Queries” feature (http://www.nlm.nih. gov/pubs/techbull/mj10/mj10_clin_query.html) is designed for specific clinical questions such as the example mentioned previously. Gathering evidence to support broader topics, such as the protocols in this book, presents the searcher with a larger challenge. Limiting searches by methodology can unwittingly eliminate the best evidence for study designs that do not lend themselves to these methods. For example, a cross-sectional retrospective design may provide the highest level of evidence for a study that examines “nurses’ perception” of the practice environment (Boltz et al., 2008). Methodological filters have other limitations, such as retrieving citations tagged “randomized controlled trials as topic” or abstracts that state a “systematic review of the literature” was conducted (which is not the same as retrieving a study that is actually a systematic review). Chapter authors were cautioned that the CINAHL database assigns publication type “systematic review” to numerous citations that upon review, we judged to be “Level V” review articles (narrative reviews or literature reviews), not necessarily the high level of evidence we would call “Level I,” (which according to our scheme are studies that do a rigorous synthesis and pooling or analysis of research results). It may not be easily discernible from an article title and abstract whether the study is a systematic review with evidence synthesis or a narrative literature review (Lindbloom, Brandt, Hough, & Meadows, 2007). These pitfalls of computerized retrieval are justification for the review by the searcher to weed false hits from the retrieved list of articles. Precision and Recall An additional challenge to an intermediary searcher is the need to balance the comprehensiveness of recall (or “sensitivity”) with precision (“specificity”) to retrieve a “useful” number of references. The Cochrane Handbook states: “Searches should seek high sensitivity, which may result in relatively low precision” (Section 6.3; Higgins & Green, 2008). Thus, retrieving a large set of articles may include many irrelevant hits. Conversely, putting too many restrictions on a search may exclude relevant studies. The goal of retrieving the relevant studies for broad topic areas required “sacrificing precision” and deferring to the chapter authors to filter false or irrelevant hits (Jenkins, 2004; Matthews et al., 1999). The iterative nature of a literature search requires that an initial set of relevant references for both broad or specific research questions serves to point authors toward best evidence as an adjunct to their own knowledge, their own pursuit of “chains of citation” (McLellan, 2001) and related records, and their clinical expertise. Thus, a list of core references on physical restraints, supplied to a chapter author, might lead to exploring citations related to wandering, psychogeriatric care, or elder abuse (Fulmer, 2002). Developing and Evaluating Clinical Practice Guidelines: A Systematic Approach 7 LEVELS OF EVIDENCE Step 3, critical appraisal of the evidence, begins with identifying the methodology used in a study (often evident from reviewing the article abstract) followed by a critical reading and evaluation of the research methodology and results. The coding scheme described in the subsequent text provides the first step in filtering retrieved studies based on research methods. Levels of evidence offer a schema that, once known, helps the reader to understand the value of the information presented to the clinical topic or question under review. There are many schemas that are used to identify the level of evidence sources. Although multiple schemas exist, they have commonalities in their hierarchical structure, often represented by a pyramid or “publishing wedge” (DiCenso, Bayley, & Haynes, 2009; McKibbon et al., 2009). The highest level of evidence is seen at the top of a pyramid, characterized by increased relevance to the clinical setting in a smaller number of studies. The schema used by the authors in this book for rating the level of evidence comes from the work of Stetler et al. (1998) and Melnyk and Fineout-Overholt (2005; See Figure 1.1). A Level I evidence rating is given to evidence from synthesized sources (systematic reviews), which can