Eduovisual
Biostatistics & Population Health
Vaccine efficacy vs effectiveness
— VE = (1 − RR) × 100% or (1 − OR) × 100%, where RR/OR compares vaccinated to unvaccinated risk of the outcome.
— A VE of 90% means 90% fewer cases among vaccinated relative to unvaccinated — NOT that 10% of vaccinated people will get sick.
— A vignette presents both an RCT result ("efficacy 95%") and a later population study ("effectiveness 68%") and asks why they differ.
— A patient or news-style quote misinterprets VE as absolute risk ("the vaccine only works 70% of the time, so 30% of vaccinated people will get sick").
— A public-health officer must choose between trial data and surveillance data for policy.
— Questions about waning immunity, variant escape, or booster timing — all are effectiveness phenomena, not efficacy failures.
— Efficacy → ideal conditions → maximum biological signal
— Effectiveness → messy reality → what the population actually experiences
— Impact → population-level disease burden averted (incorporates coverage, herd effects, indirect protection)
Board pearl: Efficacy is almost always higher than effectiveness because RCTs enroll healthier, more adherent, more homogeneous populations and measure outcomes under ideal cold-chain, dosing-interval, and follow-up conditions. If a Step 3 stem shows effectiveness > efficacy, suspect either confounding (healthy-vaccinee bias) or a different outcome definition (e.g., trial measured infection, real-world measured hospitalization).
— The "trial vs reality" gap: A new influenza vaccine showed 80% efficacy in a phase 3 trial; CDC surveillance the following winter shows 45% effectiveness. Best explanation?
— The misinterpreted percentage: A patient refuses an HPV vaccine because "it's only 90% effective, so I'll probably still get cancer." Best counseling response?
— The waning question: COVID-19 vaccine effectiveness drops from 92% at 1 month to 55% at 6 months. Mechanism?
— The strain mismatch: Influenza vaccine effectiveness is 19% this season; why?
— The herd/indirect protection puzzle: Unvaccinated children in a high-coverage community have lower disease rates than expected — what concept explains this?
— Outcome measured: infection, symptomatic disease, hospitalization, death, transmission? Each has its own VE.
— Time since vaccination: waning is built into effectiveness, not efficacy.
— Variant or strain circulating vs vaccine antigen match.
— Population characteristics: age, immunocompromise, prior infection (hybrid immunity).
— Study design: RCT (efficacy), test-negative case-control (most common for effectiveness), retrospective cohort, screening method.
— Coverage rate in the population (drives indirect/herd effects, distinct from individual VE).
Key distinction: "My vaccine didn't work" is rarely an efficacy failure in a Step 3 stem. It is almost always (1) waning immunity, (2) antigenic mismatch/variant escape, (3) immunocompromise blunting response, (4) incomplete series, or (5) misunderstanding of relative vs absolute risk. Identify which before answering.
— Seroconversion rate: proportion developing protective antibody titer (e.g., anti-HBs ≥10 mIU/mL post-hepatitis B series). A surrogate for immunogenicity, not for clinical effectiveness.
— Geometric mean titer (GMT): average antibody concentration; correlates with but does not equal protection.
— Correlate of protection: a measurable immune marker (titer threshold) statistically linked to clinical immunity (e.g., HAI titer ≥1:40 for influenza, anti-HBs ≥10 for HepB, rubella IgG positivity).
— Coverage rate (% of target population vaccinated) — the denominator for herd immunity calculations.
— Herd immunity threshold = 1 − (1/R₀). Measles R₀ ~12–18 → threshold ~92–95%. Pertussis ~92–94%. Polio ~80–86%.
— Breakthrough infection rate: cases occurring in fully vaccinated individuals — expected, not failure.
— Vaccine impact: absolute cases/hospitalizations/deaths averted in a population per year.
— A seroconverted patient with adequate titer is "immunologically vaccinated."
— Titers wane; clinical protection may persist via memory B/T cells even when antibodies are undetectable (e.g., hepatitis B — no booster needed in immunocompetent adults despite waning anti-HBs).
Board pearl: Anti-HBs ≥10 mIU/mL 1–2 months after completion of the hepatitis B series confirms response. Non-responders (<10) get a second 3-dose series; persistent non-responders after 6 total doses are considered non-responders and require HBIG after exposure. This is the prototypical Step 3 "immunogenicity surrogate" scenario.
