Reference

Glossary

Key terms used across the course: short definitions, with a link to the lesson that teaches each one in full when it is ready.

2×2 contingency table

The four-cell table behind every effect measure: two groups (treatment/control) by two outcomes (event/no event). Risk, RR, RD, NNT and OR are all just different reads of these four numbers.

Full lesson → M3 · Measures of effect: RR, OR, RD, NNT

Absence flag

The hardest red flags are things that aren't there: a missing standard-of-care comparator, an absent worst-case scenario, a dropped subgroup analysis. Readable only if you carry a full map of what an honest dossier should contain.

Full lesson → M14 · Red flags: the compressed course

Absolute risk reduction (ARR)

The actual drop in risk in percentage points (control risk − treatment risk), the same thing as the risk difference. Anchored to how many people per hundred are really affected; NNT = 1 ÷ ARR.

Full lesson → M3 · Relative vs absolute risk

Absolute vs relative effect

The same result expressed two ways: a relative change (e.g. '50% lower') versus an absolute one (e.g. '1 percentage point lower'). Relative figures sound bigger; absolute ones show real impact and depend on the baseline risk.

Full lesson → M3 · Relative vs absolute risk

Added-benefit assessment

Valuing a drug by how much better it is than the appropriate standard comparator (none/minor/considerable/major), rather than by a cost-per-QALY threshold, then negotiating price on that basis. The core of Germany's AMNOG model.

Full lesson → M12 · HTA agencies around the world

Adjustment (for confounding)

Statistical techniques that try to remove a confounder's influence after data is collected, used when randomisation isn't possible. Includes methods like propensity scores. (Full treatment in M11.)

Full lesson: M11

Adversarial setting

A room of parties with divergent goals: manufacturer (wants reimbursement), payer (wants control), clinicians/patients (want access), and the committee (wants a good decision). The assessor is the one voice without a stake.

Full lesson → M14 · In the room: defending judgement, not winning arguments

Advocacy analysis

An analysis that is technically correct yet designed to persuade, like a witness for one side, where every sentence may be true but the whole is shaped to win. Read for its direction, not for errors.

Full lesson → M14 · Reading the manufacturer's case: follow the levers

Affordability

Whether a payer's fixed budget can absorb the total cash cost of adopting a technology. Separate from cost-effectiveness: a good-value technology can still be unaffordable if the population is large.

Full lesson → M10 · Budget impact: a different question

Allocation concealment

Hiding the next treatment assignment until a patient is irreversibly enrolled, so no one can steer who ends up in which group. Distinct from both randomisation and blinding.

Full lesson → M2 · Randomisation and blinding

Assessment vs appraisal

Assessment = the technical evaluation of the evidence (what does the data show?). Appraisal = the value judgement and decision (given that, what do we do?). Some systems keep them in one body; others deliberately separate them (and pricing too).

Full lesson → M12 · HTA agencies around the world

Attrition bias

A systematic tilt from who drops out. If the people who quit a study differ from those who stay, the survivors give a misleading answer.

Full lesson → M2 · Bias, chance and confounding

Base case

The model's central set of input values, and the single ICER they produce. The reference point that sensitivity analysis varies inputs away from.

Full lesson → M9 · One-way sensitivity analysis & the tornado

Base-case reconstruction

The assessor's core move: flip every lever in a submission to its impartial, best-justified setting and re-run, then show the committee both the manufacturer's ICER and the reconstructed one. The gap is the assessment.

Full lesson → M14 · Reading the manufacturer's case: follow the levers

Baseline characteristics

A table comparing the trial groups before treatment begins, used to check that randomisation actually produced similar groups, especially important in smaller trials.

Full lesson → M2 · Randomisation and blinding

Baseline risk

How common the event is without treatment (the control-group risk). The missing number that decides whether a relative reduction translates into a large or negligible absolute benefit.

Full lesson → M3 · Relative vs absolute risk

Bias

A systematic tilt built into how a study is designed, run, or measured, pushing the result away from the truth, distinct from random luck.

Full lesson → M2 · Bias, chance and confounding

Blinding

Keeping patients, carers, or assessors unaware of who received which treatment, so their knowledge can't tilt how outcomes are delivered or measured. The main defence against information bias.

Full lesson → M2 · Randomisation and blinding

Bonferroni correction

The simplest fix for multiple testing: divide the 0.05 threshold by the number of tests, so the overall false-alarm rate stays at 5%. Run 10 tests, require p < 0.005 for each.

Full lesson: M3

Bottom-up costing

Building an episode's cost up from its individual resources, the philosophy behind micro-costing; precise about what drives cost for whom.

Full lesson → M5 · Costing: top-down vs bottom-up

Bucher method

The basic anchored indirect comparison: subtract each treatment's effect-versus-the-shared-comparator to get their relative effect.

Full lesson → M4 · Indirect comparisons and network meta-analysis

Budget impact analysis (BIA)

An analysis of the total cash cost to a payer's budget of adopting a technology across the eligible population, over a short horizon (usually 1–5 years). Answers "can we afford it?", distinct from cost-effectiveness, which asks "is it good value?"

Full lesson → M10 · Budget impact: a different question

Burden of disease

The total health lost to a condition in a population, typically measured in DALYs; a prioritisation lens rather than an allocation one.

Full lesson → M6 · The DALY and how it mirrors the QALY

Case series / case report

A description of one patient or a handful, with no comparison group. Can raise a first alarm, but cannot show cause. The bottom of the evidence ladder.

Full lesson → M2 · Types of study

Case-control study

An observational study that starts from the outcome, people with a disease (cases) versus without (controls), and looks backward at past exposure. Efficient for rare diseases.

Full lesson → M2 · Types of study

Causal inference

Drawing conclusions about cause and effect from data. In observational data (no randomisation) it means trying to recover a drug's true effect from the confounded comparison of treated vs untreated, an argument, not proof.

Full lesson → M11 · Causal inference outside the RCT

CDA-AMC (Canada)

Canada's Drug Agency (formerly CADTH, renamed 2024). A pan-Canadian reimbursement recommendation using full economic evaluation and the QALY; Quebec runs its own (INESSS). Feeds joint price negotiation.

CE mark

The EU conformity marking that lets a medical device be sold. Historically it demanded far less evidence than a drug's marketing authorisation (often observational data, surrogate outcomes), so device HTA starts from a thinner evidence base.

Censoring

When a patient is observed without the event up to some point: the study ends or they're lost to follow-up. Not missing data: it carries the truth 'survived at least this long,' which survival analysis uses.

Full lesson → M3 · Survival analysis

Central limit theorem

Averages tend toward a normal (bell) curve as the sample grows, almost regardless of the underlying data's shape.

Full lesson → M3 · The standard error

Certainty of evidence

How confident we are that the true effect lies close to the estimated effect, GRADE's four-level judgement, rated per outcome.

Full lesson → M4 · GRADE and the certainty of evidence

Chance (random error)

Misleading results that arise from pure luck, especially in small samples. Larger numbers reduce it. Distinct from bias, which a bigger sample won't fix.

Full lesson → M2 · Bias, chance and confounding

Claims data

The billing records a payer generates to pay for care. Vast and near-complete on what was billed, spanning years, but clinically thin (billing codes, not outcomes) and collected to settle payments, not answer research questions.

Full lesson → M11 · Real-world evidence: sources and why

Clinical significance

Whether an effect is big enough to actually matter to patients, separate from statistical significance, which only asks whether it's likely real. Large samples can make trivial effects statistically significant.

Full lesson → M3 · Significance and power

Clinical study report (CSR)

The full, detailed report of a trial submitted to regulators, far more complete than the journal paper, and a key HTA source when obtainable.

Full lesson: M12

Cochran's Q

A precision-weighted sum of each study's squared distance from the pooled estimate; the raw test statistic for heterogeneity.

Full lesson → M4 · Heterogeneity and I²

Cohort study

An observational study that enters people by their exposure and follows them forward in time to see who develops the outcome. Can be prospective or retrospective: the logic runs exposure → outcome either way.

Full lesson → M2 · Types of study

Collingridge dilemma

Knowledge and influence run opposite ways over time: early on you can shape a technology but know little about it; late on you know a lot but can change almost nothing. The two curves cross.

