In recent decades, indications for use of antibiotics during delivery hospitalizations have become increasingly specific, well-defined, and supported by evidence.1–3 Benefits of antibiotic administration include both reduction of maternal infection morbidity and improvement of neonatal outcomes. Guideline-based administration of antibiotics to prevent perinatal group B streptococcal (GBS) disease has been associated with reduced adverse outcomes from this cause.3–6 Reduction of infection risk after cesarean delivery with routine antibiotic prophylaxis has been established.7 Administration of antibiotics in the setting of preterm prelabor rupture of membranes (PROM) is associated with reduced neonatal morbidity and prolonged latency between presentation and delivery.1,8
Although appropriate antibiotic use may result in significant maternal and neonatal benefits, inappropriate use of these medications may result in unnecessary maternal and neonatal risks and contribute to antibiotic resistance (Tan S, Holliman R, Russell AR. Hazards of widespread use of erythromycin for preterm prelabour rupture of membranes [letter]. Lancet 2003;361:437.).9–12 Risk for early-onset neonatal sepsis with ampicillin-resistant non-GBS organisms may be associated with broad use of antenatal ampicillin.13 Antibiotic exposure in utero may be associated with increased risk for pediatric allergic disease14–16; antibiotics may alter the neonatal microbiome, which in turn may have consequences for long-term health outcomes.15
Currently, trends in use of antibiotics during delivery hospitalizations in the United States are not well characterized. Characterizing antibiotic use in specific obstetric scenarios is of importance in determining whether evidence-based recommendations are being adopted.17 Given this knowledge gap, the purpose of this study was to evaluate antibiotic trends during delivery hospitalizations in the United States.
The Premier Perspective database was used for this serial cross-sectional analysis. It is an administrative inpatient database that reports on 100% of hospitalizations for 600 individual hospitals and ambulatory surgery centers across the United States and includes approximately 15% of hospitalizations nationally. It includes information on patient demographics, hospital characteristics, medications and devices received during hospitalizations, as well as other information. Complete patient billing, hospital cost, and coding histories are contained in the database. The Perspective database is commercially available and is commonly used across medical specialties for research that evaluates inpatient medication use.18–23 The Perspective database is maintained by Premier Incorporated. On receiving data from participating hospitals, Premier undertakes an extensive, seven-part data-validation and correction process that includes more than 95 quality-assurance checks before the data are used for research.24 After validations are complete, the data are moved to the Perspective data warehouse to populate and maintain the databases for health services research.25 This database includes data from hospitals in diverse geographic regions in the United States, from both teaching and nonteaching institutions. The Columbia University Institutional Review Board deemed the study exempt given that the data were de-identified.
For this analysis, all women aged 18–54 years who underwent a delivery hospitalization from January 2006 through March 2015 were included. Delivery hospitalizations were identified based on International Classification of Diseases, 9th Revision, Clinical Modification (ICD-9-CM) billing codes, using an approach that ascertains more than 95% of deliveries (ICD-9-CM codes 650 and V27.x).26 The primary objective of this study was to estimate temporal trends in antibiotic administration without indication during vaginal delivery hospitalizations. The primary outcome was receipt of antibiotics during these hospitalizations. Evidence-based indications for antibiotic administration during vaginal deliveries included 1) preterm PROM; 2) endometritis; 3) chorioamnionitis; 4) GBS colonization; and 5) other infectious complications such urinary tract infection, pneumonia, sepsis, and other major infections (Appendix 1, available online at http://links.lww.com/AOG/B538). These indications were identified by ICD-9-CM diagnosis codes. For a diagnosis of preterm PROM, diagnosis of both PROM and preterm delivery were required. Because the development of research evidence and guidelines related to antibiotic prophylaxis for third- and fourth-degree vaginal lacerations, uterine tamponade, and manual extraction of the placenta occurred coincident to the study period,17,27,28 women with these conditions but not another indication for antibiotics were excluded from the analysis. Because antibiotics in the setting of preterm PROM and GBS colonization are administered for neonatal benefit, patients with these diagnoses were excluded if fetal death was present and there was no other maternal indication for antibiotics. Cesarean delivery was classified based on ICD-9-CM procedure and diagnosis codes. As secondary outcomes, we evaluated temporal trends in antibiotic administration for 1) all cesarean deliveries and 2) vaginal deliveries with an evidence-based indication for antibiotic administration. Temporal trends were analyzed with the Cochran-Armitage test.