— Randomize tens of thousands of participants 1:1 to vaccine vs placebo.
— Blinded outcome ascertainment (symptomatic, lab-confirmed disease per protocol).
— Pre-specified primary endpoint (often symptomatic PCR-confirmed disease ≥7–14 days after final dose).
— Per-protocol and intention-to-treat analyses.
— VE = (1 − IRR) × 100%, where IRR = incidence rate in vaccinated / incidence rate in unvaccinated.
— Or VE = (ARU − ARV)/ARU × 100% (attack rate unvaccinated minus attack rate vaccinated, divided by attack rate unvaccinated).
— Example: 8/15,000 vaccinated developed disease (AR = 0.053%); 162/15,000 placebo developed disease (AR = 1.08%). VE = (1.08 − 0.053)/1.08 = 95%.
— VE against infection (often lower)
— VE against symptomatic disease (trial primary endpoint)
— VE against severe disease/hospitalization (usually highest and most durable)
— VE against death
— VE against transmission (hardest to measure; often requires household contact studies)
Step 3 management: When a stem gives raw numbers (cases/total in each arm), calculate VE directly with the (1 − RR) formula. Do not confuse with relative risk reduction phrasing — VE is the RRR expressed as a percentage. Absolute risk reduction (ARR) and number needed to vaccinate (NNV = 1/ARR) are separate metrics often asked alongside.
— Enroll symptomatic patients presenting for testing.
— Cases = test-positive for target pathogen; controls = test-negative (presumed other respiratory illness).
— Compare vaccination status between cases and controls.
— VE = (1 − OR) × 100%.
— Strength: controls for healthcare-seeking behavior (both groups sought testing), minimizing healthy-vaccinee bias.
— Compare disease incidence in vaccinated vs unvaccinated cohorts over time.
— Vulnerable to confounding by indication, healthy-vaccinee effect, differential testing.
— VE = 1 − [(PCV/(1−PCV)) × ((1−PPV)/PPV)], where PCV = proportion of cases vaccinated, PPV = proportion of population vaccinated.
— Quick surveillance estimate; biased if denominators are inaccurate.
— Healthy-vaccinee bias: vaccinated people may be healthier, more health-literate → overestimates VE.
— Frailty/indication bias: in elderly, those vaccinated may be sicker (captured for vaccination at clinic visits) → underestimates VE.
— Differential misclassification: vaccinated may test less often or differently.
— Confounding by prior infection: hybrid immunity inflates apparent VE if not adjusted.
— Depletion of susceptibles: as the epidemic progresses, unvaccinated susceptibles get infected and removed, falsely lowering apparent VE over time.
Board pearl: The test-negative design is the go-to answer for "best study design to estimate real-world influenza or COVID-19 vaccine effectiveness." It is not an RCT — it is a case-control study nested within a care-seeking population, and it estimates effectiveness, never efficacy.
— Population heterogeneity: trials exclude immunocompromised, frail elderly, pregnant (historically), children — real world includes them all.
— Adherence: trials enforce dose intervals; real world has delayed/missed doses.
— Cold-chain breaches: trial vaccine handled perfectly; real-world storage variable.
— Variant drift: trial used circulating strain; real-world exposure may be antigenically distant.
— Waning: trial follow-up often 2–6 months; effectiveness measured years out.
— Exposure dose/intensity: real-world exposures may exceed trial challenge conditions.
— Different endpoint: trial measured symptomatic disease; surveillance measures hospitalization (VE almost always higher against severe disease).
— Indirect protection: high coverage reduces force of infection on vaccinated and unvaccinated alike.
— Hybrid immunity: prior infection plus vaccination boosts response above trial-naïve participants.
— Highest absolute benefit → highest baseline risk (elderly, immunocompromised, comorbid) even if relative VE is lower.
— NNV is lowest (best) where incidence is highest.
— Indirect protection benefits those who cannot be vaccinated (infants, immunocompromised) when surrounding coverage is high.
Key distinction: A vaccine with 50% effectiveness against a high-incidence disease (e.g., influenza in elderly, baseline 10% attack rate → ARR 5%, NNV 20) provides more absolute benefit than a vaccine with 95% efficacy against a rare disease (baseline 0.1% → ARR 0.095%, NNV ~1,053). Step 3 loves this absolute-vs-relative trap.
