Over 31% of deliveries in the United States are by cesarean.1 Surgical site infections after cesarean delivery are associated with maternal morbidity and increased medical costs.2,3 The current surgical site infection rate in U.S. hospitals varies from 1.46 to 3.82 per 100 cesarean deliveries depending on patient risk factors.4 Known risk factors for infection include extremes of age, increased body mass index (BMI, calculated as weight (kg)/[height (m)]2), premature rupture of membranes, prolonged surgical duration, and the use of staples for wound closure.5–13
Several strategies have been implemented for surgical site infection prevention, including the use of appropriate prophylactic antimicrobials, proper hair removal, and banning the use of artificial nails among surgical staff.14 Animal and human studies have noted that prophylactic antibiotics have an optimal benefit when adequate antibiotic tissue levels are present before incision15 and among patients receiving antibiotics within the hour before incision.16,17 Until recently, the prevailing practice in cesarean delivery procedures was to administer antibiotics after umbilical cord clamping because of concerns of masking neonatal sepsis.18 Several recent studies have reported that antibiotic prophylaxis before skin incision reduced postcesarean infections without affecting neonatal outcomes.19–21 In 2011, practice guidelines recommended antibiotic prophylaxis for all patients to be administered within 60 minutes before the start of cesarean delivery.22,23
A number of postcesarean infection prevention interventions were instituted at our hospital between 2003 and 2010. Time series analysis was used to estimate if those interventions had a significant effect on postcesarean surgical site infection in an 8-year cohort of patients who underwent cesarean deliveries after adjusting for underlying secular trends in patient mix.24
PATIENTS AND METHODS
The study was conducted in Barnes-Jewish Hospital in St. Louis, Missouri, a 1,250-bed tertiary care hospital with a referral base that includes eastern Missouri and western Illinois. From 2003 to 2010, there were approximately 3,600 deliveries annually at this hospital. In 2003, surgical site infection surveillance was implemented for cesarean delivery procedures. We collected monthly data on the number of cesarean delivery procedures, postcesarean surgical site infection incidence, and infection prevention practices within the hospital over an 8-year period from January 2003 to December 2010.
Several infection prevention interventions were undertaken during the study period as part of routine quality improvement. In January 2004, the anesthesiology department assumed responsibility of administering antibiotic prophylaxis for cesarean delivery cases. At this time, the hospital established the policy of administering prophylactic antibiotics before incision instead of postdelivery after umbilical cord clamping. In February 2004, a policy to ban artificial nails among surgical staff was instituted. In February 2007, there was a rapid improvement event to streamline the operating room environmental cleaning and disinfection processes. Another event of interest during the study period was the closure of the obstetric unit of another hospital in May 2006, leaving our hospital as the only obstetric care facility within St. Louis city limits. This was included in the analysis as well.
Cesarean delivery procedures were identified using the International Classification of Diseases, 9th Revision, Clinical Modification (ICD-9-CM) codes for obstetrical delivery by cesarean (74.0, 74.1, 74.2, 74.4, 74.91, and 74.99). Surgical site infection cases were identified by hospital infection prevention department routine surveillance using standard case definitions.25 We collected data on several patient characteristics to identify and adjust for secular trends in these patient factors. These factors were chosen because they are known risk factors for postcesarean surgical site infection and because they were accessible in the electronic medical records. Data on patient age, race, weight, height, and ICD-9-CM diagnosis codes were obtained retrospectively from electronic administrative and medical records. Patients with comorbidities and obstetric complications were identified based on the presence of the relevant ICD-9-CM diagnosis codes: diabetes or gestational diabetes (250, 648.0, 648.8), hypertension or preeclampsia (401–405, 642.0–642.4, 642.9), severe preeclampsia or eclampsia (642.5, 642.6, 642.7), placental abruption (641.2), premature rupture of membranes (658.1, 658.2), placenta previa (641.0, 641.1), fetal deceleration or fetal distress (659.7, 656.3) and delivery outcome, either singleton (V27.0, V27.1) or multiple birth (V27.2, V27.3, V27.4, V27.5, V27.6, V27.7). Obesity was defined as BMI 35 or greater to account for the fact that we were using weight at the time of delivery and based on previous studies that have reported BMI at term to be a risk factor for postcesarean surgical site infection.5,9 A total of 7.6% of our patient population had missing data for BMI. We determined the monthly proportion of obese patients for the remaining patients because the number of patients with missing BMI data were evenly distributed over time (data not shown). Eighty-eight (1.0%) of the patients had missing data for delivery outcome. Patients with twin or higher-order deliveries were categorized as multiple gestations.
