Pregnancy Testing in Women of Reproductive Age in US Emergency Departments, 2002 to 2006: Assessment of a National Quality Measure

Article Outline

• Abstract
• Introduction
  • Background
  • Importance
  • Goals of This Investigation
• Materials and Methods
  • Study Design
  • Setting, Selection of Participants, and Methods of Measurement
  • Data Collection and Processing
  • Outcome Measures
  • Primary Data Analysis
• Results
• Limitations
• Discussion
• Acknowledgment
• Appendix
• References
• Copyright

Study objective

We assess performance and explore definitions for a new emergency department (ED) quality measure: the proportion of women aged 14 to 50 years who have abdominal pain and receive pregnancy testing (aimed at detecting ectopic pregnancy).

Methods

We analyzed data from the National Hospital Ambulatory Medical Care Survey (NHAMCS) (2002 to 2006) to test trends and predictors of the new measure, using both restrictive and broad definitions from the International Classification of Diseases, Ninth Revision (ICD-9) and reason-for-visit codes, and determine the proportion of women with ectopic pregnancy who had undergone pregnancy testing. For comparison, we conducted a detailed chart review in 4 US hospitals among patients who visited the ED in 2006.

Results

Using a broad ICD-9 definition for inclusion in NHAMCS, 2.13 million women aged 14 to 50 years with abdominal pain visited an ED annually between 2002 and 2006. Of those, 33.0% (95% confidence interval [CI] 30.5% to 35.5%) received pregnancy testing. Testing rates were materially stable, regardless of the definition used (broad or restrictive ICD-9 or reason-for-visit code). Among women with an ICD-9 diagnosis of ectopic pregnancy, 55.6% (95% CI 43.7% to 67.6%) had a documented pregnancy test. In the chart review, among 200 women aged 14 to 50 years and with abdominal pain, 89.4% (95% CI 85.0% to 94.0%) were eligible for the measure; of those, 93.9% (95% CI 90.3% to 97.4%) received testing.

Conclusion

Analysis of national ED survey data demonstrated a large performance gap for a new pregnancy testing quality measure, whereas focused chart review at 4 sites showed a smaller gap. Given these discrepancies, additional study is recommended before the widespread implementation of the pregnancy testing measure as an assessment of ED performance.

Back to Article Outline

Introduction 

Background 

Currently, several quality measures apply directly to emergency department (ED) care and are publicly reported by the Department of Health and Human Services.¹ In 2008, the Centers for Medicare & Medicaid Services commissioned the National Quality Forum to convene a working group to review and endorse additional quality measures for ED care. One measure approved by the National Quality Forum expert panel for emergency care was “the percent of women, ages 14-50 years old, who present to the ED with a chief complaint of abdominal pain who have a pregnancy test (urine or serum) ordered in the ED.”² Exclusions include those who are postmenopausal, have had a hysterectomy, or have a documented pregnancy. The rationale for measuring pregnancy testing with a presentation of abdominal pain is to detect ectopic pregnancy, as supported by a recommendation of the American College of Emergency Physicians.³ Because pain can be the only presenting symptom of ectopic pregnancy, which is potentially lethal if misdiagnosed, the routine use of pregnancy tests in women with suggestive symptoms is generally considered standard ED practice.⁴, ⁵, ⁶, ⁷

Importance 

The deployment of quality measures in the ED has been controversial. Specifically, emergency physicians have objected to the pneumonia core measures because of the lack of evidence supporting the routine use of blood cultures and because of unintended consequences (antibiotic misuse associated with the measurement of antibiotic timing).⁸, ⁹, ¹⁰, ¹¹, ¹² Other measures, such as oxygenation assessment in pneumonia and aspirin use in acute myocardial infarction, are less controversial. However, both have ceiling effects (close to 100% of hospitals have already achieved both goals).¹³

Although the pregnancy measure has face validity, the National Quality Forum technical expert panel had similar concerns about a ceiling effect.¹⁴ To our knowledge, to date there are no published studies addressing current rates of ED pregnancy testing in this population, whether a performance gap exists, or whether current administrative data could accurately capture this measure. The National Quality Forum has not precisely defined which women with abdominal pain should receive testing, by determining the International Classification of Diseases, Ninth Revision (ICD-9)¹⁵ codes for the denominator definition.

Goals of This Investigation 

We sought to assess recent trends in the performance of this new pregnancy testing measure according to data from US ED visits in a nationwide data set, the National Hospital Ambulatory Medical Care Survey (NHAMCS). Second, we explored how to define this measure according to ICD-9 criteria and reason for visit classification, with the goal of making a sensible recommendation on measure specification. Third, we sought to assess the validity of using preexisting databases to develop quality measures by comparing performance rates in the national dataset with data from a targeted 4-hospital chart review.

