Sepsis is an increasingly important cause of maternal morbidity in developed nations. According to the Centre for Maternal and Child Enquires (formerly known as the Confidential Enquiry into Maternal and Child Health), the longest running continuous series of clinical audits in the world, deaths due to sepsis have nearly doubled over the past decade from 0.65 per 100,000 maternities during the years 2000 through 2002 to 1.13 per 100,000 for 2006 through 2008, making sepsis the leading cause of direct maternal death in the United Kingdom.1
While this increase has prompted considerable attention in the United Kingdom, very little is known about the epidemiology of sepsis in pregnancy in the United States, and there has not been a population-based epidemiologic study of sepsis during hospitalization for delivery in the United States. Estimates of frequency are based solely on retrospective reviews from tertiary care centers (from 1:1763 to 1:2517 deliveries for sepsis to 1:7654 to 1:8338 deliveries for septic shock).2–4 Data are limited regarding conditions associated with this complication in pregnancy.
The objectives of this study were to determine frequency, temporal trends, and independent associations for severe sepsis during hospitalization for delivery using the largest administrative dataset of admissions available in the United States, the Nationwide Inpatient Sample (NIS).
The University of Michigan IRB evaluated the study protocol and determined that the study was exempt from further review because it is an analysis of publicly available, de-identified data. The requirement for written informed consent was waived by the IRB.
Data for the years 1998 through 2008 were obtained from the NIS. This national all-payer, publicly available database is maintained by the Agency for Healthcare Research and Quality as part of the Healthcare Cost and Utilization Project. The NIS captures charge information on all patients admitted to a 20% sample of nonfederal acute care hospitals in the United States. Hospitals are selected based on a stratified sample accounting for 5 characteristics: bed size, ownership, location (urban or rural), teaching status, and geographical region. The sample is weighted for these 5 strata to produce a sample generalizable to the entire U.S. population. The Healthcare Cost and Utilization Project provides information about the methodology used to create the dataset including stratified sampling.a
Hospitalizations for delivery were identified using a validated algorithm based on a query of all diagnosis and procedure fields (Appendix 1, see Supplemental Digital Content 1, http://links.lww.com/AA/A596).5 For each discharge, analyzed data included patient age, race, admission source, payer, disposition, and up to 15 diagnoses and procedures codes using the International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM). Delivery center characteristics including annual delivery volume were calculated based on unique identifiers for each delivery hospital.
Among hospitalizations for delivery, the presence of sepsis was identified with the appropriate ICD-9-CM codes located in any of the primary or secondary diagnostic coding positions (Appendix 2, see Supplemental Digital Content 2, http://links.lww.com/AA/A597). The accepted definition of severe sepsis by the American College of Chest Physicians and Society of Critical Care Medicine is sepsis with acute organ dysfunction, hypotension, or hypoperfusion.6 We defined the occurrence of severe sepsis based on codes indicating acute organ dysfunction and/or hypotension concurrent with a diagnosis of sepsis (Appendix 3, see Supplemental Digital Content 3, http://links.lww.com/AA/A598), as described in previous studies.7–9 To report the severity of illness in the patients with severe sepsis over time, the rates of hemodialysis and prolonged mechanical ventilation (>96 consecutive hours) were obtained using the appropriate ICD-9-CM procedure codes. Sepsis-related death was defined by the presence of a disposition indicating death in the hospital with a concurrent diagnosis of sepsis.
The proportion of the types of organisms causing severe sepsis was identified from billing data reported for the cases for severe sepsis identified by ICD-9-CM codes (Appendix 2, see Supplemental Digital Content 2, http://links.lww.com/AA/A597). The infectious etiologies underlying the cases for severe sepsis and type of organ dysfunction among women with severe sepsis were also identified through ICD-9-CM codes.
Several states do not record race or payer type in the data submitted to the NIS. We designated missing categories for both race and payer to retain these subjects in the analysis and to maintain a representative study sample.