— Robust, often lifelong immunity after 1–2 doses.
— Contraindicated in pregnancy and significant immunocompromise.
— MMR efficacy ~93% (1 dose), 97% (2 doses) against measles; effectiveness similar — durable.
— Safe in immunocompromised; requires boosters; effectiveness more variable.
— Influenza IIV effectiveness ranges 10–60% depending on strain match.
— Excellent safety profile; conjugation creates T-cell-dependent response and memory.
— Shingrix efficacy >90% against shingles, sustained ~85% at 4 years — a standout in older adults vs live Zostavax (~50%, waning rapidly).
— Acellular pertussis: efficacy ~85% initially but effectiveness wanes substantially within 5–10 years → adolescent and adult Tdap boosters needed.
— Efficacy ~95% against ancestral strain (symptomatic disease); effectiveness against variants and over time lower, but effectiveness against severe disease remains 70–90%+ with boosters.
Board pearl: When a stem asks "why does this vaccine need a booster," the answer maps to vaccine type: polysaccharide (no memory), acellular pertussis (waning antibody), influenza (antigenic drift), tetanus (long-lived but finite memory — 10-year boosters).
— Schedule fidelity: HPV 2-dose (age 9–14) vs 3-dose (15+); shorter intervals reduce effectiveness. CDC catch-up rules matter.
— Cold chain: mRNA vaccines −80°C/−20°C; live vaccines lose potency if not refrigerated.
— Coadministration: most vaccines can be coadministered same visit, different sites. Live vaccines either same day or separated by ≥28 days.
— Concurrent immunoglobulin: defer live vaccines (MMR, varicella) 3–11 months after Ig/blood products (passive antibody blunts response).
— Acute moderate-severe illness: defer; mild illness is not a contraindication.
— Influenza: annually, ideally September–October; effectiveness measured each season.
— RSV monoclonal nirsevimab (passive immunization, not a vaccine — but tested similarly): efficacy ~75–80% against medically attended LRTI in infants.
— Maternal Tdap (27–36 weeks) and maternal RSV (32–36 weeks, seasonal): transplacental antibody — effectiveness measured in the infant.
— Insurance coverage (ACA mandates ACIP-recommended vaccines at no cost-share).
— VFC (Vaccines for Children) for uninsured/Medicaid kids <19.
— Standing orders, pharmacist administration, EHR reminders, school entry mandates — all increase coverage and thus impact, distinct from per-dose VE.
— Reduced naïve T-cell repertoire, blunted B-cell response, lower seroconversion, faster waning.
— Effectiveness typically lower than younger adults for the same vaccine.
— High-dose inactivated influenza (Fluzone HD, 60 µg HA per strain) — preferred for ≥65; ~24% more effective than standard-dose in this group.
— Adjuvanted influenza (Fluad, MF59) — alternative preferred option for ≥65.
— Recombinant influenza (Flublok) — also acceptable preferential option.
— ACIP recommends any one of these three preferentially over standard-dose IIV for ≥65.
— Shingrix (RZV) for all ≥50 (and immunocompromised ≥19): 2 doses 2–6 months apart; efficacy >90% across age bands, sustained.
— Pneumococcal: current ACIP — single dose PCV20 OR PCV15 followed by PPSV23 for adults ≥65 who haven't received PCV.
— RSV: shared clinical decision-making for adults 60–74 at increased risk; routine for ≥75.
— Hemodialysis/CKD: blunted response to hepatitis B → use higher-dose HepB formulations (e.g., 40 µg Engerix-B or Heplisav-B 2-dose). Check post-vaccination anti-HBs.
— End-stage liver disease/cirrhosis: high priority for HepA, HepB, pneumococcal, influenza — effectiveness lower but absolute benefit high.
— No renal dose adjustment for vaccines themselves.
Step 3 management: For the 70-year-old presenting for annual visit in October, the correct influenza answer is high-dose, adjuvanted, or recombinant — not standard-dose. Pair with PCV20 (if pneumococcal-naïve), RSV (≥75 or shared decision 60–74), Shingrix series status check, and Tdap/Td booster status.
— Recommended: Tdap (every pregnancy, 27–36 weeks), inactivated influenza (any trimester), COVID-19 (any trimester), RSV (32–36 weeks during seasonal window if infant won't receive nirsevimab).