Back to Article Outline

Materials and Methods 

Study Design 

We performed a cross-sectional descriptive analysis of pregnancy testing in women aged 14 to 50 years and with abdominal pain in US EDs. First, we analyzed the NHAMCS ED component from 2002 to 2006. Second, we performed a chart review at 4 EDs. Our study was approved by the University of Pennsylvania and Partners Healthcare institutional review boards.

Setting, Selection of Participants, and Methods of Measurement 

NHAMCS is a multistage probability survey of US ED and outpatient department visits to noninstitutional, general, and short-stay hospitals (excluding federal, military, and Veterans Affairs hospitals). From 2002 to 2006, the NHAMCS included 183,633 patient visit records at 408 hospital EDs, weighted to represent the 568 million ED visits in the United States during this time. We included ED visits by female patients aged 14 to 50 years and with abdominal pain as a reason-for-visit code or ICD-9 discharge diagnosis.

NHAMCS contains data fields typically available on an ED record, including patient demographics, limited clinical information, and several provider and hospital characteristics. NHAMCS data forms allow entry of up to 3 patient reasons for visit (coded with “A Reason for Visit Classification for Ambulatory Care”),¹⁶ up to 3 discharge diagnoses (coded with the ICD-9),¹⁵ and a check box for whether a pregnancy test was performed. NHAMCS data forms were coded from the ED chart by trained data abstracters. The instructions for recording the reason for visit are “enter the patient's complaint(s), symptom(s), or other reason(s) for visit in the patient's own words. The most important reason should be entered first.”¹⁷ The reason for visit is not merely a verbatim transcription of the patient's chief complaint, but a synthesis of the patient's reason for visit as documented by nurses and physicians. A complete description of the NHAMCS methodology is available at the Centers for Disease Control and Prevention's (CDC's) National Center for Health Statistics Web site.¹⁸

We conducted a chart review at 4 hospital EDs in the northeastern United States. Three are academic medical centers with emergency medicine residents, and one is a community teaching affiliate with no emergency medicine residents. ED volume in 2006 ranged from 30,000 to 57,000 visits per year. We identified all women aged 14 to 50 years who were treated in the ED and had a primary diagnosis of abdominal pain (ICD-9 789.xx) in 2006 from billing databases and reviewed the charts of 50 randomly selected patients per site. We excluded patients who left without being seen or left against medical advice. Trained research assistants entered data into a preformatted datasheet. The first 10 charts at each hospital were also reviewed by one of the investigators, and no discrepancies were identified with the data entered by research assistants.

Data Collection and Processing 

We obtained NHAMCS data from the publicly available microdata sets at the National Center for Health Statistics Web site.¹⁸

The National Quality Forum–approved performance measure does not include detailed denominator and numerator specifications, such as a set of ICD-9 codes for abdominal pain. Therefore, we first reviewed reason-for-visit and ICD-9 codes that would best identify women “who present to the ED with a chief complaint of abdominal pain.”² In reviewing codes, we aimed to identify those that included any abdominal pain regardless of location, but excluding specific diagnoses (eg, biliary colic), producing a broad definition. Next, we identified a limited set of codes that included only lower abdominal and pelvic pain, a restrictive definition. The broad definition is more sensitive and less specific for ectopic pregnancy because ectopic pregnancy more frequently causes lower abdominal symptoms, whereas the restrictive definition is more specific.¹⁹

One investigator (J.D.S.) manually reviewed all ICD-9 and reason-for-visit categories to identify potentially eligible codes and entered them into a spreadsheet. Two authors (J.D.S. and J.M.P.) independently rated each code as being part of the restrictive definition, the broad definition, or neither. After review of all codes, each code was assigned a final consensus ranking. The codes making up the restrictive and broad definitions are in Appendices E1 and E2 (available online at http://www.annemergmed.com).

Because both ED records and NHAMCS data can contain multiple reasons for visit and ICD-9 codes for each visit, a performance measure must also specify which of the listed ICD-9 codes will be included in the analysis (eg, primary code only). We defined the measure denominator to include codes from our restrictive definition that were listed as a primary discharge diagnosis code. We used this restrictive definition of abdominal pain in ICD-9 codes for all bivariate and multivariate analyses to avoid including patients with a more specific diagnosis (eg, biliary colic) as a primary diagnosis and nonspecific pain as a secondary or tertiary diagnosis. Additionally, as a sensitivity analysis we determined the size of the included population and the pregnancy testing rates for several other groups of female patients aged 14 to 50 years: (1) those with an ICD-9 code from our broad or restrictive definitions as a primary, secondary, or tertiary discharge code (Table E1, available online at http://www.annemergmed.com); (2) those with a reason-for-visit code from our consensus list (Table E2, available online at http://www.annemergmed.com) as the primary reason for visit; and (3) those with a reason-for-visit code from our consensus list as a primary, secondary, or tertiary reason-for-visit code.

