Risk of Obstetric Anal Sphincter Injury by Delivering Provider †

: Introduction and Hypothesis: Obstetric anal sphincter injuries (OASIs) complicate 5.8% of vaginal deliveries. Our objective was to assess if the primary delivery provider, a nurse-midwife versus physician obstetrician, is associated with OASIs. Methods: We performed a secondary analysis of the Consortium of Safe Labor , a multicenter, retrospective cohort study. Included were nulliparous women with singleton, vaginal delivery at ≥ 37 weeks from 2002 to 2008. Women were excluded if delivery was complicated by shoulder dystocia or from sites without midwife deliveries. Student’s t -tests, chi-squared analysis, and Fisher’s exact test were used as appropriate. Multivariable logistic regression and propensity score-matching analyses were performed. Results: Of 228,668 births at 19 sites, 2735 births from 3 sites met the inclusion criteria: 1551 physician and 1184 midwife births. Of all births, 4.2% ( n = 116) were complicated by OASIs. Physician patients were older, more often White, privately insured, with higher BMI, more medical co-morbidities, and labor inductions/augmentations. Midwife patients had higher fetal gestational age and infant birth weights (all p < 0.05). OASIs were more common in physician compared to midwife births (5.9% vs. 2.0%, p < 0.0001). This difference persisted in multivariable logistic regression. OASIs were 2.4 (95%CI 1.5–3.9) times more likely with physician delivery when controlling for maternal heart disease, episiotomy, increasing maternal age, decreasing maternal BMI, non-White race, and increasing birthweight. The AUC was 0.78. With propensity score matching, OASI rates remained higher amongst physician births (6.6% vs. 1.8%, p < 0.0001; aOR 3.8 (95%CI 2.0–7.1)). Conclusion: OASIs were more common in physician compared to midwife deliveries even when controlling for other associated factors.


Introduction
Obstetric anal sphincter injury (OASI) is defined as an injury to the anal sphincter and/or rectal mucosa sustained during vaginal delivery.OASIs complicate approximately 5.8% of all vaginal deliveries in the U.S. and there is a notable increasing trend in its incidence globally [1,2].There is significant associated short-and long-term morbidity of OASIs, including wound infection, perineal pain, dyspareunia, fecal incontinence, psychological distress, and impaired quality of life.In addition, the risk of OASIs with subsequent deliveries may be up to five times as high [2][3][4].Commonly reported risk factors are nonmodifiable, such as nulliparity, increasing maternal age, White race, operative deliveries, episiotomy, and large fetal weight [1,[5][6][7][8][9][10].Only a few studies have attempted to assess modifiable risk factors, such as the type of delivery provider.These studies have yielded varying datasets regarding the associated risk of OASIs and were limited to specific patient cohorts in Ireland and the UK [11,12].
There is a paucity of research that comparatively examines the type of delivery provider as a risk factor for OASIs in racially diverse patient populations, specifically in the United States where the obstetric care model varies from the model used across Europe.We propose that by obtaining a modern cohort of nulliparous women and controlling for factors known to be associated with OASIs, we can investigate whether having a certified nurse-midwife or physician obstetrician as the primary provider during vaginal delivery confers an associated risk of OASIs.Identifying a novel risk factor for OASIs can be used to assist delivering providers minimize the risk of OASIs during vaginal delivery.

Materials and Methods
This is a secondary analysis of the data from the Consortium on Safe Labor (CSL).The CSL is a multi-centered, retrospective cohort study which collected birth data from 228,668 deliveries in 19 hospitals within 12 institutions across the United States from 2002 to 2008.Given this nationally available database, this study was exempt from the Institutional Review Board.
Inclusion criteria for our analysis consisted of nulliparous women with successful vaginal deliveries of a cephalic, singleton fetus at ≥37 weeks of gestation.Women were excluded if their delivery was complicated by shoulder dystocia, if data regarding perineal lacerations at the time of delivery were missing, or if the site did not have midwife-led birth attendants.
Maternal variables of interest include race, ethnicity, age, education level, insurance type, pre-pregnancy BMI and weight, weight at the time of hospital admission, medical history (diabetes; hypertension; renal, heart, or thyroid disease; depression), smoking history, antenatal complications, and history of antenatal hospitalization.Intrapartum variables of interest include use of induction agents and labor augmentation, fetal distress, analgesia use, operative delivery technique, episiotomy, and length of second-stage labor.Labor induction included the use of mechanical and/or pharmacological methods including artificial rupture of membranes, mechanical dilation, misoprostol and other prostaglandins, and oxytocin.Fetal variables of interest included gestational age at the time of delivery and fetal birthweight.
The primary outcome was the presence of OASIs in physician obstetrician versus certified nurse-midwife vaginal births.We hypothesized that a higher number of OASIs would be seen amongst midwife deliveries compared to physician deliveries.A post hoc power calculation was performed indicating we have a power of 99.9% at an alpha of 0.05, given the cohort size and difference in OASIs between the groups.Our secondary outcome included assessing whether the difference in OASIs amongst the two provider types continued when adjusting for known risk factors of OASIs.
Student's t-tests, chi-squared analyses, and Fisher's exact tests were used as appropriate to assess baseline characteristics, labor factors, and frequency of OASIs within the two groups of providers.Results were presented as means ± standard deviation for continuous, normally distributed variables and frequencies (percentages) for categorical variables.Multivariable logistic regression and propensity score matching analysis were performed to adjust for characteristics associated with OASI.Factors with a p < 0.2 were fitted with backward and forward stepwise techniques to create the final model with variables associated with OASIs.
Data are presented as adjusted odds ratios (aORs) with 95% confidence intervals.SAS statistical software was used to perform the analyses (version 9.4, Institute Inc., Cary, NC, USA).This study was categorized as exempt from the George Washington University Institutional Review Board.

