a b s t r a c t

Objectives: We compared emergency physician-performed pelvic ultrasonography (EPPU) with radiology de- partment-performed pelvic ultrasonography (RPPU) in emergency department (ED) female patients requiring pelvic ultrasonography and their outcomes in relation to ED length of stay, ED readmission, and alternative diag- nosis, within a 14-day follow-up period.

Methods: This was a prospective, observational study of female patients of reproductive age who required either an EPPU or RPPU for their ED evaluation. We hypothesized that patients receiving EPPU would have a length of stay reduction greater than or equal to 60 minutes, as compared with RPPU. Statistical analyses included an independent-samples t test and multivariate regression modeling to control for factors associated with ED LOS. Results: Eighteen resident physicians performed EPPU, with 15 attending physicians supervising. Forty-eight pa- tients received only EPPU, and 84 patients received only RPPU. In univariate analysis, those who received EPPU had an ED LOS 162 minutes less than those who received RPPU (95% confidence interval, 106-209 minutes). In multivariate analysis controlling for gynecologist consultation, disposition, and pregnancy status, patients who received EPPU had an ED LOS reduction of 108 minutes when compared with RPPU (95% confidence interval, 38-166 minutes). Five patients (10%) who had received EPPU and were discharged from the ED returned to the ED within 2 weeks, but none had Alternative diagnoses.

Conclusions: Patients with EPPU had statistically and clinically significant reductions in ED LOS, even when con- trolling for disposition, gynecologist consultation in the ED, and pregnancy status. No patients in the study had an alternative diagnosis within 2 weeks of EPPU.

(C) 2014

Introduction

Background

Emergency department (ED) crowding negatively influences patient care and directly and indirectly leads to increased morbidity and mor- tality [1]. Emergency department length of stay (LOS) is a well-

? Meetings: Mediterranean Emergency Medicine Congress, Kos, Greece, September 2011. AIUM Annual Convention, Phoenix, AZ, March 2012. SAEM Annual Meeting, Chicago, IL, May 2012.

?? Conflicts of interest: None of the authors have any conflicts of interest to report.

? Author contributions: ATC and CHC have joint first authorship of this manuscript.

CHC and ATC conceived and designed the study. JCF obtained research funding and provid- ed guidance for conducting this study. ATC, CHC, and DYI supervised the research associ- ates who participated in this study. DSW, CJG, and ZJM assisted in patient recruitment and data collection, both prospective and retrospective. ATC managed the data, including quality control. ATC and CLA provided statistical advice on study design and analyzed the data. ATC, CHC, and DYI drafted the manuscript, and all authors contributed to its revi- sion. ATC and CHC take responsibility for the manuscript as a whole.

* Corresponding author.

E-mail addresses: [email protected], [email protected] (A.T. Chiem).

validated metric used to assess ED throughput as well as the risk for crowding and boarding [2]. The performance of laboratory and imaging studies contributes to increased ED LOS, with an average ultrasound study performed by the radiology department contributing to an in- crease of 56 minutes to ED LOS [3]. This is particularly important, as re- cent implementation of the Affordable Care Act has been associated with an increase in patient visits in nearly half of EDs nationwide [4].

Emergency physician-performed pelvic ultrasonography (EPPU) has been shown to be accurate in confirming Intrauterine pregnancy (IUP) and lead to ED LOS reductions. Emergency physician-performed pelvic ultrasonography has been shown to be 75% to 90% sensitive and 90% to 99% specific in the identification of IUP, with a reported ED LOS reduction of between 59 and 149 minutes [5-9]. In addition, Mateer et al [10] showed that a protocol involving pregnant patients presenting with Vaginal bleeding or lower abdominal pain and EPPU led to a reduc- tion in return ED visits for missed Ectopic pregnancy.

The American College of Emergency Physicians has recommended EPPU as first line in the identification of IUP since 2001, in the 2006 Im- aging Criteria Compendium, and in the most current guidelines from

http://dx.doi.org/10.1016/j.ajem.2014.09.006

0735-6757/(C) 2014

2008 [11,12]. In each of these publications, extended or secondary indi- cations are listed for the identification of ectopic pregnancy, ovarian cysts, Ovarian torsion, Tubo-ovarian abscess, and fibroids, as well.