We coded patient, visit, and hospital covariates a priori to identify predictors of test performance. Patient covariates included sex, age (14 to 19, 20 to 29, 30 to 39, and 40 to 50 years), race/ethnicity (recoded as white/non-Hispanic, black/non-Hispanic, other/non-Hispanic, Hispanic), and expected source of payment (private insurance/worker's compensation, Medicare/other insurance, Medicaid/State Children's Health Insurance Program, self-pay/charity). Visit covariates were ED disposition (admitted/transferred/observation, discharged), trainee provider (resident/intern versus no resident/intern), and midlevel provider seeing patient (nurse practitioner/physician assistant versus no nurse practitioner/physician assistant), and urgency (emergent/urgent versus not urgent). We excluded patients who left without being seen, left against medical advice, or were dead on admission. Hospital characteristics were region (Northeast, South, Midwest, West), urban location (in metropolitan statistical area, not in metropolitan statistical area), and hospital ownership type (voluntary nonprofit, government nonfederal, proprietary). Finally, we included year of survey (2002 to 2006) and who reviewed the chart for NHAMCS (hospital staff versus census staff) as covariates.

To estimate the proportion of patients with abdominal pain and confirmed pregnancy, we used the clinical classifications system algorithm to group discharge codes.²⁰ We defined pregnancy using clinical classifications system 11, “Complications of pregnancy; childbirth; and the puerperium,” which includes all ICD-9 codes consistent with confirmed pregnancy. We defined postpartum using clinical classifications system subgroup 11.6, “Other complications of birth; puerperium affecting management of mother,” which includes ICD-9 codes for postpartum complications. We identified patients with ectopic pregnancy by ICD-9 code 633.xx.¹⁵

Charts were reviewed by trained research assistants familiar with each ED's charting techniques, following an explicit protocol. The research assistants were instructed that the project sought to identify testing in female ED patients with abdominal pain, but they were not told of the project's specific aim. Demographic variables were age, race, and ethnicity. Potential measure exclusions were previous hysterectomy, documented infertility, and clinical evidence of pregnancy (defined as a gravid uterus or reported positive pregnancy test results before ED arrival). Diagnostic test variables were urine or serum pregnancy testing, CBC count, and pelvic or abdominal ultrasonography. Disposition variables were final disposition, chief complaint, and final diagnoses. The chart review was conducted with all documentation that would have been available at the ED visit, including nursing and physician notes, test orders, test results, and billing data. Data were entered into a preformatted and locked spreadsheet (Excel; Microsoft, Redmond, WA).

Outcome Measures 

Our primary outcome was the rate of pregnancy testing in the ED for each case definition of abdominal pain.

Primary Data Analysis 

For each case definition of abdominal pain, we calculated percentage of agreement between reviewers by using the κ statistic.²¹ We calculated the percentage of women receiving a pregnancy test and 95% confidence intervals (CIs) for NHAMCS and chart review data. We then conducted multiple-variable logistic regression to estimate the adjusted association of covariates on the primary outcome (use of pregnancy tests in patients with abdominal pain) in the NHAMCS data. The independent variables were all a priori covariates. The logistic regression model had good fit, with a c-statistic of 0.68. Bivariate and multivariate analyses were conducted only with the “restrictive” definition of abdominal pain based on the primary ICD-9 code. To determine whether pregnancy (a measure exclusion) was associated with differential testing rates among women with abdominal pain, we calculated the proportion of patients who received a pregnancy test (yes/no), stratified by a final ED diagnosis consistent with pregnancy (clinical classifications system 11). Because women with postpartum complications should be excluded from a pregnancy testing measure, we analyzed the prevalence of such diagnoses among women with abdominal pain (clinical classifications system 11.6). Finally, we analyzed the pregnancy testing rate for women with final ED diagnosis ICD-9 codes for ectopic pregnancy.

