Significant effort and resources have been directed to develop valid, standardized measures of health care quality using readily available administrative data.1,a The Agency for Healthcare Research and Quality (AHRQ) has established multiple sets of indicators for quality monitoring and improvement using hospital-level administrative data based on International Classification of Diseases, version 9, Clinical Modification (ICD-9-CM) diagnosis and procedure codes. One such set is the patient safety indicators (PSIs)b; these indicators focus on potentially preventable hospital complications after surgeries, procedures, and childbirth. Of particular interest is the experimental indicator, PSI 1, experimental quality indicator (EXP) 1, anesthesia complications. As specified by AHRQ, it measures the rate of complications per 1000 surgical discharges; surgical discharges include cesarean but not vaginal delivery.2
Given that childbirth is the number 1 reason for hospitalization, with approximately 4 million births annually in the United States,3 and that >60% of women use some form of neuraxial analgesia or anesthesia during childbirth,4 we explored the prevalence of general anesthesia and neuraxial anesthesia/analgesia complications for the standard AHRQ population and for all childbearing women, including vaginal delivery. To be completely representative of the childbirth experience, we expanded the definition of anesthesia complications to include neuraxial analgesic methods (e.g., spinal, epidural, and combined spinal–epidural analgesia). Hereafter, the term anesthesia will refer to general anesthesia and neuraxial anesthesia/analgesia, and all adverse events will be collectively referred to as anesthesia complications.
The use of general anesthesia for childbirth is uncommon and used in <5% of cases.5 Associated complications include failed intubation and aspiration of gastric contents.6 Neuraxial anesthetic methods include spinal anesthesia, which is frequently used for cesarean delivery, and epidural anesthesia/analgesia, which can be used for labor or cesarean delivery. Neuraxial anesthesia is the most common form of pain relief; approximately 3 of 5 women receive it during labor and delivery.4 Complications associated with neuraxial anesthesia can range from inadequate analgesia, resulting in persistent discomfort or severe pain to hypotensive episodes causing fetal distress or maternal respiratory or neurologic compromise. In addition, women can experience minor complications such as shivering, backache, headache secondary to dural puncture (postdural puncture headache), or more severe conditions such as epidural abscess, epidural hematoma, bradycardia, supine hypotensive syndrome, cardiac arrest, local anesthetic systemic toxicity, and rarely, equipment failure such as catheter breakage.7,8 The most common of these are maternal backache, hypotension, and postdural headache.7 Recognizing that all medical procedures have risks and benefits, both the American College of Obstetricians and Gynecologists and the American Society of Anesthesiologists have endorsed the use of neuraxial anesthesia, contending that “maternal request is a sufficient medical indication for pain relief during labor.”9
The original technical specifications for PSI 1, EXP 1 specified the denominator of the measure to include childbirth but limited it only to anesthesia complications associated with cesarean delivery. However, >60% of patients receive neuraxial analgesia during labor,4 suggesting that a large number of childbearing women would not be captured in this limited definition. Given the high utilization and potential severity of complications related to the administration of general and neuraxial anesthesia during childbirth, there are 3 objectives to this study:
- To calculate the rate of anesthesia complications as defined by AHRQ experimental indicator,
- To propose a definition of the anesthesia complication rate specific to childbirth and inclusive of vaginal delivery and neuraxial analgesia, and
- To calculate the rate of anesthesia complications in childbirth using this definition and stratified by method of delivery, and to evaluate the variation in complication rates across California hospitals to determine the relevance of tracking anesthesia complications as a potential PSI for childbirth.
The study was approved by all appropriate IRBs and complied with all patient protection criteria. The requirement for written informed consent was waived by the IRBs.
To calculate rates of anesthesia complications in California, we used the 2009 California Patient Discharge Dataset, which includes mandatory reports on all California hospital discharges.c These data are routinely validated and queried to maintain a high level of data quality. The rate of anesthesia complications was included as an EXP 1 in PSI version 3.2.d It was subsequently removed and did not appear again until version 4.310 and is present in the most recent version 4.5, published May 2013.e SAS programming code provided by AHRQ for version 3.2d was used as the basis for the calculations. There were no differences in reporting between version 3.2 and version 4.5.
