginal deliveries, overall, and by maternal race, and for forceps and vacuum‐assisted deliveries. Overall episiotomy rates showed a significant reduction from 69.6% in 1983 to 19.4% in 2000. White women's episiotomy rates declined from 79% to 32.1%, whereas black women's rates decreased from 60.5% to 11.2%. Figure 1 illustrates this decrease in episiotomy rates for vaginal deliveries overall (solid line), for blacks (large dashed lines), and for whites (small dashed lines). These two groups represent the largest racial populations studied. Cumulatively, episiotomy was performed in 43.3%, 71.7%, and 90% of SVD, vacuum‐assisted, and forceps deliveries, respectively. Figure 2 shows parallel decreasing episiotomy rate trends for overall vaginal deliveries and SVDs, some decrease since 1993 for episiotomy use in forceps deliveries, and an unchanged rate in vacuum‐ assisted deliveries. A total of 19.8% of vaginal deliveries used forceps in 1983, decreasing yearly to a nadir of 6.5% in 1991, then increasing to 11.7% in 2000, whereas 2.5% of vaginal deliveries were vacuum assisted in 1983, increasing yearly to a peak of 12.6% in 1991, then decreasing to 1.2% in 2000. Deliveries over the 18 years surveyed involved over 100 attending physicians and over 200 resident physicians, but only the attending physician could be identified by the database. Twenty of the attendings were responsible for 55.7% of the deliveries surveyed.
A univariate analysis of the association of episiotomy with clinical characteristics was performed. Decreased episiotomy rates were significantly associated with non‐white race of the mother (63.8% for whites versus 39.4% for blacks, 47.2% for Asians, and 40.6% for Hispanics, P < .001), if the mother was on Medicaid insurance (28.4% versus 47.4%, P < .001), SVDs (43.3% versus 82.8%, P < .001), in patients older than 21 years (49.1% for patients 22–34 and 49.9% for patients older than 34 versus 55.4% for age less than 21, P < .001), and if the infant was small for gestational age (41.6% versus 50.5%, P < .001). Large‐for‐gestational‐age deliveries were associated with a nonsignificant decrease in episiotomy rate (50.5% versus 45.2%, P = .09). Women having a third‐or fourth‐degree perineal laceration were significantly more likely to have received an episiotomy (85.4% versus 46.5%, P < .001).
The first multivariable model of episiotomy is presented in Table 1. This model includes year of childbirth plus all the factors from the univariate analysis except Medicaid status, which was not available until 1990. This table shows a statistically significant reduction in risk of episiotomy with increasing year of birth and maternal age greater than 21. Black race showed the most dramatic risk reduction (odds ratio [OR] 0.29, 95% confidence interval [CI] 0.28, 0.31), with less significant reductions for Asians (OR 0.71, 95% CI 0.60, 0.84) and Hispanics (OR 0.54, 95% CI 0.43, 0.68). Significant positive association with episiotomy was found in women delivered with forceps and with those who experienced third‐ or fourth‐degree perineal laceration.
A multivariable logistic regression model including insurance information is constructed on the subset of data occurring from 1990 to 2000, 18,138 vaginal births. This model shows that controlling for insurance status, the factors still significantly associated with a decreased risk of episiotomy include increasing year of childbirth (OR 0.82, 95% CI 0.81, 0.83), black race (OR 0.31, 95% CI 0.28, 0.34), Asian race (OR 0.78, 95% CI 0.65, 0.93), Hispanic race (OR 0.61, 95% CI 0.47, 0.80), maternal age 22–34 years (OR 0.49, 95% CI 0.45, 0.54), age greater than 35 years (OR 0.45, 95% CI 0.40, 0.51), and SVD (OR 0.34, 95% CI 0.30, 0.39). Receiving Medicaid insurance was also associated with a decreased episiotomy risk (OR 0.59, 95% CI 0.54, 0.64). Increased association with episiotomy was seen in forceps deliveries (OR 3.53, 95% CI 2.90, 4.28), and with third‐ or fourth‐degree lacerations (OR 3.92, 95% CI 3.44, 4.47). Non‐significant changes in episiotomy risk were seen in the women who delivered large‐for‐gestational‐age (OR 1.13, 95% CI 0.77, 1.66) or small‐for‐gestational‐age infants (OR 1.05, 95% CI 0.77, 1.43).
To try to control for parity, we selected the subset of births known to be from multiparous women. A multivariable logistic regression model with GEE is presented for this subgroup in Table 2. The model is based on 4075 multiparous births from 1990 to 2000 to enable the inclusion of insurance status in this model. Although there is some loss of power in the smaller covariate categories, the general magnitude and trend of the ORs remain the same as in the previous models.
Although the use of selective episiotomy is advocated in the literature, there is little evidence on whether the practice of performing episiotomies has actually decreased or not over the last 20 years. We found a statistically significant reduction in the overall episiotomy rate between 1983 and 2000 from 69.6% to 19.4%. The OR of 0.87 per year of childbirth in Table 1 reflects a statistically significant reduction in episiotomy rates per year from 1983 through 2000. This change in practice pattern may be largely due to the impact of a growing body of literature against routine episiotomy, including that derived from randomized controlled trials. Additionally, improved patient education and participation in decision making and changes in use of forceps and vacuum assistance probably play a role in reducing the overall number of episiotomies. No specific policy or educational initiative regarding episiotomy was implemented during the study period.
We also found that white women consistently underwent episiotomy more frequently than black women. It was initially hypothesized that the difference between races in episiotomy rates may have been due to differences in parity, prenatal care, or in the number of operative vaginal deliveries. Episiotomies remained significantly greater in white patients, however, despite multivariable logistic regression analysis controlling for these possible confounding factors. Socioeconomic and racial differences have been reported for obstetric practices and procedure rates, including cesarean delivery, epidural use, and episiotomy, with higher socioeconomic status increasing the risk for intervention. Hueston reported white race as a predictor for episiotomy (OR 2.02, CI 1.66, 2.46). He hypothesized that the increased episiotomy rate could represent a marker for patient expectations or perceived threat of a malpractice suit.11 Howard et al's study similarly found an increased episiotomy rate (34.8% versus 22.3%, P = .003) in white patients compared with black patients.12 It also showed that black primiparas were less likely to deliver with second‐degree or greater vaginal lacerations and more likely to deliver with their perineums intact, supporting anecdotal reports of lower vaginal laceration rates in black women.12 Physician perception of differing risks of severe spontaneous lacerations between racial groups may also factor into white women undergoing episiotomy more often in a misguided attempt to prevent this. It may also simply be a marker of the arbitrary and non‐scientific nature employed by the physician in determining which patients need an episiotomy.
Our study is retrospective, and it is limited by a database unable to supply information on nulliparity, epidural usage, specific infant weight, and insurance information before 1990. Secondary analyses of the multiparous subgroup were used to attempt to control for parity as best possible in assessing our results. Even in the multiparous subgroup, the association of episiotomy with race remained consistent. Physician turnover during the 18 years studied may have affected episiotomy rates; however, given that 55.7% of the deliveries were performed by 20 physicians, it is unlikely that practices of a few would have significantly affected the results.
Our findings of a decreasing rate of episiotomy between 1983 and 2000 agrees with national trends.8 Although our overall episiotomy rates appear lower than national rates, the proportion of black women in our population is larger than national demographics. Further investigation is needed to examine trends in episiotomy rates among other academic institutions and community and rural hospitals in different geographic areas.