However, these recommendations are based on studies that have 2 general limitations. The first is that a small group of emergency physi- cians who are expert in the relevant ultrasound procedures typically perform the examinations in the studies. This limits generalizability of test characteristics to emergency physicians without advanced training in ultrasonography. The second limitation is that although EPPU is rec- ommended for indications other than confirming IUP and ruling out ec- topic pregnancy, there is a paucity of studies that support these other indications for EPPU. For example, Adhikari et al [13] identified 18 cases of tubo ovarian abscess (TOA) diagnosed by EPPU in a 3-year ret- rospective study but did not report any test characteristics. In fact, a re- cent literature review assessing ultrasonography test characteristics for TOA, as performed by gynecologists and radiologists, shows much lower sensitivities and specificities than are widely cited [14].

Another limitation to these data is that they were mostly published over 10 years ago. Shih [7] published the only prospective study to evalu- ate EPPU in an adult population and impact on ED LOS in 1997, and there have been no published randomized control trials evaluating EPPU and ED LOS. Improvements in technology to the hardware and software of the machines theoretically should improve performance, although no studies exist reporting such differences. In addition, the training of emer- gency physicians in ultrasonography has expanded tremendously over the past decade. This is another posited reason for improvements in EPPU skill that requires further study. Recently, a prospective study of emergency medicine (EM) residents showed that resident performance and interpretation of Transvaginal ultrasound to diagnose IUP correlated with the interpretation of an expert ED sonologist [15].

Goals of this investigation

Prior studies support the role of EPPU in confirming IUP and its ability to reduce ED LOS. However, this is the first prospective study to examine the use of EPPU for all adult female patients with ED presentations that re- quire pelvic ultrasonography. Our hypothesis is that an EPPU will shorten ED LOS by at least 60 minutes as compared with radiology department- performed pelvic ultrasonography (RPPU). In addition, we investigated factors associated with EPPU vs RPPU selection and ED LOS as well as the rate of return to the ED after initial EPPU and discharge and whether an alternative diagnosis was discovered on subsequent visit.

Methods

Study design and setting

This was a prospective, observational study of ED patients 13 years or older that had a chief complaint and presentation that required workup with pelvic ultrasonography. The study was conducted from January to August 2011 at an urban, academic ED with 40 000 annual visits. All at- tending physicians are American Board of Emergency Medicine- certified or eligible, and approximately one-half have performed more than 800 Abdominal ultrasounds. There is a well-established ultrasound fellowship, and all of the 18 EM residents receive extensive didactic and hands-on instruction in bedside ultrasound. In addition, the ED bills for most bedside Ultrasound studies including focused abdominal sonogra- phy in trauma, biliary, deep venous thrombosis, and pelvic ultrasonogra- phy. The medical center’s institutional review board approved this study.

Selection of participants

Undergraduate research associates screened patients for eligibility from 8 AM to midnight 7 days a week, and EM physicians were responsible for obtaining informed consent. Subjects were screened for prospective enrollment if they were females older than 13 years and had a chief

complaint of lower abdominal or pelvic pain and/or vaginal bleeding or discharge, and the treating emergency physician determined that they re- quired a pelvic ultrasound as part of their ED evaluation. Patients of repro- ductive age received standard urine pregnancy testing and, if necessary, serum quantitative beta human chorionic gonadotropin and Rh factor testing. Patients who spoke languages other than English or Spanish were excluded due to lack of language translation services for consent purposes.

Intervention

The decision to perform an EPPU vs to order an RPPU was left at the discretion of the treating ED attending physician. Neither the attending nor resident physicians were blinded to the study hypothesis or outcomes. Emergency physician-performed pelvic ultrasonography was performed by the treating EM resident physician with supervision from the attending physician on shift. Sonosite M-Turbo (Sonosite, Bothell, WA) ultrasound machines were used with obstetrics examination setting and a curvilinear transducer for transabdominal and intracavitary trans- ducer for transvaginal pelvic ultrasonography. All patients received a Transabdominal ultrasound examination, and select patients also received transvaginal ultrasounds when evaluating for adnexal pathology or if deemed necessary by the attending physician. Examinations were per- formed according to the American College of Emergency Physicians Ultra- sound Imaging Criteria Compendium [12]. All attending physicians are credentialed to perform bedside ultrasonography, and EM resident physi- cians had received at least 4 hours of didactics and 8 hours of hands-on instruction in pelvic ultrasonography as part of their standard residency curriculum. In addition, EM residents take a required month-long rotation in emergency ultrasonography consisting of scan shifts and weekly qual- ity assurance image review. In addition to the core application of identifi- cation of IUP, didactic content includes identification of uterine and adnexal pathology via both phantom models and simulators.