All statistical analyses were performed using SPSS 18.0. Linear regression was used to determine if there were significant trends in patient demographic characteristics, comorbid complications, and obstetric complications over the study period. We evaluated both extremes of maternal age (18 years or younger and 35 years or older) as risk factors because younger maternal age has been shown to be associated with cesarean delivery complications.13,26
A multivariable autoregressive integrated moving average model27 was used to perform time series analysis on the 96-month cohort of patients undergoing cesarean delivery. This model analyzes temporal data as a function of its past values (autocorrelation), trend (integration), and abrupt changes in the recent past (moving average). SPSS Forecasting 19.0 Expert Modeler was used to determine the best fitting model. Monthly postcesarean surgical site infection rate was the dependent variable. The patient-related demographic and clinical variables as well as the three infection prevention interventions and the closure of the only other obstetric unit in the city were evaluated as potential explanatory variables one at a time using univariable autoregressive integrated moving average analysis. Variables that were statistically significant (P<.05) in the univariable analysis were evaluated for inclusion in a multivariable model. The stationary R2 coefficient of determination was estimated as an indicator of goodness of fit of the model and the proportion of the total variation in the data that is explained by the model. The adequacy of the autoregressive integrated moving average model was examined by estimating the Ljung-Box Q statistic28 and by assessing residuals for correlation. This study was approved by the Washington University Human Research Protection Office.
Over the 8-year period from January 2003 to December 2010, there were a total of 8,668 cesarean delivery procedures in Barnes-Jewish Hospital. Six hundred sixty (7.6%) patients were missing height data, weight data, or both (Table 1). There was no difference in age, year of surgery, or comorbidities among patients with missing BMI compared with those for whom it could be calculated (data not shown). Of the 8,008 patients for whom BMI could be calculated, 3,093 (35.7%) had BMIs of 35 or greater and the median BMI was 32.7 (range 13.4–103.0). One thousand one hundred twenty-four (13.0%) of the patients were of advanced maternal age (35 years or older) and 690 (8.0%) were 18 years old or younger at the time of surgery. Of maternal comorbidities examined in this cohort, 898 (10.4%) of the patients had diabetes or gestational diabetes, 1,523 (17.6%) had hypertension or mild preeclampsia, and 974 (11.2%) had severe preeclampsia or eclampsia. Five hundred eighty-three (6.7%) of the procedures involved patients with multiple gestations. Eight thousand two hundred five (94.7%) of the procedures were low cervical cesarean delivery procedures and 364 (4.2%) were classical cesarean delivery procedures. The median hospital stay was 4.4 days (range 1.0–116.9 days, interquartile range 4.0–5.1 days).
Temporal trends in patient demographic characteristics, comorbid conditions, and obstetric complications over the study period are shown in Table 2. Patient characteristics, including maternal race, age, diabetes (including gestational diabetes), multiple gestation, and obstetric complications, did not change significantly over the study period. However, the proportion of patients undergoing cesarean deliveries who had BMIs of 35 or greater, hypertension or mild preeclampsia, and severe preeclampsia or eclampsia increased over time.
Surgical site infections occurred in 303 (3.5%, 95% confidence interval 3.1–3.9%) of the 8,668 procedures. Figure 1 shows the monthly postcesarean surgical site infection rate during the study period and the timeline of three infection prevention interventions and the obstetric service closure event that were evaluated for potential effect on the monthly postcesarean surgical site infection rate.
Univariable autoregressive integrated moving average analysis showed a significant relationship between the incidence of postcesarean surgical site infection and a number of factors: white race, age 35 years or older, BMI 35 or greater, antibiotic prophylaxis policy, and artificial nails policy. These five predictors identified in univariable analysis were introduced together into the SPSS Expert Modeler for multivariable analysis (Table 3). The three main parameters selected when fitting an autoregressive integrated moving average model are p (the order of autocorrelation), d (the order of integration, which indicates trend), and q (the order of moving average, which accounts for the shock effects in the model). Hence, the process is denoted as autoregressive integrated moving average (p, d, q). The first parenthesis refers to the nonseasonal version and the second parenthesis refers to the seasonal version of these terms. The best fitting model for our data was autoregressive integrated moving average (0, 1, 1) (0, 0, 0), which indicates a simple exponential smoothing model with no autocorrelation, a linear trend requiring nonseasonal differencing, and lagging shock effects by 1, indicating that the mean value of the series from the last time period be considered when predicting the current value of the series. Of the five predictors that we included in the multivariable model, the prophylactic antibiotic policy was the only factor significantly associated with postcesarean surgical site infection rates. Implementation of the policy to administer prophylactic antibiotics within 1 hour before incision instead of at the time of cord clamp led to a 48% reduction in cesarean delivery surgical site infection (Δ=−5.4 surgical site infections per 100 cesarean deliveries; P<.001) (Table 3). The Ljung-Box goodness-of-fit measure (Q statistic) indicates that the model adequately captures the correlation information in the time series.