- AHRQ Population: the numerators and denominators of the measures were used exactly as defined by AHRQ. The numerator included discharges with adverse effects from anesthesia administered to patients undergoing a surgical procedure including cesarean delivery. The denominator included all surgical discharges for patients (men and women) ≥18 years or surgical discharges for women of any age if they were classified under major diagnostic category (MDC) 14, which includes pregnancy, childbirth, and the puerperium. Surgical discharges are defined by specific Medicare Severity-Diagnosis Related Groups (MS-DRG) codes. Examples of surgical discharges in MDC 14 include tubal ligation, hysterectomy, or cesarean delivery but not vaginal delivery without any additional surgical procedure. Cases with a principal diagnosis or secondary diagnosis of adverse effects or poisoning from anesthetics (present on admission) are excluded from the denominator.
To include vaginal deliveries and complications from neuraxial anesthesia in the original AHRQ PSI, modification of both the denominator and the numerator is required. Also, the designation of primary and secondary diagnosis in childbirth has different rules,11 and because of variation in the use of obstetric codes as principal and secondary diagnosis, we included women with ICD-9-CM codes designating adverse effects as principal or secondary diagnosis, including those designated as present on admission to assure capture of all cases. Given these factors, the modifications significantly changed the AHRQ specifications for the anesthesia complications indicator, and we therefore created a childbirth-specific indicator as described below.
- The numerator included the AHRQ codes listed in Table 1 plus the following:
- 668 codes (complication of administration of anesthetic or sedation in labor and delivery)
- 349 codes (reaction to spinal or spinal headache)
- The denominator included all delivery discharges as defined by MDC 14, stratified by delivery type. Cesarean deliveries were defined by MS-DRG 765, and vaginal deliveries were defined by MS-DRG 766, 767, 768, 774, and 775.11
To evaluate normative hospital-level rates and estimate the between-hospital variation for each indicator, we fitted multilevel multiple logistic regression models, using GLIMMIX procedures in SAS, version 9.212 with patient-level data clustered by hospital. We report the intrahospital correlation in outcomes calculated as a ratio of the variation associated with the hospital clustering to the total variation, which is generally referred to as the variance partition coefficient (VPC). An interpretation of the VPC is that, if it is large (e.g., >5%), there is potential for improvement in outcome by hospital structural, organizational, or practice pattern modifications. The VPC was estimated by simulation of the hospital error term distribution assumed to be normal with mean = 0 and variance estimated by the model.13,14 Hospitals with <200 annual deliveries were excluded from this analysis to facilitate a more robust estimation of model parameters. More details on the model specification and estimation are provided in Gregory et al.11
To assess potentially extreme hospitals while controlling for the familywise type I error rate α, hospitals with extremely low or high rates for anesthesia complications were defined as having risk-adjusted (for the average patient) rates significantly different from the state average using a false-discovery rate (FDR) with α = 0.05.15,16 As an alternative method to identify extreme hospital rates, we defined extreme hospitals as those statistically (with 95% confidence) below or above the lower or upper quartile of the adjusted (for the average patient) hospital-rate distribution. To assess the impact of risk adjustment on the identification of extreme hospital rates, κ coefficients were calculated for the agreement in outlier status with and without risk adjustment for each of the identification methods. The hospital rate distributions (mean, standard deviation [SD], median, minimum, and maximum) and a caterpillar plot that presents the risk-adjusted hospital rates with 95% confidence intervals (CIs) were also reported. The potentially annual preventable events in the extremely high-rate hospitals were calculated as the difference between the total observed events and the expected number of events if these hospitals had the mean risk-adjusted rate.
