If you don't remember your password, you can reset it by entering your email address and clicking the Reset Password button. You will then receive an email that contains a secure link for resetting your password
If the address matches a valid account an email will be sent to __email__ with instructions for resetting your password
Department of Medicine, Baystate Medical Center, Springfield, and Tufts University School of Medicine, Boston, MACenter for Quality of Care Research, Baystate Medical Center, Springfield, and Tufts University School of Medicine, Boston, MAClinical and Translational Science Institute at Tufts University School of Medicine, Boston, MA
Department of Medicine, Baystate Medical Center, Springfield, and Tufts University School of Medicine, Boston, MACenter for Quality of Care Research, Baystate Medical Center, Springfield, and Tufts University School of Medicine, Boston, MA
Division of Maternal-Fetal Medicine, Department of Obstetrics and Gynecology, Baystate Medical Center, Springfield, and Tufts University School of Medicine, Boston, MA
Department of Medicine, Baystate Medical Center, Springfield, and Tufts University School of Medicine, Boston, MACenter for Quality of Care Research, Baystate Medical Center, Springfield, and Tufts University School of Medicine, Boston, MAClinical and Translational Science Institute at Tufts University School of Medicine, Boston, MA
Maternal infection is a common complication of childbirth, yet little is known about the extent to which infection rates vary among hospitals. We estimated hospital-level risk-adjusted maternal infection rates (RAIR) in a large sample of US hospitals and explored associations between RAIR and select hospital features.
Study Design
This retrospective cohort study included hospitals in the Perspective database with >100 deliveries over 2 years. Using a composite measure of infection, we estimated and compared RAIR across hospitals using hierarchical generalized linear models. We then estimated the amount of variation in RAIR attributable to hospital features.
Results
Of the 1,001,189 deliveries at 355 hospitals, 4.1% were complicated by infection. Patients aged 15-19 years were 50% more likely to experience infection than those aged 25-29 years. Rupture of membranes >24 hours (odds ratio [OR], 3.0; 95% confidence interval [CI], 3.24−3.5), unengaged fetal head (OR, 3.11; 95% CI, 2.97−3.27), and blood loss anemia (OR, 2.42; 95% CI, 2.34−2.49) had the highest OR among comorbidities commonly found in patients with infection. RAIR ranged from 1.0−14.4% (median, 4.0%; interquartile range, 2.8−5.7%). Hospital features such as geographic region, teaching status, urban setting, and higher number of obstetric beds were associated with higher infection rates, accounting for 14.8% of the variation observed.
Conclusion
Obstetric RAIR vary among hospitals, suggesting an opportunity to improve obstetric quality of care. Hospital features such as region, number of obstetric beds, and teaching status account for only a small portion of the observed variation in infection rates.
Childbirth is the most common reason for hospital admission in the United States, with >4,000,000 admissions for labor and delivery occurring annually.
Although most births are uncomplicated, a small but significant number of women experience complications such as infection, trauma, and hemorrhage during childbirth.
Reducing obstetric complications has emerged as a national priority in the US, as reflected in goals established by Healthy People 20204 and the Centers for Medicare and Medicaid Services' Partnership for Patients.
and many of these infections may be preventable. Several small studies and reviews have described clinical practices that can increase the risk of infection, primarily related to cesarean deliveries.
However, little is known about the extent to which obstetric infection rates vary across hospitals or what impact structural and organizational features of a hospital may have on these rates.
To support the national goal of improving maternal outcomes following childbirth, we used hierarchical generalized linear modeling to estimate risk-adjusted maternal infection rates (RAIR) in a large sample of US hospitals. We then examined whether hospital features, such as the number of hospital beds, teaching status, geographic region, volume of deliveries, and level of implementation of electronic health records (EHR), were associated with higher rates of infection.
Materials and Methods
Study sample and data source
We conducted a cross-sectional study using Perspective, a voluntary, fee-supported database developed by Premier Inc (Charlotte, NC) that enables participating hospitals to analyze care quality and costs at their institution and to compare their performance to other institutions within the database. The database is comprised of a structurally and geographically diverse set of approximately 450 US hospitals that together account for approximately 20% of all annual hospital admissions in the United States. In addition to information derived from standard hospital discharge files (ie, Uniform Billing form-04) Perspective contains a date-stamped log of all items (eg, medications, laboratory, diagnostic tests) and therapeutic services billed to the patient or their insurer.
