Levels of Evidence
Pyramid-shaped models are often used to display the relative levels of evidence of research study designs. Several names are used for these models, including:
- Levels of evidence pyramid
- EBP pyramid
- Study design hierarchy
A research design's placement on an EBP pyramid provides information about its level of evidence. Meta analyses and systematic reviews are often listed at the top of EBP pyramids to indicate they are considered the highest levels of evidence. On the example EBP pyramid shown below, the primary study designs are randomized controlled trials, cohort studies, case control studies, and case series / case reports. Brief definitions of these study designs are available on the Study Designs tab of this box (above).
This level of evidence pyramid may be shared with you in sessions taught by informationists or librarians at the University of Michigan. Your textbooks or course materials may include other EBP pyramids or study design hierarchies. Other models may include additional study designs or information to help select high quality evidence to answer your clinical question.
Here are some additional examples:
- Levels of Evidence from Oxford Centre for Evidence-Based Medicine (March 2009)
- Hierarchy of Evidence from Duke University Medical Center Library & Archives
Primary vs. Secondary Literature
- Primary (unappraised) literature: Includes original individual studies, such as controlled trials, cohort studies, and case studies.
- Secondary (pre-appraised) literature: Analyzes and interprets groups of primary studies, including systematic reviews and meta-analyses.
Secondary Literature Study Designs
- Systematic Review: Brings together and distills the best evidence from the primary literature to answer a clinical question. Generally, this will pool the results of several RCTs or meta-analyses on the same clinical problem.
- Meta-analysis: A quantitative statistical analysis of several separate but similar experiments or studies in order to test the pooled data for statistical significance.
Key Primary Study Designs
- Randomized Controlled Study: A carefully planned experiment that studies the effect of therapy on real patients. Randomized controlled trials (RCTs) include methodologies (randomization and blinding) that reduce bias and that allow for comparison between an intervention group and a control group (no intervention). RCTs can provide sound evidence for cause and effect.
- Cohort Study: Follows patients who have a particular condition or receive a particular treatment over time, and compares them with another group who have not been affected by the condition or treatment being studied. Not as reliable as an RCT since the two groups might differ in ways other than the variable being studied.
- Case-Control Study: A study in which people who already have a condition are compared with those who do not. The researcher looks back over time to identify factors that might be associated with the condition. Often relies on medical data or patient recall and is less reliable than an RCT or cohort study because cause and effect is not necessarily established.
- Case Report: A report on the treatment of an individual patient. Because there is no control group for comparison, there is no statistical validity. A number of case reports is a Case Series.
Caveat
Level of evidence hierarchies assume the studies were conducted according to the best practices for the particular study design. If a study's methodology does not follow best practices, the level of evidence for that study will be lower. This means a well-conducted, rigorous cohort study could provide better quality evidence than a poorly conducted randomized controlled trial. This is one reason Appraise is an important step in the Evidence-Based Practice process.
Questions to Consider When Appraising Nursing Studies
1. Is the research approach appropriate?
2. Was the protection of human subjects considered?
3. Are the details of data collection clearly and logically presented?
4. Are the instrument(s) appropriate for the study both in terms of the problem and the approach?
5. Are the instrument(s) described sufficiently in terms of content, structure, validity and reliability?
6. Is the population and the method for selecting the sample adequately described?
7. Is the method for selection of the sample appropriate?
8. Is the sample size sufficient?
9. Is attrition of sample reported and explained?
10. Does the design have controls at an acceptable level for the threats to internal validity?
11. What are the limits to generalizability in terms of external validity?
Schmidt, N. A. (2012). Evidence-based practice for nurses: appraisal and application of research. 2nd ed. Sudbury, MA: Jones & Bartlett Learning.