Objective: Many meta-analyses have provided synthesised likelihood ratio data to aid clinical decision-making. However, much less has been published on how to safely combine clinical information in practice. We aimed to explore the benefits and risks of pooling clinical information during the ED assessment of suspected acute coronary syndrome. Methods: Clinical information on 1776 patients was collected within a randomised trial conducted across five South Australian EDs between July 2011 and March 2013. Bayes theorem was used to calculate patient-specific post-test probabilities using age- and gender-specific pre-test probabilities and likelihood ratios corresponding to the presence or absence of 18 clinical factors. Model performance was assessed as the presence of adverse cardiac outcomes among patients theoretically discharged at a post-test probability less than 1%. Results: Bayes theorem-based models containing high-sensitivity troponin T (hs-troponin) outperformed models excluding hs-troponin, as well as models utilising TIMI and GRACE scores. In models containing hs-troponin, a plateau in improving discharge safety was observed after the inclusion of four clinical factors. Models with fewer clinical factors better approximated the true event rate, tended to be safer and resulted in a smaller standard deviation in post-test probability estimates. Conclusions: We showed that there is a definable point where additional information becomes uninformative and may actually lead to less certainty. This evidence supports the concept that clinical decision-making in the assessment of suspected acute coronary syndrome should be focused on obtaining the least amount of information that provides the highest benefit for informing the decisions of admission or discharge.