Skip Navigation Bar
 

National Information Center on Health Services Research and Health Care Technology (NICHSR)

spacer
spacer
blue arrow
blue arrow
blue arrow
blue arrow
blue arrow
blue arrow
blue arrow
blue arrow
blue arrow
blue arrow
blue arrow
blue arrow
blue arrow
blue arrow
spacer
blue arrow facing left Previous   Next blue arrow

Introduction to Health Services Research : A Self-Study Course

Module 5: Quality Filtering and Evidence-Based Medicine and Health (page 10 of 15)
Introduction | Sampling | Assignment | Assessment | Analysis | Interpretation | Extrapolation

 

Interpretation (conclusions)

Interpretation concerns the conclusions made about the people in the study.

Interpretation concerns the conclusions made about the people in the study.

Researchers assess the strength of the association between variables (indicated by the relative risk or other measures) and the cause and effect relationship between them. Does Treatment X provide substantially more patient satisfaction and effectiveness than Treatment Y? Does short LOS cause hospital readmission? These are the types of questions considered under interpretation.

Causal Relationships

Several factors support a causal relationship. You can have more confidence assuming causation if:

  • The risk factor occurs more often in people with the specific outcome. For example, short LOS occurs primarily in those with a higher rate of hospital readmission.
  • The risk factor being studied precedes the effect. For example, smoking occurs before the development of lung cancer.
  • Changes in the risk factor produce the effect. For example, changing the LOS will change the likelihood of hospital readmission.

Other Supporting Evidence for Causation

Other supporting evidence for causation includes the:

  • Strength of the association between the factors such as LOS and hospital readmission, measured by the relative risk or other estimates.
  • Consistency of the association as demonstrated when similar outcomes are measured in different groups in different settings.
  • Biologic plausibility (in clinical studies) demonstrated by the scientific evidence that a particular factor can cause an outcome. For example, there is strong biologic evidence that cigarettes cause lung cancer whereas there is weak biologic evidence that electromagnetic fields cause cancers. Remember, however, that scientists may not be able to explain the physical foundations of a relationship that epidemiologists uncover!
  • Dose-response relationship which implies that higher levels of a risk factor will contribute more to the development of an outcome than a lower level. For example, shorter LOS would lead to higher rates of hospital readmissions, or more packs of cigarettes smoked would increase the likelihood of developing lung cancer. Evaluating dose-response relationships may be tricky, however, since some effects may increase at intermediate doses and decline at higher doses.

Evaluating Factors in Relationships Described in Articles

Evaluating all of these factors help ascertain what type of relationship exists between the variables (the risk factors and outcome). Notice how completely the researchers evaluate these criteria before they assert a conclusion at the end of their reports.

Discussion Questions

  1. As you are reading over a research paper, can you determine causal relationships before the author draws them to your attention? What clues are available from the methods used and results described in the paper?
  2. Review the section on causation above. Do a Google or other search engine search on the term "dose-response relationship." Which disciplines are represented by the hits that turn up? Did you find any articles you wanted to stop and look at? Discuss.
blue arrow facing left Previous Knowledge Check
Q15 | Q16
Next blue arrow
spacer
spacer