Risk Adjustment for Hospital-Level Analysis
Recognizing that anesthesia complications could be related to case mix, we used administrative data and ICD-9-CM codes to adjust for common clinical conditions using patient-level covariates in these models. These conditions included age, race/ethnicity, prior cesarean delivery (yes/no), preterm (<37 completed weeks of gestational age; yes/no), multiple gestation (yes/no), and a composite variable designating the presence of “other pregnancy complications”11 (yes/no). We report the odds ratios and Wald test P values for the covariates used as adjustors in the models. To assess the effect of the case-mix adjustment on the between-hospital variation, we report the VPC of the “empty models” (models without any patient covariates).
We characterized EXP 1: Anesthesia complications based on the above results, specifically comparing the rate for the AHRQ original population with the rate for the childbirth-specific population, as well as evaluating whether hospital variation in the indicator was evident.
In 2009, 508,842 deliveries occurred, of which 508,746 met the AHRQ inclusion criteria in 254 hospitals in California. Twelve (4.7%) hospitals with <200 annual deliveries (N = 1228 patients, or 0.2%) were excluded. For the remaining 242 hospitals (507,518 deliveries), the mean and SD of the annual delivery volume were 2003 (1517), with a median (range) of 1717 (212–7511). The mean age (SD) at delivery was 28 (6.3) years, and the distribution of women undergoing childbirth by race/ethnicity was as follows: Hispanic 48.7%, Caucasian 30.2%, Asian/Pacific Islander 11.3%, African American 5.7%, American Indian/Alaskan Native 3.2%, and unknown 0.8%.
Using the standard AHRQ study population, which included cesarean deliveries only, the rate (SE) of anesthesia complications was 0.13% (0.004%). This rate (SE) more than doubled to 0.31% (0.008%) for the childbirth-specific indicator. When stratified by method of delivery, anesthesia complication rates were 0.49% for cesarean delivery and 0.22% for vaginal delivery (P < 0.0001) (Table 1).
Table 2 reports the number of anesthesia complications by individual ICD-9-CM codes. The 3 most commonly observed complications were 349.0: reaction to spinal/epidural (N = 1433); 668.x: anesthesia complications in pregnancy (N = 1312); and E938.7: spinal anesthetics causing adverse effects in therapeutic use (N = 121). Codes used to modify the standard AHRQ definition (668.x and 349.0) accounted for approximately 90% of the reported anesthesia complications in the delivery study population.
The unadjusted hospital childbirth-related anesthesia complications rate mean (SD) was 0.34% (0.34%), median 0.25%, interquartile range (0.12%–0.45%), and range (0%–2.46%). No hospitals had statistically low rates. Using the FDR methodology, 18 (7.4%) hospitals were identified as having significantly high rates with risk-adjusted childbirth-related anesthesia complication rates ranging from 0.43% to 2.13%, whereas 13 hospitals (5.4%) had statistically high rates, using the alternative method of the 95% CI completely above the 75th percentile (range 0.52%–2.13%). These same 13 hospitals were also identified as extreme (with the upper quartile cutoff method) using the risk-adjusted rates obtained from an empty model without covariates and were included in the larger group of 18 hospitals with extremely high rates identified with the FDR method. However, the 18 extreme hospitals identified by the empty model differed from the 18 identified by the risk-adjusted model by 1 hospital (of the additional 5 extreme hospitals only 4 hospitals matched). Using the FDR and the 75th percentile cutoff methods to identify extreme hospitals, the κ coefficients of agreement in outlier status were 0.94 and 1, respectively. Figure 1 shows a caterpillar plot of the risk-adjusted hospital rates with 95% CIs for hospitals in the top quartile (a plot with the full sample of hospitals is available with the Supplemental Digital Content, http://links.lww.com/AA/A929). Vertical lines denote the first and third quartiles (solid) and the mean rate (dashed). Hospital index numbers followed with “+” are extreme according to the FDR method, and those followed with “++” are extreme according to both the FDR and the upper quartile cutoff methods. The number of potentially preventable events was 181 (261 observed versus 80 expected of 24,320 deliveries) for the group of 13 extreme high-rate hospitals and 223 (359 observed versus 136 expected of 41,210 deliveries) for the group of 18 extreme high-rate hospitals.