Full lesson → M13 · Early HTA, horizon scanning & patient involvement

Common comparator

A treatment (often placebo) that appears in both trials and serves as the shared anchor linking two otherwise unconnected experiments.

Full lesson → M4 · Indirect comparisons and network meta-analysis

Companion diagnostic

A test (usually for a biomarker) that exists to decide who gets a specific drug, where the drug works only in test-identified patients. Test and drug are inseparable and are assessed as a single pair.

Full lesson → M13 · HTA of diagnostics: the test-treatment pathway

Comparator

What a new treatment is measured against: usually the current standard of care, not nothing. HTA almost always asks 'better than what we already do?'

Full lesson → M1 · Framing the question: PICO

Compensatory aggregation

A weighted sum lets a high score on one criterion offset a low score on another. Usually fine, but dangerous when a low score should be disqualifying, not compensable (safety, equity floors).

Full lesson → M13 · Multi-criteria decision analysis (MCDA)

Confidence interval

A range of plausible values for the true effect: the estimate ± about 2 standard errors (for 95%). Its width shows the uncertainty; whether it excludes the no-effect value mirrors statistical significance.

Full lesson → M3 · Confidence intervals

Confounding

A hidden third factor that influences both the supposed cause and the effect, creating a misleading link, like joggers living longer partly because they're also younger and richer.

Full lesson → M2 · Bias, chance and confounding

Confounding by indication

Confounding that arises because treatment is chosen for reasons tied to prognosis: fitter (or sicker) patients get the drug non-randomly, so a naïve comparison mixes the drug's effect with the effect of who received it.

Full lesson → M11 · Causal inference outside the RCT

Consistency

In a network with closed loops, the agreement between the direct and indirect estimates of the same comparison; disagreement signals a transitivity problem.

Full lesson → M4 · Indirect comparisons and network meta-analysis

CONSORT

A reporting standard for randomised controlled trials: it governs how completely and transparently a single trial is described. The trial-level counterpart to PRISMA. Distinct from risk of bias.

Contour-enhanced funnel plot

A funnel plot overlaid with statistical-significance regions, used to judge whether a gap sits where non-significant studies would fall.

Full lesson → M4 · Publication bias and the funnel plot

Cost versus price

Cost is the resources consumed; price is what is paid for them. Costing measures resources, not payments; the two can diverge sharply.

Full lesson → M5 · Costing: top-down vs bottom-up

Cost-benefit analysis (CBA)

An evaluation that converts the health effect itself into money, allowing health to be weighed against non-health spending, the most complete and the most politically avoided in HTA.

Full lesson → M5 · The four types of economic evaluation

Cost-effectiveness acceptability curve (CEAC)

A curve built from a PSA cloud showing, for every possible threshold, the probability a technology is cost-effective (threshold on the x-axis, probability on the y-axis). Its height is certainty, not size of benefit.

Full lesson → M9 · Cost-effectiveness acceptability curves

Cost-effectiveness analysis (CEA)

An economic evaluation that measures effect in a natural clinical unit: cost per life-year, per mmHg, per event avoided, direct but not comparable across diseases.

Full lesson → M5 · The four types of economic evaluation

Cost-effectiveness plane

A chart placing a technology against its comparator: horizontal axis = incremental health (QALYs), vertical axis = incremental cost. The quadrant it lands in frames the whole decision.

Full lesson → M7 · The cost-effectiveness plane

Cost-effectiveness threshold

The maximum cost per QALY a health system treats as good value. Framed correctly it's an opportunity cost (the health the same money would produce elsewhere), not a measure of what a QALY is worth. NICE: £25,000–£35,000 (from April 2026).

Full lesson → M7 · Cost-effectiveness thresholds and willingness to pay

Cost-minimisation analysis (CMA)

An evaluation valid only when two options have proven-equal effect, so the decision reduces to which is cheaper; frequently misused when equality is assumed rather than demonstrated.

Full lesson → M5 · The four types of economic evaluation

Cost-plus pricing

Pricing a product at its production cost plus a margin, the ordinary-goods intuition. It fails for medicines, where manufacturing a dose is cheap and the price is driven by value delivered and by R&D, not by unit cost.

Full lesson → M12 · Pricing: reference and value-based

Cost-utility analysis (CUA)

An economic evaluation that measures effect in QALYs, making cost-per-QALY comparable across any disease, the HTA reference-case default.

Full lesson → M5 · The four types of economic evaluation

Coverage with evidence development (CED)

A conditional approval: fund a promising but uncertain technology provisionally while real-world evidence is collected, then confirm, renegotiate, or withdraw. Uses RWE to resolve the uncertainty that made the decision hard.

Full lesson → M11 · RWE in real decisions

Cox regression

The standard model used to estimate a hazard ratio from survival data, adjusting for other factors. It assumes proportional hazards. (More in M11.)

Full lesson: M11

Criterion (MCDA)

One explicit dimension of value in an MCDA, e.g. health gain, severity, unmet need, equity, innovation. Options are scored on each criterion; criteria must be independent to avoid double-counting.

Full lesson → M13 · Multi-criteria decision analysis (MCDA)

Cross-sectional study

An observational snapshot measuring exposure and outcome at a single point in time. Shows what travels together, but never which came first.

Full lesson → M2 · Types of study

DALY (disability-adjusted life year)

A unit of health LOST to disease, years of life lost to early death plus years lived with disability. Low is good; the mirror of the QALY.

Full lesson → M6 · The DALY and how it mirrors the QALY

Decision tree

The simplest economic model: decision nodes (choices you control) and chance nodes (outcomes nature controls) branching to leaves that carry a cost and a QALY payoff. Solved by rolling back to expected values. Fits acute, one-off, short-horizon decisions; has no time dimension.

Full lesson → M8 · Decision trees

Diagnostic 2×2 table

The four-cell table crossing test result (positive/negative) with disease status (present/absent): TP, FP, FN, TN. Every diagnostic statistic (sensitivity, specificity, PPV, NPV, LR) reads from these four cells.

Full lesson → M3 · Diagnostic tests

Differential discounting

Applying a lower discount rate to health effects than to costs, to avoid systematically penalising prevention and durable therapies.

Full lesson → M5 · Discounting and the time horizon

Direct non-medical costs

Real treatment-caused spending outside the clinic: patient travel, home adaptations, special diets.

Full lesson → M5 · Perspective: payer vs societal

Disability weight

A 0–1 weight for how much health a state destroys (0 = full health, 1 = death); derived from health-loss judgements, and NOT equal to 1 − utility.

Full lesson → M6 · The DALY and how it mirrors the QALY

Discount rate

The annual rate (e.g. NICE's 3.5%) at which future values are shrunk to present value; a high-leverage, contested parameter.

Full lesson → M5 · Discounting and the time horizon

Discounting

Converting future costs and health effects into their worth today, because a value arriving later counts for less than the same value now.

Full lesson → M5 · Discounting and the time horizon

Discrete event simulation (DES)

An individual-level model that simulates patients one at a time in continuous time, moving from event to event sampled from distributions. Each patient carries a history, so it has memory and no cycles. Flexible but data-hungry and less transparent.

Full lesson → M8 · Partitioned survival models

Distribution

The pattern of how often each value occurs across many measurements, the shape data makes when you stack it up. Read it for where values centre and how widely they spread.

Full lesson → M3 · Distributions and the normal curve

Distributional cost-effectiveness analysis (DCEA)

Instead of one population-wide sum of QALYs, DCEA models how health is distributed across groups (income, region, ethnicity…) and shows whether a technology narrows or widens health gaps.

Full lesson → M13 · Equity in HTA

Dominance

When one option beats another on both axes at once. Dominant = more effective AND cheaper (adopt); dominated = less effective AND costlier (reject). Neither needs a cost-effectiveness threshold.

Full lesson → M7 · The cost-effectiveness plane

Double-counting

Counting the same value twice in an MCDA, e.g. severity captured inside the QALY gain AND again as its own criterion. Criteria must be genuinely independent, or the weighted sum is rigged.

Full lesson → M13 · Multi-criteria decision analysis (MCDA)

Early HTA

Applying HTA thinking early in a technology's life (before evidence matures or price is set), not to judge it but to shape its development and evidence toward reimbursability. Produces direction, not a verdict.