We derived estimates of the number of U.S. hospitalizations for each diagnostic and demographic category by weighting the patient level discharge data in the NIS files using the svyset command and svy prefix in Stata Statistical Software: Release 12 (StataCorp LP, College Station, TX). The svyset command includes the opportunity to specify the sampling unit (hospid), the sampling weight (discwt), and sampling strata (Stra_final), all variables that are reported in the full NIS dataset. All statistical testing accounted for the stratified sampling design of the NIS. The logistic regression analysis was used to assess for crude temporal trends for sepsis, severe sepsis, and sepsis-related death.
A logistic regression analysis was performed to identify independent associations of severe sepsis (Appendix 4, see Supplemental Digital Content 4, http://links.lww.com/AA/A599). Covariates that have been previously shown in the literature to be associated with sepsis during pregnancy or which are clinically plausible associations were included in the model.10–12 The primary analysis evaluated associations for severe sepsis with all covariates forced into the model; these included: age, race, type of insurance, retained products of conception, preterm premature rupture of membranes (PPROM), multiple gestation, cerclage placement (rescue cerclage was defined as a cerclage placed during the current hospital admission; all cerclages placed during another admission were designated as prophylactic), diabetes mellitus (including gestational), body mass index >30 kg/m2, malignancy, human immunodeficiency virus, chronic liver disease, chronic renal disease, congestive heart failure, systemic lupus erythematous (SLE), and institutional delivery volume. A second logistic regression analysis was undertaken by adding variables using a full model fit to evaluate covariates that may increase risk for severe sepsis but could also be implicated by reverse causality; these included cesarean delivery (CD) (separated into groups of elective CD and CD during labor with vaginal deliveries serving as the referent category), stillbirth, preterm delivery, and postpartum hemorrhage. Results from the multivariable analyses were reported as adjusted odds ratios with corresponding 95% confidence intervals (CIs). Population-attributable fractions were calculated for each factor with a statistically significant adjusted odds ratio to estimate the burden of severe sepsis at the population level using adjusted odds ratios as estimates of relative risk as used in previous studies.13
Statistical analyses were performed using IBM SPSS Statistics for Windows, Release 19.0 (IBM Corp, Armonk, NY) and Stata Statistical Software: Release 12. P values <0.05 were used to define statistical significance for all analyses.
There were an estimated 44,999,260 hospitalizations for delivery in the United States during the study period. Characteristics of patients with and without severe sepsis are shown in Table 1. Sepsis complicated 1:3333 deliveries (95% CI, 1:3151–1:3540), severe sepsis complicated 1:10,823 (95% CI, 1:10,000–1:11,792) deliveries, and sepsis-related death complicated 1:105,384 deliveries (95% CI, 1:83,333–1:131,579). While the overall frequency of sepsis was stable across the study period (P = 0.95), the odds of acquiring severe sepsis increased 10% per year, with an odds ratio of 1.10 (95% CI, 1.07–1.13); sepsis-related death also increased 10% per year, with an odds ratio of 1.10 (95% CI, 1.02–1.18). These trends are depicted in Figure 1.
The multivariable logistic regression analysis of independent associations for severe sepsis is reported in Table 2. Risk factors of age >35 years, African American race, Medicaid insurance, retained products of conception, PPROM, congestive heart failure, chronic liver failure, chronic renal failure, human immunodeficiency virus infection, SLE, multiple gestation, and cerclage were all found to have independent associations for severe sepsis. In Appendix 5 (see Supplemental Digital Content 5, http://links.lww.com/AA/A600), an additional logistic regression analysis, which added to the original model those factors that may confer risk but whichmay also be associated through reverse causality, found stillbirth, preterm delivery, postpartum hemorrhage, and CD in labor to have a significant associationwith severe sepsis. Conversely, elective CD had a protective association for severe sepsis. Population-attributable fractions were calculated for factors with independent associations for severe sepsis and are reported in Table 2.
The rates of hemodialysis and mechanical ventilation more than 96 hours in women with severe sepsis were reviewed to evaluate the integrity of the increase in severe sepsis. The rates of hemodialysis did not change substantially during the study period. While the rate of mechanical ventilation decreased by 19.9% over the study period (Fig. 2), the overall increase of severe sepsis and sepsis-related death was 112% and 129%, respectively.
