Evidence-Based Practice, Research, and Quality Improvement
This domain tests whether you can appraise a study, translate findings into practice, and measure whether your own service is performing. Expect questions framed as a clinical scenario plus a research or quality problem.
Appraisal fundamentals
- Formulate answerable questions with PICO: population, intervention, comparison, outcome.
- Evidence hierarchies place systematic reviews and meta-analyses of randomized trials above single trials, cohort studies, case-control studies, and expert opinion. GRADE separates the certainty of evidence from the strength of a recommendation.
- Intention-to-treat analysis preserves randomization but dilutes effect size; per-protocol analysis risks bias from attrition.
- Check allocation concealment, blinding, and confounding before trusting an observational association.
- Heterogeneity across pooled trials is quantified with the I² statistic; high values warrant caution about pooling.
Diagnostic and outcome statistics
- Sensitivity = true positives / (true positives + false negatives); specificity = true negatives / (true negatives + false positives).
- Positive predictive value rises with prevalence; sensitivity and specificity do not.
- Absolute risk reduction is the difference in event rates; number needed to treat = 1 / absolute risk reduction.
- Confidence intervals convey precision; a p-value alone does not convey clinical importance.
Quality improvement
- The Model for Improvement uses three questions plus Plan-Do-Study-Act cycles. Track structure, process, and outcome measures in the Donabedian framework.
- Run charts and statistical process control distinguish real change from common-cause variation; a shift of six points above the median signals special-cause variation.
- Standards of care include nutrition screening within 24 hours of admission and documented reassessment; audits and root cause analyses guide correction.
Worked example
A trial reports mortality of 20% in the control arm and 15% in the treatment arm, with an absolute risk reduction of 5% and a number needed to treat of 20. If the 95% confidence interval crosses zero, the finding is not statistically significant even though the point estimate sounds impressive, so appraise the interval before changing practice.
Traps
Confusing statistical with clinical significance; treating surrogate markers as patient outcomes; assuming a screening test's positive predictive value is fixed; and claiming quality improvement success from a single data point.