Objective: To examine the accuracy of clinical diagnostic methods in assessing vaginal bleeding in pregnant women before 28 weeks’ gestation. Methods: 772 consecutive women who presented with threatened abortion before 28 weeks’ gestation were studied. Structured history and physical examination were performed on each woman as initial clinical assessment. This was followed by transvaginal sonography to determine the status of pregnancy. The accuracy of diagnoses at different clinical stages (history, physical examination, and transvaginal sonography) relative to the final diagnosis was compared using the kappa coefficient (ĸ). Results: Clinicians were unable to accurately diagnose or predict the status of the pregnancy from history alone (ĸ = 0.33; 95% CI 0.28, 0.38) or after physical examination (ĸ = 0.57; 95% CI 0.52, 0.62). Transvaginal sonography led to an accurate diagnosis of pregnancy status (ĸ = 0.96; 95% CI 0.95, 0.98) in most cases, except where an ectopic pregnancy existed. Absence of abdominal tenderness (p = 0.04), cervical excitation (p = 0.02), and incorrect identification of retained products of conception on transvaginal sonography (p = 0.01) were features of missed ectopic pregnancy. Conclusions: The clinical assessment of threatened abortion by history and physical examination is unreliable in most cases. The diagnostic accuracy is improved by the addition of transvaginal sonography. Even with the help of transvaginal sonography, a small proportion of ectopic pregnancies will be missed.