Results
Of 228,668 births at 19 sites included in the database, a total of 2735 births from 3 sites met the inclusion criteria: 1551 physician and 1184 midwife births.Physician patients were older (23 ± 5 vs. 21 ± 4 years), more often White (26.4% vs. 14.3%),commercially insured (39.1% vs. 22.8%), and had higher pre-pregnancy BMIs (25.5 ± 6.4 vs. 24.8± 5.8 kg/m 2 ; all p < 0.05; Table 1).Pre-existing medical conditions such as diabetes, heart disease, hypertension, thyroid disease, and smoking were more common in the physician cohort (all p < 0.05; Table 1).Antenatal medical complications such as gestational diabetes, antenatal hospital admission, and preeclampsia were also more common in the physician cohort (all p < 0.05; Table 1).Physician patients were more likely to undergo labor inductions (40.9% vs. 20.4%), the most common method being oxytocin infusion (37.1% vs. 18.7%), as well as labor augmentation, compared to midwife patients (28.2% vs. 16.2%,all p < 0.05; Table 2).More episiotomies and operative deliveries occurred in the physician cohort (15.5% vs. 5.2%, and 5.1% vs. 0.08%; all p < 0.05).Midwife patients had more advanced fetal gestational ages (39.7 ± 1.1 vs. 39.4 ± 1.2 weeks) and greater infant birth weights (3.3 ± 0.4 vs. 3.2 ± 0.4 kg; all p < 0.05) at the time of vaginal delivery (Table 3).Of all births, 4.24% (n = 116) were complicated by OASIs.OASIs were more common in physician compared to midwife births at 5.9% (n = 92) compared to 2.0% (n = 24), respectively (p < 0.0001).A multivariable logistic regression revealed that this difference persisted with an adjusted odds of OASIs being 2.39 times more likely with physician delivery than midwife delivery (95%CI 1.5-3.9).In this regression model, other factors associated with OASIs were maternal history of heart disease, receipt of episiotomy, increasing maternal age, decreasing maternal BMI, non-White race, and increasing fetal birthweight (Table 4).Operative delivery was not significant in the multivariable regression and thus was not included in the final model.The area under the curve for this model was found to be 0.78.Additionally, propensity score matching was conducted to ensure the strength of the regression model.A propensity score match was used to balance all potential confounders.We found 1054 matches per cohort with OASI rates of 6.6% and 1.8% for physicians and midwives, respectively.Physician was associated with 3.8-times-higher odds of OASIs relative to midwife (95% CI 2.0-7.1;p = 0.0002).

