Public Health & Biostatistics System
Public Health & Biostatistics
High-yield public health and biostatistics for USMLE Step 2/3 — confounding, bias, study design, cognitive errors, perioperative medicine, quality improvement, ethics, and military medicine, extracted from Divine Intervention and organized for exam performance.
Biostatistics Foundations
Confounding and effect modification are the two most tested statistical concepts after sensitivity/specificity — master the three criteria for a confounder and the stratification response pattern to nail any study-design question. Hardy-Weinberg and drug-ad strategy are the remaining high-yield formulas that appear reliably on every administration.
EP363
Taking the Confusion Out of Confounding
- Confounder definition: A variable that is associated with BOTH the exposure AND the outcome, and is NOT mechanistically caused by the exposure — all three criteria must be met
- Classic example: Lighter use → lung cancer (confounder = smoking; smoking correlates with lighters AND causes lung cancer; carrying a lighter doesn't make you a smoker)
- Stratification test: When you stratify by the confounder, the apparent association disappears in one subgroup — this signals confounding, not causation
- Pre-test controls: Restriction (narrow inclusion criteria, limits generalizability) and Matching (case-control pairing); both reduce but complicate study enrollment
- Randomization is the gold standard pre-test intervention — randomly distributes confounders across study arms, making groups comparable
- Post-test control: Stratified analysis reveals the hidden confounder by showing equal risk within each stratum when confounding is present
Three Criteria for a Confounder (Memorize These)
- Criterion 1: The confounder must be associated with the exposure
- Criterion 2: The confounder must be associated with the outcome
- Criterion 3: The exposure must NOT be a mechanistic cause of the confounder (directionality check)
Classic USMLE Confounder Examples
Lighters → Lung Cancer: Confounder = smoking (smokers carry lighters; smoking causes lung cancer; lighters don't make you a smoker).
Age → Obesity → CVD: Age confounds the relationship; obesity does not mechanistically cause age.
Birth order → Prader-Willi: Confounder = paternal age (older fathers have more children AND higher Prader-Willi risk).
Flu vaccine → Flu: Confounder = being a healthcare worker (HCWs get vaccinated more AND have more flu exposure).
Stratification — How to Identify Confounding on an NBME Question
They will show you a subgroup analysis. Look for two patterns that signal confounding:
- Relative risk in one subgroup drops precipitously (non-smokers: RR ~1; smokers: RR ~20) — the effect disappears in the subgroup without the confounder
- Within a single subgroup, risk is equal regardless of the exposure level (e.g., non-smokers have same lung cancer risk whether they carry 1 or 5 lighters)
Pre-Test vs Post-Test Interventions
| Timing | Method | How It Works | Key Limitation |
|---|---|---|---|
| Pre-test | Restriction | Narrow inclusion criteria to exclude confounders | Reduces sample size; limits generalizability |
| Pre-test | Matching | Pair cases/controls on confounder variables | Hard to find matching subjects; same generalizability limits |
| Pre-test | Randomization | Random assignment distributes confounders evenly | Requires large N to work; not always feasible |
| Post-test | Stratification | Analyze subgroups separately by confounder | Must know the confounder in advance to stratify |
| Post-test | Multivariate analysis | Statistical adjustment for multiple variables | Cannot adjust for unknown confounders |
Why "Correlation Does Not Imply Causation"
Because an unmeasured confounder may explain the correlation. Confounders you have not measured cannot be stratified or adjusted — this is why observational studies cannot establish causation, only RCTs can.
EP364
Effect Modification
- Effect modifier: A variable associated with the OUTCOME ONLY — not associated with the exposure (unlike a confounder, which is associated with both)
- Classic example: Alcohol → esophageal cancer (effect modifier = smoking; smoking is not caused by drinking, but smokers + drinkers have dramatically higher risk than drinkers alone)
- Asbestos + smoking: Asbestos alone → 2–3x lung cancer risk; asbestos + smoking → 60x risk. Smoking is the effect modifier — it amplifies the asbestos effect
- Stratification response: With effect modification, the association does NOT disappear in any subgroup. Instead, the strength of association differs between subgroups (strong in smokers, present but weaker in non-smokers)
- Key differentiator vs confounding: In confounding, effect disappears in one stratum. In effect modification, effect persists in all strata but at different magnitudes
Confounding vs Effect Modification — Side-by-Side
| Feature | Confounding | Effect Modification |
|---|---|---|
| Third variable associated with exposure? | Yes | No |
| Third variable associated with outcome? | Yes | Yes |
| Response to stratification | Effect disappears in one subgroup | Effect persists, but magnitude differs between subgroups |
| Goal in analysis | Eliminate or adjust for it | Report it as an important biological interaction |
| Classic USMLE example | Smoking confounds lighter → lung cancer | Smoking modifies asbestos → lung cancer (60x vs 2x) |
Exam Hook — Stratification Results
If stratification makes an effect DISAPPEAR in one group → confounding.
If stratification reveals DIFFERENT STRENGTHS of the same effect → effect modification.
The effect modifier is a biological amplifier — it makes a real relationship even stronger, not a fake one disappear.
