Evidence-based child health

Rosalind L. Smyth · Forfar & Arneil 7th Edition | EBM, Systematic reviews, RCTs, Likelihood ratios, Bayes, Diagnostic tests, Therapy, Cochrane Library
📌 Core principles: Integration of best research evidence with clinical expertise & patient values. Hierarchy of evidence (SR of RCTs > RCT > cohort > case-control). Likelihood ratios & Bayes theorem guide diagnostic reasoning. Systematic reviews (Cochrane) reduce bias.

📖 Evidence-Based Child Health: Core Concepts

🔬 Definition (Sackett)
Integration of clinical information from patient + best clinical research evidence + clinical experience. Art of medicine and objective science.
📐 Likelihood Ratio (LR)
LR+ = sensitivity/(1-specificity). LR- = (1-sensitivity)/specificity. Post-test odds = pretest odds × LR. Helps interpret diagnostic tests.
📚 Hierarchy of evidence
1a: SR of RCTs; 1b: individual RCT (narrow CI); 2a: SR of cohort; 3: case-control; 4: case-series; 5: expert opinion.
🔄 Systematic Review (Cochrane)
Rigorous methodology: prospective protocol, comprehensive search, quality assessment, meta-analysis if appropriate. Reduces bias vs narrative reviews.
🧪 Bayesian thinking
Prior probability → updated with evidence → posterior probability. Natural framework for clinical decision making (diagnosis & therapy).
📈 Key example: Celiac disease – pretest prob 10% → IgA-TTG LR+ 94 → post-test prob 91% → biopsy indicated. Negative LR- 0.06 → prob 0.7% → biopsy avoided.

🔍 Using EBM to approach a clinical symptom: 5-step framework

1
Formulate an answerable clinical question – PICO: Population (child with cough), Intervention (antibiotic), Comparison (placebo), Outcome (pneumonia resolution).
2
Search for best evidence – Cochrane Library, PubMed/MEDLINE, TRIP database, clinical guidelines. Use systematic reviews when available.
3
Critically appraise evidence – For therapy: RCT blinding, allocation concealment, intention-to-treat, effect size. For diagnosis: independent blind comparison to gold standard, spectrum of patients, reproducibility.
4
Apply to patient – Is the evidence applicable? Is my patient similar to study population? Assess benefits vs harms, patient values.
5
Evaluate performance – Monitor outcomes, update knowledge, reflect on clinical decision.
📌 Clinical pearl: Likelihood ratios are more informative than sensitivity/specificity alone. Use Fagan nomogram or online calculators to convert pretest to post-test probability.

🩺 Stepwise management: integrating EBM into child health decisions

1
Identify evidence gap – Ask focused question (e.g., “Does asthma education delivered by nurse reduce ED visits?”).
2
Access pre-appraised resources – Cochrane systematic reviews, Clinical Evidence, AAP/ SIGN guidelines.
3
Critique study design & bias – Level 1 evidence (RCT/SR) preferred. For therapy: concealment, blinding, completeness of follow-up.
4
Calculate NNT/NNH if possible – Number needed to treat = 1/ARR (absolute risk reduction). Example: if ARR=10% → NNT=10.
5
Apply shared decision-making – Present benefits/risks to family, incorporate child’s preferences, consider feasibility and costs.
6
Reassess & audit – Evaluate clinical outcomes, update practice as new high-level evidence emerges (living systematic reviews).
⚠️ Common pitfalls: Relying on narrative reviews, ignoring confidence intervals, overinterpreting p-values, therapeutic misconception, publication bias. Always check for heterogeneity in meta-analysis.

🧠 Reflex prompts: evidence-based child health in action

📊 You order a diagnostic test with LR+ = 20. Pretest probability = 30% → post-test probability?
Pretest odds = 0.3/0.7=0.43 → post-test odds = 0.43×20=8.6 → probability ≈ 90% → strong rule-in.
📉 A systematic review of RCTs shows heterogeneity (I²=85%). Next step?
Explore sources of heterogeneity (different populations, interventions), consider random-effects model, avoid pooling if clinically diverse.
🎯 Which study design best answers therapy question?
Randomized controlled trial (RCT) with concealment, blinding, and intention-to-treat analysis. Systematic review of RCTs provides highest level.
📚 “All-or-none” evidence (Level 1c)
Example: when all patients died before therapy, now some survive; or some died before, now none die. Very powerful observational evidence.
🔍 Sensitivity vs specificity – which rules out disease?
High sensitivity test (SnNout): negative result rules OUT disease. High specificity test (SpPin): positive result rules IN disease.
📌 Cochrane Library key features
2900+ systematic reviews (2006), updated quarterly, minimizes bias. Cochrane Child Health Field supports pediatric reviews.
⚠️ Why are many pediatric RCTs underpowered?
Smaller disease incidence, ethical constraints, heterogeneity of age, lack of validated pediatric outcome measures. Call for multicenter trials and networks (MCRN).
📈 Bayesian reasoning in clinical history
Each piece of history updates disease probability. Like “Dr Jenkins’ hunch” – mother’s word “strangely” raised pretest probability for meningococcal sepsis.