We performed all statistical analyses with SAS 9.3 (SAS Institute, Inc., Cary, NC). In NHAMCS, we accounted for the complex sampling design with survey procedures.²², ²³

Back to Article Outline

Results 

We identified 79 ICD-9 codes and 59 reason-for-visit codes for abdominal pain and related symptoms. After independent review and consensus discussion, we labeled 38 ICD-9 codes that would be sensitive for abdominal pain as a “broad” definition (Table E1, available online at http://www.annemergmed.com). Twenty-two of these codes that would be highly specific for abdominal pain were together labeled the “restrictive” definition (Table E1, available online at http://www.annemergmed.com). One ICD-9 code group (789.xx; “Abdominal pain, site unspecified or specified”) accounted for 81.1% of all ED visits in the restrictive definition. We had 96.2% agreement (κ=0.92) between reviewers for the “broad” definition and 92.4% agreement (κ=0.82) for the “restrictive” definition. A similar process with reason-for-visit codes yielded 35 “broad” codes that were deemed sensitive and a subset of 10 “restrictive” codes that were deemed highly specific (Table E2, available online at http://www.annemergmed.com). We had 67.8% agreement (κ=0.30) between reviewers for the “broad” definition and 94.9% agreement (κ=0.83) for the “restrictive” definition.

Projecting according to the NHAMCS sampling design, from 2002 to 2006 there were an average estimated 1.82 million (95% CI 1.59 to 2.04 million) ED visits per year by women aged 14 to 50 years and with a discharge ICD-9 diagnosis that met our restrictive definition of abdominal pain and 2.13 million (95% CI 1.86 to 2.39) by women whose diagnosis met our broader definition. Thirty-four percent (95% CI 31.2% to 36.7%) of patients with nonspecific abdominal pain by the restrictive definition and 33.0% (95% CI 30.4% to 35.5%) under the broad definition received a pregnancy test. Including secondary or tertiary ICD-9 codes for nonspecific abdominal pain within the 3 listed ICD-9 codes did not materially affect the rate of pregnancy testing (Table 1).

Table 1. Percentage of female patients aged 14 to 50 years, with abdominal pain^⁎ and receiving pregnancy test, 2002 to 2006.^†

Abdominal Pain Definition	Cases in Data Set	Estimated Total ED Visits per Year, in Thousands
		Total No.	Receiving a Pregnancy Test
			No.	% (95% CI)
ICD-9 discharge diagnosis
Restrictive definition, primary ICD-9 code	2,757	1,822	620	34.0(31.2-36.7)
Broad definition, primary ICD-9 code	3,217	2,126	702	33.0(30.5-35.5)
Restrictive definition in any ICD-9 code	3,481	2,317	784	33.8(31.4-36.2)
Broad definition in any ICD-9 code	4,062	2,684	871	32.4(30.2-34.6)
RFV
Restrictive definition, primary RFV code only	5,624	3,628	1,266	34.9(32.8-37.0)
Broad definition, primary RFV code only	7,578	4,772	1,681	35.2(33.3-37.1)
Restrictive definition in any RFV code	7,490	4,707	1,614	34.2(32.2-36.2)
Broad definition in any RFV code	9,616	5,969	2,059	34.4(32.6-36.3)

RFV, Reason for visit.

During the same period, there were on average 3.63 million (95% CI 3.20 to 4.05 million) estimated ED visits per year by women with a reason for visit that met our restrictive definition of abdominal pain and 4.77 million (95% CI 4.23 to 5.31 million) visits by women whose reason for visit met our broad definition. Among patients with a reason for visit of abdominal pain, 34.9% (95% CI 32.8% to 37.0%) of patients in the restrictive definition and 35.2% (95% CI 33.3% to 37.1%) in the broad definition received a pregnancy test (Table 1).

Unadjusted pregnancy testing rates varied by patient, hospital, and visit characteristics (Table 2). Specifically, lower testing rates were associated with increasing age, admission to hospital, Medicare insurance, provider characteristics (no resident/intern or physician assistant/nurse practitioner involvement), and hospital location in the western region. Patients whose NHAMCS record was completed by the census staff rather than hospital staff were significantly more likely to receive a pregnancy test. We found no yearly trend and no difference by race/ethnicity.

Table 2. Percentage of female patients aged 14 to 50 years, with abdominal pain^⁎ and receiving pregnancy test, by patient characteristics.

SCHIP, State Children's Health Insurance Program; NP, nurse practitioner; PA, physician assistant; MSA, metropolitan statistical area.

Percentages, odds ratios, and CIs account for NHAMCS multistage sampling design.

In adjusted analysis, 5 patient, hospital, and visit characteristics were significantly associated with pregnancy testing rates. Lower testing rates were associated with older age, Medicare insurance, being admitted, not being treated by a trainee, and being in an urban ED (Table 2). Chart review by census employees was associated with a significantly higher rate of pregnancy testing. We found no variation in testing by patient's race/ethnicity.

