|Home | About | Journals | Submit | Contact Us | Français|
To determine independent risk factors for endometritis (EMM) following low transverse cesarean section (LTCS).
We performed a retrospective case-control study from July 1999 to June 2001 in a large tertiary-care academic hospital. EMM was defined as fever beginning > 24 hours or continuing for ≥ 24 hours after delivery plus fundal tenderness in the absence of other causes for fever. Independent risk factors for EMM were determined by multivariable logistic regression. A fractional polynomial method was used to examine risk of EMM associated with the continuous variable, duration of rupture of membranes.
EMM was identified in 124/1605 (7.7%) women within 30 days after LTCS. Independent risk factors for EMM included age (odds ratio (OR) for each additional year 0.93; 95% confidence interval (CI): 0.90-0.97) and anemia/perioperative blood transfusion (OR 2.18; CI:1.30-3.68). Risk of EMM was marginally associated with a proxy for low socioeconomic status, lack of private health insurance (OR 1.72; CI: 0.99-3.00), amniotomy (OR 1.69; CI:0.97-2.95), and longer duration of rupture of membranes.
Risk of EMM was independently associated with younger age and anemia, and was marginally associated with lack of private health insurance, and amniotomy. Although duration of rupture of membranes was only marginally associated with increased risk of EMM, increased risk was observed very soon after rupture of membranes. Knowledge of these risk factors is important to guide selective use of prophylactic antibiotics during labor and heighten awareness of the risk in subgroups at highest risk of infection.
The cesarean section rate has risen steadily in the United States during the past decade, exceeding 30% in 2005.1 Endometritis (EMM) is the most common maternal infectious complication of childbirth, occurring more commonly after cesarean section than vaginal delivery.2 In a Cochrane review, the mean incidence of EMM following elective cesarean section was 7% and after non-elective or emergency operations was 30%.3
A variety of independent risk factors for post-cesarean EMM have been identified in previous studies, including no prior cesarean section,4 trial of labor,2,5 rupture of membranes > 24 hours,6 preterm4 or post-term gestation,7 low infant Apgar scores,4 young maternal age,6 antepartum infections,8 preeclampsia, meconium,9 amnioinfusion,10 postpartum anemia,4,6,8 large number of vaginal examinations,10 internal monitors,8,10 and manual placenta removal.11 Routine administration of intravenous cephalosporin plus azithromycin for prophylaxis at the time of cord clamp was temporally associated with significantly lower risk of EMM compared to use of prophylactic cephalosporin alone.5,12 The results of the study by Tita et al. are difficult to interpret however, due to an infection control intervention to promote aseptic technique conducted at the same time as addition of the azithromycin prophylaxis.5 In a randomized placebo-controlled trial administration of prophylactic cefazolin prior to skin incision was associated with significantly lower incidence of EMM than administration of cefazolin at the time of cord clamp.13 Preoperative vaginal antisepsis with povidone-iodine was also associated with significantly decreased risk of EMM in one randomized controlled study,8 but not another.14 Universal screening of all pregnant women for Group B Streptococcal infection and treatment of colonization was associated with significantly decreased risk of EMM compared to selective screening of only women with risk factors for colonization.15
The association between duration of labor and rupture of membranes with risk of EMM is difficult to determine, due to the variety of methods used to categorize duration and primarily univariate statistical methods in the older literature. More recently, rupture of membranes > 16 hours was shown to be independently associated with increased risk of EMM after all deliveries in the setting of premature rupture of membranes.16 In this study duration of ruptured membranes was categorized in 2-hour intervals, with the reference group rupture < 8 hours, and the risk of EMM was not specifically reported for cesarean section patients. Seaward reported that duration of labor > 3 hours and cesarean delivery independently increased the risk of postpartum fever.17
We performed a case-control study of EMM after low transverse cesarean section (LTCS) to determine clinically relevant independent risk factors associated with increased risk of EMM and to determine the impact of increasing duration of rupture of membranes, labor, and number of vaginal exams on risk of EMM. This knowledge is essential to develop targeted strategies to reduce the incidence of infection after cesarean section.
We performed a retrospective nested case-control study at Barnes-Jewish Hospital (BJH), a 1250-bed academic tertiary care hospital. Approval for this study was obtained from the Washington University Human Research Protection Office. Women who underwent LTCS between July 1, 1999 and June 30, 2001 were identified by International Classification of Diseases, 9th Edition, Clinical Modification (ICD-9-CM) procedure code 74.1. We focused on LTCS since low transverse uterine incisions are the preferred type of incision used for the vast majority of U.S. cesarean section deliveries.
