Some heuristics and biases are assumed to be universal for human decision-making, and may thus be expected to appear consistently and need to be considered when planning for real-life decision-making. Yet results are mixed when exploring the biases in applied settings, and few studies have attempted to robustly measure the combined impact of various biases during a decision-making process. We performed three pre-registered classroom experiments in which trained medical students read case descriptions and explored follow-up information in order to reach and adjust mental health diagnoses (∑N = 224). We tested whether the order of presenting the symptoms led to a primacy effect, whether there was a congruence bias in selecting follow-up questions, and whether confidence increased during the decision process. Our results showed increased confidence for participants that did not change their decision or sought disconfirming information. There was some indication of a weak congruence bias in selecting follow-up questions. There was no indication of a stronger congruence bias when confidence was high, and there was no support for a primacy effect of the order of symptom presentation. We conclude that the biases are difficult to demonstrate in pre-registered analyses of complex decision-making processes in applied settings.