Research Methodology, Statistics and Evidence-Based Practice
Core concepts
Study designs: RCTs (randomization, blinding, intention-to-treat), cohort, case-control, cross-sectional, and ecological. Understand bias types (selection, information, recall, confounding, publication bias) and how to control them. Measures of association: relative risk (cohort), odds ratio (case-control), absolute risk reduction, number needed to treat. Diagnostic test metrics: sensitivity, specificity, positive and negative predictive values, likelihood ratios, and how prevalence affects PPV. Understand p-values, confidence intervals, statistical significance vs clinical significance, type I and II errors, and power. Know the levels of evidence hierarchy and GRADE. Understand systematic reviews, meta-analysis, heterogeneity, and forest plots. Appraise a paper: validity, results, applicability.
Traps
PPV depends on prevalence - a highly sensitive and specific test can still have low PPV in a low-prevalence population. Odds ratio approximates relative risk only when the outcome is rare. Statistical significance is not clinical significance. Intention-to-treat analysis preserves randomization but may dilute effect size.
Rule of thumb
For a diagnostic test, remember: sensitivity rules out (SnNout), specificity rules in (SpPin). For therapy questions, calculate NNT = 1/ARR.