Additional Examples of Effect Modification
- Alcohol + ALDH2 deficiency: Asian patients with ALDH2 mutation who drink alcohol get dramatically higher rates of esophageal cancer — ALDH2 deficiency modifies the alcohol → cancer relationship
- Estrogen therapy + BRCA1 mutation: Estrogen exposure has a much stronger effect on breast cancer risk in BRCA1 carriers than in the general population
- UV radiation + xeroderma pigmentosum: UV exposure is harmful to everyone, but the relationship between UV and skin cancer is dramatically amplified in XP patients (DNA repair defect)
EP375
Hardy-Weinberg Made Easy
- Two equations: P + Q = 1 (allele frequencies); P² + 2PQ + Q² = 1 (genotype frequencies). P = dominant allele frequency, Q = recessive allele frequency
- Key rule: If they give you "disease seen in 1 in X individuals" → that is Q² (population frequency of homozygous recessive). Take square root to get Q, then calculate P = 1 − Q
- Carrier frequency = 2PQ. Always multiply by 2 — heterozygotes can be dominant-recessive OR recessive-dominant
- X-linked twist: In males (XY), there is only one X chromosome, so disease frequency = Q directly (no squaring). Only females need Q²
- Multi-step probability: Multiply probabilities: P(dad carrier) × P(mom carrier) × P(1/4 child affected) — systematic approach prevents errors
- Hardy-Weinberg assumptions: No mutation, no natural selection, no migration, random mating, large population — violations of any of these invalidate the equilibrium
The Two Hardy-Weinberg Equations
| Term | Meaning | When Given This in the Question |
|---|---|---|
| P | Frequency of dominant allele in the population | Rare — usually calculate from Q |
| Q | Frequency of recessive allele in the population | Rare — calculate from Q² |
| P² | Frequency of homozygous dominant individuals | Usually not given |
| 2PQ | Carrier frequency (heterozygotes) | This is what you usually need to find |
| Q² | Frequency of affected individuals (homozygous recessive) | This is what questions usually GIVE you ("disease in 1/X") |
Step-by-Step Worked Example — Hemophilia C in Ashkenazi Jews
Given: Disease seen in 1/25 Ashkenazi Jews. Find carrier frequency.
Step 1: Disease frequency = Q² = 1/25
Step 2: Q = √(1/25) = 1/5 = 0.2
Step 3: P = 1 − Q = 1 − 1/5 = 4/5 = 0.8
Step 4: Carrier frequency = 2PQ = 2 × (4/5) × (1/5) = 8/25 = 0.32
Multi-Generation Probability Problem
When asked "what is the chance that this unaffected son's child will be affected if he marries someone from the general population?":
- Dad's carrier probability: He is unaffected. Parents are both carriers (Aa × Aa). Non-affected offspring can be AA (1/4) or Aa (2/4). He must be unaffected, so eliminate aa. Of 3 remaining possibilities, 2 are carriers → P(dad carrier) = 2/3
- Mom's carrier probability: Unknown family history → use population carrier frequency = 8/25
- Child affected: If both parents are carriers, P(child affected) = 1/4
- Final answer: (2/3) × (8/25) × (1/4) = 16/300 = 4/75
X-Linked Recessive — Important Modification
- Males have only one X chromosome, so disease frequency in males = Q (not Q²)
- Females still need Q² (they have two X chromosomes and need both recessive alleles to be affected)
- This is why X-linked recessive diseases are much more common in males than females
EP337
How to Successfully Answer Drug Ad Questions
- Save for last: Drug ad questions have 2–3 questions attached; 38-question blocks signal a drug ad; do all other questions first to preserve time
- Read questions FIRST, then go straight to the graphs and charts — 90% of answers come from the visual data, not the text
- Assess significance immediately: For ratio-based outcomes (RR, OR), CI crosses 1 = not significant. For difference-based outcomes (means), CI crosses 0 = not significant
- Differential reading: Note what is different between arms — different sample sizes, dropout rates, duration. These are likely question hooks (intention-to-treat, power)
- Publication bias: Studies with positive results are more likely to be published — always a valid answer choice for "what limits confidence in these results?"
- Acceptable loss strategy: If drug ads cause severe anxiety, guess C or D on all and move on — they represent less than 2% of the exam
10 Rules for Drug Ad Questions
- Rule 1: Save for last — 38-question block = drug ad signal. Never do these first
- Rule 2: Drug ads are not the end of the world — at most 6 questions, <2% of the exam
- Rule 3: You have ~6 minutes total for the drug ad and its questions in a 38-item block
- Rule 4: Read the questions BEFORE reading the drug ad — gives you a focused lens
- Rule 5: Go straight to charts and graphs — 90% of answers live here
- Rule 6: Immediately determine significance from confidence intervals
- Rule 7: Good at biostats → you will crush drug ads. These are not separate skills
- Rule 8: Read with differential thinking — compare groups for numerical differences
- Rule 9: Key data points: p-values, confidence intervals, graphs, exclusion criteria, sample sizes
- Rule 10: Know your biases — publication bias, selection bias, intention-to-treat violations are common question hooks
Confidence Interval Rules
Ratios (RR, OR, HR): If 95% CI includes 1 → results NOT statistically significant.
Differences (means, proportions): If 95% CI includes 0 → results NOT statistically significant.
When they give you RBR (relative benefit reduction): This = relative risk reduction. Calculated as 1 − relative risk. If CI crosses 1 (for ratios), still not significant despite the calculated number.
Common Drug Ad Question Types
| Question Type | Where to Find Answer |
|---|---|
| What to tell patient based on study results | Check significance first — if not significant, "no clear difference" |
| What limits confidence in conclusions | Short study duration, exclusion criteria, small N, heterogeneous sample sizes |
| What biases the results | Publication bias (positive results published more), selection bias, attrition bias |
| Intention-to-treat vs per-protocol | Look for dropouts — ITT includes all randomized; per-protocol only completers |
EP520
Free 120 Discussion — Biostats, Clinical Decision-Making & Ethics
- Learning disorder: Struggles in ONE specific academic domain (reading, math, writing) with otherwise normal cognitive function and age-appropriate IQ — not global impairment
- Intellectual developmental disorder: Low IQ (<70) + pervasive impairment across multiple domains from early development — not one-subject weakness
- C. diff hand hygiene: Wash with SOAP AND WATER — alcohol gel does NOT kill C. diff or rotavirus spores. This is tested repeatedly
- Organ donation without advance directive: Use surrogate decision-maker hierarchy — spouse → adult children → parents → adult siblings. No advance directive does NOT preclude donation
- Publication bias in meta-analyses: Studies with positive results are more likely to be published than null results, skewing pooled estimates — recognize this as the correct answer for "what biases these results?"