Development of surgical site infection is a complex event with many underlying causes including patient and surgical risk factors. Methods to evaluate strategies to reduce surgical site infection rates must adjust for multiple factors that could account for changing rates such as changes in patient case mix as well as the relationship between surgical site infection rates in consecutive months. Unlike classical regression methods in which the observed data are assumed to be independent random variables, time series models take into account the possible correlation between consecutive observations and allow traditional regression and inference testing in temporal data.
Understanding and accounting for patient mix is important when surgical site infection rates are being compared across institutions, especially given increasing public reporting of data.29 In this patient population, the monthly proportion of individuals with three comorbidities increased over time: obesity, hypertension or mild preeclampsia, and severe preeclampsia or eclampsia. Hypertensive disorders have been previously reported to increase the risk of surgical site infection in patients undergoing cesarean delivery possibly by a chronic alteration of the peripheral blood supply resulting from increased vascular resistance.11,30 Higher BMI has been found to be an independent risk factor for postcesarean surgical site infection in several studies.5,6,9–13 The upward temporal trend in obesity among our patients is consistent with national trends. Among reproductive-aged women in the United States, the prevalence of obesity has more than doubled between 1976 and 2004.30,31
We used time series analysis to study the effect of a change in policy from administering prophylactic antibiotics to patients undergoing cesarean delivery at cord clamp to administering them before incision. We found that even after accounting for secular trends, the policy change led to a nearly 50% reduction in the incidence of postcesarean surgical site infection. The drop in surgical site infection incidence was sustained over a period of several years despite an increase in obesity among our patient population. A policy to ban artificial nails among surgical staff and an intervention to standardize operating room cleaning and disinfection had no effect on surgical site infection rates.
Our results support recent studies on the benefit of prophylactic antibiotic delivery before incision for preventing surgical site infection after cesarean delivery. A meta-analysis of three randomized controlled trials showed that preoperative administration of cefazolin was associated with a 53% reduction in the risk of postpartum endometritis and a 50% reduction in the risk of total infectious morbidity.19 It is interesting to note that our hospital instituted the policy of administering preincision prophylactic antibiotics to patients undergoing cesarean delivery in January 2004, when the prevailing practice was to wait until after cord clamp to administer antibiotics.
Strengths of this study include inclusion of a large cohort of patients over an 8-year study period. Also, surgical site infection data were collected independently from the study as part of routine hospital infection prevention surveillance. One limitation to our study is the use of administrative data for comorbidities, because misclassification may have occurred. Also, some measures that may have affected surgical site infection rates were unable to be assessed (eg, selection of antibiotic and timing of antibiotic administration for individual surgeries). We had no data for prepregnancy weight and used weight at the time of delivery to calculate BMI. Our study population was restricted to a single urban, academic center and the results may not be directly generalizable to other settings.
In conclusion, our study demonstrated the application of time series analysis in evaluating the effectiveness of infection prevention interventions while controlling for secular trends in patient risk factors. Our results support the administration of prophylactic antibiotics before incision in patients undergoing cesarean delivery. Future research should aim at understanding adherence to this policy and best practices for delivering prophylactic antibiotics in a timely manner in both routine and emergency obstetric situations.
1. Martin JA, Hamilton BE, Sutton PD, Ventura SJ, Mathews TJ, Kirmeyer S, et al.. Births: final data for 2007. Natl Vital Stat Rep 2010;58:1–85.
2. Declercq E, Barger M, Cabral HJ, Evans SR, Kotelchuck M, Simon C, et al.. Maternal outcomes associated with planned primary cesarean births compared with planned vaginal births. Obstet Gynecol 2007;109:669–77.
3. Olsen MA, Butler AM, Willers DM, Gross GA, Hamilton BH, Fraser VJ. Attributable costs of surgical site infection and endometritis after low transverse cesarean delivery. Infect Control Hosp Epidemiol 2010;31:276–82.
4. Edwards JR, Peterson KD, Mu Y, Banerjee S, Allen-Bridson K, Morrell G, et al.. National Healthcare Safety Network (NHSN) report: data summary for 2006 through 2008, issued December 2009. Am J Infect Control 2009;37:783–805.
5. DeVivo A, Mancuso A, Giacobbe A, Priolo AM, De DR, Maggio SL. Wound length and corticosteroid administration as risk factors for surgical-site complications following cesarean section. Acta Obstet Gynecol Scand 2010;89:355–9.
6. Johnson A, Young D, Reilly J. Caesarean section surgical site infection surveillance. J Hosp Infect 2006;64:30–5.
7. Killian CA, Graffunder EM, Vinciguerra TJ, Venezia RA. Risk factors for surgical-site infections following cesarean section. Infect Control Hosp Epidemiol 2001;22:613–7.