The Perspective database was queried for antibiotics that are commonly used during delivery hospitalizations. Use of penicillins, first through fourth generation cephalosporins, aminoglycosides, carbapenems, macrolides, fluoroquinolones, tetracyclines, as well as other drug classes was ascertained (Appendix 2, available online at http://links.lww.com/AOG/B538). Both generic and trade names were queried. Hospital, patient, obstetric, and medical characteristics associated with receipt of antibiotics during 1) cesarean delivery, 2) vaginal delivery with an evidence-based indication for antibiotic administration, and 3) vaginal delivery without an evidence-based indication for antibiotic administration were determined (Table 1). Univariable associations are presented as unadjusted risk ratios (RRs) with 95% CIs as measures of association.
We fit multivariable log-linear regression models with Poisson distribution and log link based on generalized estimating equations, including demographic, hospital, and medical and obstetric factors accounting for hospital clustering to determine which characteristics were associated with antibiotic administration for each of the three delivery scenarios. Multiple deliveries occurring to the same woman can be captured in the Perspective database if they occur at the same hospital; these models accounted for multiple deliveries to the same patient. Results are reported as adjusted RRs with 95% CIs as measures of association.
Three sensitivity analyses were performed for this study. First, hospitals are included in the Perspective database for varying durations of time. To account for confounding secondary to this changing sampling frame, we performed a sensitivity analysis restricting the population to only those hospitals that contributed data from January 2006 to December 2014. The first quarter of 2015 was not included in this sensitivity analysis because of smaller sample sizes and less precise estimates.
Second, erythromycin is rarely indicated as a first-line medication for inpatient obstetric scenarios other than for preterm PROM and for women with penicillin and cephalosporin allergies. Erythromycin eye ointment is routinely administered shortly after birth, a practice supported by the U.S. Preventive Services Task Force.1 Given that there is a possibility for misclassification of erythromycin administration appearing in maternal as opposed to neonatal drug files when administered to newborns, we performed a sensitivity analysis evaluating temporal trends in antibiotics excluding erythromycin. For this sensitivity analysis, we excluded women with penicillin allergy (ICD-9-CM code V14.0), which, aside from preterm PROM, is the most likely maternal indication for erythromycin. For the third sensitivity analysis, we repeated the second sensitivity analysis, additionally excluding scenarios in which women may have received antibiotics because their GBS status was unknown: 1) preterm labor, 2) preterm delivery, 3) PROM, and 4) delayed or prolonged labor. All analyses were performed with SAS 9.4.
Of 5,657,523 delivery hospitalizations, 120,727 (2.1%) were excluded based on the above criteria, leaving 5,536,756 delivery hospitalizations from 2006 through the first quarter of 2015 included in the analysis. This total included 2,872,286 vaginal deliveries without indication for antibiotics, 765,096 vaginal deliveries with an evidence-based indication for antibiotic administration, and 1,899,374 cesarean deliveries. The most common indication for antibiotics was cesarean delivery (33.6% of the entire cohort), followed by GBS colonization (15.8%), chorioamnionitis (1.7%), preterm PROM (1.6%), endometritis (1.2%), urinary tract infection (0.6%), and other infections (total less than 0.5%). Comparing patients undergoing vaginal delivery with an indication for antibiotics with those without, the former group were significantly more likely be aged 35 years or older, single, and black (P<.01 for all) (Table 1).
For the primary outcome, 30.6% of women with a vaginal delivery and no indication for antibiotic administration received an antibiotic during the study period. In comparison 88.9% of women undergoing cesarean delivery and 80.6% of women with a vaginal delivery and an evidence-based indication for antibiotic administration received an antibiotic. Over the study period, the proportion of women receiving antibiotics with a vaginal delivery without indication decreased 44.4%, from 38.1% in 2006 to 21.2% in 2015 (Fig. 1A). Over this period, the proportion of women receiving antibiotics with a vaginal delivery and an evidence-based indication for antibiotic administration decreased 9.1%, from 84.0% to 76.4% (Fig. 1B), and increased for cesarean delivery by 2.5%, from 87.3% to 89.5% (Fig. 1C). Trends for all three delivery categories were significant with the Cochran-Armitage test (P<.01).
For vaginal deliveries with no indication for antibiotics, the most commonly administered antibiotic was erythromycin (16.3% of patients), followed by ampicillin (7.8%) and penicillin (5.0%) (Appendix 3, available online at http://links.lww.com/AOG/B538). For cesarean delivery hospitalizations, the most common antibiotic administered was cefazolin (68.9% of patients), followed by erythromycin (14.9%) and ampicillin (9.3%). For vaginal delivery hospitalizations with an indication for antibiotics, the most commonly administered antibiotics included penicillin (35.0%), ampicillin (32.2%), and erythromycin (17.4%).