- Non-significant CI rule: If CI crosses 1 for ratios — even if the point estimate looks impressive — the result is NOT significant. Do not make clinical recommendations from non-significant data
Infection Control — Hand Hygiene Rules
| Pathogen | Hand Hygiene | Why |
|---|---|---|
| C. diff (Clostridioides difficile) | Soap and water ONLY | Spore-forming — alcohol doesn't kill spores |
| Rotavirus | Soap and water ONLY | Non-enveloped virus — alcohol less effective |
| Most other pathogens | Alcohol gel acceptable | Enveloped viruses and non-spore formers killed by alcohol |
Surrogate Decision-Making Hierarchy
- Patient's own previously expressed wishes (advance directive, living will) — highest priority
- Healthcare proxy / durable power of attorney for healthcare (if designated)
- Spouse or domestic partner
- Adult children (majority agreement if multiple)
- Parents
- Adult siblings
- Other close family/friends who know patient's wishes
- Court-appointed guardian — last resort
Key Ethics Point — No Advance Directive ≠ No Donation
Absence of an advance directive does not make a patient ineligible for organ donation. The surrogate can express the patient's known wishes. If the wife says "he wanted to be an organ donor," contact the organ bank — option B, not the ethics committee or "not a candidate."
Learning Disorder vs Intellectual Developmental Disorder
| Feature | Learning Disorder | Intellectual Developmental Disorder |
|---|---|---|
| IQ | Normal (≥70) | Below 70 |
| Scope of impairment | One specific academic area | Multiple domains (adaptive functioning, IQ, conceptual) |
| Classic presentation | "Can't read but does math well" | Late developmental milestones, can't bathe/dress independently |
| USMLE hook | Adopted child who reads poorly but has normal receptive language | IQ <70 mentioned explicitly in the question |
Epidemiology, Study Design & Diagnostic Errors
Lead-time and length-time bias are the two most counterintuitive concepts in epidemiology — both make screening look more effective than it is, but for completely different reasons. Diagnostic errors (premature closure, anchoring, confirmation bias) are now explicitly tested on Step 2 and Step 3, and the cognitive error taxonomy overlaps significantly with these heuristics.
EP458
Biases in Screening Tests — Lead Time, Length Time, Selection
- Selection bias: People who seek screening tend to have higher SES, better health literacy, and better access to care — so screened populations appear to do better regardless of the screening test's efficacy
- Lead-time bias: Screening detects disease earlier, so patients appear to "survive longer" after diagnosis — but they still die at the same age. Survival appears improved because the clock starts earlier, not because mortality actually changed
- Length-time bias: Screening preferentially catches indolent (slow-growing) disease because aggressive cancers progress too fast to be detected at a routine screening interval. Survivors of screening appear to do better because you've enriched for less aggressive disease
- Key distinguisher — why does the patient survive longer? Lead time = caught earlier (clock bias). Length time = caught less aggressive form of disease (biology bias)
- Best remedy: Randomized controlled trial with all-cause mortality as the endpoint — randomization equalizes aggressive vs. indolent disease between groups, eliminating length-time bias
Three Screening Biases — Master Table
| Bias | Mechanism | Why Survival Appears Better | Classic Example |
|---|---|---|---|
| Selection bias | Screened patients are healthier / wealthier to start | Better baseline health → better outcomes independent of screening | Wealthier patients get mammograms more; they also eat better |
| Lead-time bias | Disease detected earlier but not cured earlier | Diagnosis-to-death interval appears longer because diagnosis moved earlier; actual death date unchanged | Newborn screening for Duchenne: still dies at 25, but "survival" now looks like 25 years instead of 10 |
| Length-time bias | Slow-progressing variants stay in the "detectable" window longer | You select for biologically favorable disease — the fast-killing variants miss the screening window | Mammogram catches slow-growing breast cancer; aggressive triple-negative progresses between screens |
The One-Question Diagnostic Test for These Biases
Ask: "Why does this screened patient survive longer?"
Lead time: Because we found it sooner — same biology, just an earlier clock start.
Length time: Because we found the slow kind — the biology is inherently more favorable.
If the question says "caught it early" → lead time. If the question says "some variants grow faster" → length time.
Why RCTs Fix These Biases
Randomization puts roughly equal numbers of patients with aggressive and indolent disease into both arms. When you measure all-cause mortality (not survival from diagnosis), you see the true benefit (or lack thereof) of screening — not the artifact of catching favorable biology.
High-Yield Example Walkthrough
Huntington's disease newborn screening: Huntington's patients die at a relatively consistent age regardless of when they become symptomatic. A test that detects it at birth appears to give 40–50 years of "survival after diagnosis." This is entirely lead-time bias — the test did not add a single day of life.
EP275
Diagnostic Errors — Heuristics, Biases & Prevention
- Premature closure: Accepting a diagnosis without confirming it with appropriate testing — "jumping to conclusions." Classic hotspot: patient with exacerbation of a known disorder presents with something different and you don't test further
- Diagnostic momentum: Starting down a diagnostic path and continuing even as evidence accumulates against that diagnosis — almost like the diagnosis is rolling downhill and you can't stop it
- Confirmation bias: Ordering only tests that will CONFIRM your suspected diagnosis, ignoring data that would disprove it
- Availability heuristic: Improperly estimate probability because a diagnosis comes to mind easily (seen recently, frequently). The common pattern shows up as an uncommon disease in the past → you keep seeing that uncommon disease
- Representativeness heuristic: Classic presentation pattern recognized → assume that disease every time without considering other possibilities
- Anchoring heuristic: Settled on an early diagnosis and won't move even as contradictory evidence mounts — like premature closure but ongoing
Diagnostic Error Taxonomy
| Error Type | Core Mechanism | NBME Question Hook |
|---|---|---|
| Premature closure | Accepts diagnosis without confirmatory testing | "Physician did not order any further workup" then patient deteriorated |
| Diagnostic momentum | Keeps pursuing original diagnosis despite contradictory data | Labs and imaging are inconsistent with the diagnosis, but physician continues down that path |
| Confirmation bias | Only orders tests that could confirm the suspected diagnosis | Physician orders tests ONLY for suspected diagnosis X, ignores other possibilities |
| Availability heuristic | Sees a common presentation but calls it a rare disease they saw recently | "Physician saw many cases of [rare disease] recently" then misdiagnoses a common disease as that rare one |
| Representativeness heuristic | Classic pattern → always that disease, every time, without differential | Physician sees classic-looking case and assumes that disease without considering alternatives |
| Anchoring heuristic | First impression sticks despite accumulating contradictory evidence | Early working diagnosis is wrong; lab results don't fit; physician still sticks with it |
Availability vs Representativeness — The Critical Distinction
Availability heuristic: Common pattern was previously a RARE DISEASE → now you see rare disease everywhere.