Table 3 shows the odds ratios and 95% CIs for patient-level factors used in the multilevel, multiple logistic regression model for case-mix adjustment at the hospital level, and the percentage of variation attributable to the hospital level (VPC). Being non-Caucasian and/or preterm decreased the likelihood of an anesthesia complication, whereas having increased maternal age, prior cesarean, multiple gestations, or a combination of other pregnancy complications increased the likelihood of an anesthesia complication. The percentage of variation in complications due to hospital-level factors (VPC) was low (<0.2%) with or without adjustment for patient risk factors.
Despite childbirth being the number 1 reason for hospital admission, childbirth-specific quality indicators are neglected among current AHRQ indicators, even though some existing indicators have the potential to be pertinent and/or adapted to pregnancy.11
Given the recent increase in maternal mortality and morbidity and data suggesting that many of these adverse outcomes are preventable,17 we contend that the ability to monitor pregnancy and childbirth morbidity is critical to the safety and quality of maternity care, yet apart from PSI 17: Newborn Birth Trauma and PSIs 8 and 19: Maternal Birth Trauma, adverse events occurring during pregnancy hospitalizations have been relatively neglected in the PSI set, because pregnancy is largely excluded from the denominator of the majority of PSI.11 Furthermore, these indicators have been variously criticized since PSI 17 is pertinent to the newborn and may not be avoidable or specifically related to clinical care processes, and PSIs 8 and 19, although pertinent to the mother, focus on perineal trauma, which some clinicians feel may not be avoidable and/or, in the grand scheme of things, is of small consequence. However, increasingly, qualitative data suggest that women and clinicians do care about perineal trauma, as evidenced by the number of women who undergo cesarean delivery on maternal request and the number of clinicians who claim they would prefer cesarean delivery for themselves or their partner.18,19 We recognize that many procedures occurring during childbirth are not elective and that complications from these procedures may not always be preventable. We therefore suggest that definitions of safety and quality indicators for maternity care may need to be broader than those used for nonpregnant adults to provide an opportunity to capture and explore the morbidity that occurs during childbirth until preventability is better understood.
In this study, we explored the relevance of anesthesia complications as a potential indicator of childbirth-related patient safety. We wanted to determine whether a childbirth-specific definition of anesthesia complications could be derived that would be inclusive of patients undergoing vaginal delivery, many of whom receive neuraxial anesthesia. We narrowed the denominator to childbirth admissions only, irrespective of method of delivery. We expanded the case definition to include complications from both spinal and epidural placement so that the associated morbidity from these anesthesia methods could be tracked. The clinical relevance of this expansion is that >60% of patients receive neuraxial labor analgesia and that epidural or spinal anesthesia is the procedure of choice for cesarean delivery.3,4 Furthermore, it is important for pregnant women to understand the types of morbidities that occur and the variation in those morbidities across hospitals as increased interest is focused on patient choice for birth setting and type of care provider, satisfaction with care, and “physiologic labor” as a potential quality indicator.20, f
For the study year of interest, we found that the rate of anesthesia complications using the AHRQ specifications was 0.13%. This number does not include much of the anesthesia-related morbidity experienced during childbirth, the number 1 reason for hospital admission. By defining a childbirth-specific anesthesia complications indicator, where the denominator is all deliveries and the numerator is expanded to include 2 additional ICD-9-CM codes (668 [complication of administration of anesthetic or sedation in labor and delivery] and 349 [reaction to spinal or spinal headache]), we were able to capture a higher rate of anesthesia complications specific to childbirth (0.31%). Importantly, the additional codes accounted for approximately 90% of cases. It should be noted that although we referred to this as a rate throughout the article, for simplicity and comparative purposes, it is actually a ratio (number of complications/all childbearing women). The denominator includes all childbearing women and is not limited to childbearing women receiving analgesia or anesthesia. This is a limitation of the administrative discharge data set, which currently does not have a mechanism to track anesthesia-related procedure codes.
