In healthcare, it takes a long time for new treatments to move from clinical studies into practice: perhaps an average of 17 years [Balas et al., 2000]. Systematic review is a critical step in this research translation process because it determines what is known. To do this, a systematic review analyzes all available evidence on a particular question through a series of steps, including data extraction. The current best practice for data extraction is for two people to independently identify and extract data from each research paper. Because the data extraction step is almost always performed manually, it is very time-consuming [Tsafnat et al., 2014] yet methodological errors may cause problems with the review's conclusions [Lundh et al., 2009]. Our long-term goal is to help reviewers synthesize the literature quickly and accurately by developing a semi-automatic support system for data extraction. Towards this end, we are currently conducting an in-depth case study of a single systematic review, a Cochrane Review about oral pain relief. Through manual annotation and a content analysis of the six studies synthesized by this Cochrane Review, we will develop hypotheses about which clinical data elements can be automatically extracted. We will also develop an annotated corpus which will enable us to propose methods for automatically supporting human reviewers in data extraction. Eventually, we plan to design a semi-automated support system, and to test the two hypotheses (1) that it can reduce the time and human labor required to conduct a review and (2) that it can maintain or increase the quality of the resulting review.