Projecting according to the NHAMCS sampling design, from 2002 to 2006 there were a total of 401,000 estimated ED visits (95% CI 328,000 to 474,000 visits) per year by women aged 14 to 50 years and with a discharge diagnosis of ectopic pregnancy. Among women with a diagnosis of ectopic pregnancy, the pregnancy testing rate was 55.6% (95% CI 43.7% to 67.6%).

Among all women who met our restrictive definition of abdominal pain, 8.0% (95% CI 6.8% to 9.2%) also had an ICD-9 code consistent with pregnancy. Among this subset of women, the pregnancy testing rate was 39.7% (95% CI 32.2% to 47.3%), not significantly higher than the rate among women meeting our restrictive definition of abdominal pain who did not receive a discharge diagnosis of pregnancy (33.3%; 95% CI 30.7% to 35.9%; P=.10). There were fewer than 30 patients with a discharge diagnosis specific for a postpartum condition (clinical classifications system 11.6), too small a sample from which to calculate stable frequency estimates.

Among 200 women aged 14 to 50 years and with a final diagnosis of abdominal pain (789.xx) in 4 EDs in 2006, 89.4% (95% CI 85.0% to 94.0%) were eligible for the pregnancy quality measure (Table 3). Reasons for ineligibility were 7 patients with a previous hysterectomy, 10 who had documented infertility, and 4 who had clinical evidence of pregnancy or reported a positive pregnancy test before their ED visit. Pregnancy testing was performed in the ED for 93.9% (95% CI 90.3% to 97.4%) of eligible women, with rates ranging from 88.1% to 97.9% across the EDs.

Table 3. Results of a chart review at 4 hospitals.

Hospital	A	B	C	D	Total
ED volume	57,000	47,000	55,000	30,000
Emergency medicine residents	Y	N	Y	Y
Charts reviewed	50	50	50	50	200
Previous hysterectomy	2	3	1	1	7
Documented infertility	4	3	3	0	10
Clinical evidence of pregnancy	2	0	1	1	4
Eligible patients	42	44	45	48	179
Number receiving pregnancy test	37	42	42	47	168
Testing rate, % (95% CI)	88.1(78.3-97.9)	95.5(89.3-100)	91.3(83.2-99.4)	97.9(93.9-100)	93.9(90.3-97.4)

Back to Article Outline

Limitations 

Our study has several limitations. NHAMCS is abstracted from ED charts and is limited by the quality of charting data. However, the National Center for Health Statistics has conducted NHAMCS since 1992 with robust quality control, such as training of office staff by census field representatives and a 2-way 10% independent verification procedure, leading to average keying error rate for nonmedical items of less than 1% for items that required medical coding, discrepancy rates ranged below 1%.¹⁷ It is possible that some pregnancy tests (especially urine tests), such as those performed by triage protocol, are not recorded in the chart or are missed during NHAMCS data abstraction, which would bias our results toward a lower pregnancy testing rate and a larger performance gap. Previous studies show that NHAMCS reports a high prevalence of diagnostic test performance that appears to reflect rates in clinical practice such as ECG rates in chest pain (80%) and syncope (60%).²⁴, ²⁵

We were unable to account for all measure exclusions in NHAMCS data, specifically, women with a previous hysterectomy and those who are postmenopausal or arrive at the ED with a documented pregnancy. Additionally, we were unable to assess whether patients underwent bedside ultrasonography in the ED to confirm intrauterine pregnancy, a process that would substitute for a serologic or urine pregnancy test and should be specified as successful completion of the National Quality Forum measure. Our inability to measure these exclusions and alternate pregnancy testing methods would bias our results to report a lower pregnancy testing rate and a larger performance gap. Although documented pregnancy is a measure exclusion, we did not exclude such patients because NHAMCS does not contain data needed to determine whether or not a patient was known to be pregnant on arrival. Excluding all patients with a diagnosis of pregnancy would misclassify cases in which pregnancy status was unknown, meaning appropriate care was delivered and pregnancy was detected. Overall, including pregnant patients seems unlikely to explain the full performance gap because only approximately 8% of the patients with the restrictive definition of abdominal pain also had an ICD-9 code consistent with pregnancy.

The prevalence of hysterectomy in US women aged 18 to 49 years is approximately 9%.²⁶ The prevalence of menopause among women aged 18 to 49 years is small and not easily determined; only about 1% of women aged 40 years and younger have reached menopause, and about 10% of women have reached menopause at age 47.²⁷, ²⁸ These exclusions account for some but not all of the performance gap because our chart review showed that the exclusion rate was about 10%.