Of 4158 patients with severe sepsis, 1680 (40.4%) had a code for a specific type of organism and 2478 had an ICD-9-CM code indicating sepsis from an unspecified organism (Table 3), and 3177 (76.4%) had a concurrent infectious diagnosis for severe sepsis that was among one of the conditions for which we queried (Table 4.)
Among those patients with severe sepsis, organ dysfunction included respiratory dysfunction (34.2%), coagulation abnormalities (19.2%), renal dysfunction (16.4%), cardiovascular dysfunction (11.6%), hepatic dysfunction (10.3%), and central nervous system dysfunction (8.2%).
In this study, we used a large, representative sample of hospitalizations in the United States to estimate the frequency, independent associations, and temporal trends for severe sepsis during hospitalization for delivery. Even with adjustment for risk factors, the frequency of severe sepsis increased from 1:15,385 (95% CI, 1:12,987–1:18,519) to 1:7246 (95% CI, 6329–8333) during our 11-year study period. Thus, pregnancy-related severe sepsis is a growing source of severe maternal morbidity and mortality in the United States.
The increases in severe sepsis and sepsis-related mortality are consistent with the published UK experience.1 The increase in the United Kingdom was attributed to increased Group A β-hemolytic streptococcus and Escherichia coli infections, microbial resistance, and population factors of increasing maternal age, obesity, cardiovascular disease, migrant population, smoking, and poor overall health.1 Our U.S. data demonstrate that the increase in severe sepsis persisted despite adjusting for maternal age, presence of comorbidities, CD, multiple gestation, reliance on Medicaid, and other factors. This increase may have been due to similar factors as those identified in the United Kingdom, but not measured with administrative data such as increasing microbial resistance, obesity, smoking, substance abuse, and poor general health.
Chronic comorbid conditions such as congestive heart failure, chronic liver disease, chronic renal disease are well-established risk factors for sepsis in the general population and demonstrated associations with severe sepsis.14,15 SLE may increase the risk secondary to the use of steroids and immunosuppressive medications and immunologic abnormalities of the disease process itself.16–18 While women with these conditions are a relatively small proportion of the delivering population, the magnitude of risk associated with these conditions suggests prompt consideration of the diagnosis and treatment of sepsis in patients with these conditions.
Population-attributable fractions for all factors that had a significant association with severe sepsis were calculated to identify the most important sources of the population burden of severe sepsis in pregnancy. Importantly, none of the population-attributable fractions exceeded 5.9%, suggesting that sepsis often occurs in the absence of a recognized risk factor. This underscores the need for developing systems of care that increase sensitivity for disease detection across the entire population.
Common organisms implicated in severe sepsis in the current analysis were E coli, staphylococcus, streptococcus, and Gram-negative organisms. The underlying organism and information about antimicrobial resistance was missing from a large proportion of cases. More research on the causative organisms for maternal sepsis in the United States is warranted.
Consistent with the pattern observed for adverse maternal outcomes, advanced maternal age, African American race, and Medicaid insurance, all demonstrate independent associations with severe sepsis in multivariable analysis.19,20 Women from vulnerable socioeconomic backgrounds may have an increased risk for a range of adverse outcomes, including severe sepsis.21 Small volume hospitals demonstrated decreased risk of severe sepsis compared with larger centers; this may reflect small hospitals transferring women developing critical illness to larger centers for delivery.
Severe sepsis obstetric associations include rescue cerclage, prophylactic cerclage, retained products of conception, PPROM, and multiple gestation. Rescue cerclage conferred a greater risk than prophylactic cerclage. Information regarding whether providers evaluated for the presence of intrauterine infection, used ultrasonography before rescue cerclage placement, or a specific PPROM management strategy is not available from the NIS data. Therefore, the likelihood by which rescue cerclage and PPROM lead to maternal sepsis and the modifying effect of specific management strategies cannot be determined from this study. Finally, the higher risk of severe sepsis among women with multiple gestation may result from complications more commonly seen in this population such as PPROM, preterm labor, endometritis, postpartum hemorrhage, or gestational diabetes.22–25
The secondary multivariable analysis found preterm delivery, postpartum hemorrhage, stillbirth, and CD during labor to be associated with severe sepsis. In contrast, elective CD was found to be protective. Within the database, no temporal relation to diagnosis codes are provided; thus, we were unable to determine which diagnosis occurred first. Preterm delivery, postpartum hemorrhage, stillbirth, and CD could be associated with severe sepsis by reverse causality. For example, a patient who develops severe sepsis is more likely to have a CD. That said, CD is the most important risk factor associated with postpartum endometritis; endometritis is the etiology for 8.6% of severe sepsis cases in our dataset.26,27 Future prospective studies to define the role of intrapartum CD as a risk factor for severe sepsis are needed.