Women were included in the study if they were discharged from Jan. 1, 2008, through Dec. 31, 2009, and had an International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) principal or secondary diagnosis or procedure code for a vaginal delivery (650, 640.0x-676.9x [x = 1 or 2], or 73.59) or cesarean delivery (763.4, 669.71, 74.x [x = 0-2, 4], or 74.99). We excluded discharges for ectopic and molar pregnancies and for pregnancies ending in spontaneous or elective abortion because we were interested in exploring intrapartum/peripartum infections. We also excluded patients who were transferred from or to another institution, because we did not have information about the clinical course or treatments prior to admission or subsequent outcomes, and women age <15 or >44 years because 15-44 years is a common age range for childbearing. In addition we excluded hospitals that recorded <100 deliveries over the 2-year study period to provide stable estimates of infection rates, and because these institutions do not routinely provide obstetric care. Permission to conduct the study was obtained from the institutional review board at Baystate Medical Center in Springfield, MA.
Obstetric infection
A delivery was considered complicated by infection if the patient received ≥1 diagnoses consistent with infection using a broad set of ICD-9-CM codes that have been used in earlier studies of infections associated with childbirth
(Appendix; Supplementary Table 1). We excluded ICD-9-CM infection codes with a fifth digit of 3, which indicates an antepartum condition, because we were most interested in risk-adjusted infection rates occurring in the intrapartum/peripartum period as well as the association between these risk-adjusted infection rates and hospital features. We organized infection codes into groups of related diagnoses for descriptive purposes (Table 1). Each infection code was counted toward the overall frequency of each type of infection. When calculating hospital-level infection rates, a patient was considered to either have experienced or not experienced an infectious complication regardless of the number of infection codes associated with a single delivery.
TABLE 1Frequency of maternal infections by category of infection
Infection
n (%)
Any infection below
40,605 (4.1)
Puerperal infection
20,519 (2.1)
Maternal pyrexia
16,067 (1.6)
Surgical site infection
3523 (0.4)
Infection of genitourinary tract
1964 (0.2)
Sepsis
1319 (0.1)
Other maternal infection
1456 (0.2)
Goff. Patterns of obstetric infection rates. Am J Obstet Gynecol 2013.
We recorded patient demographics (age, gender, race/ethnicity, marital status, and insurance status) and conditions that might confer elevated risk for obstetric infection. We used 2 complementary methods to identify maternal comorbidities and pregnancy-specific conditions that could influence a patient's risk of infection. The presence of any of 29 comorbidities was computed using Elixhauser Comorbidity Software, version 3.1, developed by the Agency for Healthcare Research and Quality.
These conditions were originally developed to predict risk for cesarean delivery, but have also been used for risk adjustment for infection rates in obstetric patients.
For conditions that appeared in both sets, such as hypertension and substance abuse, we created combined indicators for patients identified by either method. Gestational diabetes and diabetes existing prior to pregnancy were assessed separately because they confer different risk for infection.
Using data from the American Hospital Association (AHA) annual survey and Premier Inc, we noted each hospital's geographic location, number of hospital beds, number of obstetric beds, number of deliveries in the 2-year period, whether the hospital was located in an urban or rural setting, teaching status, and whether a hospital reported full implementation of EHR. Four questions on the AHA Annual Survey (2008)
2012 Measures maintenance technical report: acute myocardial infarction, heart failure, and pneumonia 30-day risk-standardized readmission measures. Yale New Haven Health Services Corp/Center for Outcomes Research and Evaluation.
were used to define a hospital's level of implementation. The questions encompassed EHR use related to patient-level health information, results management, order entry management, and decision support. A hospital was categorized as having a fully implemented EHR if all 4 domains were reported as “fully implemented.”
Statistical analysis
We evaluated the association of patient demographics, maternal comorbidities, pregnancy-related conditions, and structural and organizational hospital features with the presence of “any infection” using χ2 statistics. We used this composite measure of infection to assess hospital infection rates because it allowed for inclusion of rare diagnoses while reducing the risk that variation in coding practices across hospitals would result in biased rate estimates. Using a model-building strategy that retained factors with P < .05, or those that were theoretically important to obstetric infections, we employed hierarchical generalized linear modeling to model the log odds of experiencing infection related to childbirth adjusting for patient demographics, maternal comorbidities, and pregnancy-specific conditions that could increase risk of infection, while including a random hospital effect. Conditions, such as diabetes existing prior to pregnancy, which did not meet the significance criterion for inclusion in the model but were clinically important were forced into the model. Selected interaction terms were evaluated. From the final model, we calculated hospital-specific RAIR as the ratio of predicted (using hospital random effect) to expected (using average hospital effect) events multiplied by the overall unadjusted infection rate, a form of indirect standardization that is used in hospital outcomes measurement initiatives sponsored by the Centers for Medicare and Medicaid Services.