Full lesson → M13 · Early HTA, horizon scanning & patient involvement

Economic evaluation

The comparison of two or more options by both their costs and their effects, the family of methods (CMA, CEA, CUA, CBA) that differ only in how they measure effect.

Full lesson → M5 · The four types of economic evaluation

Economic model

A structured set of assumptions that turns trial evidence into the lifetime costs and QALYs a decision needs, bridging the gaps a trial leaves open (time, endpoint, evidence, population).

Full lesson → M8 · Why a model? Extrapolation beyond the trial

Effect modifier

A patient characteristic that changes the size of a treatment's effect; if it's distributed differently across trials, it breaks an indirect comparison.

Full lesson → M4 · Indirect comparisons and network meta-analysis

Effective sample size

After reweighting patients (as in MAIC), the equivalent number of unweighted patients the analysis is really based on, often far smaller than the raw count.

Full lesson → M4 · Indirect comparisons and network meta-analysis

Effectiveness

How well a treatment works in routine, real-world practice: ordinary patients, imperfect adherence, measured against the care people already receive. What HTA needs to know.

Full lesson → M1 · Efficacy vs effectiveness

Efficacy

How well a treatment works under ideal, controlled conditions: selected patients, perfect adherence, expert centres. What a regulatory trial is built to show.

Full lesson → M1 · Efficacy vs effectiveness

Efficacy vs effectiveness

Efficacy = can an intervention work under ideal, controlled conditions (an RCT's question). Effectiveness = does it work in real patients under routine care (real-world evidence's question). The real-world effect is usually smaller, the efficacy-effectiveness gap.

Full lesson → M11 · Real-world evidence: sources and why

Egger's test

A regression test for funnel-plot asymmetry; it detects small-study effects in general and is unreliable with fewer than 10 studies or for odds ratios.

Full lesson → M4 · Publication bias and the funnel plot

Electronic health records (EHR)

Clinical notes and results from routine care. Clinically rich (labs, diagnoses, prescriptions) but messy, incomplete, and unstandardised, captured for delivering care, not for analysis.

Full lesson → M11 · Real-world evidence: sources and why

Endpoint

The outcome a study measures to judge whether a treatment worked; the first choice that governs every downstream number.

Full lesson → M6 · Endpoints: hard vs surrogate

EPAR

European Public Assessment Report, the EMA's published assessment of a medicine, a source of trial data beyond journal publications.

Full lesson: M12

EQ-5D

The standard utility instrument: a patient describes their state across five dimensions, and a country-specific value set converts that description into a utility.

Full lesson → M6 · Where utility numbers come from

Equity (in HTA)

Fairness in how health is distributed: who gains, who loses, whether gaps between groups widen or close, as an explicit concern alongside efficiency. Standard cost-per-QALY is blind to it: it maximises total health and ignores distribution.

Full lesson → M13 · Equity in HTA

Equity weight

An explicit multiplier that values a QALY gained by the worse-off more than one gained by the better-off (e.g. ×1.5), formalising prioritarianism. Openly buys some equity at the cost of some efficiency.

Full lesson → M13 · Equity in HTA

EU HTA Regulation (HTAR)

Regulation (EU) 2021/2282, operational from January 2025. Creates a permanent EU framework for joint HTA work (chiefly the Joint Clinical Assessment), pooling the clinical evaluation while leaving value, pricing, and reimbursement to member states.

Full lesson → M12 · EU HTA Regulation & Joint Clinical Assessment

EUnetHTA

The earlier, voluntary, project-based cooperation between European HTA bodies, now replaced by the permanent legal structure of the EU HTA Regulation.

European Medicines Agency (EMA)

The EU body that assesses whether a medicine's benefits outweigh its risks and grants marketing authorisation. It decides whether a drug may be sold, which is a separate question from HTA's 'is it better, and is it worth it?'

Evidence to decision

The step from graded evidence to an actual recommendation, which in HTA also weighs cost, budget impact, and equity, not just certainty.

Full lesson: M14

Expected value

The probability-weighted average of a set of outcomes: Σ (probability × payoff). Rolling back a decision tree means replacing each chance node with its expected value.

Full lesson → M8 · Decision trees

Expected value of perfect information (EVPI)

The expected value of perfect information: the most it would be worth to eliminate all parameter uncertainty before deciding. Equals the probability of a wrong decision times its consequence, scaled by the patient population.

Full lesson → M9 · The value of information (EVPI)

Expected-value decision rule

The principle that a cost-effectiveness decision is made on the mean (expected) net benefit of the PSA cloud, not on whether its probability of being cost-effective clears some bar like 50% or 95%.

Full lesson → M9 · Cost-effectiveness acceptability curves

Expert witness role

In the committee room you shift from author (who controls the document) to expert witness (who is questioned). Your job is to inform the committee's decision as a disinterested adviser, not to argue for an outcome as a party.

Full lesson → M14 · In the room: defending judgement, not winning arguments

Explanatory trial

The classic RCT: randomised under ideal, controlled conditions (narrow patients, strict protocol, often placebo) to measure efficacy: can it work? The contrast to a pragmatic trial, which measures whether it works in practice.

Full lesson → M11 · Pragmatic trials

External control arm

Using real-world or historical data on untreated patients as the comparison for a single-arm study, often the only option in diseases too rare for a randomised trial, but carrying full confounding risk.

Full lesson → M11 · RWE in real decisions

External validity

Whether a model's outputs match data it was NOT built from: a different trial, a registry, or the same trial's mature follow-up. Reproducing the source data is not external validation.

Full lesson → M8 · Model validation

External validity (generalisability)

How well a study's result carries over to patients outside the trial: older, sicker, or otherwise different from its sample. A true result can still fail to generalise.

Full lesson → M2 · Internal vs external validity

Extrapolation

Projecting an outcome (usually a survival curve) beyond the end of the trial data, out to the full lifetime horizon. A leap beyond the evidence: many curves fit the observed data yet diverge in the unobserved tail.

Full lesson → M8 · Why a model? Extrapolation beyond the trial

Face validity

Whether a model's structure and outputs are clinically plausible to an expert. Often the only check that catches a confidently-computed but false assumption.

Full lesson → M8 · Model validation

Fact–judgement separation

Marking where the evidence ends ('the ICER is £34,000') and the value call begins ('acceptable given severity'), so a decision-maker can accept the facts while applying their own values, the decision that's rightfully theirs.

Full lesson → M14 · Writing the recommendation: the decision-maker's tool

FDA (US Food and Drug Administration)

The US medicines regulator. Like the EMA, it grants marketing authorisation (deciding whether a drug may be sold), which is separate from any decision to fund or reimburse it.

Financial-based agreement

A managed-entry deal that shares BUDGET risk: the drug's value is accepted, but total cost is uncertain. Includes simple/confidential discounts, price-volume agreements, spend caps, and dose/duration caps. Says nothing about whether the drug works.

Full lesson → M12 · Risk-sharing and managed entry agreements

Fixed-effect model

A meta-analysis model assuming all studies estimate one single true effect, differing only by chance.

Full lesson → M4 · Fixed vs random effects

Forest plot

A chart showing each study's effect estimate as a square with its confidence interval as a horizontal whisker, plotted against a null line, the standard display for meta-analyses. Square size reflects study weight; the diamond at the bottom shows the pooled result.

Full lesson → M4 · Reading a forest plot

Formulary listing

The moment a reimbursement 'yes' becomes operational: the drug enters the funded list, with its indication, any subgroup restrictions, and its (often confidential) price specified, the point at which a patient can actually be prescribed it at public expense.

Full lesson → M12 · The reimbursement process

Friction-cost approach

A method that values lost productivity only until a sick worker is replaced, yielding a figure many times smaller than the human-capital approach.

Full lesson → M5 · Perspective: payer vs societal

Funnel plot

A scatter plot of each study's effect against its precision, used to eyeball whether an evidence base looks complete and symmetric.

Full lesson → M4 · Publication bias and the funnel plot

Global Burden of Disease (GBD)

The large study measuring health lost to disease worldwide in DALYs; source of standard life expectancies and disability weights.

Full lesson → M6 · The DALY and how it mirrors the QALY

GRADE

A structured system for rating how much certainty we have in an effect estimate for a given outcome, on four levels from High to Very Low.

