Article type
Year
Abstract
Introduction/Objective: To define the optimal content of electronic database reports of randomized clinical trials (RCTs) based on a task analysis of systematic reviewing, and to compare this content list to published recommendations for RCT reporting. Background: RCT reporting recommendations (e.g., CONSORT) must limit their scope and number to what can be reported within the space constraints of paper-based text articles. We have proposed that RCTs be reported into structured databases called trial banks, in addition to being reported in text. We present here a task-based rationale for the information that these trial banks should contain, if they are to satisfy completely the information needs of systematic reviewers.
Methods: Based on a literature review and on our own systematic-review experience, we identified the reasoning tasks for systematic reviewing and organized them into a hierarchy of tasks and subtasks. For tasks that could be completed with several different methods, we specified subtasks for each method. We then specified the necessary and sufficient trial information for completing each task. We also documented the rationale for the tasks and their information requirements.
Results: We identified 4 top-level tasks of systematic reviewing, and decomposed them into 62 subtasks that induced 132 unique requirements for trial information. Task 1: Trial retrieval: For accurate electronic retrieval of RCTs, the clinical content of RCTs must be indexed in a controlled clinical vocabulary. Task 2: Judging a trial's internal and external validity: We identified 42 subtasks for this task (e.g. Determine potential for bias due to funding source) that required a total of 126 trial-information items (e.g., funding sources and their types). Because no standard methods exist for accomplishing most of these subtasks, this task decomposition was method neutral. Task 3- Quantitative synthesis: This task had 2 subtasks each for the fixed and random effects and meta-regression methods, and a total of 5 required trial items. Task 4: Contextual interpretation: This task had 18 general information requirements for placing an RCT in its proper scientific, epidemiological, socioeconomic, and ethical context. In comparison to the information requirements of 18 published trial-critiquing instruments and reporting guidelines, 95% (106/112) of the unique items required by the instruments were also required by our task analysis. Conversely, only 65% (87/132) of the items required by our analysis was required by the instruments.
Discussion: Our task analysis of systematic reviewing explicitly relates RCT reporting requirements to the tasks and information needs of systematic reviewers. It justifies the inclusion of 132 unique items of trial information into RCT reports. These recommendations overlap with, but are more extensive than other published recommendations. If RCTs were reported into electronic trial banks in accordance with these reporting recommendations, then systematic reviewers would have direct electronic access to all the trial information they need for completing the tasks of systematic reviewing.
Methods: Based on a literature review and on our own systematic-review experience, we identified the reasoning tasks for systematic reviewing and organized them into a hierarchy of tasks and subtasks. For tasks that could be completed with several different methods, we specified subtasks for each method. We then specified the necessary and sufficient trial information for completing each task. We also documented the rationale for the tasks and their information requirements.
Results: We identified 4 top-level tasks of systematic reviewing, and decomposed them into 62 subtasks that induced 132 unique requirements for trial information. Task 1: Trial retrieval: For accurate electronic retrieval of RCTs, the clinical content of RCTs must be indexed in a controlled clinical vocabulary. Task 2: Judging a trial's internal and external validity: We identified 42 subtasks for this task (e.g. Determine potential for bias due to funding source) that required a total of 126 trial-information items (e.g., funding sources and their types). Because no standard methods exist for accomplishing most of these subtasks, this task decomposition was method neutral. Task 3- Quantitative synthesis: This task had 2 subtasks each for the fixed and random effects and meta-regression methods, and a total of 5 required trial items. Task 4: Contextual interpretation: This task had 18 general information requirements for placing an RCT in its proper scientific, epidemiological, socioeconomic, and ethical context. In comparison to the information requirements of 18 published trial-critiquing instruments and reporting guidelines, 95% (106/112) of the unique items required by the instruments were also required by our task analysis. Conversely, only 65% (87/132) of the items required by our analysis was required by the instruments.
Discussion: Our task analysis of systematic reviewing explicitly relates RCT reporting requirements to the tasks and information needs of systematic reviewers. It justifies the inclusion of 132 unique items of trial information into RCT reports. These recommendations overlap with, but are more extensive than other published recommendations. If RCTs were reported into electronic trial banks in accordance with these reporting recommendations, then systematic reviewers would have direct electronic access to all the trial information they need for completing the tasks of systematic reviewing.