2012 Measures maintenance technical report: acute myocardial infarction, heart failure, and pneumonia 30-day risk-standardized readmission measures. Yale New Haven Health Services Corp/Center for Outcomes Research and Evaluation.
Our primary model included all deliveries, and we stratified by vaginal or cesarean delivery in a secondary analysis.
We then evaluated the bivariate associations of structural and organizational hospital features with RAIR using analysis of variance and t tests. Lastly, we modeled RAIR across hospitals as a function of structural and organizational hospital features and estimated the proportion of variation in RAIR attributable to hospital features.
Results
Study sample
From the initial sample of 1,038,555 deliveries at 424 hospitals, 3913 were excluded due to presence of an ICD-9-CM code for ectopic or molar pregnancy or spontaneous or induced abortion, 29,888 due to transfer into or out of the hospital or unknown discharge status, 3140 for maternal age <15 or >44 years, and 425 because the delivery occurred at a hospital (n = 69) with <100 deliveries during the 2-year study period. Our final sample included 1,001,189 deliveries at 355 hospitals (Figure 1).
FIGURE 1Flow diagram depicting exclusions and final sample size
The majority of women (75%) were between ages 20-34 years, 50% were married, 25% were black or Hispanic, and 42% had a public form of health insurance such as Medicaid (Table 2). Cesarean deliveries accounted for 39% of the deliveries included in the study. The most commonly identified maternal comorbidities and pregnancy-specific conditions included cesarean delivery during a previous pregnancy (18.3%), advanced maternal age (≥35 years) (15.7%), hypertension (8.1%), and preterm delivery (7.6%) (Table 2). Maternal mortality was 0.01% and median length of stay was 2 days (interquartile range [IQR], 2−3) for vaginal deliveries and 3 days (IQR, 3−4) for cesarean deliveries.
Of the deliveries included in the study, 40,605 (4.1%) were complicated by infection. Puerperal infections were the most common, affecting 2.1% of deliveries, followed by maternal pyrexia (1.6%) and surgical site infections (0.4%). Genitourinary tract infections (0.2%) and sepsis (0.1%) were relatively uncommon (Table 1). Of the deliveries complicated by infection, maternal mortality was 0.06% and median length of stay was 3 days (IQR, 2−3) for vaginal deliveries and 4 days (IQR, 3−5) for cesarean deliveries.
Among the hospitals, 28% were teaching hospitals, 77% were in an urban setting, 43% were in the south region, and 28% had >30 obstetric beds. Relatively few hospitals (19%) reported complete implementation of EHR (Table 3).
TABLE 3Association between hospital features and risk-adjusted maternal infection rates
Patient demographics, maternal comorbidities, and pregnancy-specific conditions considered in modeling RAIR are shown in Table 2. Adjusted odds ratios (OR) from the final main effects model for RAIR are shown in Supplementary Table 2. Age was strongly associated with risk for infection; when compared to patients ages 25-29 years, patients ages 15-19 years had 59% higher risk for infection, and patients ages 35-44 years had 29% lower risk. Maternal comorbidities and pregnancy-specific conditions that were most strongly associated with infection (OR, >2.0) and occurred in >5% of deliveries with infection included rupture of membranes >24 hours (OR, 3.0; 95% confidence interval [CI], 3.24−3.5), blood loss anemia (OR, 2.54; 95% CI, 2.47−2.62), and unengaged fetal head (OR, 3.11; 95% CI, 2.97−3.27). Although other risk factors such as cerebral hemorrhage were strongly associated with infection, they occurred infrequently. Interactions of the patient demographic variables (eg, age, race, insurance, and marital status) with comorbid and pregnancy-specific conditions were included in a final model used to estimate RAIR.
Separate models for vaginal and cesarean deliveries gave results similar in magnitude and direction for most risk factors (Supplementary Table 2).
Hospital risk-adjusted infection rates
Unadjusted hospital infection rates ranged from 0.0−12.3% (median, 3.2%; IQR, 2.0−4.6%). However, after adjusting for differences in patient case mix, the RAIR ranged from 1.0−14.4% (median, 4.0%; IQR, 2.8−5.7%) (Figure 2). Women delivering at hospitals at the 75th percentile of infection rates had a 2-fold risk of experiencing an infection as compared to women delivering at a hospital at the 25th percentile.