Example: You had a patient with epigastric pain → back → nausea who turned out to have DKA (not pancreatitis). Now you call every similar case DKA, even though pancreatitis is far more common.
Representativeness heuristic: Classic pattern of disease X → you always call it X, never considering anything else.
Example: Epigastric pain → back → nausea → alcoholic = acute pancreatitis, every time, without ruling out DKA, AAA, etc.
Strategies to Reduce Diagnostic Errors
- Heuristic awareness: Know the three heuristics (availability, representativeness, anchoring) and recognize when you're using one
- Diagnostic timeout: Periodically "start fresh" — especially when patient not improving as expected. Ask: does all data fit my current diagnosis?
- Worst case scenario medicine: Always ask what is the most life-threatening cause first — rule it out, then move down the differential
- Systematic approach: Use checklists and standardized frameworks — reduces cognitive shortcuts that lead to errors
- Ask "why": Don't just label — understand pathophysiology. Reduces anchoring because you think through mechanism, not just pattern
- Bayes theorem: Consider pre-test probability before ordering. High-risk patient with chest pain → CT angiogram directly. Low-risk patient → D-dimer first. Prevents both over-testing and under-testing
Systems Factors That Increase Diagnostic Error Risk
- Pressure to perform (Press Ganey metrics): Rushing through patients in ED → premature closure → more errors
- Fatigue and cognitive overload: Impairs slow, deliberate System 2 thinking; increases reliance on fast, error-prone System 1 heuristics
- EMR-facilitated confirmation bias: Easy to order only the tests that confirm → doesn't prompt you to rule out alternatives
EP530
Cognitive Errors and The USMLEs
- Availability error: Improperly estimate probabilities because a diagnosis more easily comes to mind — due to recent or frequent exposure. "I've seen so much COVID-19 that I keep diagnosing flu as COVID-19"
- Representation error: Classic presentation in a non-classic population — you see the textbook picture but ignore disease prevalence (e.g., diagnosing multiple myeloma in a 2-year-old)
- Premature closure: Jumping to conclusions without confirming diagnosis with appropriate testing. Key signal: question stem says no workup was done, then patient deteriorated
- Anchoring bias: You stick with the first thing that comes to mind even as mounting evidence suggests it's wrong. Key signal: the question gives you contradictory data (normal BNP, no edema) but the physician still calls it CHF
- Confirmation bias: You think of X and only order tests to confirm X — never consider alternatives. Key signal: physician selects only confirmatory tests for one disease
- Attribution error: Negative stereotype applied to a patient's new complaints — "Oh, they're always complaining." Patient with illness anxiety presents with real disease, you dismiss it as anxiety again
- Affective error: Personal feelings toward patient (frustration, attachment) affect the quality of clinical decision-making. Key signal: physician is described as frustrated or angry with non-compliant patient
Complete Cognitive Error Taxonomy
| Error | Core Mechanism | Key USMLE Signal |
|---|---|---|
| Availability error | Something comes to mind easily because it's recent or frequent | "Physician treated many COVID patients" → misses flu; something recently seen → over-diagnosed |
| Representation error | Classic presentation in the wrong population (ignores prevalence) | Multiple myeloma in a toddler; CLL in a child; "slam dunk" presentation but wrong demographic |
| Premature closure | Jump to conclusion without diagnostic confirmation | Physician makes diagnosis, does no testing, patient deteriorates |
| Anchoring bias | Stick with first diagnosis despite accumulating contrary evidence | Normal BNP + no crackles + no edema → still called "CHF exacerbation" |
| Confirmation bias | Only order tests to confirm, never to rule out | Physician orders a specific set of tests only for disease X; never considers alternatives |
| Attribution error | Negative stereotype about patient based on past behavior | Illness anxiety patient presents with real disease → physician dismisses as "in their head again" |
| Affective error | Personal feelings (frustration, favoritism) affect clinical care | Non-compliant patient has new complaint → physician doesn't evaluate thoroughly due to frustration |
The Common Mix-Ups — How to Keep Them Straight
Availability vs Anchoring: In availability, the wrong diagnosis comes from frequent/recent exposure. In anchoring, the wrong diagnosis persists despite contradictory evidence in the current case.
Attribution vs Affective: Attribution = stereotype based on the patient's HISTORY (they always present this way). Affective = your FEELINGS about the patient (frustration, attachment).
Premature closure vs Anchoring: Premature closure = never did the workup. Anchoring = did the workup, findings don't fit, still sticking with it.
Active Error vs Latent Error vs Near Miss
| Term | Definition | Classic Example |
|---|---|---|
| Active error | Mistake made at the bedside by the clinician caring for the patient at that moment | Giving the wrong drug dose while writing orders |
| Latent error | No mistake yet, but the system is set up in a way that makes errors likely — a mistake waiting to happen | Two look-alike medications stored next to each other in the medication Pyxis |
| Near miss | An error occurred, but harm did not reach the patient | Wrong dose ordered, caught by pharmacist before administration |
Screening & Preventive Medicine — Perioperative
Perioperative medicine is one of those topics that is almost entirely pattern recognition — it's scenario-based and cannot be reasoned through. Divine calls this out explicitly: you either know the management or you don't. The 15+ distinct scenarios covered in EP613 represent a high-density yield zone for Step 2 and Step 3.
