Study design and setting
This is a prospective, observational study enrolling a convenience sample of adult participants presenting with dyspnea to an urban Rwandan emergency department. We enrolled participants presenting to the Emergency Department at University Teaching Hospital of Kigali (UTH-K) over a 1-year period from January 2017 to January 2018. Ethical approval was obtained from the UTH-K Ethical Committee and the Institutional Review Board (IRB) of the College of Medicine and Health Sciences/University of Rwanda and the Partners Healthcare IRB prior to study initiation. Written consent was obtained from participants prior to participation or next of kin if the patient was too sick to consent.
UTH-K, the largest tertiary care hospital in Rwanda, is 500-inpatient bed hospital, which provides speciality surgical and medical services, emergency and critical care, as well as diagnostic services (laboratory and imaging) to approximately 2 million people of Kigali and serves as the country’s main referral hospital. This emergency department receives 25,000 visits per year mostly acutely ill patients, including trauma as well as non-trauma patients. Patients at UTH-K are either directly brought in the ED from home, street, or transferred from peripheral hospitals in Rwanda. All adult patients pass through the ED and are received by either emergency and critical care residents or general practitioners. Imaging modalities including X-ray and computed tomography (CT) scans when available. The emergency department has access to ultrasound equipment immediately at the point-of-care, and ultrasound training is a formal curricular component and required competency of the emergency medicine residency-training program.
Participants
The study included a convenience sample of participants triaged as Yellow, Orange or Red according to the Triage Early Warning Score/SAT-TEWS [14] who complained of breathlessness. Participants were adults, age > 16 years. All others were excluded. Patients were only enrolled when the principal investigator (PI) was present in the department; which was mainly during daytime hours any day of the week.
Data collection
The PI, an emergency medicine physician with focused ultrasound training consistent with international recommendations [15] performed the ultrasound scans (including Heart–Lung-IVC “triple scan”, FASH and DVT studies) after history and physical exam by the treating medical team, but before other diagnostic studies were completed.
Ultrasonography features of pneumonia, pneumothorax, hemothorax, DVT, pulmonary embolism, ADHF, acute COPD or asthma exacerbation, pulmonary edema, pericardial effusion, pleural effusion and extra-pulmonary tuberculosis were recorded.
After initial evaluation by ED staff, general practitioners and residents/registrars, the PI asked the clinical team to generate their top 3 presumed diagnoses, and rank their confidence in the leading diagnosis on a Likert scale (1–5) after which ultrasound was then completed by the PI. The PI did not perform an independent physical exam but some findings were evident on visual inspection during the ultrasound (tachypnea, work of breathing, diaphoresis, abdominal distention, etc.), so the PI was not overall blinded to the clinical findings. Ultrasound findings were communicated to the treating team, after which the treating team reassessed post-ultrasound presumed diagnoses, ranked their confidence in the leading diagnosis. Follow-up was done to obtain diagnosis at discharge for the enrolled participants. The hospital diagnosis at discharge was used as the standard for comparison in determining the accuracy of the pre- and post-ultrasound diagnoses. The lead sonographer received dedicated ultrasound training over a 3-year period, was enrolled in an ultrasound fellowship and had passed an ultrasound observed structured clinical skills assessment prior to leading this study. Another ultrasound-fellowship trained investigator for quality assurance purposes reviewed random samples of 5% studies and there was no discordance with interpretation. A SonoSite M-Turbo® ultrasound machine (FUJIFILM SonoSite, Inc, Bothell, WA) was utilized to scan patients, and ultrasound clips were saved on a flash drive and then stored on an external hard disk. We used REDCap (Research Electronic Data Capture), a web-based electronic capture to enter and store the data [16].
Statistical analysis
The primary objective of the study was to determine the proportion of cases presenting with acute dyspnea in which ultrasound findings change the clinician’s diagnosis for the patient. Based on some prior literature and our prior experience in this setting, we estimated this to be about 40% of cases. We wanted to ensure that the margin of error around our estimate is less than 10%. We used the standard formula for calculating the sample size needed to estimate a proportion in a large population of patients: n = (z
2a/2
* p * (1 − p))/d2). In our case, using the z distribution with an alpha of 0.05, the sample size needed was calculated to be 93 patients. We planned to oversample by about 5% to allow for missing follow up data in patients.
Descriptive statistics, including proportions for categorical data and means for normally distributed continuous data, were obtained using Microsoft Excel 2011 to identify characteristics of the dyspneic patients and the accuracy of physician diagnosis. Categorical variables were compared using Pearson Chi-square or Fisher’s exact test, as appropriate. Finally, student’s t test was used for comparing the means for normally distributed data when appropriate (significance was set at p-value < 0.05).
Ultrasound protocols
Heart and IVC
For each patient enrolled, four cardiac views were obtained (parasternal long, parasternal short, subcostal and 4-chambers views). A visual estimate of the LV function was completed and classified as normal, decreased, or hyperdynamic. Pericardial effusion was assessed, and the quality (simple or complex) and the size of the effusion (qualitatively small or large) were noted. We defined sonographic tamponade whenever pericardial effusion was associated with RA collapse and/or RV collapse during diastole plus a plethoric IVC. The IVC was evaluated by visual estimate and recorded as “flat” and “volume responsive” when collapsing was visualized; it was recorded “plethoric” and “non-volume responsive” when it was visualized to be full without any variation during both cycles of respiration; it was reported to be “normal” when neither flat or plethoric [13].
Lungs
Lung ultrasound was performed following a 4-quadrant view protocol (Fig. 1) [11]. We started with 2 anterior views between vertical parasternal line (PSL) and the anterior axillary line (AAL). We then coupled with 2 views between the AAL and the posterior axillary line (PAL). The same protocol was applied to both the right and the left lung. Lung ultrasound was completed using a phased probe coupled with a high-frequency probe. The following features were recorded on both sides (right and left): lung sliding, pleural effusion small/large and simple/complex, B-lines focal/diffuse, subpleural/hepatization consolidation, and static/dynamic bronchograms.
Fash
Multi-organ ultrasonography included a FASH scan evaluating 6 zones (Fig. 2) which were described to have potential significant prediction for extra-pulmonary TB in HIV patients [17, 18]. FASH was incorporated into the scan protocol considering the local TB burden and clinical presentation of extra-pulmonary TB being very non-specific and difficult to diagnose, in addition to data on para-aortic nodes predicting pulmonary TB [19]. A low frequency curvilinear and a high-frequency linear probe were used for this application. Possible positive findings were: pleural effusion, pericardial effusion, hepatic and splenic microabscesses, ascites, and para-aortic lymphadenopathies.
DVT evaluation
A bedside ultrasound was also conducted to assess for lower extremity deep venous thrombosis. We used the two-point (two zone) DVT compression examination [20]. The region of common femoral vein (CFV) and the greater saphenous vein (GSV) bifurcation and the region of popliteal vein trifurcation were scanned for compressibility; a positive DVT scan was recorded whenever the vein was not compressible with enough pressure to deform the artery.