— Contraindicated (live): MMR, varicella, LAIV, yellow fever (unless travel risk outweighs), smallpox.
— Maternal vaccine effectiveness is measured in the infant (transplacental IgG); maternal RSV vaccine ~70–80% effective against severe infant RSV disease in first 6 months.
— Maternal antibody can blunt early infant response (why MMR is delayed to 12 months).
— Live vaccines avoided <12 months for most; rotavirus is a notable exception (oral, age-restricted to start by 15 weeks, complete by 8 months).
— HPV: 2 doses if initiated 9–14, 3 doses if 15+ or immunocompromised — effectiveness near 100% against vaccine-type infection if completed before sexual debut.
— Live vaccines generally contraindicated: HIV with CD4 <200, solid organ transplant, active chemo, high-dose steroids (≥20 mg prednisone ≥14 days), biologics like rituximab.
— Inactivated vaccines safe but less effective — give anyway; consider additional doses.
— Shingrix (recombinant) is preferred and now recommended ≥19 with immunocompromise.
— Hematopoietic stem cell transplant: restart entire vaccine series starting 6–12 months post-transplant.
— Anti-CD20 (rituximab): defer vaccines until ≥6 months after last dose if possible; antibody response markedly blunted.
Key distinction: Asymptomatic HIV with CD4 ≥200 → MMR and varicella can be given. CD4 <200 → live vaccines contraindicated. This threshold is repeatedly tested.
— Local pain, low-grade fever, myalgia 24–48 hours post — evidence of immune response, not adverse effect requiring evaluation.
— Anaphylaxis: ~1–5 per million doses; observe 15 min post-vaccination (30 min if prior allergy history).
— Guillain-Barré: historically associated with 1976 swine flu vaccine (~1 excess case/100,000); modern influenza vaccines ~1 excess case/million — risk lower than from natural influenza infection.
— Myocarditis with mRNA COVID-19 vaccines: highest in males 12–29 after dose 2; risk far lower than myocarditis from COVID-19 infection itself.
— Intussusception with rotavirus: small excess risk (~1–5 per 100,000) — age cutoffs strictly enforced.
— Thrombosis with thrombocytopenia syndrome (TTS) with adenoviral-vector COVID vaccines (J&J): rare; led to preferential mRNA recommendation.
— Primary failure: no immune response (e.g., HepB non-responder).
— Secondary failure: response wanes below protective threshold (pertussis, mumps).
— Breakthrough infection: expected statistical event, not failure — calculate using VE.
— VAERS (Vaccine Adverse Event Reporting System) — passive surveillance; anyone can report; hypothesis-generating, not causal.
— Vaccine Safety Datalink (VSD) — active surveillance in integrated health systems.
— VICP (Vaccine Injury Compensation Program) — federal no-fault compensation for covered vaccines.
Board pearl: VAERS data cannot prove causation — it has no denominator and accepts unverified reports. Step 3 may test recognition that "VAERS shows X cases" is not equivalent to "vaccine caused X." VSD with concurrent unexposed comparison is the appropriate causal-inference tool.
— Anaphylaxis post-vaccination → epinephrine IM, ED transport, allergy referral, VAERS report, document contraindication.
— Suspected vaccine-preventable disease in a vaccinated patient → confirm with PCR/serology, genotype strain (variant or vaccine-type?), report to local health department.
— Outbreak in undervaccinated cluster (measles in a school with <95% MMR coverage) → mass vaccination campaign, exclusion of unvaccinated, contact tracing, post-exposure MMR within 72 hours or IG within 6 days.
— Declining effectiveness (e.g., influenza VE <30% mid-season) → public-health messaging shifts toward antivirals, NPIs; does not mean stop vaccinating.
— Variant emergence with immune escape → consider updated vaccine composition (influenza annual reformulation, COVID-19 strain updates).
CCS pearl: In a CCS case with suspected measles, the correct sequence is (1) isolate (airborne precautions, negative-pressure room), (2) confirm (measles IgM, PCR from NP swab and urine), (3) report to public health immediately — do not wait for confirmation, (4) post-exposure prophylaxis for susceptible contacts (MMR ≤72h or IG ≤6 days), (5) supportive care plus vitamin A in children. Reporting is a scored action.
— VE = RRR expressed as percent.
— ARR = attack rate unvaccinated − attack rate vaccinated.