EP613
Perioperative Medicine — Social Sciences
- Oral hypoglycemics preoperatively: Hold ALL oral antidiabetics before surgery. Use insulin instead. Goal glucose <180 mg/dL perioperatively. Metformin risk: lactic acidosis with surgical stress, renal/hepatic impairment
- Anticoagulation bridge: Hold warfarin and antiplatelet agents (aspirin, clopidogrel) preoperatively. Bridge with heparin if anticoagulation required (heparin effect reverses quickly). Exception: coronary stent patients — continue aspirin to prevent stent restenosis
- Atlantoaxial instability screening: Rheumatoid arthritis, Down syndrome, ankylosing spondylitis → all require preoperative cervical spine X-ray before intubation/procedures involving neck manipulation
- Stress dose steroids: Any patient on chronic corticosteroids (e.g., giant cell arteritis, Addison's disease) requires perioperative stress-dose steroids — chronic steroid use suppresses the HPA axis, preventing appropriate cortisol surge during surgical stress
- Post-MI elective surgery: Wait at least 60 days after MI before elective surgery. Less than 60 days = very high risk of reinfarction and death
- Herbal supplements: Stop ALL herbal supplements (ephedra, ginkgo, ginseng, kava, St. John's wort, echinacea) at least 1 week before surgery
Complete Perioperative Scenario Reference
| Clinical Scenario | Perioperative Management |
|---|---|
| Diabetes on oral medications | Hold all oral antidiabetics. Use insulin perioperatively. Goal glucose <180 mg/dL |
| Warfarin / anticoagulants | Hold before surgery; bridge with heparin if needed (reversible, short-acting) |
| Aspirin / clopidogrel (no stent) | Hold several days before surgery to reduce bleeding risk |
| Coronary stent (bare metal or DES) | Continue aspirin perioperatively — stent thrombosis risk outweighs bleeding risk |
| Rheumatoid arthritis + intubation | Preoperative cervical spine X-ray — screen for atlantoaxial instability (C1-C2 subluxation) |
| Down syndrome + intubation | Same — atlantoaxial instability screening required |
| Ankylosing spondylitis + intubation | Same — cervical X-ray required |
| Chronic steroid use (≥6 months) | Stress dose steroids perioperatively — HPA axis suppressed, can't mount cortisol response |
| Addison's disease | Stress dose steroids — primary adrenal insufficiency cannot surge cortisol |
| Herbal supplements | Hold ALL herbal supplements ≥1 week before surgery |
| Prior MI + elective surgery | Wait ≥60 days post-MI. Less = high risk reinfarction/death |
| Patient on beta blocker (well-tolerated) | Continue perioperatively — shown to decrease cardiac events and mortality |
| Uncontrolled hypertension | Postpone elective surgery if BP >180/110; control BP first |
| Smoker + elective surgery | Stop smoking 4–8 weeks before surgery — impairs wound healing, increases aspiration/respiratory complication risk |
| Asthma (well-controlled) | Preoperative inhaled beta-2 agonists + possible steroids; avoid non-selective beta-blockers |
| Asthma (poorly controlled) | Delay elective surgery until asthma controlled |
| COPD | Preoperative arterial blood gases (ABG) required |
| Chronic alcohol use | Increased aspiration risk perioperatively; high risk of aspiration pneumonitis |
| GLP-1 agonist use | Increased aspiration risk — slows gastric emptying; hold before surgery |
| Diabetic gastroparesis | Increased aspiration risk |
| Hiatal hernia | Increased aspiration risk |
| Obesity | Screen for OSA with STOP-Bang questionnaire |
| MEN2 (pheochromocytoma) + surgery | Alpha blockade FIRST, then beta blockade — prevent hypertensive crisis from catecholamine surge |
| Hyperthyroidism (Graves') + surgery | Preoperative beta blocker — prevents thyroid storm. Beta blockers inhibit 5'-deiodinase → ↓ T4→T3 conversion |
| Thyroid surgery | Check PTH and serum calcium intraoperatively — inadvertent parathyroid destruction causes hypocalcemia |
| Dental procedure + prosthetic valve / prior IE | Endocarditis prophylaxis 30–60 min before: amoxicillin (cell wall inhibitor) |
| Any surgical procedure | Antibiotic prophylaxis: cefazolin 30–60 min before incision |
| Malignant hyperthermia | Triggered by inhaled anesthetics or succinylcholine; treat with dantrolene (ryanodine receptor antagonist) |
High-Risk Aspiration Scenarios
Anything that impairs consciousness, mental status, or GI motility increases aspiration risk:
Alcohol use disorder · Stroke · Brain tumor · Chronic opioid use · GLP-1 agonists · Diabetic gastroparesis · Hiatal hernia · NPO violation (not fasting)
Malignant Hyperthermia — Key Points
Triggers: Inhaled halogenated anesthetics (halothane, desflurane, sevoflurane) or succinylcholine
Mechanism: Ryanodine receptor mutation → uncontrolled Ca²⁺ release from sarcoplasmic reticulum → muscle rigidity + hyperthermia
Inheritance: Autosomal dominant
Treatment: Dantrolene (blocks ryanodine receptor)
Healthcare Systems & Quality Improvement
Quality improvement buzzwords are tested as discrete vocabulary items — you need to know not just what PDSA and Six Sigma mean, but the subtle differences between QC, QA, and QI, and between lean (waste) and Six Sigma (defects). These distinctions are the favorite NBME trap in this content area.
EP230
Quality and Safety — USMLE Content Series
- Quality Control (QC): Daily, ongoing review of a product or process to ensure it meets standards — happens frequently (e.g., resident checks intern's notes every day). Retrospective only
- Quality Assurance (QA): Less frequent quality control — periodic audit rather than daily review. Still retrospective. Think of it as "spot checking" to make sure quality hasn't degraded
- Quality Improvement (QI): Identifies a problem and designs an intervention to fix it. Can be retrospective OR prospective. The I stands for Improvement AND Interventional
- PDSA Cycle: Plan-Do-Study-Act — the scientific method applied to QI. Set a plan, execute it, analyze results, then act to sustain or modify. Ongoing cycle
- Lean model: Eliminate WASTE (inefficiencies in the process). Originated at Toyota. Not about defective products — about removing steps that don't add value
- Six Sigma: Eliminate DEFECTS — target fewer than 3.4 defects per million operations. Uses DMAIC methodology. Data-driven (unlike PDSA which is hypothesis-driven)
QC vs QA vs QI — Master Table
| Term | Frequency | Timing | Purpose | Classic Analogy |
|---|---|---|---|---|
| Quality Control (QC) | Daily / continuous | Retrospective only | Ensure product meets standards in real-time | Resident checks intern's notes every day |
| Quality Assurance (QA) | Periodic (weekly/monthly audit) | Retrospective only | Confirm quality hasn't slipped when not being watched | Resident checks intern's notes once a week |
| Quality Improvement (QI) | Project-based | Retrospective OR prospective | Identify problem → design intervention → measure impact | Med school gives all students laptops + UWorld to improve board scores |
Lean vs Six Sigma — The Critical Distinction
Lean model: Eliminate WASTE (unnecessary steps, inefficiencies) — inspired by Toyota Production System. Ask: "Is this step adding value?"