Discussion
The overall rate of OASIs amongst our study population was 4.24%, which is similar to the approximate rate found in the full CSL cohort at 5.8% [1].We discovered that in this cohort, physicians were serving a slightly older, Whiter, and more affluent population than the midwives.Physician patients were also more likely to have medical co-morbidities documented and undergo labor induction and/or augmentation compared to certified nurse-midwife patients.Even when controlling for these population differences and known risk factors for OASIs, we found that OASIs at the time of vaginal delivery were more likely to occur with an obstetrician physician as the delivering provider when compared to certified nurse-midwives.This is one of the first studies to examine the delivering provider as a risk factor for OASIs in the U.S.Only a few studies have examined the delivery provider as a risk factor for OASIs, which were based in England and Ireland, respectively.One prospective, observational study compared rates of OASIs in a hospital obstetric unit to those in several freestanding midwifery-led units [12].Women who delivered in the midwifery-led units more frequently had an intact perineum at the time of vaginal delivery.In comparison, a recent study in a university-affiliated district hospital found that the risk of OASIs was twice as likely in the midwifery-led units compared to the hospital's obstetrician-led units [11].Not only do these data conflict, but likely are not generalizable to a diverse patient population and an American treatment paradigm.
We suspect there are a few differences between physicians and midwives that have led to the differences in OASIs.First, as mentioned above, the patient populations, at least in this cohort, are different.Physician patients in our study had more pre-existing and antenatal medical comorbidities, which may have led to more intrapartum complications or higher risk of maternal/fetal distress requiring an expedited delivery.There was a higher frequency of fetal distress and intrapartum magnesium use within the physician cohort comparatively, which may support this theory.Although, interestingly, midwife patients had more meconium-stained fluid, which can traditionally be a sign of fetal distress.
Secondly, there are also major differences in practice style.Our findings also reaffirm the pattern of decreased use of labor induction and augmentation agents as well as delivery interventions amongst midwives when compared to physicians [13].A study exploring the effect of midwives on perinatal outcomes in the U.S. demonstrated that midwife practice favors waiting for spontaneous onset of labor, hospital admission once active labor is reached, and more conservative use of pharmacologic and/or surgical interventions during the progression of labor, which is consistent with our findings [14].Our analysis demonstrates that with the use of fewer intrapartum interventions, we see a lower frequency of OASIs specifically amongst the midwife patients.It is possible that our findings reflect and support limited labor or birth interventions, a core aspect of midwifery philosophy, as a tool to reduce OASIs.
A third major consideration for the decreased OASIs amongst midwife providers is that they may be underdiagnosed.A 2012 survey found only 34% of midwives reported they were confident in OASI assessment at the time of delivery and just 22% indicated they felt prepared to repair the laceration [15].An additional survey conducted in 2020 revealed the overall accuracy of perineal laceration identification amongst the midwife respondents ranged from 49 to 99% and a high frequency of OASI misidentification was accompanied by subsequent inappropriate use of the OASI severity grading scale [16].Although these surveys took place in the U.K. where the midwifery model differs from the U.S., there is a possibility that OASIs may be underrepresented amongst our midwife patient cohort.
A final consideration is the presence and participation of physician trainees (fellows, residents, and medical students) in patient care at the time of delivery.This is a delivery factor impacting the physician cohort in our study which we were unable to measure.As the institutions associated with this database are teaching hospitals, we must consider the possibility that by allowing physician trainees particularly those early in their careers to participate in vaginal deliveries, this may increase the risk of more advanced obstetric lacerations.While there is not a considerable amount of research comparatively examining the role of physician trainees and the incidence of OASIs, a 2016 retrospective cohort study examined the role of midwife experience.Highly experienced midwives, those with more than 10 years of experience, had the lowest incidence of OASIs when compared to midwives deemed moderately experienced or inexperienced, with between 2 and 10 years and less than 2 years of experience, respectively [3].
A major strength of this study is our large, racially, and socioeconomically diverse patient population from multiple centers.This is more reflective of the general population of the U.S. than previous studies examining the frequency of OASIs amongst varying provider types.In addition, we identified a risk factor that has not been fully assessed for OASIs.It is common knowledge that operative deliveries and episiotomies increase the risk.However, our study findings ask an important question about the way physicians manage labor and how that increases the risk of OASIs.
A major limitation is the exclusion of data from CSL sites with an absence of certified nurse-midwife providers.Although there has been a steady increase in the rate of midwifeattended hospital births in the United States since 1975, reaching a peak of 8.6% in 2016, a majority of sites included in the CSL did not include midwives as lead birth attendants at the time of delivery [14,17].Secondly, absent or missing data from the database and inconsistent documentation of factors of interest such as length of induction and length of labor were additional limitations which prevented full examination of these factors.In addition, in our analysis, we were unable to determine if patients underwent intrapartum transfers from midwifery to physician care.This commonly occurs on Labor and Delivery wards due to evolving complications and may impact the risk of OASIs.Lastly, the only midwife deliveries considered in this database were hospital deliveries, which may not be reflective of midwifery care in birth-centers and home births where the incidence of OASIs may differ.Although these limitations may affect the generalizability of our results, there

Table 1 .
Baseline maternal demographics by provider type.

-pregnancy BMI (kg/m 2 ) *
* Variable had more than 5% of missing data during the analysis.HS: High school; BMI: Body Mass Index; HELLP: Hemolysis, Elevated Liver enzymes, Low Platelet count.Diabetes: Includes both Type I and Type II Diabetes Mellitus.Data are presented as mean (±standard deviation) or frequency (%).

Table 2 .
Hospital admission and labor factors by provider type.

Table 3 .
Delivery variables by provider type.

Table 4 .
Multivariable logistic regression of factors associated with OASIs.