— A 95% VE with ARR of 1% means 1 in 100 cases prevented per vaccinated person at trial-era incidence.
— Drives cost-effectiveness; varies massively with disease incidence.
Key distinction: A vaccine that prevents severe disease but not infection or transmission can have very high individual VE-against-hospitalization yet fail to produce herd immunity. COVID-19 vaccines exemplify this — high VE for severe disease, lower and waning VE for infection/transmission, so herd immunity through vaccination alone was never achievable.
— Expected breakthrough (statistical): if VE is 90%, 10% of vaccinated exposures still result in disease relative to unvaccinated baseline.
— Incomplete series: HPV single dose, Shingrix single dose, HepB 1–2 doses — partial protection only.
— Waning immunity: pertussis effectiveness drops ~10%/year after DTaP; mumps effectiveness wanes over decades.
— Antigenic drift/shift: influenza H3N2 mismatch years.
— Variant immune escape: SARS-CoV-2 Omicron vs ancestral-strain vaccine.
— Immunocompromise blunting response: transplant, chemo, anti-CD20.
— Cold-chain failure: vaccine inactive at administration.
— Wrong vaccine for the pathogen: pneumococcal vaccine does not cover all serotypes; serogroup B meningococcal requires separate vaccine from ACWY.
— Misdiagnosis: not actually the vaccine-preventable disease — confirm with lab testing.
— Primary vaccine failure: never seroconverted (e.g., HepB non-responder — ~5–10% of healthy adults).
— Misunderstanding VE as absolute risk.
— Conflating association (VAERS report) with causation.
— Distrust of pharmaceutical industry or government.
— Religious/philosophical objection.
— Cost/access barriers (usually solvable — ACA, VFC).
— Each requires a different motivational interviewing response.
Board pearl: A fully MMR-vaccinated patient who develops measles is more likely to have been exposed to a high-dose source plus modest waning than to represent vaccine failure — and they typically have milder, atypical disease and lower transmission. Confirm with measles-specific IgM and PCR; report regardless.
— Td/Tdap: every 10 years; Tdap once in adulthood (or every pregnancy for women).
— Influenza: annual.
— COVID-19: per current ACIP — generally annual, more frequent for immunocompromised and ≥65.
— Pneumococcal: PCV20 alone OR PCV15 + PPSV23 in ≥65 or high-risk adults.
— HepB: no routine booster in immunocompetent responders; check titers and boost in hemodialysis and healthcare workers as indicated.
— Meningococcal: 5-year boosters for ongoing-risk adults (asplenia, complement deficiency, microbiologists, recurrent exposure).
— Use minimum intervals (not "restart series" — vaccines do not need restarting if interrupted, with rare exceptions like oral typhoid).
— HPV catch-up through age 26 routine, shared decision-making 27–45.
— Adult MMR: 1 dose if born ≥1957 without immunity evidence; 2 doses for healthcare workers, students, travelers.
— Influenza (annual), COVID-19 (current ACIP), Td/Tdap status, MMR/varicella history, HepB if at risk, HepA if at risk, HPV through 26 (45 with SDM), Shingrix ≥50, pneumococcal ≥65 or high-risk, RSV ≥75 (or 60–74 SDM), meningococcal if indicated, travel vaccines as appropriate.
Step 3 management: Every preventive visit is a vaccine opportunity. CCS-style: order indicated vaccines at the visit, document refusals with reason, schedule second doses, set EHR reminders. "Missed opportunity to vaccinate" is a scored quality measure.
— Observe 15 min (30 min if prior allergic reaction history) for anaphylaxis.
— Counsel on expected reactogenicity (24–48 h soreness, low fever) and when to seek care (anaphylaxis signs, persistent high fever, severe neurologic symptoms).
— Document lot number, site, date in immunization registry (IIS).
— HepB post-series in healthcare workers, dialysis, infants of HBsAg+ mothers.
— Rubella in pregnancy planning (and postpartum MMR if non-immune).
— Measles/varicella titers in healthcare workers without documentation.
— Routine post-vaccination titer checks otherwise not recommended.
— Explain VE as relative reduction: "95% efficacy means among people exposed, vaccinated individuals are 95% less likely to get sick than unvaccinated. It does not mean 5% of vaccinated people will get sick."
— Use absolute numbers when helpful: "Out of 1,000 vaccinated people exposed, we'd expect X cases vs Y cases unvaccinated."
