Table 2

CLARITY tool for risk of bias in cohort studies: acceptable standards for each domain

DomainCriteria for determination of low risk of bias
Selection of exposed and non-exposed cohorts drawn from the same population We will consider the exposed and non-exposed cohorts to be selected from the same population and at low risk of bias if both are drawn from the same database of patients presenting at the same points of care over the same time frame.
Assessment of exposure We will consider secure records (eg, hospital or pharmacy records) or repeated patient interviews (or other forms of ascertainment) asking about the exposure to have a low risk of bias. Examples of ascertainment of exposure with a higher risk of bias will include structured interviews at a single point in time, written self-report and retrospective self-report (which may be subject to recall bias).
Confidence that the outcome of interest was not present at the start of the study We will consider that there is a low risk of bias resulting from the outcome of interest being present at the start of the study if there is a complete description of all prognostically important baseline (pre-exposure) characteristics for patients included in the cohort.
Matching or adjustment for all plausible prognostic variables We will consider studies to be at a low risk of bias if the study has used a comprehensive matching method or has statistically adjusted for al plausible prognostic variables. Studies that match or adjust for a minority of plausible prognostic variables or use not matching or adjustment at all will be considered to be at a high risk of bias. We will not consider statements of no differences between groups that are not substantiated by quantifiable data to be sufficient for establishing comparability.
Confidence in the assessment of the presence or absence of prognostic factors We will consider studies to be at a low risk of bias if they include an interview or self-completed survey of all participants, or if there is reproducibility or documentation of the accuracy of abstraction of prognostic data. We will consider studies that extract prognostic information from a database with no documentation of quality of abstraction of prognostic variables to be at high risk of bias.
Confidence in the assessment of outcome We will consider studies that use independent blinded assessment or record linkage to a secure record (eg, hospital or pharmacy record) to be a low risk of bias. Studies that use unblinded independent assessment or self-report will be considered to be at higher risk of bias and studies with no description of how outcome was assessed will be considered to be at high risk of bias.
Adequacy of follow-up We will consider studies to be at a low risk of bias if there is no missing outcome data, if the reasons for missing outcome data are unlikely to be related to the true outcome, or if the missing outcome data is balanced in numbers across intervention groups, with similar reasons for missingness across groups. We will also consider studies to be at low risk of bias if missing data have been imputed using appropriate methods.
Balance of co-intervention between groups We will consider studies to be at a low risk of bias of most or all relevant co-interventions that might influence the outcome of interest are documented to be similar in the exposed and unexposed groups.
  • CLARITY, Clinical Advances Through Research and Information Translation.