Methods of measurement

After informed consent was obtained, each patient received either an EPPU or RPPU at the discretion of the attending physician. A research associate accessed the electronic medical record (EMR) system to ex- tract data needed for the study onto a data collection sheet. This includ- ed demographic data, chief complaint, Triage time, location of pelvic ultrasound study (EPPU, RPPU, or both), type of ultrasound performed (transabdominal/endovaginal), disposition time, and discharge diagno- sis. This was then imputed to an Excel spreadsheet that was placed on a secured server. Emergency department LOS was calculated as the dura- tion from the initial triage time to the disposition time. Emergency phy- sician-performed pelvic ultrasonography interpretations were extracted from the EMR provider note for the visit, under a specific ul- trasound section. All EPPU images were archived and reviewed for image adequacy and accuracy by the ED ultrasound director, who is the senior investigator on this study. Radiology department-performed pelvic ultrasonography interpretations were extracted from the attending radiologist interpretation in the EMR system.

At least 2 weeks after initial ED visit, research associates accessed the EMR to collect data that were not available during the ED visit, such as whether patients received a gynecologist (ObGyn) consult in the ED or pending lab results. Consult and disposition data were apparent at this later time via ObGyn consult notes or discharge notes in the EMR that corresponded to the date of visit in the ED. For patients who were discharged from the ED, researchers also recorded any Subsequent ED visits and repeat imaging studies within 2 weeks as well as whether the studies led to an alternative diagnosis than that of the original ED visit.

Data collection and processing

There were 2 components to data collection: (1) the point-of-care collection of ultrasound interpretation as well as the clinical and

demographic data and (2) the chart review to assess the subjects’ clini- cal course during and after the index ED visit. This chart review followed strict criteria involving training of 2 research associates as well as data verification by a study investigator in 10% of these chart reviews [16]. Research associates collected data via extraction from the EMR system onto a paper questionnaire. This was then imputed into a Microsoft Excel 2011 (Microsoft, Redmond, WA) database that was placed on a se- cured server. The database was then imported into SPSS version 19 (IBM, Armonk, NY) for data analysis.

Primary data analysis

Before study enrollment, a power analysis was performed to deter- mine the total number of patients needed to evaluate the study ques- tion. Based on prior studies, we estimated the ED LOS of patients who required pelvic ultrasonography to be 300 minutes, with an SD of 100 minutes. To detect a difference of at least 60 minutes, using an ? of .05 and a ? of .2, and a conservative ratio of RPPU to EPPU of 2, we calculated a sample size of 132 patients.

To perform parametric testing, ED LOS data, reported in minutes, were log transformed to fit a normal distribution for analysis and trans- formed back for reporting. An independent-samples t test was used to test for difference in mean LOS between EPPU and RPPU. We used linear Regression modeling to control for the effects of ObGyn consultation in the ED, pregnancy status, disposition, and location of ultrasound (ED vs radiology) on ED LOS.

In addition, to assess for interactive effects, we performed odds ratio analyses between EPPU vs RPPU and ObGyn consultation, disposition, and pregnancy status. Factors showing significant association with type of ultrasound performed underwent subgroup analysis to assess for true differences between LOS and type of ultrasound in these sub- groups. As we only performed 3 tests for interaction in our subgroup analysis, we did not pursue additional statistical measures to decrease the chance of a type I error [17].

Results

Characteristics of study subjects

We screened 203 patients and enrolled 157 (Fig. 1). Of those screened but not enrolled, we were unable to obtain consent from 18 patients. Fourteen patients had inadequate study data (eg, improper medical re- cord number listed), 3 patients did not speak English or Spanish, and 1 pa- tient did not have a pregnancy test result reported. Ten of the patients

who had EPPU (12%) had images that were deemed inadequate by the ED ultrasound director and were also excluded. Of the 157 patients in- cluded in the study, 48 (30%) received EPPU, 84 (54%) received RPPU, and 25 (16%) received both. Patients who received both EPPU and RPPU were excluded from the analyses, but we did examine the correlation between interpretations to elucidate the possible reasons for obtaining both studies in sensitivity analysis. All 18 resident physicians performed at least 1 EPPU for the study, and 15 attending physicians supervised. For the EPPU studies, there was no discrepancy in interpretation between the treating physicians and the ED ultrasound director.

Main results

Patients who received an EPPU had an ED LOS of 265 minutes (range, 80-1021 minutes), and patients who received an RPPU had an ED LOS of 427 minutes (range, 145-1411 minutes). Overall, there was a statistical- ly significant reduction in the LOS in the EPPU group compared with the RPPU group of 162 minutes, (95% confidence interval [CI], 106-209 mi- nutes). In Linear regression analysis, when controlling for pregnancy sta- tus, ObGyn consult in the ED, and admission rate, there was still a statistically significant reduction in LOS for the ED ultrasound group of 108 minutes (95% CI, 38-166 minutes).