Temporal trends for the primary outcome were similar in the adjusted vs the unadjusted analysis. Adjusted risk for receipt of antibiotics during vaginal delivery hospitalizations without an indication for antibiotics was lower in 2015 vs 2006 (adjusted RR 0.56, 95% CI 0.55–0.57) (Table 2). Factors associated with increased antibiotic receipt during vaginal delivery hospitalizations in both the unadjusted and adjusted models included black vs white race (adjusted RR 1.26, 95% CI 1.25–1.27, respectively), urban vs rural hospitals (adjusted RR 1.36, 95% CI 1.35–1.38, respectively), hospitals in the South vs the Northeast (adjusted RR 1.42, 95% CI 1.41–1.43, respectively), multiple gestation (adjusted RR 1.47, 95% CI 1.44–1.51, respectively), and postpartum hemorrhage (adjusted RR 1.33, 95% 1.31–1.34, respectively).
In the adjusted model for receipt of antibiotics during vaginal delivery hospitalizations with an indication for antibiotics, risk was lower in 2015 vs 2006 (adjusted RR 0.92, 95% 0.91–0.92), similar to the unadjusted models (Appendix 4, available online at http://links.lww.com/AOG/B538). For this outcome, Midwest vs Northeast hospital location was associated with increased antibiotic receipt in both the unadjusted and adjusted models (RR 1.24, 95% CI 1.23–1.24, adjusted RR 1.21, 95% 1.21–1.22, respectively); other factors were not major predictors of antibiotic receipt. For the adjusted model for receipt of antibiotics during cesarean delivery hospitalizations, Northeast region was associated with decreased antibiotic administration in the adjusted and unadjusted models; other factors were not major predictors of antibiotic receipt (Appendix 5, available online at http://links.lww.com/AOG/B538).
Results of the three sensitivity analyses aligned with those of the primary outcome analysis. When the analysis was limited to hospitals that contributed data throughout the entire January 2006–March 2015 study period similar, trends in antibiotic administration were similar: antibiotic use during vaginal delivery hospitalizations without an indication for antibiotic administration decreased from 37.7% to 19.4% from 2005 to 2014, decreased from 81.6% to 75.8% for vaginal delivery hospitalizations with an indication for antibiotics, and decreased from 87.7% to 87.6% for cesarean deliveries (Appendices 6–8, available online at http://links.lww.com/AOG/B538). In the second sensitivity analysis restricted to women without a penicillin allergy undergoing vaginal delivery without an indication for antibiotic administration, a decrease in antibiotic administration was also noted. Excluding erythromycin and women with penicillin allergies, 20.2% of women undergoing vaginal delivery without an indication for antibiotics received antibiotics in 2006 vs 15.1% in 2015, a 25.2% decrease (Fig. 2). Additionally, excluding women with PROM, preterm delivery, preterm labor, and delayed or prolonged labor to account for antibiotic administration to women with unknown GBS status, use of inappropriate antibiotics decreased from 17.5% in 2006 to 13.1% in 2015, a 25.4% decrease (Appendix 9, available online at http://links.lww.com/AOG/B538).
Our study demonstrated that, for women undergoing vaginal delivery without an indication for antibiotic receipt, use of antibiotics decreased more than 40%, with sensitivity analysis demonstrating significantly decreased rates of antibiotic use. These findings support that evidence-based rationales for antibiotic use are increasingly being adopted into clinical practice. This change in clinical practice may be beneficial in reducing antibiotic resistance and reducing risk of adverse reactions to unnecessary medications and may have important downstream health effects.
According to the Centers for Disease Control and Prevention, 20–50% of all antibiotics prescribed during acute care hospitalizations in the United States are unnecessary or inappropriate2; this antibiotic use has led to increased antibiotic resistance. Evidence-based guidelines have been established to characterize specific indications, timing, duration, and type of antibiotic use in labor and delivery.29 Childbirth represents one of the most common indications for acute care hospitalizations, and less frequent use of inappropriate antibiotics would represent an important public health achievement. Further research regarding care quality related to antibiotic management is indicated based on these findings. Specifically, informatics data from hospital-system electronic medical records may be able to provide more granular detail in demonstrating appropriate and timely administration of antibiotics based on individual diagnoses. Although the overall trends in our analysis demonstrated improvements in indication-based antibiotic use, it is possible that there may be important variation in care in specific obstetric scenarios.
Our study has several strengths. First, it has a large sample size, including more than 5 million delivery hospitalizations, and it is reflective of the practice patterns of a large proportion of deliveries within the United States. The database provides information on a wide range of demographics and different hospital settings and is representative of a geographically and clinically diverse sample. Our analysis was strengthened by a sensitivity analysis restricted to hospitals contributing data over the entire period, demonstrating that trends were not a result of the changing sampling frame of hospitals in the database.