Full lesson → M4 · GRADE and the certainty of evidence

Half-cycle correction

An adjustment for the timing error in cycle-based models: accruing a cycle's costs and QALYs from start-of-cycle occupancy overcounts, end-of-cycle undercounts. Centring the accounting on each cycle's midpoint (the trapezoidal rule) fixes it.

Full lesson → M8 · Half-cycle correction

Hard endpoint

An outcome that matters directly to the patient: death, stroke, quality of life, as opposed to a stand-in marker.

Full lesson → M6 · Endpoints: hard vs surrogate

HAS (France)

France's HTA body. Rates clinical benefit first: whether to reimburse (SMR) and how much of an improvement over existing options (ASMR), with the ASMR level driving price negotiation.

Hazard

The instantaneous rate of an event among those who have survived to that moment, a 'speed' of events right now, not a cumulative total. The building block of the hazard ratio.

Full lesson → M3 · Hazard ratios

Hazard ratio (HR)

A single-number summary comparing the event rate (hazard) between two groups over time; HR < 1 means benefit. It states a relative rate, not extra time or extra survivors, and assumes a constant effect (proportional hazards).

Full lesson → M3 · Hazard ratios

Health inequality (vs inequity)

A health inequality is any difference in health between groups; a health inequity is a difference that is unfair or avoidable. Not every inequality is unjust, but the ones efficiency deepens, along lines of disadvantage, usually are.

Full lesson → M13 · Equity in HTA

Health state

A described condition of health (e.g. moderate disability with pain) assigned a single utility value for QALY calculation.

Full lesson → M6 · The QALY as a construct

Heterogeneity

Variation among study results beyond what chance (sampling error) alone would produce, i.e. real differences in the underlying effect.

Full lesson → M4 · Heterogeneity and I²

Heterogeneity (between patients)

Genuine variation between patients (e.g. by age or risk), a real feature of the world, addressed with subgroups. Not the same as parameter uncertainty, which is about how well we know a value.

Full lesson → M9 · Probabilistic sensitivity analysis

Hierarchy of evidence

A ranking of study designs by how well they guard against being fooled: anecdote at the bottom, systematic reviews of trials at the top. A starting presumption of strength, not a verdict.

Full lesson → M2 · The hierarchy of evidence

Highly Specialised Technologies (HST)

NICE's separate appraisal route for treatments of very rare conditions, using a much higher willingness-to-pay threshold (~£100,000–300,000 per QALY) and tolerating more uncertainty, but with narrow entry criteria, so most orphan drugs don't qualify.

Full lesson → M12 · Orphan drugs and rare diseases

HITAP (Thailand)

Thailand's HTA agency, a leading example from a middle-income country, with methods and a threshold adapted to its context, and wide influence on HTA in lower- and middle-income countries.

Horizon scanning

Systematically identifying technologies coming down the line before they arrive, so systems can prepare. Part of looking forward rather than assessing after the fact.

Full lesson → M13 · Early HTA, horizon scanning & patient involvement

HTA agency

A body that runs health technology assessment for a health system. Not one model but a family: differing on whether they use a cost-per-QALY threshold, and on who assesses evidence, decides, and sets price.

Full lesson → M12 · HTA agencies around the world

HTA report

The written assessment that turns analysis into a decision-ready recommendation. Not documentation of what you did, but a decision tool for an overloaded reader: ordered by importance, calibrated for uncertainty, with fact and judgement kept separate.

Full lesson → M14 · Writing the recommendation: the decision-maker's tool

Human-capital approach

A method that values lost productivity at the person's full expected earnings for the whole time they're off work, potentially decades.

Full lesson → M5 · Perspective: payer vs societal

The percentage of total variation across studies attributable to real heterogeneity rather than chance. A proportion, not a magnitude.

Full lesson → M4 · Heterogeneity and I²

ICER

The incremental cost-effectiveness ratio: ΔCost ÷ ΔEffect, in £ per QALY (the extra cost of one additional unit of health versus the comparator).

Full lesson → M7 · The ICER — calculation and interpretation

ICER (Institute for Clinical and Economic Review, US)

An independent, non-profit US body with NO decision-making power: it publishes value-based benchmark prices that inform payers' negotiations. (Not to be confused with the ICER cost-effectiveness metric from Module 7; unrelated, same initials.)

Impartiality as credibility

In an adversarial hearing, your authority comes not from being right but from being disinterested. Impartiality is pure perception: it lasts only as long as you visibly behave like someone with no stake, and 'winning' the argument spends it.

Full lesson → M14 · In the room: defending judgement, not winning arguments

Imprecision

A GRADE downgrade domain: the estimate is too uncertain to act on: a wide confidence interval, too few events, or a CI crossing the decision threshold.

Full lesson → M4 · GRADE and the certainty of evidence

Incidence

How often new cases appear over a period of time, the rate at which people develop a condition. What a cohort, following people forward, can measure. (Full treatment in M3.)

Full lesson: M3

Incidence vs prevalence

Prevalence = everyone who already has the condition (a built-up pool, often a one-off surge when a drug launches). Incidence = new cases each year (a steady recurring inflow). The mix sets the shape of the budget stream.

Full lesson → M10 · Population, uptake, and horizon

Incremental cost and effect (ΔCost, ΔEffect)

The difference a new technology makes versus its comparator: extra cost (ΔCost, £) and extra health (ΔEffect, QALYs). Every cost-effectiveness result is built from this pair of differences.

Full lesson → M7 · The cost-effectiveness plane

Indirect costs

The economic value of work not done because of illness: sick leave, reduced output, early retirement; usually the largest and most contested cost bucket.

Full lesson → M5 · Perspective: payer vs societal

Indirect treatment comparison

Estimating how two treatments compare when no trial tested them head-to-head, by using a comparator both were separately tested against.

Full lesson → M4 · Indirect comparisons and network meta-analysis

Indirectness

A GRADE downgrade domain: any gap between the evidence you have and the question you're asking: wrong population, comparator, or a surrogate instead of a clinical outcome.

Full lesson → M4 · GRADE and the certainty of evidence

Informal care costs

The economic value of unpaid care from family and friends, real even though no invoice is raised.

Full lesson → M5 · Perspective: payer vs societal

Information (measurement) bias

A systematic tilt from how things are measured. If one group is measured differently from another, the comparison is distorted. Recall bias is one variety.

Full lesson → M2 · Bias, chance and confounding

Intention-to-treat (ITT)

Analysing trial participants in the group they were randomised to, regardless of what they actually did, preserving randomisation's protection against bias. (Full treatment later, in M4.)

Full lesson: M4

Internal validity

How confident we can be that a study's result is true for the people actually in it: that the effect is real and not an artefact of bias or confounding.

Full lesson → M2 · Internal vs external validity

Internal validity (verification)

Checking a model computes what it claims to, free of implementation bugs: via zero/extreme-input tests, probability sums, and reproducing hand calculations. Says nothing about whether the assumptions are true.

Full lesson → M8 · Model validation

International reference pricing (IRP)

External reference pricing: a country pegs a drug's price to a basket of other countries' prices (average, lowest, or median). Because only list prices are visible, IRP references prices no one actually pays, driving launch sequencing and resistance to cutting headline prices.

Full lesson → M12 · Pricing: reference and value-based

Inverted pyramid

Ordering a report by descending importance: recommendation and key conditions first, support beneath, detail last. Because the reader stops when they have enough, order determines which findings actually reach them.

Full lesson → M14 · Writing the recommendation: the decision-maker's tool

IQWiG + G-BA (Germany)

Germany's AMNOG system, deliberately separated: IQWiG assesses a drug's added benefit vs the appropriate comparator, the G-BA rates it, and the insurance funds (GKV-SV) negotiate price. No QALY threshold.

Joint Clinical Assessment (JCA)

Under the EU HTA Regulation, a single shared scientific assessment of a technology's relative clinical effectiveness and safety, done once at EU level, while cost-effectiveness, value, price, and reimbursement stay national. Non-binding.

Full lesson → M12 · EU HTA Regulation & Joint Clinical Assessment

Kaplan-Meier curve

A step-shaped survival curve: it drops at each event in proportion to the patients still at risk, and only marks (doesn't drop at) each censoring. The standard way to display time-to-event data.