Tests of interaction and subgroup analysis

Overall, 47% of patients were pregnant, with 77% of patients who re- ceived EPPU and 24% of patients who received RPPU being pregnant. In tests of association, pregnant patients were 10.8 times more likely to re- ceive an EPPU (odds ratio [OR], 10.8; 95% CI, 4.7-25) (Fig. 2). The ObGyn consult rate in the ED was 25%, with 32% of RPPU patients receiving ObGyn consult, compared with 9% of EPPU patients. Patients who re- ceived an ObGyn consult in the ED were 5.2 times more likely to receive an RPPU (OR, 5.2; 95% CI, 1.7-16). Overall admission rate was 15%, with 9% of EPPU patients and 20% of RPPU patients being admitted. Patients who were admitted were 2.8 times more likely to receive RPPU (OR, 2.8; 95% CI, 0.9-8.8).

In subgroup analysis, pregnant patients who received EPPU had a statistically significant reduction in LOS of 140 minutes (95% CI, 59- 101 minutes) (Fig. 3). In nonpregnant patients, those who received EPPU had a statistically significant reduction in LOS of 112 minutes (95% CI, 7-193 minutes). However, there were only 11 nonpregnant pa- tients who received EPPU vs 64 who received RPPU. In patients who re- ceived an ObGyn consult, those who had EPPU had a nonstatistically significant reduction in LOS of 105 minutes. However, there were only

Fig. 1. Study screening and enrollment flowchart.

90

80 77

70

60

Pregnancy and EPPU

OR 10.8 (95% CI, 4.7, 25)

Pregnant (%)

Table 1

Discharge diagnosis based on type of pelvic ultrasound obtained

Diagnosis EPPU, n (%) RPPU, n (%) Both, n (%) Total, n (%)

IUP 24 (50) 9 (11) 7 (28) 40 (25)

Ectopic pregnancy - - 1 (4) 1 (1)

50

40

30

20

10 9

0

ObGyn and RPPU OR 5.2 (95% CI, 1.7, 16)

32

24

20

9

47

OBGYN (%)

25 Admission (%)

15

Admission and RPPU OR 2.8 (95% CI, 0.9, 8.8)

Molar pregnancy - 1 (1) - 1 (1)

Cornual pregnancy - - 1 (4) 1 (1)

early pregnancy - 1 (1) - 1 (1)

Leiomyoma/fibroid 1 (2) 10 (12) - 11 (7)

Ovarian cyst/mass - 12 (14) 3 (12) 15 (10)

Ovarian torsion - 2 (2) - 2 (1)

PID/TOA 1 (2) 4 (5) - 5 (3)

Abdominal/pelvic pain^a 7 (15) 21 (25) 4 (16) 32 (20)

Vaginal bleeding^b 3 (6) 13 (15) - 16 (10)

n = 48 n = 84 n = 157

EPPU RPPU Overall

Fig. 2. Proportion of pregnancy, ObGyn consult in the ED, and admission in EPPU group, RPPU group, and overall. Odds ratios are listed for pregnancy and EPPU, ObGyn consult in the ED and RPPU, and admission and RPPU.

4 patients who received EPPU and ObGyn consult. In patients who did not receive an ObGyn consult, those who had EPPU had a statistically significant reduction in LOS of 139 minutes (95% CI, 80-186 minutes). Subgroup analyses were not performed in patients who were admitted or discharged, as there was no statistically significant association be- tween admission status and type of ultrasound performed.

Diagnoses

Of the EPPU group, 24 patients were diagnosed with IUP (50%); 10 patients, with threatened abortion (21%); and 7 patients, with abdomi- nal or pelvic pain (15%). Of the RPPU group, 21 patients were diagnosed with abdominal or pelvic pain (25%); 13 patients, with vaginal bleeding (15%); 12 patients, with Ovarian mass or cyst (14%); 10 patients, with fi- broids (12%); and 9 patients, with IUP (11%). There were 2 cases of ovar- ian torsion by RPPU and 2 cases of ectopic (ectopic and cornual/ interstitial) in the group who received both EPPU and RPPU. These and less common diagnoses are listed in Table 1. In addition, 2 weeks after the initial ED visit, there was no change in or alternative diagnoses in the 44 patients who had EPPU and were discharged.

