First Trimester Point of Care Ultrasound: Imaging Features and Review Behaviors Associated With Diagnostic Accuracy.

Abstract

We aimed to identify the most diagnostically challenging features in first-trimester point-of-care ultrasound (FT-POCUS) images. We also sought to determine the physician image review behaviors associated with increased diagnostic accuracy. We conducted a multicenter prospective cross-sectional study in a convenience sample of emergency physicians in the United States and Canada. The web-based intervention included 400 FT-POCUS cases acquired via the transabdominal or transvaginal approach. Participants reviewed FT-POCUS cases to identify pregnancy-related imaging findings. We captured clickstream-level data with each case encounter, including the correctness of a participant's response and physician image review behaviors. We enrolled 317 participants, who collectively generated 16,295 case interpretations. The most diagnostically challenging imaging findings included eccentrically located gestational sac and endometrial collection/heterogeneous uterine material (p < 0.001 for all comparisons). Participants who reported "definite" certainty, as opposed to "probable," demonstrated a significantly higher odds of getting the diagnosis of intrauterine pregnancy (IUP) present or absent correct (OR = 4.48; 95% CI 4.00, 5.01) and a lower odds of time spent reviewing cases (OR = 0.46; 95% CI 0.40, 0.51). Those who reviewed a higher proportion of available views per case were more likely to accurately identify a fetal heartbeat (OR = 1.51; 95% 1.34, 1.69), multiple IUPs (OR = 1.33; 95% CI 1.10, 1.61), and adnexal structures (OR = 1.11; 95% CI 1.04, 1.17), but less likely to correctly identify an IUP (OR = 0.93; 95% CI 0.88, 0.99) and endometrial fluid collection/heterogeneous uterine material (OR = 0.96; 95% CI 0.92, 0.99). Emergency physicians interpreting FT-POCUS images encountered specific diagnostic challenges that may increase risks to patient safety. We found that higher diagnostic confidence correlated with greater diagnostic accuracy and efficiency. Reviewing a larger proportion of available images improved diagnostic accuracy for some findings, but not for others.