Nonetheless, our findings suggest that administrative data can be used to track childbirth-related anesthesia complications. Furthermore, our findings support current published reports that anesthesia complications are indeed rare.21 This perhaps explains why this exploratory indicator has not been further advanced as a PSI for the general population.g However, the rate of childbirth-specific anesthesia complications was more than twice that measured by the PSI as defined by AHRQ for the general population, lending support for additional studies to validate anesthesia complications as a potential PSI for childbirth. Although the aggregate rate for all deliveries (0.31%) appears low, it is within range or exceeds other national PSIs that are currently endorsed by AHRQ and others such as PSI 9: Hemorrhage or Hematoma Rate; PSI 12: Pulmonary Embolism Deep Vein Thrombosis Rate; PSI 13: Wound Dehiscence Rate; and PSI 15: Accidental Puncture or Laceration Rate,11,22–24 which have observed rates of 5.86, 4.51, 1.85, and 2.45 per 1000 eligible patients, respectively.h
The AHRQ PSIs were designed to screen for potentially avoidable complications that can be prevented by health system interventions. To be considered a good quality indicator, the Institute of Medicine specified that certain criteria need to be met.25 In the current study, the number of cases was sufficient to enable hospital-level monitoring, and the proposed expanded definition for anesthesia complications specific to childbirth met the 4 AHRQ/Institute of Medicine evaluation criteria for quality indicators as illustrated below:
- Importance: Given that childbirth is the leading cause for hospital admissions and a majority of laboring women use a form of anesthesia, the potential burden is large. Childbirth-related anesthesia complications are meaningful to patients, payers, and policy makers. Furthermore, the high rate of childbirth-related complications suggests that this proposed childbirth-specific measure is important to account for complications experienced by women who deliver vaginally and/or receive neuraxial anesthesia during childbirth.
- Scientific acceptability: To be applicable to childbirth, the original measure had to be modified to account for women who deliver vaginally and/or experience an adverse reaction to neuraxial anesthesia. This enhancement resulted in higher case ascertainment. Our analysis suggests that the variation attributable to the hospital was low judging by the VPC values, which likely reflect the rarity of the event. However, we were able to demonstrate sufficient variation at the hospital level to identify “outlier” hospitals with higher than expected rates of anesthesia complications. Depending on the methodology chosen (13 or 18 outlier hospitals), 80 or 136 adverse events per year, respectively, may be preventable in the State of California if these outlier hospitals were able to decrease their adverse event rates to the level of the mean-adjusted risk rate. Additional work is needed to establish the reliability of this measure and to determine if there are clinical practices within these outlier hospitals that are susceptible to system-level change.
- Feasibility: The rate of childbirth-specific anesthesia complications can be derived from administrative data, at low cost, and allows for pertinent subgroup analysis (by region, hospital, delivery method, patient demographics). Notably, the effect of case-mix adjustment on the results was minimal and based on this analysis is not recommended (data not shown). Although medical chart abstraction is the “gold standard,” it is not a feasible option, primarily due to the high volume of hospital deliveries. Therefore, the use of administrative data for this proposed indicator is both acceptable and recommended.
- Usability: The rate of childbirth-specific anesthesia complications is an adaptation of an existing AHRQ experimental indicator and is easily interpretable by clinicians and patients.
Although hospital discharge data are provided by nearly all hospitals and thus allow nearly universal monitoring, they also have inherent weaknesses, which may include coding errors and under- or overreporting of medical conditions or complications.26 Hospitals without full-time anesthesia services may also report fewer complications because fewer procedures are undertaken.27 Limitations and variations in the number of diagnoses and procedures documented may also bias reported rates.28 Finally, AHRQ criterion validity, sensitivity, and positive-predictive value have reported variability between 29% and 56% and 44% and 74%.28
Despite these limitations, we suggest further exploration of a childbirth-specific anesthesia complication PSI. Specifically, we have modified the definition to include complications that occur among both cesarean and vaginal deliveries and that occur from both types of childbirth-related anesthesia (regional and general). The calculated rate of morbidity from childbirth-related anesthesia complications exceeds that of the AHRQ general population. In spite of low VPC values suggesting that little variation in hospital rates is explained by hospital factors, we have been able to identify a group of hospitals whose childbirth-specific anesthesia complication rate was significantly higher than the 75th percentile, allowing for benchmarking at the hospital level, and a potential opportunity for improvement. Institutions with high anesthesia complications in pregnancy could trend these data by condition, provider, or provider type (e.g., nurse anesthetist or anesthesiologist) and direct quality improvement efforts as appropriate.