Sensitivity analysis

Of the 25 patients who received both EPPU and RPPU, 17 (68%) were pregnant, 7 (28%) received an ObGyn consult in the ED, and 3 (12%)

250

139

112

105

140

Mean Reduction in LOS (min) with 95% CI bars

200

150

100

50

0

Pregnant patients Non-pregnant OBGYN consult No OBGYN consult (n = 57) patients (n = 75) (n = 31) (n = 101)

Fig. 3. Subgroup analyses showing mean reduction in ED LOS in pregnant and nonpreg- nant patients as well as those who received and did not receive ObGyn consult in the ED. Subgroup analysis was not performed based on admission vs discharged patients, as there was statistically significant level of interaction detected.

Threatened abortion 10 (21) 5 (6) 3 (12) 18 (11)

Fetal demise^c 1 (2) 6 (7) 5 (20) 12 (8)

Other^d 1 (2) - 1 (4) 2 (1)

Total 48 84 25 157

^a Diagnoses that were related to abdominal or renal etiology or nonspecific.

^b Vaginal bleeding as discharge diagnosis without definitive pathology seen on pelvic ultrasound.

^c Includes spontaneous, inevitable, and missed abortion.

^d One diagnosis was syncope and EPPU was performed to rule out ectopic pregnancy, whereas the other diagnosis was amenorrhea.

were admitted. Of the EPPU interpretations, 21 (84%) agreed with the RPPU interpretation. These included 5 cases of IUP where the presence of subchorionic hemorrhage was added to the RPPU interpretation but not the EPPU interpretation. Of the 5 EPPU interpretations with abnor- mal IUP, all had RPPU interpretations that indicated a more specific rea- son for abnormal IUP (eg, “fetal anencephaly”). Of the discordant cases, 1 EPPU interpretation was of IUP, but the RPPU was interstitial pregnan- cy; 2 had no EPPU pathology listed but showed ovarian cysts; 1 EPPU case documented adnexal mass but had a normal RPPU; and 1 EPPU case documented nondiagnostic IUP (NDIUP), whereas the RPPU inter- pretation was of IUP (Table 2).

Discussion

This is the first prospective study of a general adult female population that demonstrates a clinically and statistically significant reduction in LOS for patients who receive an EPPU as opposed to RPPU. We also found a strong association between type of ultrasound study and pregnancy sta- tus as well as ObGyn consult performed in the ED. However, when con- trolling for these factors in regression analysis as well as in examination of subgroups, the ED LOS reduction associated with EPPU still persisted.

There are key differences between our study and the only other pro- spective study that evaluates EPPU and ED LOS, by Shih [7] in 1997. Shih

[7] enrolled only pregnant patients and assessed the impact of ObGyn consultation on ED LOS, whereas our study enrolled all patients who re- quired pelvic ultrasonography and assessed ObGyn consultation and disposition status on ED LOS. Our study also used a more generalizable group of emergency physicians (18 resident physicians and 15 attend- ing physicians) with a wide range of ultrasonography experience. In contrast, in the previous study, although 14 emergency physicians prac- ticed at the institution, there is no explicit account of the number of emergency physicians that participated in the study. Another key differ- ence is that 13 of 14 physicians in the prior study did not undergo EM residency training, whereas all physicians in our study were or are cur- rently being trained in EM and, therefore, benefit from formal ultraso- nography training.

The ED LOS associated with EPPU was consistent with prior retrospec- tive studies that demonstrated a similar reduction of 60 to 140 minutes [5-9]. Interaction, however, was suggested between selection of EPPU vs RPPU and pregnancy status as well as need for ObGyn consult in the ED. This was expected, as many emergency physicians consider EPPU usually in the context of identifying IUP in pregnant patients. In addition to this indication, the EPPUs in our study were able to diagnose patients with