In interpreting study results, there are several important limitations. First, with administrative data, which are used mainly for billing purposes, there is concern for misclassification and under-ascertainment of both medications and diagnoses. Specifically, in our analysis there was a larger than anticipated proportion of women receiving erythromycin. We thus performed a sensitivity analysis excluding erythromycin among women undergoing vaginal delivery with no other indication for antibiotics; this analysis found that, excluding erythromycin, the decrease in the proportion of women receiving antibiotics without an indication was attenuated. Although the high rate of erythromycin use is likely a result of neonatal administration assigned to the maternal drug file, we were not able to evaluate more granular, chart-level data to validate actual medication receipt. It is possible that, because of erythromycin misclassification, the true decreased incidence of unindicated antibiotic use may be of lesser magnitude than in the primary analysis.
Second, a significant proportion of women undergoing cesarean or vaginal delivery with an indication for antibiotics did not receive an antibiotic; further research with more granular data is indicated to determine to what degree our findings may be a result of suboptimal care compared with under-ascertainment of diagnoses in administrative data. To some degree, trends noted could have been the result of improved ascertainment and documentation over the study period, with indications being more commonly captured in administrative data; likewise, some women who received antibiotics may have had an indication for receipt not captured in coding. However, that there was a relatively large decrease in unindicated antibiotic administration over the study period supports that these findings are in fact representative of change in clinical practice.
A third limitation is that our results reflect practices up to 2015; there could have been subsequent improvement in obstetric antibiotic administration since that time that is not examined in our study. Additionally, given that the data derive from an administrative database, we could not perform chart review to analyze the causes of noncompliance with antibiotic guidelines. Finally, we are not able to evaluate a number of important characteristics related to appropriate use of antibiotics, including timing of administration relative to diagnosis of individual conditions; whether the antibiotics were administered antepartum, intrapartum, or postpartum; whether individual drug dosages and duration of therapy were appropriate; and whether medications were continued on an outpatient basis when indicated.
Over the study period, administration of antibiotics during vaginal delivery hospitalizations without an indication for antibiotic use declined significantly. This trend represents a meaningful improvement in obstetric care. Hospital-system level clinical studies further characterizing antibiotic use with granular informatics data are indicated to further analyze care quality and optimization with regard to use of antibiotics.
1. Curry SJ, Krist AH, Owens DK, Barry MJ, Caughey AB, et al. Ocular prophylaxis for gonococcal ophthalmia neonatorum: US Preventive Services Task Force reaffirmation recommendation statement. JAMA 2019;321:394–8.
2. CDC. Core elements of hospital antibiotic stewardship programs. Atlanta, GA: US Department of Health and Human Services, CDC; 2014. Available at: https://www.cdc.gov/antibiotic-use/healthcare/implementation/core-elements.html
. Retrieved August 6, 2019.
3. Prevention of group B streptococcal early-onset disease in newborns. ACOG Committee Opinion No. 782. American College of Obstetricians and Gynecologists. Obstet Gynecol 2019;134:e19–40.
4. Verani JR, McGee L, Schrag SJ; Division of Bacterial Diseases National Center for Immunization and Respiratory Diseases, Centers for Disease Control and Prevention. Prevention of perinatal group B streptococcal disease—revised guidelines from CDC, 2010. MMWR Recomm Rep 2010;59:1–36.
5. Van Dyke MK, Phares CR, Lynfield R, Thomas AR, Arnold KE, Craig AS, et al. Evaluation of universal antenatal screening for group B streptococcus. N Engl J Med 2009;360:2626–36.
6. Lukacs SL, Schrag SJ. Clinical sepsis in neonates and young infants, United States, 1988-2006. J Pediatr 2012;160:960–5.e1.
7. Smaill FM, Gyte GM. Antibiotic prophylaxis versus no prophylaxis for preventing infection after cesarean section. The Cochrane Database of Systematic Reviews 2010, Issue 1. Art. No.: CD007482. doi:10.1002/14651858.CD007482.pub2.
8. Mercer BM, Miodovnik M, Thurnau GR, Goldenberg RL, Das AF, Ramsey RD, et al. Antibiotic therapy for reduction of infant morbidity after preterm premature rupture of the membranes. A randomized controlled trial. National Institute of Child Health and Human Development Maternal-Fetal Medicine Units Network. JAMA 1997;278:989–95.