Name: Samia El Haj Ibrahim, MPH.
Contribution: This author helped design the study and prepare the manuscript.
Attestation: Samia El Haj Ibrahim approved the final manuscript.
Name: Moshe Fridman, PhD.
Contribution: This author helped design the study, conducted the data analysis, and prepare the manuscript.
Attestation: Moshe Fridman approved the final manuscript. Moshe Fridman attests to the integrity of the original data and the analysis reported in this manuscript.
Name: Lisa M. Korst, MD, PhD.
Contribution: The author helped design the study, analyze the data, and prepare the manuscript.
Attestation: Lisa M. Korst approved the final manuscript.
Name: Kimberly D. Gregory, MD, MPH.
Contribution: This author helped design the study and prepare the manuscript.
Attestation: Kimberly D. Gregory approved the final manuscript, attests to the integrity of the original data and the analysis reported in this manuscript, and is the archival author.
This manuscript was handled by: Cynthia A. Wong, MD.
The authors thank the reviewers and editor for their insightful comments and suggestions on how to make the presentation clearer and more meaningful.
a National Quality Forum. Available at: www.nqf.org. Accessed October 24, 2013.
b Agency for Healthcare Research and Quality. Available at: www.qualityindicators.ahrq.gov. Accessed October 24, 2013.
c OSHPD Patient Discharge Data Set. Available at: www.oshpd.ca.gov. Accessed October 24, 2013.
d Agency for Healthcare Research and Quality: Patient Safety Indicators. Software Documentation, version 3.2. Available at: www.qualityindicators.ahrq.gov/Downloads/Software/SAS/V32/psi_sas_documentation_v32.pdf. Accessed October 24, 2013.
e Agency for Healthcare Research and Quality Experimental Quality Indicators #1, Technical Specifications, Rate of Complications of Anesthesia, version 4.5. Available at: www.qualityindicators.ahrq.gov. Accessed October 24, 2013.
f The American Congress of Obstetricians and Gynecologists. Available at: www.acog.org/About_ACOG/ACOG_Departments/Patient_Safety_and_Quality_Improvement/reVITALize_Obstetric_Data_Definitions. Accessed October 24, 2013.
g Agency for Healthcare Research and Quality. Available at: www.qualitymeasures.ahrq.gov/browse/archive.aspx?type=2. Accessed October 24, 2013.
h Agency for Healthcare Research and Quality. Benchmark Data Tables. May 2013. Available at: www.qualityindicators.ahrq.gov/Downloads/Modules/PSI/V45/Version_45_Benchmark_Tables_PSI.pdf. Accessed October 24, 2013.
