American Journal of Emergency Medicine Oct 2018 Epub - Pubmed Link
1. POCUS can decrease the number of etiologies on the differential diagnosis in emergency department patients with chest pain or shortness of breath.
2. Ultrasound and chest xray had comparable test characteristics for the pathologies seen with the exception of pneumonia where ultrasound is more sensitive, chest xray more specific.
Welcome to a good ol’ fashioned shootout! Ultrasound versus chest xray in patients who are coming to the emergency department for chest pain or shortness of breath. Let’s back up a bit first - at this point, we are pretty sure of a few things. A) Lots of people come to the emergency department with these complaints and their diagnoses range from very benign to “already dead”. B) Almost everyone gets a chest xray. C) These chest xrays are not always useful D) Ultrasound has been shown to be accurate for a number of the most important diagnostic considerations. Sooooo, if we connect the dots here we can see that the investigators wanted to see how point of care ultrasound would fare in this undifferentiated emergency department population. Can it help close the differential diagnosis? Could it be more accurate than its arch frenemy the chest xray? We shall see.
How does POCUS influence the differential diagnosis in patients in the emergency department that present with chest pain or shortness of breath?
How accurate is POCUS compared to chest xray for the final diagnosis in this population?
Single academic center in US
Inclusion:
Emergency department patient with chief complaint of chest pain or shortness of breath AND had a chest xray ordered
Over 18 years old
Exclusion
Referred from outside hospital with known diagnosis
EP was aware of results of any other diagnostic imaging prior to data collection
Pregnancy
Patient “unavailable” to complete exam
Unable to consent including clinical instability
Prospective observational study
1 Emergency physician caring for patient selected possible diagnoses from a predetermined list
Asthma/COPD
Acute coronary syndrome
Pulmonary embolism
Pericarditis
Aortic Dissection
Malignancy
Musculoskeletal
Upper respiratory illness
Other - GI, anxiety, nonspecific chest pain
Pneumothorax
Pleural effusion
Pneumonia
Pulmonary edema
Pericardial effusion
*Last 5 considered assessable with US
2 Focused heart and lung ultrasound performed by physician who was blinded to additional imaging, but unblinded to other clinical information
3 The POCUS results were disclosed to treating physician. Treating physician then re-surveyed regarding differential diagnosis.
4 Final composite diagnosis was obtained by retrospective chart review
Power calculation: needed 128 patients to achieve 80% power to detected 0.5 difference in mean differentials.
Ultrasound fellows or fellowship trained physicians (not caring for the patient)
Phased array probe (for cardiac component) and Curvilinear for Pulmonary
Cardiac Protocol
Needed at least 2 of the four basic views
Assessed for pericardial effusion, decreased LV systolic function, RV strain
Thoracic Protocol
Six zones - anterior, lateral, posterior bilaterally
Assessed for lung sliding, pleural effusion, consolidation, B-lines
Pneumonia = lobar consolidation or multiple patchy consolidation with air bronchograms
Isolated interstitial consolidation (without other artifacts) did not count as pneumonia
Pulmonary Edema = ≥3 B-lines in both lung fields
Pneumothorax = lack of pleural sliding
Pleural Effusion = Spine sign
5 Min Sono -Pericardial Effusion
5 Min Sono -Right Heart Strain
5 Min Sono - Pleural Effusions
The Evidence Atlas - Pulmonary
N = 128 patients
Mean age 64
55% male
34% chest pain
37% shortness of breath
29% both
27% (35/129) ultimately diagnosed with conditions that could be assessed with POCUS
68 (of 147 diagnoses) of atypical chest pain and “Other” diagnosis
Primary Outcomes
Change in Differential Diagnosis:
Before POCUS: median of 5 diagnoses (IQR 3-6, average of 2 were assessable by US) → After POCUS: 3 (IQR 2-4)
Relative Sensitivities
Pneumonia (note: 30 POCUS with indeterminant pneumonia studies)
Sensitivity Chest xray 38% (0.13-0.70) vs POCUS 89% (0.54-1.00)
Specificity Chest xray 96% (0.90-0.99) vs 74% (0.64-0.82)
All other sensitivities were similar and had overlapping confidence intervals.
Pneumothorax
Sensitivity Chest xray 100% vs POCUS 100%
Specificity Chest xray 100% vs POCUS 100%
Pleural Effusion
Sensitivity Chest xray 100% vs POCUS 100%
Specificity Chest xray 80% vs POCUS 71%
Pulmonary Edema
Sensitivity Chest xray 81% vs POCUS 78%
Specificity Chest xray 92% vs POCUS 93%
**No pericardial effusions or atelectasis cases
Other Findings
36% had abnormal cardiac ultrasound
57% had abnormal lung ultrasound
Enlarged aortic root was most commonly indeterminant (in 23 cases)
Agreement Between CXR and Ultrasound - Kappa Values for each diagnosis
Pneumonia 1
Pneumothorax 0.4
Ultrasound seemed to decrease differential of non-sonographic diagnoses as well. Authors hypothesize it could have been positive findings on POCUS made other diagnoses less likely.