Six Sigma: Eliminate DEFECTS (errors in the final product) — target 3.4 defects per million. Uses DMAIC. Ask: "How often is our product wrong?"
You can have an efficient (lean) process that still produces occasional defects. You can have a zero-defect (Six Sigma) process that's very inefficient. They solve different problems.
PDSA Cycle — Steps
- Plan: Identify the problem and design an intervention (e.g., give every med student a laptop with UWorld)
- Do: Implement the intervention
- Study: Analyze the results (did USMLE scores improve?)
- Act: If effective, institutionalize it; if not, modify and repeat the cycle
DMAIC (Six Sigma Methodology)
| Step | Meaning |
|---|---|
| Define | Define the problem (e.g., med students take 4 hours to pre-round on 2 patients) |
| Measure | Quantify the problem with data |
| Analyze | Find root causes — why is this happening? |
| Improve | Implement a solution (e.g., create a pre-rounding checklist) |
| Control | Monitor and sustain the improvement over time |
Key DMAIC advantage over PDSA: DMAIC is data-driven. You study the process before proposing a solution. PDSA is more hypothesis-driven (plan first, then test).
Other Key Quality & Safety Buzzwords
| Term | Definition |
|---|---|
| Hawthorne effect | People behave differently when they know they are being watched. Observed behavior ≠ usual behavior |
| Weber effect | When adverse events are tracked and monitored, their incidence decreases over time. Tracking itself drives improvement |
| Safety champion | The designated person responsible for quality and safety at an institution |
| High-value care equation | Value = (Service × Quality) / Cost. Residents are high-value: good service + improving quality + fixed low cost |
| FMEA (Failure Mode & Effects Analysis) | PROSPECTIVE process — identifies how a process might fail before it happens, and analyzes the effects of those failures |
| Root cause analysis | RETROSPECTIVE process — after an adverse event, works backward to find the root cause. Same concept as FMEA, but after the fact |
| Swiss cheese model | Adverse events result from multiple system failures stacking up. Each layer of defense (like a slice of cheese) has holes; if holes align, harm reaches the patient. Multiple redundant safeguards reduce this risk |
FMEA vs Root Cause Analysis — The Timing Difference
FMEA: Prospective — we haven't made the mistake yet, but we're analyzing how we could.
Root cause analysis: Retrospective — the mistake happened, now we're finding out why.
Both use the same logic; timing of application is the differentiator.
Ethics, Communication & Palliative Care
Palliative care and end-of-life ethics are now explicitly tested, and the scenarios are almost entirely vocabulary-driven — knowing what a POLST form is, the difference between palliative sedation and euthanasia, and the doctrine of double effect separates an average from an excellent score on this content area.
EP268
Palliative Care 2 — USMLE Content Series
- Palliative care vs hospice: Palliative care can occur at ANY stage of illness alongside curative treatment. Hospice is end-of-life focused — patient must have ≤6 month prognosis and forgo curative treatment to qualify
- POLST / MOLST form: Physician Orders for Life-Sustaining Treatment — a legally binding medical order (not just a preference document). More detailed than a DNR — specifies CPR, mechanical ventilation, artificial nutrition preferences
- Doctrine of double effect: A treatment intended to relieve suffering (e.g., high-dose morphine) that may unintentionally shorten life is ethically permissible — intent is to relieve suffering, not to cause death
- Palliative sedation: Sedation to relieve intractable suffering at end of life — ethically distinct from euthanasia (intent is comfort, not death). Permissible if patient/surrogate consent obtained
- Advance directives: Living will (patient's written wishes) vs healthcare proxy (person designated to decide). Healthcare proxy takes precedence over family members without designation when the patient lacks capacity
- Capacity vs competency: Capacity is a clinical determination (physicians assess). Competency is a legal determination (courts decide). You assess capacity at the bedside; you don't wait for a judge unless there's a dispute
Palliative Care vs Hospice — Critical Distinctions
| Feature | Palliative Care | Hospice |
|---|---|---|
| Stage of illness | Any stage — can occur alongside curative treatment | Terminal illness with ≤6-month prognosis |
| Curative treatment | Continues alongside palliative care | Must be foregone to qualify for hospice |
| Goal | Improve quality of life, manage symptoms | Comfort and dignity at end of life |
| Setting | Hospital, outpatient, home | Home, hospice facility, or nursing home |
| Insurance | Covered like any care | Medicare hospice benefit requires formal enrollment |
Advance Directive Types
| Document | What It Is | Key USMLE Point |
|---|---|---|
| Living will | Patient's written statement of wishes for specific situations (e.g., "no mechanical ventilation if vegetative") | Guides care only when patient lacks capacity |
| Healthcare proxy / DPOA-HC | Designates a person to make decisions when patient lacks capacity | Overrides family members who were NOT designated |
| DNR / DNI order | Physician order specifying no CPR or intubation | Stays in effect in hospital; must have separate out-of-hospital DNR for paramedics |
| POLST / MOLST | Detailed physician medical order covering CPR, ventilation, artificial nutrition, hospitalization preferences | More comprehensive and actionable than a DNR — travels with the patient |
Doctrine of Double Effect — The Classic USMLE Scenario
A terminally ill cancer patient is in severe pain. The physician prescribes escalating doses of morphine, knowing that at the required dose, the medication may suppress respiration and hasten death. This is ethically permissible under the doctrine of double effect because: (1) the act itself (pain relief) is not inherently wrong, (2) the intent is to relieve suffering, not cause death, and (3) the benefit (pain relief) is proportionate to the foreseen harm.