Table 2

Sensitivity analysis of subjects who received both EPPU and RPPU

Study no.	Chief complaint	EPPU interpretation	RPPU interpretation	Agreement	Pregnant	ObGyn consult	Admitted	Discharge diagnosis
7	Vaginal bleeding	IUP	IUP with subchorionic hemorrhage	Yes	Yes	Yes	No	Threatened abortion
14	Vaginal bleeding	IUP	IUP with subchorionic	Yes	Yes	No	No	Threatened abortion
17	Pelvic pain	No pathology	hemorrhage Ovarian cyst	No	No	No	No	Ovarian cyst
25	Abdominal pain	IUP	Interstitial pregnancy	No	Yes	No	No	Interstitial pregnancy, UTI
28	Abdominal pain	No pathology	No pathology	Yes	Yes	No	No	Early pregnancy vs
33	Pregnant with nausea	Abnormal IUP	IUP with anencephaly	Yes	Yes	Yes	No	spontaneous abortion Fetal anencephaly
34	Abdominal pain	Adnexal mass	No pathology	No	No	No	No	Abdominal pain
58	Pelvic pain	IUP	IUP with subchorionic	Yes	Yes	No	No	Threatened abortion
60	6-wk EGA with vomiting	NDIUP	hemorrhage NDIUP	Yes	Yes	No	No	Hyperemesis gravidum
62	Pelvic pain	Abnormal IUP	IUP with decreased fetal	Yes	Yes	No	No	IUP, abdominal pain
64	Abdominal pain	No IUP	heartbeat No IUP	Yes	Yes	Yes	Yes	Ectopic pregnancy
65	Pelvic pain	NDIUP	IUP	No	Yes	No	No	Hypokalemia, diarrhea, IUP
67	Pelvic pain,	No IUP	Adnexal cyst, no torsion	No	No	No	No	Amenorrhea
69	LMP 3 mo ago 6-wk EGA with	IUP	IUP with subchorionic	Yes	Yes	No	No	IUP, abdominal pain
71	vaginal bleeding Vaginal bleeding	IUP	hemorrhage IUP with subchorionic	Yes	Yes	No	No	Threatened abortion
75	Sent from gynecologist	Abnormal IUP	hemorrhage Interstitial pregnancy	Yes	Yes	Yes	No	Interstitial pregnancy,
	clinic for spontaneous							spontaneous abortion
79	abortion in interstitial pregnancy 9-wk EGA with	IUP	IUP	Yes	Yes	No	No	IUP, abdominal pain,
	pelvic pain							dehydration, ketosis, resolved
81	Pelvic pain	No pathology	No pathology	Yes	No	No	No	Dysuria and frequency
109	Fever and vaginal discharge	Free fluid around ovary	Free fluid around ovary	Yes	No	Yes	Yes	Sepsis, concern for TOA
112	12-wk EGA with vaginal bleeding	Abnormal IUP	Abnormal IUP	Yes	Yes	No	No	Spontaneous abortion, vaginal bleeding
126	Pelvic pain	Adnexal cyst	Adnexal cyst, no torsion	Yes	No	No	No	Pelvic pain, ovarian cyst
128	Vaginal bleeding	Abnormal IUP	Fetal demise	Yes	Yes	Yes	No	Fetal demise, missed abortion
138	8-wk EGA with	No IUP	No IUP	Yes	Yes	No	No	Spontaneous abortion
147	vaginal bleeding Pelvic pain	Pelvic free fluid	Complex pelvic	Yes	No	Yes	Yes	TOA
152	Abdominal pain	Ovarian cyst	fluid collection Ovarian cyst	Yes	No	No	No	Ovarian cyst

Abbreviations: UTI, urinary tract infection; EGA, estimated gestational age; LMP, last menstrual period.

fibroids, ovarian cysts, ectopic pregnancy, and TOA. Although there are no studies that explicitly assess preference, the study authors note that gyne- cologists and other consultants typically prefer a “formal” Radiology study. This may be due to need for documentation purposes, inability of the con- sultant to interpret bedside ultrasound images, and/or preference for ra- diologist interpretation.

Although patients who received both EPPU and RPPU were not part of the hypothesis testing, sensitivity analysis was performed on this subgroup resulting in similar pregnancy rate as patients who received an EPPU and yet a similarly high consult rate as those who received an RPPU. Admission rate for this subgroup fell between the EPPU and RPPU group at 12%. It is unclear why a subsequent RPPU was ordered for patients who had IUP on EPPU, in cases where the RPPU interpreta- tion was IUP with subchorionic hemorrhage. Possibly, this surrounding anechoic or hypoechoic area around the IUP was deemed abnormal by the attending ED physician, and/or further risk stratification was sought via measurement. This is a controversial issue in the ObGyn literature, as 1 meta-analysis showed a 2-fold association between subchorionic he- matoma and spontaneous abortion as well as stillbirth [18], but a recent observational study showed no association [19]. For patients who had either an NDIUP or abnormal IUP per EPPU, the RPPU agreed and refined the interpretation. This is important because ED physicians were able to recognize normal from abnormal IUPs and appropriately sought a RPPU to further characterize. One caveat is in the 2 interstitial pregnancy cases, in which the EPPU characterized as IUP and abnormal IUP. Inter- stitial pregnancies are technically both intrauterine and ectopic, as they

are located in the horn or cornu of the uterus where the relatively thin myometrium and surrounding Sampson arteries place patients at high risk for hemorrhage as the gestation progresses [20]. A myometrium thickness or endomyometrial mantle less than 5 mm around the gesta- tion sac is considered diagnostic of interstitial pregnancy [21]. Emergen- cy physician-performed pelvic ultrasonography was also able to identify pathology in 4 of 6 nonpregnant patients who had adnexal pa- thology on RPPU. Reasons for a subsequent RPPU may include the need to better characterize the abnormality as well as possibly per ObGyn consult request in the 2 cases for admission for TOA.

