Artificial Intelligence-Based Assessment of Indices of Right Ventricular Function.

OBJECTIVES: Echocardiographic assessment of right ventricular (RV) function is based largely on visual estimation of tricuspid annulus and motion of the free wall. Regional strain analysis has provided an objective measure of myocardial performance assessment, but is limited in use by vendor-specific software. The study was designed to investigate statistical correlation between RV region-specific strain and echocardiographic parameters of RV function using a vendor-neutral RV-specific strain assessment program. DESIGN: This is a retrospective study. SETTING: Tertiary hospital. PARTICIPANTS: One hundred seven patients undergoing coronary artery bypass graft, valve repair or replacement, or a combination of procedures. INTERVENTION: None. MEASUREMENTS AND MAIN RESULTS: One hundred seven patients underwent comprehensive echocardiographic of RV function intraoperatively. Off-line analysis of global, longitudinal, and septal strain was performed using a vendor-neutral software. The 2 values were compared statistically. All pairs demonstrated strong statistical significance; the strongest relationships were between (1) RV fractional area change (FAC) (%)-RV longitudinal strain (r2 = 0.83, p < 0.001), and (2) tricuspid annular plane systolic excursion (mm)-lateral S' velocity (cm/s) (r2 = 0.80, p < 0.001). The weakest correlations were (1) RV FAC (%)-lateral S' velocity (cm/s) (r2 = 0.37, p < 0.001), and (2) lateral S' velocity (cm/s)-RV longitudinal strain (r2 = 0.40, p < 0.001). CONCLUSION: RV function can be assessed objectively by strain analyses across different platforms using the artificial intelligence-based vendor-neutral strain analysis software. There is a statistically significant correlation between strain values and conventional 2-dimensional echocardiographic parameters of RV function.

Summative Objective Structured Clinical Examination Assessment at the End of Anesthesia Residency for Perioperative Ultrasound.

While standardized examinations and data from simulators and phantom models can assess knowledge and manual skills for ultrasound, an Objective Structured Clinical Examination (OSCE) could assess workflow understanding. We recruited 8 experts to develop an OSCE to assess workflow understanding in perioperative ultrasound. The experts used a binary grading system to score 19 graduating anesthesia residents at 6 stations. Overall average performance was 86.2%, and 3 stations had an acceptable internal reliability (Kuder-Richardson formula 20 coefficient >0.5). After refinement, this OSCE can be combined with standardized examinations and data from simulators and phantom models to assess proficiency in ultrasound.

Simulation-based Assessment to Reliably Identify Key Resident Performance Attributes.

BACKGROUND: Obtaining reliable and valid information on resident performance is critical to patient safety and training program improvement. The goals were to characterize important anesthesia resident performance gaps that are not typically evaluated, and to further validate scores from a multiscenario simulation-based assessment. METHODS: Seven high-fidelity scenarios reflecting core anesthesiology skills were administered to 51 first-year residents (CA-1s) and 16 third-year residents (CA-3s) from three residency programs. Twenty trained attending anesthesiologists rated resident performances using a seven-point behaviorally anchored rating scale for five domains: (1) formulate a clear plan, (2) modify the plan under changing conditions, (3) communicate effectively, (4) identify performance improvement opportunities, and (5) recognize limits. A second rater assessed 10% of encounters. Scores and variances for each domain, each scenario, and the total were compared. Low domain ratings (1, 2) were examined in detail. RESULTS: Interrater agreement was 0.76; reliability of the seven-scenario assessment was r = 0.70. CA-3s had a significantly higher average total score (4.9 ± 1.1 vs. 4.6 ± 1.1, P = 0.01, effect size = 0.33). CA-3s significantly outscored CA-1s for five of seven scenarios and domains 1, 2, and 3. CA-1s had a significantly higher proportion of worrisome ratings than CA-3s (chi-square = 24.1, P < 0.01, effect size = 1.50). Ninety-eight percent of residents rated the simulations more educational than an average day in the operating room. CONCLUSIONS: Sensitivity of the assessment to CA-1 versus CA-3 performance differences for most scenarios and domains supports validity. No differences, by experience level, were detected for two domains associated with reflective practice. Smaller score variances for CA-3s likely reflect a training effect; however, worrisome performance scores for both CA-1s and CA-3s suggest room for improvement.