Our study findings should be interpreted in light of its design. The NIS does not capture many clinical details, such as Acute Physiology and Chronic Health Evaluation (APACHE II) scores or Sequential Organ Failure Assessment (SOFA) scores, measures of severity of critical illness. Certain conditions such as obesity are undercoded. In addition, while sepsis can occur during the antepartum or postpartum period, we focused our analysis on patients admitted to the hospital for delivery. To analyze associations, we sought a sample of otherwise healthy controls. Because the NIS only captures hospitalizations, this analysis is only possible during the delivery hospitalization. Administrative data cannot be verified and may result in underestimation or overestimation of severe sepsis cases. It is also possible that a patient could have suffered acute end-organ injury first, and subsequently developed sepsis; this patient would be categorized as having severe sepsis in our study. Nevertheless, independent of whether the end-organ injury was causally related to sepsis or the sepsis was superimposed on preexisting organ dysfunction, clinical experience suggests that these patients are a particularly high-risk group whose prognosis is worse than those who have sepsis in the absence of end-organ injury. Furthermore, our operational definition of severe sepsis as diagnostic codes indicating sepsis along with codes indicating end-organ injury is a standard approach in the intensive care unit epidemiology literature.7–9,15
It is possible that changes in coding practice increased documentation of severe sepsis without an actual increase in clinical acuity. The study period coincided with the initiation of the Surviving Sepsis Campaign (2004)28 and a landmark paper by Rivers et al. (2001)29 highlighting the importance of early goal-directed therapy. Increased intensity in sepsis coding has demonstrated a shift in the primary diagnostic code from the underlying etiology (e.g., pneumonia with sepsis listed as a secondary diagnosis) to sepsis as the primary diagnosis.30 To attenuate this error, a sepsis composite was considered which documented across any of the diagnostic positions rather than strictly the primary position.
In conclusion, maternal severe sepsis and sepsis-related deaths are increasing in the United States. Severe sepsis often occurs in the absence of a recognized risk factor and underscores the need for developing systems of care that increase sensitivity for disease detection across the entire population. Patients with conditions of congestive heart failure, chronic liver disease, chronic renal disease, and SLE have an increased risk of severe sepsis. Physicians should enhance surveillance in patients with risk factors for developing infection and institute early treatment when signs of sepsis are emerging.
Name: Melissa E. Bauer, DO.
Contribution: Melissa Bauer contributed to the study design and prepared the manuscript.
Attestation: Melissa Bauer approved the final manuscript, attests to the integrity of the original data and the analysis reported in this manuscript, and is the archival author.
Name: Brian T. Bateman, MD, MSc.
Contribution: Brian Bateman contributed to the study design and critically edited the manuscript.
Attestation: Brian Bateman approved the final manuscript and attests to the integrity of the original data and the analysis reported in this manuscript.
Name: Samuel T. Bauer, MD.
Contribution: Samuel Bauer was the obstetrical consultant and critically edited the manuscript.
Attestation: Samuel Bauer approved the final manuscript.
Name: Amy M. Shanks, MS.
Contribution: Amy Shanks contributed to study design, performed statistical analyses, and edited the manuscript.
Attestation: Amy Shanks approved the final manuscript and attests to the integrity of the original data and the analysis reported in this manuscript.
Name: Jill M. Mhyre, MD.
Contribution: Jill Mhyre contributed to the study design and critically edited the manuscript.
Attestation: Jill Mhyre approved the final manuscript and attests to the integrity of the original data and the analysis reported in this manuscript.