We attempted to control for selection bias by regression modeling as well as with subgroup analysis. In our regression model, there was still a clinically and statistically significant ED LOS reduction. In the same sub- groups that were associated with a particular ultrasound exam (EPPU vs RPPU), there was still an ED LOS reduction for the EPPU group. With the exception of patients who received an ObGyn consult in the ED, all ED LOS reductions were statistically significant.

In addition, there were no missed alternative diagnoses in patients receiving EPPU as assessed by 2-week follow-up through review of re- turn visits to our ED. Emergency physician-performed pelvic ultraso- nography interpretation did not differ between treating physicians and ED ultrasound director, suggesting that EPPU can be accurate in the hands of many emergency physicians. However, this study was not designed to explicitly assess diagnostic accuracy of EPPU performed by a large group of physicians with a wide range of ultrasonography ex- perience and skill. Rather, our data suggest that EPPU may be

considered for other indications aside from confirming IUP and that EPPU is associated with a meaningful reduction in ED LOS that main- tained clinical and statistical significance even in subgroup analysis.

Limitations

The main limitation of this study is the observational design in which provider and consultant decision-making factors were not adequately addressed in the analysis. This leads to selection bias, as implicit factors may influence the type of ultrasound obtained, which may affect the LOS. For example, although pregnant patients were 10 times more likely to receive EPPU, there are likely more factors than pregnancy status that influenced provider selection of EPPU for these patients. This is also like- ly true for the association between RPPU and those who received ObGyn consults in the ED. In addition, our physicians were not blinded to the study hypothesis and, therefore, may have expedited patient manage- ment in those receiving EPPU via a Hawthorne effect. This study also did not incorporate other clinical data, such as presence of systemic in- flammatory response syndrome criteria or presence of prior significant gynecologic or obstetric history.

However, EPPU was able to identify pathology outside the usual in- dication for IUP, including ovarian cysts and pelvic free fluid as well as abnormal IUPs. This argues against the notion that ED LOS was associat- ed with pregnancy status or relatively low complexity presentations.

Our convenience sampling of hours outside of midnight to 8 AM may also have led to selection bias. However, this would likely have the effect of minimizing the differences in ED LOS, as RPPU would likely take lon- ger during these hours. In patients who were excluded, there was no other indication that the reasons for exclusion may be associated with other selection biases.

We also did not assess physician attitudes or cognitive factors asso- ciated with the decision to pursue EPPU vs RPPU. A future study may in- corporate a questionnaire that can be administered to physicians in real time that addresses these issues.

Although we used a large group of resident and attending physicians to perform EPPU, this was a single-center study, and our results may not be generalizable to other institutions. This is especially true of insti- tutions where ultrasound education and quality assurance may not be as robust.

In terms of our definition of ED LOS (triage time to disposition time), we incorporated both the waiting room time and the treatment time, but not the boarding time [22]. At our institution, triage time to ED rooming time is mandated to be less than 30 minutes, so any differences in triage time between groups receiving EPPU and RPPU should have minimal influence on overall LOS. In addition, patients with more acute diagnoses requiring ObGyn consult and RPPU are more likely to be given a higher emergency severity index score and roomed earlier. Therefore, the ED LOS differences would be underestimated with this as- sociated bias. We also did not incorporate actual ED boarding time, that is, duration between disposition and actual exit from the ED, as there are many factors associated that are not related to the study question.

In our study, we determined that there were no alternative diagno- ses at 2-week follow-up. Because of our method of determining this through EMR review, we cannot say that this is actually reflective of missed diagnoses as patients may have gone to other facilities or ObGyn offices for follow-up care.

Because of limitations associated with our observational study de- sign, we attempted to use both subgroup analysis and linear regression in an effort to reduce biases associated with our findings. However, a study incorporating a randomized control design would best help to mitigate these biases.

Conclusion

Our prospective study demonstrates a significant reduction in ED LOS for patients who received an EPPU compared with patients who re- ceived an RPPU, even when controlling for factors such as pregnancy status, ObGyn consultation in the ED, and final disposition. This is the first study to assess EPPU in the context of expanded indications for use and not just for first trimester bleeding. A randomized clinical trial is needed to confirm our results and to control for potential selection bias associated with selection of RPPU vs EPPU and ED LOS.

