Much of the recent research in DNA computing has focused on designing better overall techniques for computation, or implementing the techniques in simulation or a wet-lab in order to show the viability of these techniques for solving small SAT problems. In this paper, we examine a major obstacle to using DNA computing to solve larger, real-world SAT problems for which the correct answer is not already known. In particular, we ask the following question: Given the results of a DNA computation, how does one determine the answer to the underlying SAT problem, and how does one examine the confidence of this answer? We examine this question in detail for selection-based DNA computing, and show-that it is non-trivial to answer. We then introduce a method we call "decision thresholds" for answering it which can be applied to any variation of selection-based DNA computing. Furthermore, we provide an example by applying this method to the technique of using a network of microreactors employing negative selection of ssDNA.