9. Bland ML, Vermillion ST, Soper DE, Austin M. Antibiotic resistance patterns of group B streptococci in late third-trimester rectovaginal cultures. Am J Obstet Gynecol 2001;184:1125–6.
10. Mercer BM, Carr TL, Beazley DD, Crouse DT, Sibai BM. Antibiotic use in pregnancy and drug-resistant infant sepsis. Am J Obstet Gynecol 1999;181:816–21.
11. Rao GA, Mann JR, Shoaibi A, Bennett CL, Nahhas G, Sutton SS, et al. Azithromycin and levofloxacin use and increased risk of cardiac arrhythmia and death. Ann Fam Med 2014;12:121–7.
12. Martinez de Tejada B. Antibiotic use and misuse during pregnancy and delivery: benefits and risks. Int J Environ Res Public Health 2014;11:7993–8009.
13. Towers CV, Carr MH, Padilla G, Asrat T. Potential consequences of widespread antepartal use of ampicillin. Am J Obstet Gynecol 1998;179:879–83.
14. Risnes KR, Belanger K, Murk W, Bracken MB. Antibiotic exposure by 6 months and asthma and allergy at 6 years: findings in a cohort of 1,401 US children. Am J Epidemiol 2011;173:310–8.
15. Kuperman AA, Koren O. Antibiotic use during pregnancy: how bad is it? BMC Med 2016;14:91.
16. Stensballe LG, Simonsen J, Jensen SM, Bonnelykke K, Bisgaard H. Use of antibiotics during pregnancy increases the risk of asthma in early childhood. J Pediatr 2013;162:832–8.e3.
17. Use of prophylactic antibiotics in labor and delivery. ACOG Practice Bulletin No. 199. American College of Obstetricians and Gynecologists. Obstet Gynecol 2018;132:e103–19.
18. Fang MC, Maselli J, Lurie JD, Lindenauer PK, Pekow PS, Auerbach AD. Use and outcomes of venous thromboembolism prophylaxis after spinal fusion surgery. J Thromb Haemost 2011;9:1318–25.
19. Kulik A, Rassen JA, Myers J, Schneeweiss S, Gagne J, Polinski JM, et al. Comparative effectiveness of preventative therapy for venous thromboembolism after coronary artery bypass graft surgery. Circ Cardiovasc Interv 2012;5:590–6.
20. Prabhakaran S, Herbers P, Khoury J, Adeoye O, Khatri P, Ferioli S, et al. Is prophylactic anticoagulation for deep venous thrombosis common practice after intracerebral hemorrhage? Stroke 2015;46:369–75.
21. Ritch JM, Kim JH, Lewin SN, Burke WM, Sun X, Herzog TJ, et al. Venous thromboembolism and use of prophylaxis among women undergoing laparoscopic hysterectomy. Obstet Gynecol 2011;117:1367–74.
22. Wright JD, Lewin SN, Shah M, Burke WM, Lee SM, Sun X, et al. Quality of venous thromboembolism prophylaxis in patients undergoing oncologic surgery. Ann Surg 2011;253:1140–6.
23. Zacharia BE, Youngerman BE, Bruce SS, Hershman DL, Neugut AI, Bruce JN, et al. Quality of postoperative venous thromboembolism prophylaxis in Neuro-oncologic surgery. Neurosurgery 2017;80:73–81.
24. Stulberg JJ, Delaney CP, Neuhauser DV, Aron DC, Fu P, Koroukian SM. Adherence to surgical care improvement project measures and the association with postoperative infections. JAMA 2010;303:2479–85.
25. Swanson SJ, Meyers BF, Gunnarsson CL, Moore M, Howington JA, Maddaus MA, et al. Video-assisted thoracoscopic lobectomy is less costly and morbid than open lobectomy: a retrospective multiinstitutional database analysis. Ann Thorac Surg 2012;93:1027–32.
26. Kuklina EV, Whiteman MK, Hillis SD, Jamieson DJ, Meikle SF, Posner SF, et al. An enhanced method for identifying obstetric deliveries: implications for estimating maternal morbidity. Matern Child Health J 2008;12:469–77.
27. WHO guidelines for the management of postpartum haemorrhage and retained placenta. Geneva (Switzerland): World Health Organization; 2009. p. 19–20.
28. Duggal N, Mercado C, Daniels K, Bujor A, Caughey AB, El-Sayed YY. Antibiotic prophylaxis for prevention of postpartum perineal wound complications: a randomized controlled trial. Obstet Gynecol 2008;111:1268–73.
29. Sullivan SA, Soper D. Antibiotic prophylaxis in obstetrics. Am J Obstet Gynecol 2015;212:559–60.