Full lesson → M3 · Survival analysis

Learning curve

The improvement in a device's results as operators gain experience. The effect is a curve, not a point: early trials undersell a device, expert-centre trials oversell what an average hospital will get in year one.

Full lesson → M13 · HTA of medical devices

Likelihood ratio (LR)

How much a test result updates the odds of disease: LR+ = sensitivity ÷ (1 − specificity); LR− = (1 − sensitivity) ÷ specificity. LR+ > 10 or LR− < 0.1 makes a meaningful difference. Does not depend on prevalence.

Full lesson → M3 · Diagnostic tests

Linked evidence

Valuing a diagnostic by stitching together separate evidence (the test's accuracy studies plus the treatment's own trials) across a modelled pathway, because one study rarely runs all the way from test to health outcome.

Full lesson → M13 · HTA of diagnostics: the test-treatment pathway

List price vs effective price

List price is the public headline price; effective price is the real, lower price paid after a confidential discount. Decisions are made on the effective price, but public ICERs and budget impacts are often computed on the list price.

Full lesson → M10 · Affordability and the payer's decision

Log-rank test

A statistical test for whether two survival curves differ overall. Commonly paired with the hazard ratio. (More in M3.)

Full lesson: M3

Logistic regression

A statistical model for yes/no outcomes that naturally produces odds ratios, which is part of why odds ratios are so common in the literature. (More in M11.)

Full lesson: M11

Managed entry / risk-sharing agreement (MEA / RSS)

A contract that lets a payer say 'yes, conditionally' to an uncertain drug by sharing a specific risk: financial (budget/volume) or outcomes (does it work?). Also called a risk-sharing scheme (RSS). Umbrella for price-volume, spend caps, payment-by-results, and coverage-with-evidence.

Full lesson → M12 · Risk-sharing and managed entry agreements

Manufacturer submission

The economic analysis a company submits to get its technology reimbursed. Not fraud and not a neutral report; it's advocacy: competent, usually error-free, but built so that every permissible choice favours the drug.

Full lesson → M14 · Reading the manufacturer's case: follow the levers

Marketing authorisation

A medicines regulator's decision (EMA in Europe, FDA in the US) that a drug is safe and effective enough to be sold. It permits sale; it does NOT mean any health system will pay for it. Reimbursement is a separate, later process.

Full lesson → M12 · The reimbursement process

Markov assumption (memorylessness)

The rule that a patient's next move depends only on their current state, not on how they got there or how long they've been there. Keeps the model small; often its central simplification.

Full lesson → M8 · Markov models

Markov model

A model that represents a disease as a small set of health states, with patients moving between them each cycle by transition probabilities. Its defining feature is memorylessness: where a patient goes next depends only on their current state. Suited to chronic, recurring disease.

Full lesson → M8 · Markov models

Matching-adjusted indirect comparison (MAIC)

An indirect comparison that reweights one trial's individual patient data so its population characteristics match the competitor trial's reported averages before comparing.

Full lesson → M4 · Indirect comparisons and network meta-analysis

Mean (average)

The arithmetic average, the balance point of a distribution. Fully informative for symmetric data, but on skewed data it gets dragged toward the long tail and can mislead.

Full lesson → M3 · Variation and uncertainty

Median

The middle value when data is lined up in order: half above, half below. More honest than the mean for skewed data like costs or survival, because outliers don't drag it.

Full lesson → M3 · Distributions and the normal curve

Median survival

The time at which the survival curve crosses 50%: half the patients have had the event by then. Preferred over the mean (survival is skewed); reported as 'not reached' if the curve never falls to 50%.

Full lesson → M3 · Survival analysis

Mediator

A variable on the causal path from treatment to outcome (e.g. the drug lowers blood pressure, which prevents strokes). Unlike a confounder, you must NOT adjust for a mediator: doing so removes part of the drug's real effect.

Full lesson → M11 · Causal inference outside the RCT

Medical device

A health technology that isn't a drug. Unlike a molecule, its effect isn't a stable, isolated property; it's a system of device × operator × version × time (operator-dependence, learning curves, fast versioning), which is why it can't be assessed like a drug.

Full lesson → M13 · HTA of medical devices

MeSH

Medical Subject Headings, the controlled vocabulary used to tag records in MEDLINE, so a search can match a paper by its assigned tag even when its text used different words.

Meta-analysis

The statistical combination of results from multiple studies into a single pooled estimate, weighting each study by its precision rather than averaging them equally.

Full lesson → M4 · Meta-analysis: pooling

Methodological uncertainty

Uncertainty over which analytical conventions to apply: perspective, discount rate, time horizon, comparator. Contested choices about how the analysis is framed, explored by scenario rather than distribution.

Full lesson → M9 · Scenario analysis

Micro-costing

A bottom-up costing method that builds an episode's cost from its individual resource components, one by one, precise but expensive.

Full lesson → M5 · Costing: top-down vs bottom-up

Model validation

Checking that a model is credible, not just that it runs. Four distinct questions: face (clinically plausible?), internal/verification (arithmetic correct?), external (matches data it wasn't built from?), and cross (agrees with independent models?).

Full lesson → M8 · Model validation

Monte Carlo simulation

Repeatedly drawing random samples from input distributions and running a model on each, to build up the distribution of its output. The engine behind probabilistic sensitivity analysis.

Full lesson → M9 · Probabilistic sensitivity analysis

Moving target problem

Devices iterate (v1→v2→v3, often ~every 18 months) faster than a rigorous trial can report, so by the time strong evidence arrives, it describes a discontinued version. Rigour guarantees the evidence is out of date.

Full lesson → M13 · HTA of medical devices

Multi-criteria decision analysis (MCDA)

Deciding across several dimensions of value at once: list explicit criteria, score each option on each, weight how much each matters, and aggregate into one value score. It makes visible the value-weighting that cost-per-QALY hides.

Full lesson → M13 · Multi-criteria decision analysis (MCDA)

Multiple comparisons (multiplicity)

Testing many hypotheses at once: many outcomes, subgroups, or time points, which multiplies the chance of a false alarm. Run enough tests and 'significant' results appear from pure noise.

Full lesson → M3 · Multiplicity and p-hacking

Narrative review

An expert summary of a field based on the author's own judgement of which studies matter, with no pre-declared rule for inclusion.

Full lesson → M4 · Systematic reviews

Natural history (of disease)

How an untreated condition progresses over time: survival, symptoms, stages. A descriptive question with no treatment comparison, so real-world evidence answers it well (and often is the only source).

Full lesson → M11 · RWE in real decisions

Natural units

Real clinical measures of effect used in CEA: life-years gained, mmHg lowered, events avoided, legible to clinicians but not convertible across conditions.

Full lesson → M5 · The four types of economic evaluation

Net benefit

Benefit expressed in money minus cost in money; positive means fund it. The natural output of CBA and, via the threshold, an alternative to the ICER.

Full lesson: M7

Net benefit (NMB / NHB)

Cost-effectiveness expressed as a difference instead of a ratio, using the threshold to convert cost and health into one unit. Net monetary benefit (NMB) is in £; net health benefit (NHB) is in QALYs; a positive value means cost-effective.

Full lesson → M7 · Net benefit

Network meta-analysis

A generalisation of indirect comparison across a whole web of treatments at once, combining direct and indirect evidence and estimating every treatment against every other.

Full lesson → M4 · Indirect comparisons and network meta-analysis

NICE (England)

England's HTA agency. The archetype of the threshold-plus-QALY model: an explicit £20,000–£30,000 per QALY range (£25,000–£35,000 from 2026), recommends, and the NHS implements.

Normal distribution (bell curve)

A symmetric, bell-shaped distribution that appears throughout nature, described completely by two numbers: its mean and its standard deviation. Follows the 68–95–99.7 rule.

Full lesson → M3 · Distributions and the normal curve

Null hypothesis

The default 'nothing is going on' assumption a study tests against: no effect, no difference, any gap is just chance. Statistics works by trying to embarrass it, not prove the alternative.

Full lesson → M3 · The p-value

Number needed to treat (NNT)

How many patients you must treat to prevent one event: NNT = 1 ÷ risk difference. Small NNT = powerful; large NNT = many treated per benefit. The key bridge from effect size to cost.