Demographic, pharmacy and laboratory data were obtained from the BJH Medical Informatics database. Data on potential risk factors was collected from the surgical admission medical records, including all physician, nursing, and operative notes (Table 1).
Potential EMM cases were identified using ICD-9-CM diagnosis codes for EMM (670.02, 670.04) during the original surgical admission or an inpatient or emergency department readmission within 30 days of surgery and/or receiving prolonged duration of antibiotics beyond a prophylactic course post-surgery, as described previously.18,19 EMM was defined as fever ≥ 38°C beginning > 24 hours or continuing at least 24 hours after delivery plus fundal tenderness, with no other recognized cause for fever.20 Control patients were selected using a random number generator from the LTCS subjects during the study period who were not diagnosed with either EMM or surgical site infection (SSI). The control patients were described previously in a study of risk factors for SSI after LTCS.21 Medical records were reviewed for all case and control patients to determine if they met the EMM case definition within 30 days after operation.
Univariate and multivariable logistic regression models were used to identify independent risk factors for EMM. Continuous variables were compared using the Mann-Whitney U test. A multivariable logistic regression model was performed using backwards selection, including all variables with p < 0.15 in the univariate analysis or a priori clinical significance. All continuous variables with p < 0.15 were evaluated using a fractional polynomial approach in order to preserve the continuous nature of covariates suspected to have a non-linear relationship with EMM.22 To determine the final multivariable model, we used the SAS macro of Meier-Hirmer to determine the appropriate transformation of each continuous covariate.23 Since covariate values of zero preclude logarithm and negative power transformations, continuous covariates with values of 0 were transformed by adding 1. Odds ratios are presented for relevant exposure categories, using the mean of each category as the reference point and 95% confidence intervals calculated as described by Royston and Sauerbrei.24 After identification of the main effects, clinically relevant interactions between dichotomous variables were tested, with p < 0.05 the criterion for inclusion in the model. Model fit was assessed using the C statistic. All tests were two-tailed and p < 0.05 was considered significant. Analyses were performed with SAS 9.1 (SAS Institute Inc., Cary, NC) and SPSS version 14.0 (SPSS, Inc., Chicago, IL).
1,759 patients underwent cesarean section at Barnes-Jewish Hospital between July 1, 1999 and June 30, 2001. Among the 1,605 (91.2%) patients with low transverse uterine incisions, 124 (7.7%) patients with EMM within 30 days after operation were identified. 310 control patients without EMM or surgical site infection were randomly selected for comparison. One hundred seven (24.7%) women had elective cesarean sections, 254 (58.5%) had urgent, and 73 (16.8%) had emergent procedures. The majority of patients (71.4%) had a primary cesarean section.
The univariate results of risk factors for EMM are presented in Tables 2 and and3.3. Women with EMM were more likely to be younger, non-white, unmarried, and to have Medicare, Medicaid, or no health insurance (lack of private health insurance). Younger age was also associated with fewer prenatal visits (p = .005, one-way ANOVA), as was lack of private health insurance (p < .001, one-way ANOVA). Younger age was also associated with slightly higher gestational age at delivery (p = .015, Spearman's rho). Patients with Neisseria gonorrheae or Chlamydia infection during pregnancy and with higher ASA class were also more likely to develop EMM. Obstetric factors associated with an increased risk of EMM included preeclampsia, clinical chorioamnionitis, labor (spontaneous or induced), amnioinfusion (transcervical infusion of saline into the uterus to dilute meconium-stained fluid or relieve cord compression due to low amniotic fluid volume), use of internal fetal monitors, more vaginal exams, and spontaneous or artificial rupture of membranes (amniotomy). Surgical factors associated with a decreased risk of EMM included elective surgery, manual placenta removal, and bilateral tubal ligation. Manual removal of the placenta (compared to cord traction) was significantly associated with the type of attending physician, with private physicians significantly less likely to remove the placenta manually compared to university faculty (Chi-square, p < .001).
