Table 2

Framework for considering study quality and likelihood of bias6

Study featureQualities sought
Sample of patientsConsecutive or randomly selected sample, recruited as a single cohort unclassified by disease state, recruited from clinical setting and point in referral process where test would be used, selection and referral process fully described, clinical and demographic characteristics fully described.
Reference diagnosisMethod and tests described in detail, positive and negative diagnoses clearly described, diagnosis likely to be close to truth, available for all patients based on same tests and information in all patients. Blinding procedures used to prevent knowledge of result of experimental test influencing the reference diagnosis, made before treatment started.
Experimental testApplication of test described in detail, positive and negative test results clearly described, blinding procedures used to ensure that test is undertaken without knowledge of reference diagnosis, test undertaken before treatment started, results reported for all patients, including those with “grey zone” results, “gold standard” test in all patients (EEG, or detailed description of clinical criteria where EEG not undertaken)