8. Mitt P, Lang K, Peri A, Maimets M. Surgical-site infections following cesarean section in an Estonian university hospital: postdischarge surveillance and analysis of risk factors. Infect Control Hosp Epidemiol 2005;26:449–54.
9. Olsen MA, Butler AM, Willers DM, Devkota P, Gross GA, Fraser VJ. Risk factors for surgical site infection after low transverse cesarean section. Infect Control Hosp Epidemiol 2008;29:477–84.
10. Opoien HK, Valbo A, Grinde-Andersen A, Walberg M. Post-cesarean surgical site infections according to CDC standards: rates and risk factors. A prospective cohort study. Acta Obstet Gynecol Scand 2007;86:1097–102.
11. Schneid-Kofman N, Sheiner E, Levy A, Holcberg G. Risk factors for wound infection following cesarean deliveries. Int J Gynaecol Obstet 2005;90:10–5.
12. Tran TS, Jamulitrat S, Chongsuvivatwong V, Geater A. Risk factors for postcesarean surgical site infection. Obstet Gynecol 2000;95:367–71.
13. Ward VP, Charlett A, Fagan J, Crawshaw SC. Enhanced surgical site infection surveillance following caesarean section: experience of a multicentre collaborative post-discharge system. J Hosp Infect 2008;70:166–73.
14. Bratzler DW. The Surgical Infection Prevention and Surgical Care Improvement Projects: promises and pitfalls. Am Surg 2006;72:1010–6.
15. Burke JF. The effective period of preventive antibiotic action in experimental incisions and dermal lesions. Surgery 1961;50:161–8.
16. Classen DC, Evans RS, Pestotnik SL, Horn SD, Menlove RL, Burke JP. The timing of prophylactic administration of antibiotics and the risk of surgical-wound infection. N Engl J Med 1992;326:281–6.
17. Steinberg JP, Braun BI, Hellinger WC, Kusek L, Bozikis MR, Bush AJ, et al.. Timing of antimicrobial prophylaxis and the risk of surgical site infections: results from the Trial to Reduce Antimicrobial Prophylaxis Errors. Ann Surg 2009;250:10–6.
18. Prophylactic antibiotics in labor and delivery. Practice Bulletin No. 47. American College of Obstetricians and Gynecologists. Obstet Gynecol 2003;102:875–82.
19. Costantine MM, Rahman M, Ghulmiyah L, Byers BD, Longo M, Wen T, et al.. Timing of perioperative antibiotics for cesarean delivery: a metaanalysis. Am J Obstet Gynecol 2008;199:301.e1–6.
20. Lamont RF, Sobel JD, Kusanovic JP, Vaisbuch E, Mazaki-Tovi S, Kim SK, et al.. Current debate on the use of antibiotic prophylaxis for caesarean section. BJOG 2011;118:193–201.
21. Sullivan SA, Smith T, Chang E, Hulsey T, Vandorsten JP, Soper D. Administration of cefazolin prior to skin incision is superior to cefazolin at cord clamping in preventing postcesarean infectious morbidity: a randomized, controlled trial. Am J Obstet Gynecol 2007;196:455.e1–5.
22. Use of prophylactic antibiotics in labor and delivery. Practice Bulletin No. 120. American College of Obstetricians and Gynecologists. Obstet Gynecol 2011;117:1472–83.
23. Antimicrobial prophylaxis for cesarean delivery: timing of administration. Committee Opinion No. 465. American College of Obstetricians and Gynecologists. Obstet Gynecol 2010;116:791–2.
24. Zeger SL, Irizarry R, Peng RD. On time series analysis of public health and biomedical data. Annu Rev Public Health 2006;27:57–79.
25. Horan TC, Andrus M, Dudeck MA. CDC/NHSN surveillance definition of health care-associated infection and criteria for specific types of infections in the acute care setting. Am J Infect Control 2008;36:309–32.
26. Olsen MA, Butler AM, Willers DM, Gross GA, Devkota P, Fraser VJ. Risk factors for endometritis after low transverse cesarean delivery. Infect Control Hosp Epidemiol 2010;31:69–77.
27. Box GEP, Jenkins GM, Reinsel GC. Time series analysis: forecasting and control. 3rd ed. Englewood Cliffs (NJ): Prentice Hall; 1994.
28. Ljung GM, Box GEP. On a measure of lack of fit in time series models. Biometrika 1978;65:297–303.
29. Association of Professionals in Infection Control and Epidemiology. Available at: www.apic.org
. Retrieved January 17, 2012.
30. Cardoso Del Monte MC, Pinto Neto AM. Postdischarge surveillance following cesarean section: the incidence of surgical site infection and associated factors. Am J Infect Control 2010;38:467–72.
31. Rasmussen K, Yaktine A. Weight gain during pregnancy: reexamining the guidelines. Washington (DC): The National Academies Press Institute of Medicine; 2009.