Inclusion Criteria

Basic criteria for original or review articles:

  • in English
  • about humans
  • about topics that are important to the clinical practice of medicine, nursing, rehabilitation, and other health professions, other than descriptive studies of prevalence
  • analysis of each article consistent with the study question.

Studies of prevention or treatment must meet these additional criteria:

  • random allocation of participants to comparison groups
  • ≥ 10 patients per group (assessed for outcome)
  • ≥ 1 specified outcome assessed in ≥ 80% of those randomized at any given follow-up point
  • outcome measure of known or probable clinical importance
  • subgroup analyses must be preplanned, with groups analyzed as they were randomized; analysis must test for interaction between ≥ 2 subgroups.

Studies of diagnosis must meet these additional criteria:

  • inclusion of a spectrum of participants where some, but not all, have the disorder or derangement of interest
  • inclusion of ≥ 100 participants, with ≥ 50 participants with the disease and ≥ 50 participants without the disease
  • objective diagnostic ("gold") standard (e.g., laboratory test not requiring interpretation) OR current clinical standard for diagnosis (e.g., a venogram for deep venous thrombosis), preferably with documentation of reproducible criteria for subjectively interpreted diagnostic standard (i.e., report of statistically significant measure of agreement beyond chance among observers)
  • each participant must receive both the new test and some form of the diagnostic standard
  • interpretation of diagnostic standard without knowledge of test result
  • interpretation of test without knowledge of diagnostic standard result.
Diagnostic tests may also be tested in randomized trials, in which case the criteria for prevention or treatment apply.

Studies of differential diagnosis must meet these additional criteria:

  • a cohort of patients who present with a similar, initially undiagnosed but reproducibly defined, clinical problem
  • explicit description of clinical setting, including the referral filter
  • ascertainment of diagnosis for ≥ 80% of patients using a reproducible diagnostic work-up strategy for all patients and follow-up until patients are diagnosed or follow-up ≥ 1 month for acute disorders or ≥ 1 year for chronic or relapsing disorders.

Studies of prognosis must meet these additional criteria:

  • inception cohort of patients at a similar and early point in the course of a disease or condition, all initially free of the outcome of interest
  • prospective standardized data collection specifically for the purpose of the study (i.e., identifying and following up an inception cohort of patients with a disease)
  • follow-up of ≥ 80% of patients until the occurrence of a major study endpoint or to the end of the study.

Studies of clinical prediction guides must meet these additional criteria:

  • purpose is to validate or compare a rule/index/scale/model that combines ≥ 2 factors into some type of score/ranking that assigns individual patients to different levels of risk for a specific outcome (diagnosis, prognosis, treatment responsiveness) based on the presence/absence of these factors
  • data for the prediction guide must be available before data on the outcome that it is predicting
  • the guide must be generated in one or more sets of real (not hypothetical) patients (derivation or development cohort)
  • the guide must be validated in another set of real (not hypothetical) patients (validation cohort); internal bootstrapping is not acceptable as validation
    • studies validating a previously derived clinical prediction guide must provide a reference for derivation of the guide and must assess ≥ 1 of the outcomes assessed in the derivation cohort
    • prediction guides developed using individual patient data from ≥ 1 study do not require separate validation
  • study must provide information on how to apply the prediction guide in individual patients or cite a reference to this information.

Studies of etiology of harm from medical interventions must meet these additional criteria:

  • explicit purpose is to assess adverse effects of an intervention
  • prospective data collection with clearly identified comparison groups for those at risk for the outcome of interest (in descending order of preference: quasi-randomized controlled trial, nonrandomized controlled trial, cohort study with case-by-case matching or statistical adjustment to create comparable groups, nested case–control study)
  • blinding (masking) of observers of outcomes to exposures (criterion assumed to be met if outcome is objective, e.g., all-cause mortality or objective test)
  • if harm reported, relative risk (RR) or hazard ratio (HR) or equivalent ≥ 2.0, with a lower 95% CI that excludes 1.5
  • if no harm reported, upper 95% CI of RR or HR or equivalent excludes 1.5.
Randomized controlled trials assessing adverse effects are evaluated using criteria for studies of prevention or treatment.

Studies of quality improvement or continuing education must meet these additional criteria:

  • random allocation of participants or units to comparison groups
  • ≥ 10 patients per group (assessed for outcome)
  • ≥ 1 specified outcome assessed in ≥ 80% of those randomized at any given follow-up point
  • outcome measure of known or probable clinical or educational importance.

Studies of the economics of health care programs or interventions must meet these additional criteria:

  • the economic question addressed must be based on comparison of alternatives in real patients
  • alternate diagnostic or therapeutic services or quality improvement activities must be compared on the basis of both the outcomes produced (effectiveness) and resources consumed (costs)
  • evidence of both effectiveness and costs reported in single randomized controlled trial that passes criteria for prevention or treatment
  • results should be presented in terms of the incremental or additional costs and outcomes of one intervention over another
  • where uncertainty exists in the estimates or imprecision in the measurement, a sensitivity analysis should be done.

Systematic review articles must meet these additional criteria:

  • explicit statement of the clinical topic
  • identifiable description of the methods, including the databases searched and inclusion and exclusion criteria for selecting articles for detailed review; reviews of prognosis must have "inception cohort" as an inclusion criteria
  • ≥ 1 major database searched.

Reviews that explicitly do not include ≥ 1 study that meets the above-noted criteria for treatment, diagnosis, differential diagnosis, prognosis, clinical prediction guides, etiology of harm from medical interventions, quality improvement, or economics are excluded.

Individual patient data meta-analyses must meet these additional criteria:

  • includes data combined at a patient level from ≥ 2 original studies about prevention, treatment, diagnosis, differential diagnosis, prognosis, clinical prediction, etiology, quality improvement, or economics of health care.

These criteria are subject to modification if, for example, it is found feasible to apply higher standards that increase the validity and applicability of studies for clinical practice. The objective of the criteria screen is to include only the very best literature, consistent with a reasonable number of articles "making it through the filter."

Updated 14 November 2013