Development, Application and Evaluation of Risk of Bias Criteria for Case-Crossover and Time-Series Studies of Air Pollution and Health

Abstract

Systematic review and meta-analysis methods are increasingly being applied to environmental health literature. However, these methods have not been routinely applied within the context of formal risk assessments, despite common aims and practices, including systematic identification, analysis and summary assessment of the weight of evidence linking exposures and outcomes. A particular gap in these practices is the availability of standardized criteria for assessing risk of bias in studies of environmental exposures, operationalized in a form that can be efficiently and reliably applied by reviewers / risk assessors to a potentially large number of primary studies. Building on the Navigation Guide systematic review methodology, we developed, applied and evaluated risk of bias criteria applicable to time series and case-crossover studies linking air pollution and cardiovascular and respiratory morbidity, operationalizing them in DistillerSR™ systematic review software in the context of a systematic review of health effects of nitrogen dioxide. Risk of bias domains comprised: selection bias and generalizability, exposure assessment, confounding, outcome assessment, completeness of outcome data, selective outcome reporting, conflict of interest and other sources of bias. Risk of bias criteria were developed through literature review and expert consultation and evaluated with respect to content and face validity, inter-rater agreement and completion time. Our findings address the feasibility and reliability of our risk of bias criteria for time series and case-crossover studies linking air pollution and cardiovascular and respiratory morbidity. These criteria may provide a promising tool in the context of both systematic review and risk assessment. PROSPERO registration number CRD42018084497.