10-Day USMLE Biostatistics & Epidemiology Study Plan

How to Use This Plan

• Daily time: 3-4 hours focused study
• Resources: First Aid (Biostatistics chapter), Sketchy/Dirty Medicine videos, UWorld question blocks (filter by Biostatistics/Epidemiology), and Park's Textbook of Preventive and Social Medicine for deeper reading
• Rule: Do 10-20 practice questions at the end of each day on that day's topic

Day 1 - Study Design Foundations

• Types of studies: observational vs. experimental
• Descriptive studies: case report, case series, cross-sectional
• Analytical observational studies:
  • Cohort (prospective and retrospective): follows exposed vs. unexposed
  • Case-control (retrospective): starts with outcome, looks back at exposure
  • Cross-sectional: prevalence snapshot
• Experimental studies: RCT (gold standard), community trial, crossover trial
• Ecological study (unit = population, not individual)
• When to use each design - classic USMLE vignette triggers

• Case-control → calculate Odds Ratio (OR)
• Cohort → calculate Relative Risk (RR)
• Cross-sectional → prevalence; cannot establish causation
• RCT → only study that can establish causation
• Retrospective cohort ≠ case-control (still calculates RR)

Day 2 - Measures of Disease Frequency

• Incidence = new cases / population at risk / time
• Prevalence = all existing cases / total population at a point in time
• Relationship: Prevalence ≈ Incidence × Duration
• Attack rate (outbreak setting)
• Mortality rate vs. morbidity rate
• Crude vs. specific vs. adjusted rates
• Standardized mortality ratio (SMR)
• Years of potential life lost (YPLL)
• Disability-adjusted life years (DALY) and Quality-adjusted life years (QALY)

• A new treatment that prolongs survival without curing will increase prevalence without changing incidence
• Prevalence > incidence for chronic diseases; prevalence ≈ incidence for acute diseases
• Emigration of sick patients → prevalence decreases (doesn't mean disease is less common)

Day 3 - Measures of Association & Risk

• Relative Risk (RR) = incidence in exposed / incidence in unexposed
• Odds Ratio (OR) = (a×d) / (b×c) from a 2×2 table
• When OR approximates RR: rare disease assumption
• Absolute Risk Reduction (ARR) = risk in control - risk in treatment
• Relative Risk Reduction (RRR) = ARR / risk in control = 1 - RR
• Number Needed to Treat (NNT) = 1 / ARR
• Number Needed to Harm (NNH) = 1 / Attributable Risk Increase
• Attributable Risk (AR) = incidence in exposed - incidence in unexposed
• Population Attributable Risk (PAR)
• 2×2 contingency table mastery - build and fill from any vignette

• RR = 1 → no association; RR > 1 → risk factor; RR < 1 → protective
• NNT and NNH are the most clinically relevant measures for Step 2/3 but appear on Step 1
• OR overestimates RR when disease is NOT rare

Day 4 - Screening, Sensitivity & Specificity

• Sensitivity (SnOut) = TP / (TP + FN) → rules OUT disease if negative
• Specificity (SpIn) = TN / (TN + FP) → rules IN disease if positive
• Positive Predictive Value (PPV) = TP / (TP + FP) → affected by prevalence
• Negative Predictive Value (NPV) = TN / (TN + FN) → affected by prevalence
• Sensitivity vs. specificity trade-off (ROC curve)
• Effect of changing cutoff threshold on sensitivity/specificity
• Effect of disease prevalence on PPV and NPV
• Likelihood ratios (LR+ and LR-)
• Criteria for a good screening test (Wilson & Jungner criteria)
• Serial vs. parallel testing:
  • Parallel → increases sensitivity and NPV (rules out)
  • Serial → increases specificity and PPV (rules in)

• PPV increases with higher prevalence; NPV decreases with higher prevalence
• Sensitivity and specificity are fixed properties of the test (not affected by prevalence)
• "Best screening test" = high sensitivity; "best confirmatory test" = high specificity

Day 5 - Bias, Confounding & Validity

• Selection bias: sampling bias, Berkson's bias (hospital-based controls), healthy worker effect, non-response bias, loss to follow-up bias
• Information (measurement) bias:
  • Recall bias (case-control studies)
  • Observer/interviewer bias
  • Hawthorne effect
  • Lead-time bias, length-time bias (screening studies)
• Confounding: a third variable associated with both exposure and outcome
• Methods to control confounding: randomization, restriction, matching, stratification, multivariate analysis
• Effect modification (interaction): different from confounding; do NOT adjust
• Internal validity vs. external validity (generalizability)
• Placebo effect, nocebo effect
• Ecological fallacy

• Lead-time bias → appears to improve survival but only because diagnosis is earlier
• Length-time bias → screening catches slower-growing (less aggressive) tumors, making screened patients appear to survive longer
• Recall bias is biggest threat in case-control studies
• Randomization is the best way to control for both known and unknown confounders