Full lesson → M3 · Measures of effect: RR, OR, RD, NNT

Observational study

A study that watches what happens without assigning treatment: patients (or doctors) choose. Useful and often large, but open to confounding because the groups differ from the start.

Full lesson → M2 · Types of study

Odds

Events divided by non-events (not by everyone). Distinct from risk (events ÷ everyone): a 1% risk is odds of about 1-to-99. The basis of the odds ratio.

Full lesson → M3 · Measures of effect: RR, OR, RD, NNT

Odds ratio (OR)

The ratio of the odds of an event in two groups: (a/b)÷(c/d). Close to relative risk when events are rare, but always further from 1 than RR; it overstates the effect more as events get common. The only effect measure a case-control study can give.

Full lesson → M3 · Odds and odds ratios

One-way sensitivity analysis

Varying a single input across its plausible range while holding all others at the base case, to see how far the ICER moves. Its limit: it can't capture inputs varying jointly.

Full lesson → M9 · One-way sensitivity analysis & the tornado

Operator-dependence

A device's effect depends on who uses it: effect = device × skill. Unlike a drug, the operator is part of the intervention, wrecking a trial's ability to isolate 'the device' and making blinding impossible.

Full lesson → M13 · HTA of medical devices

Opportunity cost

The health a fixed budget was already producing elsewhere, given up to fund something new. In HTA the real cost of a treatment is this displaced health, not the money spent.

Full lesson → M1 · Opportunity cost

Opportunity loss

The net benefit forgone when a decision turns out wrong for the true state of the world, the gap between what the best choice would have delivered and what the actual choice did. Zero when the decision was right.

Full lesson → M9 · The value of information (EVPI)

Orphan drug

A drug for a rare disease, a regulatory category defined by a population threshold (EU: disease affecting ≤5 in 10,000; US: <200,000 patients), carrying incentives like market exclusivity and tax credits. Small population + high R&D → very high price.

Full lesson → M12 · Orphan drugs and rare diseases

Outcomes-based agreement

A managed-entry deal that shares PERFORMANCE risk: payment is tied to whether the drug actually delivers. Includes payment-by-results and coverage-with-evidence. Elegant in theory, but demands a measurable outcome and is costly to run.

Full lesson → M12 · Risk-sharing and managed entry agreements

Overall survival (OS)

Time until death from any cause, a hard endpoint in oncology, and the benchmark PFS is meant to stand in for.

Full lesson → M6 · Endpoints: hard vs surrogate

Overhead costs

Shared running costs (management, cleaning, heating, IT, the building) that can't be traced to one patient and must be allocated across episodes. In costing, 'indirect costs' means this, distinct from the productivity sense.

Full lesson → M5 · Costing: top-down vs bottom-up

P-hacking

Trying many analyses of the same data: different outcomes, subgroups, cut-offs, and reporting only those that reach p < 0.05. Usually not fraud, but it manufactures false 'findings' from noise.

Full lesson → M3 · Multiplicity and p-hacking

P-value

The probability of getting a result at least as extreme as the one observed, if there were truly no effect. How often chance alone would fake your result, not the probability the null is true.

Full lesson → M3 · The p-value

Parameter uncertainty

Uncertainty about the true value of a model input because evidence is limited. Shrinks as better data arrives. Distinct from heterogeneity (real variation between patients).

Full lesson → M9 · Probabilistic sensitivity analysis

Partitioned survival model

A model that builds health-state occupancy directly from two survival curves: progression-free = PFS, dead = 1 − OS, progressed-but-alive = OS − PFS. No transition probabilities. Dominant in oncology; its weakness is that the curves are extrapolated independently.

Full lesson → M8 · Partitioned survival models

Patient involvement

Bringing patients' and carers' knowledge (what outcomes matter, what quality of life is like) into HTA. Most powerful early, where it can shape which endpoints a trial even measures.

Full lesson → M13 · Early HTA, horizon scanning & patient involvement

Patient registry

A real-world data source that deliberately follows a defined cohort (a disease or a drug). Clinically deep and validated, but narrow, costly, and often selective about who's included.

Full lesson → M11 · Real-world evidence: sources and why

Payer perspective

The narrowest analytical boundary: only costs and savings falling on the health budget are counted.

Full lesson → M5 · Perspective: payer vs societal

Payment-by-results

An outcomes-based agreement where the payer pays only for patients in whom the drug works, by a pre-agreed definition of 'works'. Needs a fast, clean, measurable outcome, rare in practice, and administratively heavy.

Full lesson → M12 · Risk-sharing and managed entry agreements

PBAC (Australia)

Australia's reimbursement committee, and the pioneer: Australia was the first country to make economic analysis a formal requirement for public reimbursement (1993). QALY-centric; recommends to the Minister, then price is negotiated.

Performance & detection bias

Subtler tilts in how a treatment is delivered (performance) or how outcomes are looked for and recorded (detection), often tied to lack of blinding. Part of formal risk-of-bias grading. (More in M4.)

Full lesson: M4

Perspective

The boundary an analysis draws around whose costs and savings are counted, the choice that decides what goes in the numerator, from payer-only to societal.

Full lesson → M5 · Perspective: payer vs societal

PICO

A four-slot frame for any clinical question: Population, Intervention, Comparator, Outcome. How HTA turns a vague 'does it work?' into a question evidence can actually answer.

Full lesson → M1 · Framing the question: PICO

Placebo

An inactive dummy treatment made to look identical to the real one, so patients (and often clinicians) can't tell who got what. The device that makes blinding possible. Its surgical equivalent is a sham procedure.

Full lesson → M2 · Randomisation and blinding

Point estimate

A single best-guess number from a study (like '40% of patients responded'), an estimate of a hidden true value, not the truth itself. Incomplete without a measure of its uncertainty.

Full lesson → M3 · Variation and uncertainty

Pooling

Combining the results of several studies into one weighted estimate, the core computation of a meta-analysis.

Full lesson → M4 · Meta-analysis: pooling

Population EVPI

Per-patient EVPI multiplied by the number of future patients the decision covers over its lifetime, the ceiling on what any research to resolve the uncertainty could be worth.

Full lesson → M9 · The value of information (EVPI)

Population vs sample

The population is the full, hidden truth (everyone the result is about); a sample is the handful actually studied. A study measures the sample to estimate the population.

Full lesson → M3 · Variation and uncertainty

PPV & NPV

Positive predictive value: how often a positive test is truly positive. Negative predictive value: how often a negative test is truly negative. Both depend on prevalence, unlike sensitivity and specificity.

Full lesson → M3 · Diagnostic tests

Pragmatic trial

A randomised trial run in realistic, routine-practice conditions (broad patients, usual-care comparator, patient-relevant outcomes) to measure effectiveness. Because it randomises, it defeats confounding, unlike observational real-world evidence.

Full lesson → M11 · Pragmatic trials

Pre- and post-test probability

Pre-test probability is the chance of disease before testing, usually the prevalence or a clinician's prior estimate. A test result updates this to a post-test probability via likelihood ratios or the 2×2 table.

Full lesson → M3 · Diagnostic tests

Pre-specification (& pre-registration)

Declaring a trial's primary endpoint and analysis plan in advance (ideally registered publicly) so the main test keeps its honest false-alarm rate and 'findings' can't be fished out after seeing the data.

Full lesson → M3 · Multiplicity and p-hacking

Precision

In literature searching, the fraction of retrieved records that are actually relevant (relevant ÷ all retrieved). Distinct from diagnostic specificity.

Full lesson → M4 · Search strategy

Precision

How tightly an estimate is pinned down: higher precision means a smaller standard error and a narrower interval.

Full lesson → M3 · The standard error

Prediction interval

The range in which the true effect of a future study is expected to fall, combining uncertainty in the mean with the between-study spread (τ²).

Full lesson → M4 · Heterogeneity and I²

Preference-based measure

A way of valuing health that reveals preferences through a trade-off (TTO, SG), as opposed to a simple rating.

Full lesson → M6 · Where utility numbers come from

Present value

What a future cost or benefit is worth in today's terms, after discounting: PV = value ÷ (1+r)^t.

