Setting
This prospective observational study was conducted in the emergency department (ED) at 2 large, urban, tertiary-care centers serving approximately 80,000 and 100,000 patients per year, a majority of whom (> 90%) are African American. The study received approval by the Institutional Review Board (IRB).
Selection of participants
From February, 2015 to July, 2017 a convenience sample of eligible ED patients with a primary diagnosis of AHF treated with intravenous (IV) therapy were recruited during hours, where a study team member was available to perform a POC echocardiogram. Whenever possible, we attempted to enroll patients prior to administration treatment. The inclusion criteria were: 18 years of age or greater, primary admitting diagnosis of AHF (determined by the attending emergency physician) being treated with IV therapy, and the ability to provide informed written consent. Exclusion criteria were: the lack of immediate availability of a study team member capable of performing the echocardiograms, the need for emergent, resuscitative intervention (e.g., cardiopulmonary resuscitation, endotracheal intubation, and cardioversion), rapid atrial fibrillation/flutter or any other tachyarrhythmia requiring rate or rhythm control, heart rate persistently greater than 120 beats/min, poor image quality precluding speckle-tracking analysis, plans for emergent percutaneous coronary intervention (PCI) from the ED, pregnancy, incarceration, and plans for to another institution.
Potentially eligible subjects were identified by trained research assistants after discussion with the treating physician to enquire about the primary diagnosis and treatment plan so that all efforts could be made to enroll the patient, obtain consent and perform the baseline POC echocardiogram prior to the initiation of therapy. We allowed for a 20-min window from the initiation of therapy until the baseline POC echocardiogram—if images were unable to be obtained within this window, the patient was not eligible for enrollment.
Study design
At the time of enrollment, study subjects underwent a baseline POC echocardiogram that was performed by a study team member consisting of 2 ultrasound fellowship-trained emergency physicians with expertise in POC echocardiography and speckle-tracking, as well as a certified cardiac sonographer who is an employee of our department’s division of clinical research. Vital signs and therapeutic interventions were recorded at the time of the baseline echocardiogram. Baseline demographic data, comorbidities, and home medications were also recorded at this time. Following the 1st echocardiogram, patients underwent usual care for AHF; a 2nd POC echocardiogram was performed 23 h following enrollment. Vital signs, laboratory findings, and further therapies and interventions up to this point were also recorded. Following the 2nd POC echocardiogram, subjects were tracked in-hospital until discharge at which time arrangements were made for a 30-day follow-up telephone interview to ascertain post-discharge outcomes (Fig. 1).
Point-of-care echocardiograms
All POC echocardiograms were performed using the same protocol with a portable ultrasound system (Vivid q by GE, Milwaukee, WI). Study echocardiograms were limited and included only the apical imaging window utilizing the 4-chamber, 2-chamber, and long axis views. After the recording of these ultrasound clips, the POC echocardiogram was considered complete and stored both in the ultrasound system and transmitted to our EchoPAC (GE Milwaukee, WI, USA) workstation for subsequent offline analysis by an investigator who was blinded to treatment, all clinical and laboratory variables and outcome.
Strain analysis
Offline analysis was conducted according to the recommendation of the American Society of Echocardiography and the European Society of Cardiovascular Imaging using commercially available software (EchoPAC version 110.1.1 BT 11, GE Milwaukee, WI, USA). We calculated GLS from the baseline POC echocardiogram and again from the 2nd POC echocardiogram performed 23 h following enrollment using the embedded Automated Function Imaging (AFI) tool. Using the longitudinal strain tracings (Fig. 2), LV MDI was subsequently calculated for baseline and repeat POC echocardiograms by calculating standard deviation of the time-to-peak longitudinal (negative) strain for each of the segments analyzed. All of the strain analyses were performed by a single operator (MF) who was blinded to all of the clinical, radiological and laboratory data at the time of analysis. The automated tracings provided by the AFI software were utilized without any manual adjustments unless it was noted by the operator that the automatic tracings of the endocardial border were grossly inappropriate. Only echocardiograms with frame rates between 40 and 80 frames per second were included for analysis. Echocardiograms that had more than 1 non-trackable segment in a single imaging plane were excluded from analysis as GLS is unable to be calculated when any of the imaging planes are missing > 1 segmental strain. Our prior study demonstrated a very high inter-rater reliability between sonographers of with a concordance correlation coefficient of 0.993 [10].
Primary outcomes and data analysis
All statistical analyses were performed using SAS 9.4. Categorical variables were compared using Chi squared, or Fisher’s exact test in cases were expected cell counts was insufficient, and continuous measures were compared using the paired T test and Wilcoxon Rank sum (both approaches were consistent in categorizing results at the 0.05 significance level). Logistic regression was performed to investigate the relationship between GLS, MDI, and our primary outcome which was readmission or death within 30 days of discharge. We assessed initial, final, and percent change to investigate whether absolute measure, or directional trends in MDI and GLS were useful predictors of patient outcome. Single and multiple logistic regressions were performed using measures of MDI and GLS individually, and in combination. The c-statistic [area under receiver–operator characteristic curve (AUC)] was reported as a measure of logistic regression performance, and Wald’s Chi square was used to evaluate predictor significance in each model.