The large success of evidence based medicine (EBM) has led to the growing expectation on evidence based policy (especially from researchers). It is hard to argue against using evidence effectively, but there might be difficulty in doing this in policy making. Parkhurst (2017) makes a good discussion. Some reviews and summaries are here and here.
One large difference in applying the method of EBM to policy making is its difference in objectives. There are not much conflict in the aim and objectives in medicine: patients, doctors and insurers all want better quality of care. However, this is not the same in policies. There are many dimensions of issues and the interests of actors may turn against each other.
For example, the American Medical Association argued that "laws that regulate abortion should be evidence-based and designed to improve women's health", while the main arguement in the public was about women's right to make decision regarding abortion. The discussion from AMA, although based on evidence, was not grasping the point.
A common approach to improve evidence based policy is to bridge the gap by simply training individuals: making research more understandable and increasing political demand for research. However, the problem might lay deeper. There might be a certain bias lying under each side, making the understanding difficult. Evidence advocates are often worried about "technical bias" in which evidence is misused or manipulated for political reasons. On the other hand, policy makers are concerned about "issue bias" in which appeals to evidence serve to obscure key social values or impose political priority in unrepresentative ways.
Politics should be portrayed as a process by which social values are translated into government priorities (rather than a barrier), and there should be an institutionalised system that govern the use of evidence in bettwer ways, based on considerations on the political nature and origins of these biases. This will require explicit normative principles such as scientific fidelity to counter technical bias and democratic representation to counter issue bias to guide institutional changes.