Day 6 - Statistical Tests & Hypothesis Testing

• Null hypothesis (H₀) and alternative hypothesis (H₁)
• Type I error (α): rejecting a true null (false positive) - probability = p-value threshold (0.05)
• Type II error (β): failing to reject a false null (false negative)
• Power = 1 - β (probability of correctly detecting a true effect)
• Factors that increase power: larger sample size, larger effect size, higher event rate, lower variability, higher α threshold
• p-value: probability of getting results as extreme as observed if H₀ is true; p < 0.05 = statistically significant
• Confidence Interval (CI): if 95% CI for RR or OR includes 1 → not significant; if CI for difference includes 0 → not significant
• Choosing the right statistical test:
  • t-test: compare means of 2 groups (continuous data)
  • ANOVA: compare means of ≥3 groups
  • Chi-square: compare proportions (categorical data)
  • Pearson correlation: linear relationship between 2 continuous variables
  • Meta-analysis: pools results from multiple studies

• Narrow CI = more precise; wide CI = less precise (small sample)
• Statistical significance ≠ clinical significance
• A very large study can have statistically significant but clinically meaningless results

Day 7 - Prevention, Disease Control & Surveillance

• Levels of prevention:
  • Primary: prevent disease before it occurs (vaccination, lifestyle)
  • Secondary: early detection and treatment (screening, Pap smear)
  • Tertiary: reduce disability from established disease (rehab, dialysis)
• Primordial prevention: prevent risk factors from emerging (societal level)
• Herd immunity and threshold calculations
• Quarantine vs. isolation
• Notifiable diseases and surveillance systems
• Epidemic curve shapes: common source (point, continuous, intermittent) vs. propagated
• Outbreak investigation steps (USMLE loves this sequence)
• Endemic, epidemic, pandemic definitions
• Zoonoses and transmission types
• Vaccine schedules and types (live attenuated vs. inactivated)
• Cold chain concept

• Herd immunity threshold = 1 - (1/R₀)
• Point source epidemic → sharp peak, all cases within 1 incubation period
• Propagated epidemic → multiple peaks, increasing over time

Day 8 - Clinical Epidemiology & Evidence-Based Medicine

• Hierarchy of evidence: Systematic review/meta-analysis > RCT > Cohort > Case-control > Cross-sectional > Case report
• Meta-analysis: forest plots, heterogeneity (I² statistic), funnel plots for publication bias
• Systematic review vs. narrative review
• Clinical practice guidelines development
• Absolute vs. relative risk in communicating results
• Pre-test and post-test probability (Bayes' theorem conceptually)
• Intention-to-treat (ITT) vs. per-protocol analysis
• Blinding: single, double, triple-blind
• Allocation concealment vs. randomization
• Crossover design: each subject is their own control
• Factorial design
• Surrogate endpoints vs. clinical endpoints

• ITT analysis: all randomized patients analyzed in their original group (even if they crossed over); minimizes bias
• Double-blind: both patient AND investigator are blinded
• Funnel plot asymmetry → publication bias
• Forest plot: diamond crossing the line → not significant

Day 9 - Special Topics: Clinical Decision Making & Pharmacoepidemiology

• Decision analysis: decision trees, expected utility
• Cost-effectiveness analysis vs. cost-benefit analysis
• Pharmacovigilance: phase I-IV clinical trials
  • Phase I: safety and dosage (healthy volunteers, small)
  • Phase II: efficacy and side effects (patients, medium)
  • Phase III: comparison to current standard, large RCT
  • Phase IV: post-marketing surveillance (rare adverse effects)
• Post-marketing surveillance and reporting systems (MedWatch)
• Drug safety: FDA pregnancy categories (historical), black box warnings
• Survival analysis: Kaplan-Meier curves, log-rank test, hazard ratio
• Censoring in survival analysis
• Cox proportional hazards model
• Regression to the mean
• Standardization: direct vs. indirect
• Matching in case-control studies

• Phase III → pivotal trial for FDA approval
• Phase IV catches rare adverse effects (e.g., agranulocytosis with clozapine identified post-approval)
• Kaplan-Meier curves: when lines diverge and stay apart = durable benefit; crossing lines = effect reversal

Day 10 - Full Integration, High-Yield Review & Practice Block

• All formulas on one sheet: sensitivity, specificity, PPV, NPV, RR, OR, ARR, NNT, attributable risk
• 2×2 table practice from memory - build 5 tables
• Quick mnemonic review: SnNout, SpPin, Lead-time vs. length-time bias

• Prevalence ↑ → PPV ↑, NPV ↓ (most commonly tested)
• A screening program is started → measured survival improves even without treatment benefit (lead-time bias)
• Cohort in a hospital → Berkson's bias
• Workers appear healthier than general population → healthy worker effect
• OR vs. RR: OR in case-control; RR in cohort; OR approximates RR only for rare disease
• p = 0.04 and CI 95% [0.3 to 2.1] for OR → which takes priority? The CI (includes 1 → not significant)

• 40-question mixed biostatistics + epidemiology UWorld block (timed)
• Full review of every wrong answer
• Note any formula or concept you missed

• Write your own 1-page cheat sheet from memory
• List every bias with its classic study setting
• List every statistical test with its data type

Topic Priority Summary

Daily Schedule Template

Recommended Resources by Day

• First Aid for USMLE Step 1 - Biostatistics & Epidemiology chapter (primary quick reference)
• Park's Preventive and Social Medicine - deeper reading for Day 7 (epidemiology/surveillance)
• UWorld - filter by Behavioral Sciences > Biostatistics after each day
• Sketchy Biostats or Boards & Beyond Biostats - video reinforcement for Days 3-6
• Anki - use premade USMLE biostats deck (Zanki or AnKing) throughout