Evidence Based Medicine

1. Evidence Based Medicine Dr Chris Martin

2. Objectives Understand the basic structure of a research study and its analysis. Be able to read and evaluate a research paper. Be able to find evidence.

3. Structure of research studies Quantitative v qualitative studies Questions - null hypothesis, Bias and the methods to minimize it: confounding; controlling; randomization; sample size. Statistical tests, �p� values, �95% confidence intervals�.

4. Questions Are Cox IIs better than standard NSAIDs? How do you stop people attending A&E inappropriately? Why do GPs manage shingles so badly? Do statins reduce the risk of death from CVD one year after a first MI?

5. Null hypothesis Question: Do statins reduce the risk of death from CVD one year after a first MI? Null hypothesis: Statins do not increase the risk of CVD death one year after a first MI.

6. Task 1 What is the corresponding null hypothesis to this research question? Does new drug A reduce the severity and duration of vertigo?

7. Task 1 answer New drug A does not reduce the duration or severity of vertigo.

8. Types of research Quantitative Qualitative

9. Quantitative research Answers questions involving numerically measureable outcomes that can be statistically analyzed. Does drug A reduce the severity and duration of vertigo?

10. Qualitative research Answers subjective questions that explore the meaning or our understanding of certain problems. �What are the barriers to the adoption of guidelines for managing heart failure in secondary care?�

11. Quantitative study design Randomized controlled trials. Case-control studies. Crossover trials. Before and after studies. Surveys Cohort studies.

12. Qualitative studies Semi-structured interviews Focus groups Textual analysis Observational studies Questionnaires

13. Controlled This means that the intervention is compared to something � usually a placebo or �normal treatment�.

14. Randomization Reduces selection bias Evenly distributes confounding factors between intervention and control groups.

15. Bias Systematic errors in a research study that lead to a distortion of the results of the study. Observer bias (eliminated by blinding) Sampling bias (eliminated by randomization) Publication bias (eliminated by publishing everything) .....

16. Confounding The situation when both the postulated �cause� and �effect� under investigation are correlated with a third (possibly unknown) factor that gives the erroneous impression that the postulated �cause� leads to the �effect�. Does wearing a skirt cause breast cancer?

17. Statistical significance Statistical test Any mathematical procedure that determines the likelihood of a difference between two outcome measures is not due to chance.

18. �p� The result usually expressed as �p� � the probability that the difference is due to chance. The lower �p� is the more likely it is a difference reflects a real effect and not chance. By convention a result is considered statistically significant if p<0.05 (the probability of the effect being due to chance is less than 5%).

19. Confidence interval (CI) The range of values that would contain 95% of all results that occurred by chance. Think of it as the range of results that would contain the central 95% of results if repeated and experiment 100 times. Example: the average reduction in duration of pain in those subjects taking aspirin was 0.5 hours (95%CI: -0.1 to2.6)

20. Sample size Increased sample size increases the precision of estimation of effect size - the bigger the number, the greater the chance of finding a small effect. Increased sample size reduces the likelihood that random effects bias the results.

21. Sample size calculation Decide on the minimum clinically significant difference. Decide on the level of significance (p value) http://statpages.org/proppowr.html

22. Lighten up!