Full lesson → M5 · Discounting and the time horizon

Prevalence

How common a disease is in the population being tested. It doesn't affect a test's sensitivity or specificity, but it critically determines PPV: the same test gives mostly false positives when the disease is rare.

Full lesson → M3 · Diagnostic tests

Price–volume agreement

A deal linking price to the number of patients treated, e.g. a lower price (or rebate) once volume passes a threshold. Limits total spend while protecting the headline unit price.

Full lesson → M10 · Affordability and the payer's decision

Prioritarianism

The view that a unit of health counts for more when it goes to the worse-off. Not equalising outcomes, and not pure maximising: it weights the same QALY up for the more disadvantaged. The logic behind equity weights.

Full lesson → M13 · Equity in HTA

PRISMA

Preferred Reporting Items for Systematic reviews and Meta-Analyses. A reporting standard consisting of a 27-item checklist and a flow diagram that accounts for every record from search to inclusion.

Full lesson → M4 · PRISMA and study flow

Probabilistic sensitivity analysis (PSA)

Giving every uncertain input a probability distribution, then drawing from all of them at once and running the model thousands of times (Monte Carlo). The output is a cloud of ICERs on the cost-effectiveness plane, capturing joint uncertainty one-way analysis misses.

Full lesson → M9 · Probabilistic sensitivity analysis

Prognostic factor

A patient characteristic that affects outcomes regardless of treatment; anchoring on a common comparator cancels differences in these for free.

Full lesson → M4 · Indirect comparisons and network meta-analysis

Progression-free survival (PFS)

Time until a tumour grows or the patient dies; a common oncology surrogate for overall survival, often weakly correlated with it at trial level.

Full lesson → M6 · Endpoints: hard vs surrogate

Propensity score

A method for observational data: model each patient's probability of receiving the treatment from their measured characteristics, then match or weight to balance treated and untreated groups on those characteristics. Handles measured confounding only, never unmeasured.

Full lesson → M11 · Causal inference outside the RCT

Proportional hazards

The assumption behind a single hazard ratio: that the treatment's relative effect stays constant over time. Violated when survival curves cross, converge, or separate only late, and then one HR misleads.

Full lesson → M3 · Hazard ratios

Publication bias

The distortion that arises when studies with positive or significant results are more likely to be published than null ones, so the visible literature overstates an effect.

Full lesson → M4 · Publication bias and the funnel plot

QALY (quality-adjusted life year)

A unit of health combining length and quality of life: one year in full health = 1 QALY, computed as time × utility.

Full lesson → M6 · The QALY as a construct

QUADAS-2

A risk-of-bias tool for diagnostic accuracy studies.

Full lesson → M4 · Risk of bias: ROBINS-I and QUADAS-2

Random-effects model

A meta-analysis model assuming the true effect varies across studies, estimating the mean of that distribution.

Full lesson → M4 · Fixed vs random effects

Randomisation

Assigning participants to groups by chance, so the groups are alike in every respect except the treatment. The mechanism that lets a trial rule out confounding.

Full lesson → M2 · Randomisation and blinding

Randomised controlled trial (RCT)

A study that assigns participants to treatment or control purely by chance, making the groups otherwise identical so nothing hidden can confound the comparison.

Full lesson → M2 · Types of study

Rare disease

A disease affecting very few people: defined administratively (EU: ≤5 in 10,000; US: <200,000). The same disease can be 'rare' in one jurisdiction and not another; the threshold is a legal construct, not a fact of nature.

Full lesson → M12 · Orphan drugs and rare diseases

Real-world evidence (RWE)

Evidence on a technology's use and effects drawn from routine practice (registries, claims data, electronic health records) rather than controlled trials. Answers "does it work?" (effectiveness), not "can it work?" (efficacy).

Full lesson → M11 · Real-world evidence: sources and why

Recall bias

A distortion where people with a disease remember past exposures more thoroughly than those without, a particular threat to case-control studies that rely on memory.

Full lesson → M2 · Bias, chance and confounding

Red flags (in a dossier)

A recurring warning signature on the surface of a dossier that points to a likely problem, the compressed pattern-recognition of an experienced assessor. A signal that says 'look here,' not a verdict that says 'this is wrong.'

Full lesson → M14 · Red flags: the compressed course

Reference case

An HTA agency's mandated recipe for how a submission must be built, including which perspective is required for the base case.

Full lesson → M5 · Perspective: payer vs societal

Reference pricing

Setting a drug's price by external comparison rather than by value: internal (pegged to comparable drugs in the country's therapeutic cluster) or external/IRP (pegged to a basket of other countries' prices).

Full lesson → M12 · Pricing: reference and value-based

Regression to the mean

The tendency of extreme measurements to drift back toward the average on remeasurement, a cousin of the winner's curse, and a frequent source of illusory 'improvements.'

Reimbursement process

The path a drug travels from finished trials to being funded, a sequence of independent gates (submission, assessment, appraisal, decision & negotiation, listing, review), each of which can stop it for a different reason. Separate from, and after, marketing authorisation.

Full lesson → M12 · The reimbursement process

Relative (clinical) effectiveness

How much better (or worse) a technology performs than its comparator on clinical outcomes, the scientific question the JCA answers. Distinct from cost-effectiveness, which weighs that benefit against price.

Full lesson → M12 · EU HTA Regulation & Joint Clinical Assessment

Relative risk (RR)

How many times the risk changes: treatment risk ÷ control risk. RR = 0.5 means half the risk. A ratio; it says nothing about how many people are affected without the baseline risk.

Full lesson → M3 · Measures of effect: RR, OR, RD, NNT

Relative risk reduction (RRR)

The proportional drop in risk: how much smaller the treated risk is relative to control (e.g. 2%→1% is a 50% RRR). Blind to baseline: the same RRR can mean a huge or trivial absolute benefit.

Full lesson → M3 · Relative vs absolute risk

Reporting bias

The broader family of distortions from selective reporting, whole studies (publication bias) or selectively reported outcomes within a study, that skews the available evidence.

Full lesson → M4 · Publication bias and the funnel plot

Researcher degrees of freedom

The many points in an analysis where the method permits a range of honest, defensible choices (comparator, endpoint, horizon, extrapolation…). Advocacy lives in taking the favourable option at each one.

Full lesson → M14 · Reading the manufacturer's case: follow the levers

Residual (unmeasured) confounding

The bias left after adjusting for measured confounders, caused by confounders that were never recorded. Invisible inside the study (balance on measured variables says nothing about unmeasured ones) and impossible to remove without randomisation.

Full lesson → M11 · Causal inference outside the RCT

Restricted mean survival time (RMST)

The average time alive up to a set time point, an alternative survival summary that doesn't assume proportional hazards, increasingly used when a single hazard ratio can't be trusted.

Review protocol

A pre-registered plan for a systematic review that declares the question, search strategy, and inclusion rules before any results are seen. Registered on PROSPERO or a similar registry.

Full lesson → M4 · PRISMA and study flow

Risk difference (RD)

The absolute change in risk: control risk − treatment risk, in percentage points. Unlike relative risk, it reflects how many people per hundred are actually affected.

Full lesson → M3 · Measures of effect: RR, OR, RD, NNT

Risk of bias

A structured judgement of whether a study's design left room for its result to be systematically distorted, assessed across defined domains. Distinct from reporting quality and from whether bias actually occurred.

Full lesson → M4 · Risk of bias: RoB 2

RoB 2

The Cochrane risk-of-bias tool for randomised trials, assessing five domains; the overall judgement equals the weakest domain.

Full lesson → M4 · Risk of bias: RoB 2

ROBINS-I

A risk-of-bias tool for non-randomised studies of interventions, with a strong focus on confounding.

Full lesson → M4 · Risk of bias: ROBINS-I and QUADAS-2

ROC curve

Receiver operating characteristic, a plot of sensitivity vs 1 − specificity as the diagnostic threshold moves. Shows the trade-off; the area under the curve (AUC) summarises how well a test separates cases from non-cases.

Full lesson → M3 · Diagnostic tests

Rule of rescue

The moral imperative to help an identifiable person facing death, even when the same resources would statistically help more anonymous people elsewhere. A personal, deontological pull that the impersonal, utilitarian cost-effectiveness threshold cannot capture.