Capacity Assessment — Bedside
Four elements required for a patient to have decision-making capacity:
- Understanding: Patient understands the medical situation and the proposed treatment
- Appreciation: Patient appreciates how the information applies to their own situation
- Reasoning: Patient can reason through options and weigh pros/cons
- Expression: Patient can consistently communicate a choice
Capacity vs Competency
Capacity: Clinical determination made by the treating physician at the bedside. Can fluctuate (patient has capacity in the morning, loses it at night from delirium).
Competency: Legal determination made by a court. Stable — you are legally competent or not.
On the USMLE: The physician assesses capacity. The court determines competency. Never say "refer to court to determine if patient has capacity" — that's wrong. Assess it yourself.
Ethics Principles — Quick Reference
| Principle | Meaning | Classic USMLE Application |
|---|---|---|
| Autonomy | Patient's right to decide their own care | Competent adult can refuse any treatment, including life-saving treatment |
| Beneficence | Acting in the patient's best interest | Recommending chemotherapy that will likely improve survival |
| Non-maleficence | Do no harm | Avoiding unnecessary procedures with high complication risk |
| Justice | Fair distribution of resources and care | Organ allocation by waitlist priority, not social value |
| Confidentiality | Patient information protected | Can break confidentiality for imminent harm to identifiable third party (Tarasoff rule), mandatory reportable diseases, child abuse |
| Informed consent | Patient must understand risks, benefits, alternatives before consenting | Must include option to refuse; emergency exception exists (patient unconscious, no time) |
Social Determinants & Special Populations — Military Medicine
Military medicine is explicitly named in the USMLE content outline, making it testable and non-negotiable. The PTSD content overlaps heavily with psychiatry (treatment, pharmacology, CBT modalities) while the infectious disease component (deployment vaccines, Q fever, leishmaniasis, leptospirosis) is genuinely unique to this population cluster.
EP204
The USMLE and the Military — Part 1 (TBI & PTSD)
- TBI classification by GCS: Mild (GCS 13–15) = concussion; Moderate (GCS 9–12); Severe (GCS ≤8). Most vulnerable brain regions: anterior temporal lobes + orbital frontal cortex
- TBI management pearls: N-acetylcysteine (preventive). Tranexamic acid (TXA) within 3 hours to reduce mortality. Hyperventilation = fastest way to acutely lower ICP (short-term only — causes cerebral vasoconstriction and ischemia if prolonged). Do NOT give corticosteroids in TBI — increases mortality
- Post-concussive syndrome: Chronic headaches, dizziness, sensitivity to light/sound, anxiety, sleep problems after mild TBI. Second-impact syndrome: second TBI during recovery from first → catastrophic outcome
- PTSD treatment: SSRIs (sertraline, fluoxetine, paroxetine) + SNRIs (venlafaxine). CBT modalities: prolonged exposure therapy (also for OCD, GAD, phobias) + cognitive processing therapy. Prazosin for nightmares. NO benzodiazepines (worsens outcomes in veterans)
- PTSD neurobiology: Smaller hippocampus. Exaggerated dexamethasone suppression test response. Elevated urine catecholamines + paradoxically LOW cortisol (inverse of normal stress response)
- Acute stress disorder vs PTSD: Same symptoms — if <1 month = acute stress disorder (no SSRIs, only psychotherapy). If ≥1 month = PTSD (add SSRIs). Debriefing is NOT effective and may be harmful
TBI — Classification and Key Management
| Severity | GCS Score | Also Called | Key Management Points |
|---|---|---|---|
| Mild TBI | 13–15 | Concussion | Avoid second impact (second impact syndrome = catastrophic); watch for post-concussive syndrome |
| Moderate TBI | 9–12 | — | CT head first; MRI if CT negative; rehab after acute phase |
| Severe TBI | ≤8 | — | Intubation + mechanical ventilation; ICP management; neurosurgery if hematoma |
ICP Management in TBI
- Head of bed elevation — improves cerebral perfusion
- Hyperventilation — fastest acute ICP reduction. Blow off CO₂ → cerebral vasoconstriction → ↓ ICP. Short-term only — prolonged causes ischemia
- Mannitol — osmotic diuretic; caution in CHF
- Hypertonic saline — caution: can cause severe hypernatremia
- Normothermia — do NOT let patient become hyperthermic (increases brain metabolic demand)
- Decompressive craniectomy — if hematoma compressing brain
- Do NOT give corticosteroids — shown to increase mortality in TBI (unlike vasogenic edema from tumor, where steroids help)
Most Tested TBI vs Tumor Steroid Rule
Brain tumor edema: Corticosteroids HELP — reduce vasogenic edema around tumor (dexamethasone is standard).
Traumatic brain injury: Corticosteroids WORSEN — shown to increase death in TBI. Never give steroids for TBI.