This manuscript was handled by: Cynthia A. Wong, MD.
The authors thank Mary Lou V. H. Greenfield, MPH, MS, RN (Senior Research Associate, Department of Anesthesiology, University of Michigan Health System, Ann Arbor, MI), Tiara Forsyth, BS (Research Assistant, Department of Anesthesiology, University of Michigan Health System), and Chloe Powell, BSE (Research Assistant, Department of Anesthesiology, University of Michigan Health System) for their support in the preparation of this manuscript.
a Overview of the Nationwide Inpatient Sample (NIS). Available at: http://www.hcup-us.ahrq.gov/reports/methods/2003_2.jsp. Accessed February 5, 2013.
1. Cantwell R, Clutton-Brock T, Cooper G, Dawson A, Drife J, Garrod D, Harper A, Hulbert D, Lucas S, McClure J, Millward-Sadler H, Neilson J, Nelson-Piercy C, Norman J, O’Herlihy C, Oates M, Shakespeare J, de Swiet M, Williamson C, Beale V, Knight M, Lennox C, Miller A, Parmar D, Rogers J, Springett A. Saving Mothers’ Lives: Reviewing maternal deaths to make motherhood safer: 2006–2008. The Eighth Report of the Confidential Enquiries into Maternal Deaths in the United Kingdom. BJOG. 2011;118(Suppl 1):1–203
2. Kilpatrick SJ, Matthay MA. Obstetric patients requiring critical care. A five-year review. Chest. 1992;101:1407–12
3. Mabie WC, Barton JR, Sibai B. Septic shock in pregnancy. Obstet Gynecol. 1997;90:553–61
4. Mabie WC, Sibai BM. Treatment in an obstetric intensive care unit. Am J Obstet Gynecol. 1990;162:1–4
5. Kuklina EV, Whiteman MK, Hillis SD, Jamieson DJ, Meikle SF, Posner SF, Marchbanks PA. An enhanced method for identifying obstetric deliveries: implications for estimating maternal morbidity. Matern Child Health J. 2008;12:469–77
6. Bone RC, Balk RA, Cerra FB, Dellinger RP, Fein AM, Knaus WA, Schein RM, Sibbald WJ. Definitions for sepsis and organ failure and guidelines for the use of innovative therapies in sepsis. The ACCP/SCCM Consensus Conference Committee. American College of Chest Physicians/Society of Critical Care Medicine. Chest. 1992;101:1644–55
7. Bateman BT, Schmidt U, Berman MF, Bittner EA. Temporal trends in the epidemiology of severe postoperative sepsis after elective surgery: a large, nationwide sample. Anesthesiology. 2010;112:917–25
8. Dombrovskiy VY, Martin AA, Sunderram J, Paz HL. Rapid increase in hospitalization and mortality rates for severe sepsis in the United States: a trend analysis from 1993 to 2003. Crit Care Med. 2007;35:1244–50
9. Martin GS, Mannino DM, Eaton S, Moss M. The epidemiology of sepsis in the United States from 1979 through 2000. N Engl J Med. 2003;348:1546–54
10. Acosta CD, Bhattacharya S, Tuffnell D, Kurinczuk JJ, Knight M. Maternal sepsis: a Scottish population-based case-control study. BJOG. 2012;119:474–83
11. Kramer HM, Schutte JM, Zwart JJ, Schuitemaker NW, Steegers EA, van Roosmalen J. Maternal mortality and severe morbidity from sepsis in the Netherlands. Acta Obstet Gynecol Scand. 2009;88:647–53
12. van Dillen J, Zwart J, Schutte J, van Roosmalen J. Maternal sepsis: epidemiology, etiology and outcome. Curr Opin Infect Dis. 2010;23:249–54
13. Rockhill B, Newman B, Weinberg C. Use and misuse of population attributable fractions. Am J Public Health. 1998;88:15–9
14. Bertoni AG, Saydah S, Brancati FL. Diabetes and the risk of infection-related mortality in the U.S. Diabetes Care. 2001;24:1044–9