There were no differences in preoperative hemoglobin or hematocrit values between EMM cases and controls, however, EMM cases had significantly lower postoperative hemoglobin and hematocrit values. Estimated blood loss during surgery was not related to risk of EMM, but perioperative blood transfusion (both leukocyte-depleted and non-leukocyte depleted blood products) was associated with increased risk of EMM.
Patients with premature rupture of membranes were significantly more likely to develop EMM than patients who ruptured at or after the onset of labor (44 (41.5%) vs. 80 (24.4%); p = 0.001). Of the 106 (24.4%) women with premature rupture of membranes, there was a marginal association between shorter duration of rupture before labor and EMM (median duration, 112 minutes for cases vs. 203 minutes for controls; p = 0.099). Compared to women who ruptured at or after the onset of labor, women with premature rupture of membranes were marginally more likely to have a clinical diagnosis of chorioamnionitis (25 (23.6%) vs. 53 (16.2%); p = 0.083). Overall, patients with clinical chorioamnionitis had significantly increased risk of developing EMM compared to patients without chorioamnionitis in univariate analysis. There was no association between subclinical chorioamnionitis identified solely by the pathology reports and EMM (Table 3).
Independent risk factors for EMM identified in the multivariable analysis included younger age and anemia and/or blood transfusion (Table 4). Anemia and/or blood transfusion was associated with a more than two-fold increased risk of EMM. We combined postoperative anemia and blood transfusion into one variable for analysis since some women with multiple transfusions did not have blood collected at the suspected nadir. In addition, some women with very low hemoglobin values refused blood transfusion, so neither variable alone captured all women with severe anemia. Lack of private health insurance, amniotomy, duration of rupture of membranes, and clinical chorioamnionitis were associated with marginally increased risk of EMM, controlling for labor. Age was modeled as a linear variable due to the absence of convincing evidence of nonlinearity (Table 4). Duration of rupture of membranes was modeled using the term 1/√(duration of rupture in minutes + 1) using the fractional polynomial approach. As shown in Figure 1, the odds of EMM increased immediately after rupture of membranes, but did not increase appreciably with increased duration of rupture, adjusting for the other variables in the model. We tested internal fetal monitors rather than amniotomy in a separate model, but it was not significantly associated with EMM after adjusting for duration of rupture of membranes and the other variables (p = .251).
Our results show that younger age and perioperative anemia or transfusion independently increased the risk of EMM following LTCS in women at an urban tertiary-care university-affiliated hospital. Clinical chorioamnionitis, amniotomy, duration of membrane rupture, and lack of private health insurance were marginally associated with increased risk of EMM.
We determined that the risk of EMM increased in linear fashion with decreasing age. Very young age has been reported to increase the risk of EMM in two studies,25,26 although the reason for this increased risk is unknown. Despite the use of our multivariate analysis in our study, residual confounding may exist because of undiagnosed sexually transmitted infections or undocumented Group B Streptococcal vaginal colonization, among other factors. This may be preferentially true in younger women, who had significantly fewer prenatal visits than older women and therefore less opportunity for diagnostic testing.
In univariate analysis, rupture of membranes was associated with significantly increased risk of EMM, but after inclusion of amniotomy and other variables in the multivariable model duration of rupture was only marginally significant. Amniotomy was associated with 1.7-fold increased odds of EMM, and is biologically plausible, due to the potential for introduction of organisms from the vagina and cervix into the amniotic fluid during the procedure. In previous studies including multivariate analysis, rupture of membranes has been analyzed as a binary or categorical variable.6,16 The studies analyzing rupture as a categorical variable are more difficult to interpret, since in most of these studies the reference category consisted of women without ruptured membranes plus women with ruptured membranes of defined duration (e.g., < 8 hours), compared to women ruptured for longer durations (e.g., > 8 hours). In our study the odds of EMM increased approximately 1.7-fold within 1 hour after rupture of membranes and remained at that level with increasing duration of rupture. Thus our results are more consistent with increased risk of EMM associated with rupture per se, regardless of the duration.
Duration of labor and binary labor (yes/no) were not associated with increased risk of EMM after controlling for duration of rupture of membranes. Rupture of membranes allows for ascending spread of bacteria into the previously sterile amniotic fluid, and thus this variable is more biologically plausible in the model than labor per se. Internal monitors was also excluded from the multivariate model after controlling for duration of ruptured membranes, although it was associated with over 3-fold higher risk of EMM in univariate analysis.