Acknowledgments

The authors would like to thank Mr Marcus Williams for his assistance in collecting data for this study. The authors also like to thank Dr Shahram Lotfipour and the Emergency Medicine Research Associates Program for their help in enrolling patients during the study.

References

1. Bernstein SL, Aronsky D, Duseji R, et al. The effect of emergency department crowding on clinically oriented outcomes. Acad Emerg Med 2009;16:1-10.
2. McCarthy ML, Ding R, Pines JM, et al. Comparison of methods for measuring crowding and its effects on length of stay in the emergency department. Acad Emerg Med 2011;18(12):1269-77.
3. Kocher KE, Meurer WJ, Desmond JS, et al. Effect of testing and treatment on emer- gency department length of stay using a national database. Acad Emerg Med 2012;19(5):525-34.
4. ACEP. emergency visits up since implementation of ACA. Accessed on 07/14/14 from http://newsroom.acep.org/ACEP-Emergency-Visits-Up-Since-Implementation-of-ACA; 2014.
5. Burgher SW, Tandy TK, Dawdy MR. Transvaginal ultrasonography by emergency physicians decreases patient time in the emergency department. Acad Emerg Med 1998;5(8):802-7.
6. Blaivas M, Sierzenski P, Piecque D, Lambert M. Do emergency physicians save time when locating a live intrauterine pregnancy with bedside ultrasonography? Acad Emerg Med 2000;7(9):988-93.
7. Shih CHY. Effect of emergency physician-performed pelvic sonography on length of stay in the emergency department. Ann Emerg Med 1997;29:348-52.
8. Mateer JR, Aiman EJ, Brown MH, et al. Ultrasonographic examination by emergency physicians of patients at risk for ectopic pregnancy. Acad Emerg Med 1995;2:867-73.
9. McRae A, Edmonds M, Murray H. Diagnostic accuracy and clinical utility of emergen- cy department targeted ultrasonography in the evaluation of first-trimester pelvic pain and bleeding: a systematic review. Can J Emerg Med 2009;11(4):355-64.
10. Mateer JR, Valley VT, Aiman EJ, et al. Outcome analysis of a protocol including bed- side endovaginal sonography in patients at risk for ectopic pregnancy. Ann Emerg Med 1996;27:283-9.
11. American College of Emergency Physicians. ACEP emergency ultrasound guidelines

    -2008. Accessed on 7/12/14 from http://www.acep.org/Clinical-Practice-Manage- ment/Ultrasound; 2008.

    American College of Emergency Physicians. Emergency Ultrasound imaging criteria compendium. Accessed on 7/12/14 from http://www.acep.org/Clinical-Practice- Management/Ultrasound; 2006.

12. Adhikari S, Blaivas M, Lyon M. Role of transvaginal ultrasonography in the diagnosis of tubo-ovarian abscess in the emergency department. J Emerg Med 2008;34(4):429-33.
13. Lee DC, Swaminathan AK. Sensitivity of ultrasound for the diagnosis of tubo-ovarian abscess: a case report and literature review. J Emerg Med 2011;40(2):170-5.
14. Macvane CZ, Irish CB, Strout TD, Owens WB. Implementation of transvaginal ultra- sound in an emergency department residency program: an analysis of resident in- terpretation. J Emerg Med 2012;43(1):124-8.
15. Kaji AH, Schriger D, Green S. Looking through the retrospectoscope: reducing bias in emergency medicine chart review studies. Ann Emerg Med 2014;64(3):292-8.
16. Lagakos SW. The challenge of subgroup analyses-reporting without distorting. N Engl J Med 2006;354(16):1667-9.
17. Tuuli MG, Norman SM, Odibo AO, et al. Perinatal outcomes in women with subchorionic hematoma: a systematic review and meta-analysis. Obstet Gynecol 2011;117(5):1205-12.
18. Stamatopoulos N, Lu C, Infante F, et al. Does the presence of subchorionic haematoma increase the risk of miscarriage? Ultrasound Obstet Gynecol 2013;42:54.
19. Rastogi R, Gi M, Rastogi N, Rastogi V. Interstitial ectopic pregnancy: a rare and diffi- cult clinicosonographic diagnosis. J Hum Reprod Sci 2008;1(2):81-2.
20. Doane B, Perera P. Emergency ultrasound identification of a cornual ectopic preg- nancy. West J Emerg Med 2012;13(4):315.
21. Ding R, McCarthy ML, Desmond JS, et al. Characterizing waiting room time, treat- ment time, and boarding time in the emergency department using quantile regres- sion. Acad Emerg Med 2010;17(8):813-23.