The chart review was limited to 4 hospitals in the northeastern United States; the results may not be generalizable across US EDs.

Back to Article Outline

Discussion 

We analyzed a newly approved National Quality Forum quality measure, pregnancy testing for women with abdominal pain, and found a clinically important discrepancy between performance in a nationally representative database (NHAMCS) and a chart review at EDs. In the national database, only about one third of women aged 14 to 50 years and with a diagnosis of abdominal pain received a pregnancy test in the ED, regardless of the exact denominator specifications. In an explicit chart review at EDs, more than 90% of eligible patients received a pregnancy test. Although our analysis of the measure in NHAMCS does not include several important exclusions from the National Quality Forum–approved performance measure, this is unlikely to explain the discrepancy.

There are several possible reasons for the large performance difference between data sets. First, the 4 EDs may not be representative of the average US ED. In these hospitals, virtually all attending physicians are emergency medicine board certified or prepared, and three fourths of the hospitals have emergency medicine residents. Because the majority of ED patients nationwide are not treated in urban, academic EDs, actual US performance on this quality measure may be lower.

Second, methodological differences may explain some of the differences in findings. Our chart review was specifically designed to look for pregnancy testing in women with abdominal pain, so the sensitivity to detect pregnancy testing may have been higher than that of the NHAMCS. However, the NHAMCS is a structured chart review that has been conducted for 15 years by the National Center for Health Statistics, with defined methods and strict quality control. Pregnancy testing is abstracted individually as one of a list of diagnostic tests. Abstracters are trained to look for wording of “pregnancy test” or “HCG” (human chorionic gonadotropin). Given these clear guidelines, a systematic problem in the NHAMCS about abstracting pregnancy tests seems unlikely. Nevertheless, 2 findings raise questions about the internal validity of pregnancy testing data in NHAMCS. First, pregnancy testing was documented among only 55% of patients with a diagnosis of ectopic pregnancy. We would expect this number to be close to 100%, but we are not convinced that such a small sample size (146 unweighted observations) is suitable to explain the accuracy of all pregnancy test charting in the NHAMCS. Of note, some patients with ectopic pregnancy diagnosed outside the ED may be referred to the ED before treatment and may not undergo pregnancy testing because they already have a diagnosis. Second, we found that patients reviewed by census staff reviewers were more likely to receive a pregnancy test than patients reviewed by hospital employees, presumably because census reviewers perform a more thorough review of the chart. Although this raises a question about the quality of standard hospital staff chart reviews, there was still a large performance gap among census-reviewed charts, with fewer than half of eligible women receiving pregnancy tests.

A secondary finding in our study was that there were no major differences in performance rates related to the measure definition, which is important because developing precise specifications is critical. Specifications include a numerator (those who received the test/treatment), denominator (those who should have received the test/treatment), appropriate exclusions determined by detailed chart review, and data source and administrative codes for data collection.²⁹ The approved National Quality Forum measure includes a description of the numerator, denominator, and exclusions but lacks specific ICD-9 codes on which to base administrative data collection.³⁰ We defined 2 potential denominator definitions (one restrictive and one broad) and found similar testing rates in each population in the NHAMCS data. According to these results, we would suggest that the National Quality Forum adopt our restrictive definition for the measure denominator, or just ICD-9 code 789.xx, which accounted for more than 80% of all cases meeting our restrictive definition. It is reassuring that testing rates do not vary among the different definitions because slight technical differences in measure specifications should not result in dramatically different performance rates. Second, we believe a more restrictive denominator definition is clinically sensible because a higher proportion of such patients should have received a pregnancy test. For example, the broad group included both upper and lower abdominal pain, whereas the restrictive group included only lower abdominal pain. Although upper abdominal pain is a possible presentation of ectopic pregnancy, lower abdominal pain is more common and there is a stronger indication for a pregnancy test in the latter group. Therefore, the restrictive definition is less likely to give us false positives (ie, those who are measured but should not necessarily have received the test).

Several factors were associated with differences in pregnancy testing rates after multivariate adjustment. These may reflect rational clinical reasoning if the variable is associated with a lower likelihood that the patient could be pregnant, as fertility declines with increasing age and admission to the hospital may be a marker for a chronic illness, many of which reduce fertility. Also, being treated by an intern or resident was associated with higher testing rates, which could reflect either higher testing rates or better documentation by trainees.

There was not an increasing trend of pregnancy testing from 2002 to 2006. The stability of performance further supports adoption of the new National Quality Forum measure as a means to drive performance improvement.

