Development of a framework to identify research gaps from systematic reviews
ABSTRACT Our objective was to develop a framework to identify research gaps from systematic reviews.
We reviewed the practices of (1) evidence-based practice centers (EPCs), and (2) other organizations that conduct evidence syntheses. We developed and pilot tested a framework for identifying research gaps.
Four (33%) EPCs and three (8%) other organizations reported using an explicit framework to determine research gaps. Variations of the PICO (population, intervention, comparison, outcomes) framework were most common. We developed a framework incorporating both the characterization of the gap using PICOS elements (also including setting) and the identification of the reason(s) why the gap exists as (1) insufficient or imprecise information, (2) biased information, (3) inconsistency or unknown consistency, and (4) not the right information. We mapped each of these reasons to concepts from three common evidence-grading systems.
Our framework determines from systematic reviews where the current evidence falls short and why or how the evidence falls short. This explicit identification of research gaps will allow systematic reviews to maximally inform the types of questions that need to be addressed and the types of studies needed to address the research gaps.
[Show abstract] [Hide abstract]
ABSTRACT: The aim of this study was to identify and prioritize research gaps to help decrease maternal mortality. We conducted a two-stage survey. We provided participants (Cochrane Collaboration experts) with a list of 319 problem/population, intervention, comparison, and outcome questions built from 178 Cochrane systematic reviews. Questions were classified according to causes of maternal death. Respondents of the first round refined the research questions and prioritized them by eliminating those that were considered of low priority, according to four criteria. They also included additional questions. In the second round, respondents prioritized 62 questions. The overall response rates for the first and second rounds were 47% (73 of 155) and 17% (363 of 2,121), respectively. Participants ranked 62 of the research questions as "very relevant." Approximately 20% of all questions that were identified in Cochrane reviews and two-third of questions of the second round were considered of "very high priority." More women (235) than men (128) participated in the survey. We did not find statistically significant differences when comparing the groups of very relevant questions by the type of respondent, income, country, and round. We identified research priorities by mapping and improving the understanding of research needs in low- and middle-income settings internationally.Journal of clinical epidemiology 01/2014; 67(3):314-24. DOI:10.1016/j.jclinepi.2013.10.007 · 5.48 Impact Factor
[Show abstract] [Hide abstract]
ABSTRACT: Introduction Choice of outcomes is critical for clinical trialists and systematic reviewers. It is currently unclear how systematic reviewers choose and pre-specify outcomes for systematic reviews. Our objective was to assess the completeness of pre-specification and comparability of outcomes in all Cochrane reviews addressing four common eye conditions. Methods We examined protocols for all Cochrane reviews as of June 2013 that addressed glaucoma, cataract, age-related macular degeneration (AMD), and diabetic retinopathy (DR). We assessed completeness and comparability for each outcome that was named in ≥25% of protocols on those topics. We defined a completely-specified outcome as including information about five elements: domain, specific measurement, specific metric, method of aggregation, and time-points. For each domain, we assessed comparability in how individual elements were specified across protocols. Results We identified 57 protocols addressing glaucoma (22), cataract (16), AMD (15), and DR (4). We assessed completeness and comparability for five outcome domains: quality-of-life, visual acuity, intraocular pressure, disease progression, and contrast sensitivity. Overall, these five outcome domains appeared 145 times (instances). Only 15/145 instances (10.3%) were completely specified (all five elements) (median = three elements per outcome). Primary outcomes were more completely specified than non-primary (median = four versus two elements). Quality-of-life was least completely specified (median = one element). Due to largely incomplete outcome pre-specification, conclusive assessment of comparability in outcome usage across the various protocols per condition was not possible. Discussion Outcome pre-specification was largely incomplete; we encourage systematic reviewers to consider all five elements. This will indicate the importance of complete specification to clinical trialists, on whose work systematic reviewers depend, and will indirectly encourage comparable outcome choice to reviewers undertaking related research questions. Complete pre-specification could improve efficiency and reduce bias in data abstraction and analysis during a systematic review. Ultimately, more completely specified and comparable outcomes could make systematic reviews more useful to decision-makers.PLoS ONE 10/2014; 9(10):e109400. DOI:10.1371/journal.pone.0109400 · 3.53 Impact Factor
Infection Control and Hospital Epidemiology 08/2014; 35(S1):S1-S67. DOI:10.1086/676882 · 3.94 Impact Factor