FIGURE 2Distributions of unadjusted and risk-adjusted hospital infection rates
Secondary analysis of RAIR for cesarean deliveries revealed infection rates ranging from 1.5−18.4% (median, 5.4%; IQR, 3.9−7.7%). Vaginal delivery infection rates ranged from 0.05−13.3% (median, 3.1%; IQR, 2.1−4.6%) (Supplementary Table 2). RAIR for vaginal and cesarean deliveries were strongly correlated (Spearman r = 0.69, P < .0001).
Association between structural and organizational hospital features and risk-adjusted infection rates
RAIR was associated with a number of hospital features (Table 3). Hospitals in the west region had the highest mean RAIR (5.3; 95% CI, 4.7−5.9) while those in the south region had the lowest mean RAIR (4.0; 95% CI, 3.7−4.4). Larger hospitals (≥400 beds) had higher rates (5.3; 95% CI, 4.8−5.8) than smaller hospitals (<200 beds) (4.1; 95% CI, 3.7−4.4). Higher RAIR was also observed in hospitals with a greater number of obstetric beds (≥30) (5.4; 95% CI, 4.9−5.9) when compared to hospitals with fewer obstetric beds (<15) (3.8; 95% CI, 3.5−4.2) (Table 3). Teaching hospitals had higher RAIR (5.4; 95% CI, 4.9−5.9) compared to nonteaching hospitals (4.3; 95% CI, 4.0−4.6). In a multiple regression model, structural and organizational hospital features explained approximately 14.8% of the observed variation in risk-adjusted infection rates (Table 4).
TABLE 4Risk for higher hospital risk-adjusted infection rate attributable to structural and organizational hospital features (n = 355)
Interpretation: adjusting for other factors in model, hospitals in west region have risk-adjusted maternal infection rates averaging 1.51% higher than those in south region; hospitals with <15 obstetric beds have rates averaging 1.39% lower than hospitals with ≥30 obstetric beds; teaching hospitals have rates averaging 0.83% higher than nonteaching hospitals.
SE
P value
Region
West
1.51
0.32
< .0001
Northeast
0.85
0.39
.0280
Midwest
0.52
0.31
.0936
South (reference)
0
—
—
No. of obstetric beds
Unknown
−0.76
0.54
.1593
<15
−1.39
0.33
< .0001
15-29
−0.71
0.31
.0206
≥30 (reference)
0
—
—
Teaching status
Teaching hospital
0.83
0.28
.0036
Nonteaching hospital (reference)
0
—
—
Goff. Patterns of obstetric infection rates. Am J Obstet Gynecol 2013.
a Interpretation: adjusting for other factors in model, hospitals in west region have risk-adjusted maternal infection rates averaging 1.51% higher than those in south region; hospitals with <15 obstetric beds have rates averaging 1.39% lower than hospitals with ≥30 obstetric beds; teaching hospitals have rates averaging 0.83% higher than nonteaching hospitals.
In this study of >1 million deliveries at 355 hospitals across the United States, approximately 4.1% of women experienced an infection during hospital admission for childbirth. We observed substantial variation in hospital infection rates that persisted even after adjustment for differences in patient case mix across institutions, with hospitals at the 75th percentile having RAIR twice that of hospitals at the 25th percentile. Although several structural and organizational hospital features were associated with higher RAIR, together these features explained only a small fraction of the observed variation in infection rates, consistent with other estimates of the impact such factors have on patient outcomes.
Our study provides important additions to the literature on obstetric infection rates by including a large sample of US hospitals, estimating hospital-specific RAIR, and examining variation in these rates across hospitals. In a study by Berg et al,
obstetric infection rates were compared for 2 time periods, 1993 through 1997 and 2001 through 2005, using the National Hospital Discharge Survey, but this study did not report hospital-specific rates. Srinivas et al
also explored trends in maternal complication rates, but used 2 state databases for the primary analysis and did not explore variation at the hospital level. Gregory et al
developed a method to measure uncomplicated or “ideal” deliveries. However, this unique approach does not allow for comparison of rates of undesirable outcomes such as infection.
Our study also contributes to the literature on how structural and organizational hospital characteristics may impact obstetric infection rates. Although this area has been studied for other conditions, such as acute myocardial infarction,
demonstrated an association between the volume of deliveries and lower composite complication rates. In that study, hospitals with approximately 100-1600 deliveries annually had the lowest unadjusted infection rates. This is similar to our findings of lower RAIR being associated with lower number of deliveries over 2 years (100-2149). Studies in disciplines other than obstetrics have compared care quality and patient outcomes at teaching hospitals and nonteaching hospitals. These studies have generally shown lower mortality and better overall patient outcomes in the teaching hospitals.
