0091 Evaluating the epidemiological evidence – a comparison of frameworks for assessing individual studies

    loading  Checking for direct PDF access through Ovid

Abstract

Introduction

Frameworks and tools developed to evaluate observational studies of environmental and occupational exposures for systematic reviews draw from similar efforts in the field of healthcare management to achieve increased transparency in assessment conclusions.

Methods

This presentation compares approaches developed by U.S. and European government and academic institutions that evaluate risk of bias and sensitivity for observational studies of environmental and occupational exposures. An international collaborative project to adapt a risk of bias tool developed by the Cochrane Collaborative (Risk of Bias in Non-Randomised Studies of Interventions) that will address the varied study designs and exposure assessment methods used in these types of studies is described.

Results

Several commonalities are identified, including the use of signalling questions to guide evaluations, outcome-specific rather than study-specific evaluations, and avoidance of numeric scores. All frameworks evaluate participant selection/attrition/exclusion, confounding, reliability and validity of exposure and outcome assessments, and selective reporting. Less consistently used domains are analytic methods, sensitivity, and conflict of interest. The frameworks use different approaches to derive an overall conclusion about risk of bias or confidence.

Conclusion

As with narrative reviews, structured frameworks depend heavily on expert judgement requiring the involvement of reviewers with the correct discipline-specific expertise. The transparency of the overall evidence integration in a systematic review depends on the knowledgeable and clear presentation of study evaluation conclusions.

Conclusion

The views expressed are those of the author and do not necessarily represent the views or policies of the U.S. Environmental Protection Agency.

Related Topics

    loading  Loading Related Articles