Evidence-Based Practice, Research Methods, and Biostatistics
What this domain tests
Critical appraisal, study design, and interpretation of diagnostic and treatment statistics.
Core rules
- PICO frames a clinical question: population, intervention, comparator, outcome.
- Evidence hierarchy: systematic reviews and RCTs above cohort, case–control, cross-sectional, and case reports.
- Randomization, allocation concealment, blinding, and intention-to-treat reduce bias.
- Sensitivity = true positives divided by all diseased. Specificity = true negatives divided by all healthy. PPV and NPV depend on prevalence.
- A p-value indicates compatibility with the null hypothesis, not the size or importance of an effect. Confidence intervals show precision.
- NNT = 1 divided by absolute risk reduction.
- Confounding, selection bias, and publication bias can distort conclusions.
Worked example
A periodontal screening test has sensitivity 90% and specificity 80%. In 1,000 patients with 10% prevalence: 100 have disease, 90 test positive; 900 are healthy, 180 test false-positive. PPV = 90/(90+180) = 33%. In a high-prevalence clinic, PPV rises. This illustrates why a good test can mislead in a low-prevalence population.
Traps
- Treating a statistically significant result as clinically important.
- Confusing sensitivity with PPV.
- Assuming association from a cross-sectional study proves causation.
- Ignoring confidence intervals and relying only on p-values.