Table 1.
Evidence Levels | Study Design |
---|---|
I | Randomized well-designed trials with low alpha and beta error* |
II | Randomized trials with high alpha and beta error* |
III | Nonrandomized concurrently controlled studies |
IV | Nonrandomized historically controlled studies |
V | Case series |
Adapted from Sackett9
Alpha error refers to the probability (generally set at 95% or greater) that a significant outcome (e.g., p < 0.05) is not a result of chance occurrence. Beta error refers to the probability (generally set at 80% to 90% or greater) that a nonsignificant result (e.g., p > 0.05) is the correct conclusion of the study or studies. The estimation of beta error is generally the result of a power analysis. The power analysis includes a sample size analysis to project the size of the study population necessary to ensure that significant differences will be observed if actually present.