In contrast, we found that teaching hospitals had higher infection rates. Although health care−associated infections are a common measure of quality and patient safety,
infection has not been a commonly selected outcome when comparing teaching and nonteaching hospitals, limiting comparison to other studies. It is possible that differences in the gestational age of patients delivered at teaching vs nonteaching hospitals might serve as an unmeasured confounder, however gestational age was not available in the data used for this study.
This study had a number of strengths. First, we included a large number of patients drawn from a diverse group of hospitals, enhancing the generalizability of our findings. Second, we used a composite measure of infection rather than individual infection codes to reduce the risk that the variation in the rates we observed could be explained by differences in the coding practices at individual hospitals. Third, while “present on admission” codes
were not commonly used during the time period studied, we limited diagnoses to those most likely to occur during admission for childbirth by using a subset of fifth-digit codes. This made the infections identified more likely to be associated with factors related to hospital practices at the time of the delivery, but may have caused us to miss some pertinent infections. Fourth, we adjusted for a large number of maternal comorbidities and pregnancy-specific conditions found to be associated with the risk of maternal infection, thereby reducing the chance that variation in RAIR across hospitals reflected differences in patient case mix. Additionally, the use of multilevel regression modeling accounted for the natural clustering of patients within hospitals. Cesarean delivery is a known risk factor for postpartum endometritis and rates of cesarean deliveries vary across hospitals. By including both modes of delivery in the primary analysis, our hospital-level infection risk estimates were not influenced by local preferences for cesarean deliveries and offer a more patient-centric view of the risk of infection associated with the choice of hospital.
The first potential limitations of this study are the accuracy of ICD-9-CM coding and inability to capture outpatient codes. These are inherent limitations of all studies using administrative data. Prior validation studies of obstetric ICD-9-CM codes demonstrate variation in estimates of sensitivity and specificity of codes for a number of conditions, with codes for surgical procedures generally being the most reliable.
Although we attempted to mitigate some of the potential limitations of coding accuracy by using a composite of multiple diagnosis codes, some of our study's findings may be partly explained by differences in documentation and coding across institutions. For example, our surgical site infection rate was lower than rates found in other studies.
This may be due in part to the lack of a specific ICD-9-CM code to identify obstetric surgical site infections or ability to capture only infections identified during the admission. Although the code for maternal pyrexia is associated with infection, this code may also be used for fever from a noninfectious source, such as blood transfusion. A final example of limitations due to coding accuracy relates to obesity codes, which have been found to have high specificity but low sensitivity in obstetrics.
The second potential study limitation was our focus on infections that became apparent during the index hospitalization; overall infection rates are undoubtedly higher than we estimated. Third, although we attempted to adjust for a large number of maternal comorbidities and pregnancy-specific factors that could influence the risk of infection, some of the variation in infection rates may be related to limitations on our ability to fully account for the patient-level risks. Fourth, some of the unexpected findings related to the comorbidity rates may be explained by unmeasured factors, such as preterm births excluded due to transfers of preterm deliveries. Finally, the overrepresentation of southern hospitals in the perspective dataset means interpretation of geographic differences in RAIR must be made with care. This geographic oversampling may explain why the rate of cesarean deliveries found in this sample was higher than national averages.
Centers for Disease Control and Prevention Recent trends in cesarean delivery in the U.S. National Center for Health Statistics. Data Brief no. 35, March 2010.
In conclusion, we found that risk-adjusted infection rates following childbirth vary considerably across hospitals, and that key structural and organizational hospital features explain only a modest amount of this variation. To support large-scale efforts to improve the quality of obstetric care, additional research is needed to identify organizational factors and clinical strategies that enable some hospitals to achieve lower infection rates.
2012 Measures maintenance technical report: acute myocardial infarction, heart failure, and pneumonia 30-day risk-standardized readmission measures. Yale New Haven Health Services Corp/Center for Outcomes Research and Evaluation.
The project described was supported by grant number KL2 RR025751 from the National Center for Research Resources and grant number KL2 TR000074 from the National Center for Advancing Translational Sciences , National Institutes of Health (S.L.G.).
The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.
The authors report no conflict of interest.
Cite this article as: Goff SL, Pekow PS, Avrunin J, et al. Patterns of obstetric infection rates in a large sample of US hospitals. Am J Obstet Gynecol 2013;208:456.e1-13.
00137-3&id=gr1.jpg){kind=link}
00137-3&id=gr2.jpg){kind=link}