— Address waning: "Protection against severe illness stays high; protection against any infection drops over months — that's why we recommend a booster."
— Address breakthrough: "Even with a great vaccine, some breakthrough is expected and usually milder."
Board pearl: For pregnant patients found to be rubella non-immune prenatally, do not vaccinate during pregnancy (live vaccine, contraindicated). Administer MMR postpartum before discharge — this is a recurring Step 3 / CCS scored item and a classic missed-opportunity scenario.
— Federal law (National Childhood Vaccine Injury Act, 1986) requires the appropriate Vaccine Information Statement (VIS) be provided before each dose of covered vaccines. Document VIS edition date and date given.
— Verbal consent typically sufficient; some states require written consent.
— Adolescents: minor consent laws for vaccines vary by state — some allow minors to consent to HPV, HepB, meningococcal independently.
— Vaccine-preventable diseases (measles, pertussis, etc.) — reportable to public health, often within 24 hours.
— Certain serious AEs after covered vaccines — providers must report to VAERS (federally mandated for events listed in the VAERS Table of Reportable Events).
— All 50 states require vaccines for school entry; exemptions (medical always; religious/philosophical vary).
— Healthcare worker vaccination requirements (influenza, HepB, MMR, varicella, Tdap) — refusal may have employment consequences but is legally complex.
— Respect autonomy while clearly recommending vaccination.
— Document the conversation, the specific vaccine refused, and the risk discussed (the AAP "Refusal to Vaccinate" form is a model).
— Do not dismiss families solely for refusal in most ethical frameworks (AAP recommends continued engagement).
— Adverse outcomes from missed vaccination disproportionately affect under-resourced communities — address access barriers (transportation, cost, language).
Step 3 management: Document refusal with specific risks discussed, schedule a follow-up to revisit, and capture any opportunity to vaccinate at hospital discharge, ED visit, or urgent care encounter — never wait for the "perfect" wellness visit.
Board pearl: When the answer choices include both "vaccine efficacy" and "vaccine effectiveness," the choice hinges on study design: randomized → efficacy; observational/surveillance → effectiveness. This single discrimination decides many Step 3 biostat items.
"In a phase 3 trial, 4/10,000 vaccinated and 80/10,000 placebo developed disease. What is vaccine efficacy?"
→ VE = (1 − 4/80) × 100% = 95%.
Parent says "90% effective means my child has a 10% chance of disease." Best response?
→ Explain VE as relative reduction; provide absolute risk context; recommend vaccination.
RCT showed 95%; post-marketing surveillance shows 70%. Best explanation?
→ Real-world population heterogeneity, waning, variant drift, or cold-chain — not "the trial was wrong."
"Best design to estimate seasonal influenza vaccine effectiveness?"
→ Test-negative case-control.
Unvaccinated infants have lower pertussis rates in a high-coverage community.
→ Herd/indirect protection (community immunity).
HIV CD4 180, due for MMR.
→ Contraindicated; defer until CD4 ≥200 sustained.
COVID-19 effectiveness drops from 92% to 55% over 6 months. Mechanism?
→ Waning humoral immunity ± variant escape; severe-disease VE remains higher.
Observational study shows vaccinated have 80% lower all-cause mortality. Concern?
→ Healthy-vaccinee bias / confounding by indication — implausibly large effect on non-vaccine-related outcomes.
VAERS reports increase after a new vaccine — does this prove causation?
→ No; need VSD or controlled study with denominator.
Hospitalized 68-year-old with CAP, never had pneumococcal vaccine.
→ Order PCV20 at discharge, document, arrange PCP follow-up.
Key distinction: Step 3 rewards identification of bias, study design, and management of the next step — not just the calculation. Always ask: what design produced this number, and what does that imply about its interpretation?
Vaccine efficacy is the relative risk reduction measured under the controlled, randomized conditions of a phase 3 trial, while vaccine effectiveness is the analogous real-world reduction measured in observational, often test-negative case-control studies — and the gap between them reflects population heterogeneity, waning immunity, variant drift, cold-chain and adherence imperfections, not vaccine failure.
Board pearl: If the stem mentions randomization, blinding, or a phase 3 trial → efficacy. If it mentions surveillance, registries, EHR data, or test-negative design → effectiveness. The number after the percent sign is almost never the most important part of the answer — the design and its biases are.