We found that pregnancy testing was performed for only about one third of female patients aged 14 to 50 years and with abdominal pain in the national data set but more than 90% of the time in northeastern academic EDs. This discrepancy could be partially due to biases in either data set. In the NHAMCS, an inability to exclude women without childbearing potential (eg, hysterectomy) would bias our results toward a lower pregnancy testing rate and a larger performance gap, although our chart reviews in 4 academic hospitals may not represent the diversity of ED practice across the United States and bias results toward higher performance. In the NHAMCS, we found no difference between pregnancy testing rates with several definitions of abdominal pain. According to these results and clinical reasoning, the National Quality Forum should consider adopting a more restrictive definition for this measure because it may minimize the number of false-positive results. Before widespread implementation in quality measurement programs, the measure should receive further field testing in a diverse set of EDs across the United States.

Back to Article Outline

The authors acknowledge the valuable contributions of research assistants Geoffrey Fauchet, BA, Christopher Delaney, BS, Jacqueline Moreau, BA, and Joshua Isserman, MS. Thanks to Jeanne Zimmermann, BA, and Paul Guttry, BA, for thoughtful editorial suggestions.

Back to Article Outline

Appendix 

Table E1. ICD-9 discharge codes reviewed for ED pregnancy test quality measure.

Projections based on counts of fewer than 30 records are unstable. We present the actual number of cases in the NHAMCS for relative frequency.

Table E2. Reason-for-visit codes reviewed for ED pregnancy test quality measure.

Projections based on counts of fewer than 30 records are unstable. We present the actual number of cases in the NHAMCS for relative frequency.

NOS, Not otherwise specified; HCG, human chorionic gonadotropin.