Negative POCUS, Positive CXR (13% of cases)
5 with atelectasis (2 discharged home with COPD, 1 discharged next day with post-viral cough, 2 required work up for chest pain)
1 with spiculated lesion
1 with mild pulmonary edema (did not require treatment)
Looking at diagnoses that changed significantly between pre and post-POCUS
Pneumonia 77% → 49%
Pulmonary edema 71% → 29%
Pericardial effusion 51% → 4%
Pleural effusion 50% → 22%
Pneumothorax 43% → 1%
Asthma/COPD 39% → 28%
POCUS took an average of 10 minutes to be completed
The usual things: Highly trained physicians doing the ultrasounds. Also - not having the treating physician doing the ultrasound is contrary to the intended use of POCUS (integrating the information into the clinical background). Limited sample size. Not all pathologies represented. Might hesitate to do a POCUS first strategy based on a population where not all of these pathologies were seen.
Is measuring the overall number of pathologies on the differential an accurate measure of impact of a diagnostic test? Maybe after doing ultrasound, you can rule out somethings, but add some things at the same time. Furthermore, although closing the differential is helpful to providers, we don’t know if these were appropriate changes in the differential (because the final diagnosis may not accurately reflect this - see below). Also, there were overlapping interquartile ranges for the data for number of differential diagnoses, so that casts doubt on a true difference of significance.
Composite diagnosis - this is a problem with regard to the reported accuracies. The gold standard was determined from a final diagnosis and treating providers may have only selected the single leading diagnosis, not necessarily all of the diagnoses. This also could be biased based on either the chest xray AND/OR the POCUS. If it favored the chest xray, would really bias the accuracy of the ultrasound. There is evidence that this influences the results: they found pericardial effusions but none listed in the final diagnosis, pleural effusion is listed as final composite diagnosis in 1 patient (despite pleural effusions being found in at least 31 of the patients). We don’t know how many patients were admitted or discharged and we don’t know who determined this final diagnosis. It would likely matter if a patient was observed over several days, had additional testing/data points collected and the composite diagnosis from the time after the ED visit was used as opposed to a composite diagnosis that draws only on data from the ED visit/EP. Using an inaccurate final composite diagnosis (or diagnoses) leads to inaccurate sensitivities and specificities for POCUS and CXR.
Other Little Things
No comparison of how chest xray affected the differential. What is a good number to get here?
Significant number of indeterminant studies (both CXR and POCUS).
68% of the patients did not end up with a diagnosis (considered “atypical chest pain” or “other diagnosis”).
Some portions of the methodology are vague. For example, were the CXRs portable (single view) or 2 view?
Did not include PE as diagnosis assessable by POCUS.
Methods states that kappa scores would be calculated for interrater reliability between POCUS user and an expert. Ultimately, kappa only reported for agreement between POCUS/CXR and composite diagnosis
CXR vs POCUS - my time vs someone elses. This is a big point that is seldom discussed. “Ultrasound people” often advocate strongly for POCUS first strategy for many complaints. The argument is that this is faster, safer, and just as accurate as xray in a lot of scenarios. What is rarely adjusted for in this equation is the provider time is takes to perform the POCUS versus the alternative. Certainly, there are ways to continue to be efficient with ultrasound, but for single provider in a busy center, it is hard to argue with the fact that it is a lot less effort to order a test that someone else does. A solution is getting more comfortable with ultrasound and working it into protocols that can be done quickly and add something over traditional work ups.
Authors state, “In no patients did the chest xray provide new or actionable information.” They suggest that it would be reasonable to have an ultrasound first approach with chest xray only if positive finding. You could just as easily make the claim to only do a CXR if a negative ultrasound. This is really only considering the positive findings of chest xray. It may have been very valuable to have negative chest xrays in narrowing the differential. The authors also state, “Given that ultrasound can be done faster than CXR at a lower cost and with no radiation, an ultrasound-first approach for evaluation of patients with chest pain or shortness of breath with a follow-up CXR only in patients with any positive findings on lung ultrasound may prove to be a reasonable approach.” I think this statement is a reach. Let’s be honest - there are times that a chest xray can pick up things that an ultrasound would not. Aortic pathology comes to mind - ultrasound is not that sensitive and chest xray can occasionally help. If the authors meant to speak only to the abilities of POCUS in detecting the POCUS assessable pathologies, then this should have been clearly stated. Their study design asks clinicians to pick from 14 possible diagnoses which include a number of pathologies in which CXR may provide helpful information and POCUS may not. The authors soften it a bit and make it more reasonable in the article with the final statement, “Considering the benefits of POCUS in reducing cost of care, the speed of care delivery, and as a radiation free test, these data may suggest that the incorporation of ultrasound into the initial evaluation of ED patients with chest pain or shortness of breath may reduce the need for CXR in some cases and provide complementary information in others.” I think this is a fair statement.
Very similar article published around the same time: Point-of-care Ultrasonography for Detecting the Etiology of Unexplained Acute Respiratory and Chest Complaints in the Emergency Department: A Prospective Analysis. Cureus Aug 2018
59 patients in this study, US performed by medical students, Gold standard was diagnosis at discharge,
Sensitivity of POCUS 79% vs 67% chest xray
Specificity 71% vs 100% chest xray.
P-values were not significant for these comparisons. However, more sensitive, less specific - similar to this study. They did overall test characteristics for all comers, rather than by etiology like in the main study we discussed.
1. POCUS can decrease the number of etiologies on the differential diagnosis in emergency department patients with chest pain or shortness of breath.
2. Ultrasound and chest xray had comparable test characteristics for the pathologies seen with the exception of pneumonia where ultrasound is more sensitive, chest xray more specific.
Expert Reviewer for this Post