Objective: To evaluate the predictive value of first-trimester ultrasound and biochemical markers in determining pregnancy outcomes among women with symptoms of threatened miscarriage (TM). Design: Prospective cohort study Setting: Accident and Emergency Department of the national public hospital in Malta Sample: In total, 118 singleton pregnancies between 5 +0 and 12 +6 weeks’ gestation, confirmed as viable via transvaginal ultrasound, were recruited between January 2023 and June 2024, using convenience sampling. Methods: Ultrasound and biochemical markers were measured and other clinical and sociodemographic parameters were collected via questionnaire. Univariate logistic regression identified individual predictors of pregnancy loss. Multivariate logistic regression (MLR) and random forest (RF) modelling were applied to assess combined predictive performance. Main outcome measures: Comparison of MLR and RF in predicting pregnancy outcome in TM. Results: Among 118 women with TM, 77% progressed to live birth and 23% experienced pregnancy loss. Multivariate logistic regression identified progesterone, cervical length, mean gestational sac diameter (MGSD), trophoblast thickness, sFlt-1:PlGF ratio, and maternal age as significant predictors of outcome. Higher progesterone, cervical length, MGSD and sFlt-1:PlGF ratio were associated with reduced risk, while maternal age over 35 raised the likelihood of loss. The model achieved 82.7% accuracy (AUC = 0.89). Random forest modelling further improved accuracy to 93.1% (AUC = 0.97), confirming the combined predictive value of ultrasound and biochemical markers. Conclusion: Ultrasound and biochemical markers offer predictive value in TM. Machine learning models, particularly random forest, may enhance early risk stratification in clinical settings.