15. Angus DC, Linde-Zwirble WT, Lidicker J, Clermont G, Carcillo J, Pinsky MR. Epidemiology of severe sepsis in the United States: analysis of incidence, outcome, and associated costs of care. Crit Care Med. 2001;29:1303–10
16. Bosch X, Guilabert A, Pallarés L, Cerveral R, Ramos-Casals M, Bové A, Ingelmo M, Font J. Infections in systemic lupus erythematosus: a prospective and controlled study of 110 patients. Lupus. 2006;15:584–9
17. Navarro-Zarza JE, Alvarez-Hernández E, Casasola-Vargas JC, Estrada-Castro E, Burgos-Vargas R. Prevalence of community-acquired and nosocomial infections in hospitalized patients with systemic lupus erythematosus. Lupus. 2010;19:43–8
18. Iliopoulos AG, Tsokos GC. Immunopathogenesis and spectrum of infections in systemic lupus erythematosus. Semin Arthritis Rheum. 1996;25:318–36
19. Mhyre JM, Bateman BT, Leffert LR. Influence of patient comorbidities on the risk of near-miss maternal morbidity or mortality. Anesthesiology. 2011;115:963–72
20. Centers for Disease Control and Prevention (CDC). . State-specific maternal mortality among black and white women—United States, 1987–1996 MMWR Morb Mortal Wkly Rep. 1999;48:492–6
21. Bryant AS, Worjoloh A, Caughey AB, Washington AE. Racial/ethnic disparities in obstetric outcomes and care: prevalence and determinants. Am J Obstet Gynecol. 2010;202:335–43
22. Schaaf JM, Mol BW, Abu-Hanna A, Ravelli AC. Trends in preterm birth: singleton and multiple pregnancies in the Netherlands, 2000-2007. BJOG. 2011;118:1196–204
23. Mercer BM, Crocker LG, Pierce WF, Sibai BM. Clinical characteristics and outcome of twin gestation complicated by preterm premature rupture of the membranes. Am J Obstet Gynecol. 1993;168:1467–73
24. Alexander JM, Leveno KJ, Rouse D, Landon MB, Gilbert SA, Spong CY, Varner MW, Caritis SN, Harper M, Wapner RJ, Sorokin Y, Miodovnik M, O’Sullivan MJ, Sibai BM, Langer O, Gabbe SGEunice Kennedy Shriver National Institute of Child Health and Human Development (NICHD) Maternal-Fetal Medicine Units Network (MFMU). . Cesarean delivery for the second twin. Obstet Gynecol. 2008;112:748–52
25. Wen SW, Demissie K, Yang Q, Walker MC. Maternal morbidity and obstetric complications in triplet pregnancies and quadruplet and higher-order multiple pregnancies. Am J Obstet Gynecol. 2004;191:254–8
26. Declercq E, Barger M, Cabral HJ, Evans SR, Kotelchuck M, Simon C, Weiss J, Heffner LJ. Maternal outcomes associated with planned primary cesarean births compared with planned vaginal births. Obstet Gynecol. 2007;109:669–77
27. Burrows LJ, Meyn LA, Weber AM. Maternal morbidity associated with vaginal versus cesarean delivery. Obstet Gynecol. 2004;103:907–12
28. Dellinger RP, Carlet JM, Masur H, Gerlach H, Calandra T, Cohen J, Gea-Banacloche J, Keh D, Marshall JC, Parker MM, Ramsay G, Zimmerman JL, Vincent JL, Levy MMSurviving Sepsis Campaign Management Guidelines Committee. . Surviving Sepsis Campaign guidelines for management of severe sepsis and septic shock. Crit Care Med. 2004;32:858–73
29. Rivers E, Nguyen B, Havstad S, Ressler J, Muzzin A, Knoblich B, Peterson E, Tomlanovich MEarly Goal-Directed Therapy Collaborative Group. . Early goal-directed therapy in the treatment of severe sepsis and septic shock. N Engl J Med. 2001;345:1368–77
30. Lindenauer PK, Lagu T, Shieh MS, Pekow PS, Rothberg MB. Association of diagnostic coding with trends in hospitalizations and mortality of patients with pneumonia, 2003-2009. JAMA. 2012;307:1405–13