1. Farquhar M Patient Safety and Quality: An Evidence-Based Handbook for Nurses. 2008 Rockville, MD Agency for Healthcare Research and Quality
2. Agency for Healthcare Research and Quality. AHRQ Quality Indicators—Guide to Patient Safety Indicators. 2003 Rockville, MD Agency for Healthcare Research and Quality
3. Martin JA, Hamilton BE, Ventura SJ, Osterman MJK, Mathews TJ. Births: final data for 2011. Natl Vital Stat Rep. 2013;62:1–16
4. Osterman MJK, Martin JA. Epidural and spinal anesthesia use during labor: 27-state reporting area, 2008. Natl Vital Stat Rep. 2011;59:1–13, 16
5. Bucklin AB, Hawkins JL, Anderson JR, Ullrich FA. Obstetric anesthesia workforce survey. Anesthesiology. 2005;103:645–53
6. Hobbs A, Cockerham R. General anesthesia for operative obstetrics. Anaesth Intens Care. 2013;14:346–9
7. Afolabi BB, Lesi FE. Regional versus general anaesthesia for caesarean section. Cochrane Database Syst Rev. 2012;10:CD004350
8. Wlody D. Complications of regional anesthesia in obstetrics. Clin Obstet Gynecol. 2003;46:667–78
9. American College of Obstetricians and Gynecologists. . Analgesia and cesarean delivery rates. ACOG Committee Opinion. No. 339. Obstet Gynecol. 2006;107:1487–8
10. Agency for Healthcare Research and Quality. Patient Safety Indicators Software Instruction, SAS, Version 4. 2011 Rockville, MD Battelle
11. Gregory KD, Korst LM, Lu MC, Fridman M. AHRQ patient safety indicators: time to include hemorrhage and infection during childbirth. Jt Comm J Qual Patient Saf. 2013;39:114–22
12. SAS Institute. . The GLIMMIX procedure. SAS/STAT 9.2 User’s Guide. 20102nd ed. Cary, NC SAS In:
13. Goldstein H Multilevel Statistical Models. 2003 London, UK Hodder Arnold
14. Goldstein H, Browne WJ, Rasbash J. Partitioning variation in multilevel models. Understanding Statistics. 2002;1:223–31
15. Jones HE, Ohlssen DI, Spiegelhalter DJ. Use of the false discovery rate when comparing multiple health care providers. J Clin Epidemiol. 2008;61:232–40
16. Benjamini Y, Hochberg Y. Controlling the false discovery rate: a practical and powerful approach to multiple testing. J R Stat Soc B. 1995;57:289–300
17. Amnesty International. Deadly Delivery: The Maternal Health Care Crisis in the USA. 2010 New York, NY: Amnesty International
18. Al-Mufti R, McCarthy A, Fisk NM. Survey of obstetricians’ personal preference and discretionary practice. Eur J Obstet Gynecol Reprod Biol. 1997;73:1–4
19. Klein MC. Obstetrician’s fear of childbirth: how did it happen? Birth. 2005;32:207–9
20. Institute of Medicine and National Research Council. An Update on Research Issues in the Assessment of Birth Settings—Workshop Summary. 2013 Washington, DC The National Academics Press
21. Kuklina EV, Meikle SF, Jamieson DJ, Whiteman MK, Barfield WD, Hillis SD, Posner SF. Severe obstetric morbidity in the United States: 1998-2005. Obstet Gynecol. 2009;113:293–9
22. Koch CG, Li L, Hixson E, Tang A, Phillips S, Henderson JM. What are the real rates of postoperative complications: elucidating inconsistencies between administrative and clinical data sources. J Am Coll Surg. 2012;214:798–805
23. Henderson KE, Recktenwald AJ, Reichley RM, Bailey TC, Waterman BM, Diekemper RL, Storey PE, Ireland BK, Dunagan WC. Clinical validation of the AHRQ postoperative venous thromboembolism patient safety indicator. Jt Comm J Qual Patient Saf. 2009;35:370–6
24. Zrelak PA, Utter GH, Sadeghi B, Cuny J, Baron R, Romano PS. Using the Agency for Healthcare Research and Quality patient safety indicators for targeting nursing quality improvement. J Nurs Care Qual. 2012;27:99–108
25. Hurtado MP, Swift EK, Corrigan JM Envisioning the National Health Care Quality Report. 2001 Washington, DC National Academy Press
26. Thomas EJ, Petersen LA. Measuring errors and adverse events in health care. J Gen Intern Med. 2003;18:61–7
27. Marmor TR, Krol DM. Labor pain management in the United States: understanding patterns and the issue of choice. Am J Obstet Gynecol. 2002;86:173–80
28. Romano PS, Mull HJ, Rivard PE, Zhao S, Henderson WG, Loveland S, Tsilimingras D, Christiansen CL, Rosen AK. Validity of selected AHRQ patient safety indicators based on VA National Surgical Quality Improvement Program data. Health Serv Res. 2009;44:182–204