Back to Article Outline

References 

1. Hospital Compare: a quality tool provided by Medicare. http://www.hospitalcompare.hhs.govAccessed December 10, 2008
   • View In Article
2. National Quality Forum. NQF endorses measures to address care coordination and efficiency in hospital emergency departments [press release]. National Quality Forum Web site http://www.qualityforum.org/News_And_Resources/Press_Releases/2008/NQF_ENDORSES_MEASURES_TO_ADDRESS_CARE_COORDINATION_AND_EFFICIENCY_IN_HOSPITAL_EMERGENCY_DEPARTMENTS.aspxAccessed September 10, 2009
   • View In Article
3. Clinical policy: critical issues for the initial evaluation and management of patients presenting with a chief complaint of nontraumatic acute abdominal pain. Ann Emerg Med. 2000;36:406–415
   • View In Article
   • Abstract
   • Full Text
   • Full-Text PDF (53 KB)
4. Hedges JR, Kaib JJ, Armao JC. Detection of ectopic pregnancy in an outpatient population: the role of the beta-HCG level. J Emerg Med. 1984;2:85–93
   • View In Article
   • MEDLINE
   • CrossRef
5. Ramoska EA, Sacchetti AD, Nepp M. Reliability of patient history in determining the possibility of pregnancy. Ann Emerg Med. 1989;18:48–50
   • View In Article
   • Abstract
   • Full-Text PDF (229 KB)
   • CrossRef
6. Kaplan BC, Dart RG, Moskos M, et al. Ectopic pregnancy: prospective study with improved diagnostic accuracy. Ann Emerg Med. 1996;28:10–17
   • View In Article
   • Abstract
   • Full Text
   • Full-Text PDF (724 KB)
   • CrossRef
7. Kohn MA, Kerr K, Malkevich D, et al. Beta-human chorionic gonadotropin levels and the likelihood of ectopic pregnancy in emergency department patients with abdominal pain or vaginal bleeding. Acad Emerg Med. 2003;10:119–126
   • View In Article
   • MEDLINE
   • CrossRef
8. Walls RM, Resnick J. The Joint Commission on Accreditation of Healthcare Organizations and Center for Medicare and Medicaid services community-acquired pneumonia initiative: what went wrong?. Ann Emerg Med. 2005;46:409–411
   • View In Article
   • Full Text
   • Full-Text PDF (75 KB)
   • CrossRef
9. Pines JM. Measuring antibiotic timing for pneumonia in the emergency department: another nail in the coffin. Ann Emerg Med. 2007;49:561–563
   • View In Article
   • Full Text
   • Full-Text PDF (61 KB)
   • CrossRef
10. Wachter RM, Flanders SA, Fee C, et al. Public reporting of antibiotic timing in patients with pneumonia: lessons from a flawed performance measure. Ann Intern Med. 2008;149:29–32
    • View In Article
11. Kanwar M, Brar N, Khatib R, et al. Misdiagnosis of community-acquired pneumonia and inappropriate utilization of antibiotics: side effects of the 4-h antibiotic administration rule. Chest. 2007;131:1865–1869
    • View In Article
    • MEDLINE
    • CrossRef
12. Welker JA, Huston M, McCue JD. Antibiotic timing and errors in diagnosing pneumonia. Arch Intern Med. 2008;168:351–356
    • View In Article
    • CrossRef
13. Jha AK, Li Z, Orav EJ, et al. Care in U.S. hospitals—the hospital quality alliance program. N Engl J Med. 2005;353:265–274
    • View In Article
    • CrossRef
14. National Quality Forum. NQF member and public comments requested on the draft report “National voluntary consensus standards for emergency care—phase II: hospital-based emergency department care.”. National Quality Forum Web site http://www.qualityforum.org/projects/ongoing/emergency/comments2/index.aspAccessed December 10, 2008
    • View In Article
15. 2006 ICD-9-CM volume 1: diagnosis codes. ICD9Data.com Web site http://www.icd9data.com/2006/Volume1/default.htmAccessed December 10, 2008
    • View In Article
16. Schneider D, Appleton L, McLemore T. A reason for visit classification for ambulatory care. Vital Health Stat 2. 1979;2:i-vi, 1-63
    • View In Article
17. Public-Use Data File Documentation (2006 National Hospital Ambulatory Medical Care Survey). Hyattsville, MD: National Center for Health Statistics; 2006;ftp://ftp.cdc.gov/pub/Health_Statistics/NCHS/Dataset_Documentation/NHAMCS/doc06.pdfAccessed September 10, 2009
    • View In Article
18. Centers for Disease Control and Prevention. Ambulatory health care data. National Hospital Ambulatory Medical Care Survey (NHAMCS) Web site http://www.cdc.gov/nchs/about/major/ahcd/ahcd1.htmAccessed December 10, 2008
    • View In Article
19. Ramakrishnan K, Scheid DC. Ectopic pregnancy: forget the “classic presentation” if you want to catch it sooner. J Fam Pract. 2006;55:388–395
    • View In Article
    • MEDLINE
20. Agency for Healthcare Research and Quality. HCUP clinical classifications software (CCS) for ICD-9-CM (Healthcare Cost and Utilization Project (HCUP). 2000-2003). http://www.hcup-us.ahrq.gov/toolssoftware/ccs/ccs.jspAccessed January 11, 2009
    • View In Article
21. Landis JR, Koch GG. The measurement of observer agreement for categorical data. Biometrics. 1977;33:159–174
    • View In Article
    • MEDLINE
    • CrossRef
22. Hing E, Gousen S, Burt C. Guide to using masked design variables to estimate standard errors in public use files of the National Ambulatory Medical Care Survey and the National Hospital Ambulatory Medical Care Survey. Inquiry. 2003;40:401–415
    • View In Article
    • MEDLINE
    • CrossRef
23. SAS documentation of survey procedures. http://support.sas.com/documentation/cdl/en/statugsurveylogistic/61836/PDF/default/statugsurveylogistic.pdfAccessed January 11, 2009
    • View In Article
24. Pezzin LE, Keyl PM, Green GB. Disparities in the emergency department evaluation of chest pain patients. Acad Emerg Med. 2007;14:149–156
    • View In Article
    • CrossRef
25. Schuur JD, Justice A. Measuring quality of care in syncope: case definition affects reported electrocardiogram use but does not bias reporting. Acad Emerg Med. 2009;16:40–49
    • View In Article
    • CrossRef
26. Saraiya M, Lee N, Blackman D, et al. Self-reported papanicolaou smears and hysterectomies among women in the United States. Obstet Gyencol. 2001;98:269–278
    • View In Article
27. Henderson KD, Bernstein L, Henderson B, et al. Predictors of the timing of natural menopause in the Multiethnic Cohort Study. Am J Epidemiol. 2008;167:1287–1294
    • View In Article
    • CrossRef
28. Luborsky JL, Meyer P, Sowers MF, et al. Premature menopause in a multi-ethnic population study of the menopause transition. Hum Reprod. 2003;18:199–206
    • View In Article
    • MEDLINE
    • CrossRef
29. National Quality Forum. NQF measure submission form version 3.0. National Quality Forum Web site http://www.qualityforum.org/pdf/fmMeasureSubmissionVer3.0-2008-08-25.docAccessed December 10, 2008
    • View In Article
30. National Quality Forum. Current NQF-endorsed measures. National Quality Forum Web site http://www.qualityforum.org/pdf/Btblendorsedmeasurescurrent.xls2008Accessed December 10, 2008
    • View In Article