PTSD — Complete Pharmacology and Psychotherapy
| Treatment Category | Agents / Modalities | Notes |
|---|---|---|
| First-line pharmacotherapy | SSRIs: sertraline, fluoxetine, paroxetine; SNRI: venlafaxine | Must have symptoms ≥1 month (otherwise acute stress disorder — no SSRIs) |
| Nightmares (adjunct) | Prazosin (alpha-1 blocker) | Reduces adrenergic nighttime surges |
| DO NOT USE | Benzodiazepines | Worsens outcomes in veterans with PTSD; avoid |
| CBT — prolonged exposure therapy | Gradual exposure to trauma triggers in safe environment | Also first-line CBT for OCD, GAD, specific phobias |
| CBT — cognitive processing therapy | Discuss trauma in detail; address maladaptive thoughts | Thought-based; challenges distorted cognitions about the traumatic event |
| NOT recommended | Debriefing | Not shown to help; may be harmful in some studies |
PTSD Neurobiology (Testable Details)
- Smaller hippocampal volume on MRI
- Exaggerated response to dexamethasone suppression test (cortisol suppressed more than normal)
- Elevated urine catecholamines (hyperadrenergic state)
- Paradoxically LOW cortisol in urine (unlike acute stress where cortisol rises)
Military-Specific Epidemiology
- ~1 in 4 deployed veterans develops PTSD
- Most common PTSD comorbidity: alcohol use disorder
- Military sexual trauma — much more common in female service members; major PTSD risk factor
- Veterans have higher rates of homelessness; #1 comorbidity in homeless veterans = substance use disorder (primarily alcohol)
- ~45% of homeless veterans have PTSD
- Psychiatric problems in military members peak many decades after the original trauma exposure
- Children of deployed service members: higher rates of depression, anxiety, behavioral problems
- Better family resilience correlates with: pre-deployment preparation AND increased communication during deployment
EP231
Military Medicine Part 2 — Infections, Vaccines & Occupational Hazards
- Required deployment vaccines: MMR, Td/Tdap, influenza, polio, hepatitis A, typhoid, varicella. Region-specific additions: yellow fever, smallpox, anthrax
- Q fever (Coxiella burnetii): Rickettsial infection. Exposure to livestock (cattle, sheep, goats) — birthing fluids and placenta are highly infectious. Presents with fever, headache, atypical pneumonia. Treat with doxycycline
- Cutaneous leishmaniasis: Sandfly-transmitted protozoa (Leishmania). Painless skin ulcer with raised borders — classic description "volcano crater." Central Asia, Middle East conflict zones. Treatment: sodium stibogluconate or liposomal amphotericin B
- Leptospirosis: Exposure to water contaminated with rat urine. Classic in flood-prone or tropical deployment zones. Weil's disease (severe form): jaundice + AKI + hemorrhage + uveitis. Treat with penicillin or doxycycline
- Melioidosis (Burkholderia pseudomallei): Soil and water in Southeast Asia. Can present as pneumonia, septicemia, or localized infection. Treat with IV ceftazidime or meropenem (acute), then TMP-SMX (eradication)
- Gulf War Illness: Constellation of medically unexplained symptoms (fatigue, cognitive difficulties, musculoskeletal pain) in veterans of the 1991 Gulf War. Etiology uncertain — multifactorial. Not a psychiatric diagnosis
Military Deployment Vaccines
| Vaccine | Status | Notes |
|---|---|---|
| MMR | Required all deployments | Live attenuated — contraindicated in pregnancy and immunocompromised |
| Td / Tdap | Required all deployments | Tdap given once; Td booster every 10 years |
| Influenza | Required all deployments | Annual — inactivated preferred in military |
| Polio (IPV) | Required all deployments | Inactivated — do NOT use OPV (not available in US) |
| Hepatitis A | Required all deployments | 2-dose series; endemic in many conflict zones |
| Typhoid | Required all deployments | Oral (live, 4 doses) or IM (inactivated, 1 dose) |
| Varicella | Required if not immune | Live vaccine — document immunity before deployment |
| Yellow fever | Region-specific (Africa, South America) | Live vaccine; one dose = lifelong immunity |
| Smallpox (vaccinia) | Specific high-risk deployments | Live virus; complications: myopericarditis (young males), progressive vaccinia (immunocompromised) |
| Anthrax | Bioterrorism risk deployments | Multi-dose series + post-exposure prophylaxis with ciprofloxacin |
Military-Specific Infections — Master Table
| Infection | Organism | Exposure | Key Features | Treatment |
|---|---|---|---|---|
| Q fever | Coxiella burnetii (obligate intracellular) | Livestock birth fluids, placenta, aerosol — cattle, sheep, goats | Fever, headache, atypical pneumonia; no rash (unlike other Rickettsial); may cause endocarditis chronically | Doxycycline |
| Cutaneous leishmaniasis | Leishmania spp. (protozoa) | Sandfly bite — Middle East, Central Asia | Painless ulcer with raised indurated borders ("volcano crater"); can progress to mucocutaneous or visceral (kala-azar) forms | Sodium stibogluconate or liposomal amphotericin B |
| Leptospirosis | Leptospira interrogans (spirochete) | Water or soil contaminated with rat/animal urine — flooding, tropical | Biphasic: flu-like illness then Weil's disease (jaundice + AKI + hemorrhage + uveitis) | Penicillin G (severe) or doxycycline (mild) |
| Melioidosis | Burkholderia pseudomallei (gram-negative) | Soil and water contact — Southeast Asia, Northern Australia | Pneumonia, abscess formation, septicemia; "Vietnam time bomb" — reactivates decades later | IV ceftazidime or meropenem (acute); TMP-SMX (eradication phase) |
| Brucellosis | Brucella spp. | Unpasteurized dairy, livestock, laboratory exposure | Undulant (wave-like) fever, sweats, arthralgia, lymphadenopathy; no rash | Doxycycline + rifampin (6 weeks minimum) |
Q Fever — High-Yield Differentiator
Q fever is the only Rickettsial disease that does NOT present with a rash. All others (Rocky Mountain spotted fever, typhus, ehrlichiosis) have rashes. Q fever also has a unique exposure history — livestock birthing, NOT tick bites (unlike most other Rickettsial diseases). Doxycycline remains the treatment regardless of Rickettsial subtype.
Gulf War Illness
- Affects veterans of the 1991 Persian Gulf War (Operation Desert Storm)
- Symptoms: chronic fatigue, cognitive difficulties ("Gulf War brain"), musculoskeletal pain, headaches, rashes, gastrointestinal disturbances
- Etiology is multifactorial and debated: possible role of organophosphate pesticides, nerve agent exposure, pyridostigmine (anti-nerve agent prophylaxis), smoke from oil well fires
- It is NOT a psychiatric diagnosis and is NOT malingering — this is important for the USMLE
- Recognized by the VA as a service-connected condition
Social Determinants of Health — Military Populations
Housing insecurity: Veterans have disproportionately high rates of homelessness. Homeless veterans: #1 comorbidity = substance use disorder (alcohol most common).
Mental health burden: PTSD, depression, suicide risk all elevated in returning service members. Suicide rate among veterans significantly higher than age-matched civilian population.
Family impact: Children of deployed service members: ↑ depression, ↑ anxiety, ↑ behavioral problems. Two protective factors: pre-deployment preparation AND communication during deployment.