Severe anemia or transfusion of packed red blood cells was associated with 2-fold increased odds of EMM in the multivariate model. Anemia has previously been reported as an independent risk factor for EMM in one study.6 Severe anemia was not solely a proxy for excessive blood loss during surgery, since estimated volume of blood loss was not associated with increased risk of EMM in our study population. We and others have reported that transfusion increases the risk of a variety of hospital-acquired infections.27-30 The reason for the association of severe anemia with increased risk of EMM is not entirely clear, but may in part be due to residual confounding with underlying severity of illness. In addition, some investigators have suggested that transfusion increases the risk of infection via immunomodulation and release of bioactive mediators from contaminating allogeneic white blood cells.27,31 Almost 2/3 of the transfused patients in our study received only leukodepleted blood. Interestingly, there was no difference in risk of EMM associated with leukodepleted blood transfusions vs. non-leukodepleted. Our finding supports some recent studies questioning the impact of leukodepletion on nosocomial infection rates associated with transfusion.32,33
Chorioamnionitis was associated with significantly increased odds of EMM in univariate analysis but was not formally associated with increased risk of EMM in the multivariate model. In previous publications reporting increased risk of EMM associated with chorioamnionitis, investigators have controlled for duration of rupture or labor as categorical or binary variables, which most likely resulted in residual confounding.5,17
Non-private health insurance, our proxy for low socioeconomic status, was associated with marginally increased risk of EMM. Reasons for this association may include higher rates of sexually transmitted infections, higher risk of Group B Streptococcal colonization, and less likelihood of treatment of these infections due to fewer prenatal care visits. In addition, micronutrient and vitamin insufficiency may in part explain this relationship. Women with Medicaid coverage and African-American women have increased likelihood of prepregnancy anemia, and are less likely to report multivitamin use than non-Hispanic white women and those with private insurance.34,35 Higher body mass index has also been shown to be associated with worse diet quality during pregnancy.36 A variety of micronutrients play critical roles in immune responses,37,38 so it is plausible that micronutrient deficiencies associated with poor quality diet and lower multivitamin usage could be associated with increased risk of EMM in women with low socioeconomic status.
In contrast to some previous work, we did not find an association between Group B Streptococcus colonization and development of EMM. This may be due to our inability to identify all women with Group B Streptococcus colonization based on retrospective review of hospital records. It is also possible we did not detect risk of EMM associated with Group B Streptococcus colonization due to abrogation of this risk by antibiotic prophylaxis during labor. Over one-quarter of the women in our study received ampicillin, penicillin, or clindamycin before incision (alone or in combination with cephalosporin prophylaxis). Over half of the women with documented Group B Streptococcal colonization received anti-GBS prophylaxis, and another 20% of colonized women were treated for chorioamnionitis. Dumas recently reported that antibiotic prophylaxis against Group B Streptococcus was associated with significantly decreased risk of EMM after vaginal delivery.39 Thus it is possible that antibiotic prophylaxis and therapy in our study was successful at decreasing the risk of EMM due to Group B Streptococcus colonization.
A limitation of this study is the analysis of older data. During the time period of this study, routine administration of prophylactic antibiotics for cesarean section was performed at cord clamp rather than before incision. In our population almost half of the women received antibiotics before incision, and receipt of prophylactic antibiotics at cord clamp was not associated with increased risk of EMM. Although earlier administration of prophylactic antibiotics should theoretically decrease the risk of EMM, it would not be expected to alter the relationship of other variables with risk of EMM.
In summary, perioperative anemia/transfusion and decreasing age were independently associated with increased odds of EMM, while lack of private health insurance, amniotomy, and rupture of membranes were associated with marginally increased odds of infection. While most of these factors are not easily modifiable, knowledge of the increased risk associated with specific factors can be used to tailor antibiotic prophylaxis regimens and heighten surveillance for signs of EMM in the post-operative period in women at highest risk of infection.
We gratefully acknowledge Zohair Karmally, Cherie Hill and Stacy Leimbach for assistance with data collection and management. This work was supported in part by grants from the Centers for Disease Control and Prevention (Prevention Epicenter Program, UR8/CCU715087) and the National Institutes of Health (K01AI065808, MAO and K24AI06779401, VJF).
Presented in part at the Annual Clinical Meeting of the American College of Obstetricians and Gynecologists, San Francisco, CA, May 2005.
None of the authors have any conflicts of interest.