Full lesson → M12 · Orphan drugs and rare diseases

Sampling distribution

The spread of an estimate across many hypothetical repeats of the same study; its own SD is the standard error.

Full lesson → M3 · The standard error

Sampling variation

The wobble in a result caused purely by which individuals happened to land in your sample. It isn't an error; it's built into sampling, and it shrinks as the sample grows.

Full lesson → M3 · Variation and uncertainty

Scenario analysis

Rebuilding a model under a different defensible assumption (a different extrapolation, perspective, or structure) and recomputing from scratch, then comparing verdicts. The tool for structural and methodological uncertainty, which PSA can't capture.

Full lesson → M9 · Scenario analysis

Search strategy

A PICO question translated into database language: each concept becomes a block, synonyms joined with OR within a block, blocks joined with AND between them.

Full lesson → M4 · Search strategy

Selection bias

A systematic tilt from who gets into a study. If the people studied aren't representative, the result is skewed before any data is collected.

Full lesson → M2 · Bias, chance and confounding

Sensitivity

The fraction of truly diseased patients who test positive, a test's ability to find cases. A highly sensitive test misses few cases: good for ruling a disease out ('SnOut').

Full lesson → M3 · Diagnostic tests

Sensitivity analysis

Testing how much a model's result changes when its uncertain inputs are varied. Deterministic (one input at a time) or probabilistic (all at once); the umbrella term for the whole of uncertainty analysis.

Full lesson → M9 · One-way sensitivity analysis & the tornado

Signal, not verdict

A red flag says 'look here,' not 'this is wrong.' Every flag has an innocent explanation beside the guilty one, so spotting it is fast but resolving why it's there is the real appraisal. Treating a flag as proof is prejudice.

Full lesson → M14 · Red flags: the compressed course

Signature (of a problem)

A red flag isn't the problem itself but its trace on the surface of a dossier: where the form of the writing gives away the substance. Experts read how something is presented, not just what it says.

Full lesson → M14 · Red flags: the compressed course

Skewed distribution

A lopsided distribution with a long tail on one side, common for healthcare costs, length of stay, and survival time. The mean is pulled toward the tail, so the median often describes it better.

Full lesson: M5

Small-study effects

The tendency for smaller studies to show systematically different (often larger) effects than larger ones, a symptom with several possible causes, not only publication bias.

Full lesson → M4 · Publication bias and the funnel plot

Societal perspective

The widest analytical boundary: every cost wherever it falls (including lost productivity and informal care) is counted.

Full lesson → M5 · Perspective: payer vs societal

Societal preferences

Health-state values elicited from the general public rather than from patients; the reference-case basis for utilities in most HTA agencies.

Full lesson → M6 · Where utility numbers come from

Specificity

The fraction of truly healthy patients who test negative, a test's ability to clear the innocent. A highly specific test rarely raises false alarms: good for ruling a disease in ('SpIn').

Full lesson → M3 · Diagnostic tests

Spend cap

A negotiated ceiling on a payer's total annual spend on a technology; above it, the manufacturer absorbs the cost. Fixes affordability without changing the unit price or the ICER.

Full lesson → M10 · Affordability and the payer's decision

Standard deviation (SD)

A measure of how spread out individual values are around their average, the width of the distribution. Acts as a unit of distance from the mean. Distinct from the standard error.

Full lesson → M3 · Distributions and the normal curve

Standard error

How much an estimate (e.g. a mean) would wobble from one repeat of a study to the next, the SD of the sampling distribution. SE = SD / √n.

Full lesson → M3 · The standard error

Standard Gamble (SG)

A method that values a health state by the risk of death someone would accept for a chance at full health; utility = the success probability at which they're indifferent.

Full lesson → M6 · Where utility numbers come from

Standard of care

The treatment patients would otherwise receive today: the care a new technology must be compared against, not a placebo or an outdated option.

Full lesson → M1 · Framing the question: PICO

Statistical power

A study's ability to detect a real effect if one exists, the chance it comes back 'significant' when the drug truly works. Driven mainly by sample size. Low power makes 'not significant' uninformative.

Full lesson → M3 · Significance and power

Statistical significance

A result is called 'statistically significant' when its p-value falls below a chosen threshold (by convention 0.05), i.e. it would be fairly unusual by chance alone. A flag, not a verdict, and not a measure of importance.

Full lesson → M3 · The p-value

Structural uncertainty

Uncertainty about whether the model itself is built the right way: its shape, its extrapolation, its number of health states. A discrete choice between models, not a value, so it can't be sampled in a PSA.

Full lesson → M9 · Scenario analysis

Subgroup analysis

Looking for an effect within a slice of the trial population (e.g. older patients, a biomarker group). Hypothesis-generating, never confirmatory, and a prime source of multiplicity when many subgroups are examined.

Full lesson → M3 · Multiplicity and p-hacking

Summary of findings table

A GRADE output table giving, per critical outcome, the effect estimate and its certainty rating side by side.

Full lesson → M4 · GRADE and the certainty of evidence

Surrogate endpoint

A faster, cheaper marker (LDL, blood pressure, tumour shrinkage, PFS) measured in place of a hard endpoint, valid only when properly validated.

Full lesson → M6 · Endpoints: hard vs surrogate

Surrogate outcome

A short-term, easy-to-measure marker (like a blood-sugar reading) used as a stand-in for an outcome patients actually care about (like avoiding heart attacks). Whether the stand-in holds up is its own question.

Full lesson: M6

Surrogate validation

Establishing that a treatment's effect on a surrogate reliably predicts its effect on the hard endpoint, the trial-level arrow, not mere patient-level correlation.

Full lesson → M6 · Endpoints: hard vs surrogate

Survival analysis

Methods for outcomes measured as time-to-event (survival, time to relapse) rather than yes/no. Built to handle patients observed for different lengths and those who haven't yet had the event.

Full lesson → M3 · Survival analysis

Systematic review

A review that applies a pre-specified, reproducible rule for finding and selecting studies, so the set of included studies is fixed before any results are known.

Full lesson → M4 · Systematic reviews

Target population

The number of patients who will actually be treated with a technology, the output of a funnel: covered population × prevalence × eligible fraction × uptake. The key driver of budget impact.

Full lesson → M10 · Population, uptake, and horizon

Tariff

A published, standardised payment or average cost per activity, usually someone's top-down costing, reused across the system.

Full lesson → M5 · Costing: top-down vs bottom-up

Test statistic

A number summarising how far the data sits from 'no effect,' measured in standard errors (e.g. z = difference ÷ SE). The bigger it is, the more extreme the result, and the smaller the p-value.

Full lesson → M3 · The p-value

Test-treatment pathway

The chain from a diagnostic test to a health outcome: test → result → decision → treatment → outcome. A test's value is the difference in outcome this pathway produces; it lives downstream, not in the test itself.

Full lesson → M13 · HTA of diagnostics: the test-treatment pathway

Test-treatment RCT

The gold-standard design for valuing a diagnostic: randomise patients to a whole test-and-treat strategy versus an alternative, and measure the health outcome at the end. Captures the full pathway, but is large, long, and rare.

Full lesson → M13 · HTA of diagnostics: the test-treatment pathway

Test–treat pathway

The clinical flow: test, interpret, treat (or don't). HTA models this pathway end-to-end: what the test costs, how it changes treatment decisions, and what those decisions cost and deliver. (More in M9.)

Full lesson: M9

Time horizon

How far into the future an analysis counts costs and effects; must be long enough (often a lifetime) to capture everything that differs between options.

Full lesson → M5 · Discounting and the time horizon

Time preference

The preference for benefits now over benefits later; one of the contested justifications for discounting health.

Full lesson → M5 · Discounting and the time horizon

Time Trade-Off (TTO)

A method that values a health state by how many years of life someone would give up to live the rest in full health; utility = healthy years accepted ÷ years in the state.

Full lesson → M6 · Where utility numbers come from

Top-down costing

Deriving an average cost per patient by taking a large total (e.g. a hospital's annual spend) and dividing by the number of cases.

Full lesson → M5 · Costing: top-down vs bottom-up

Tornado diagram

A chart stacking one-way sensitivity results as horizontal bars (each the ICER's low-to-high range for one input), widest at the top, so the value drivers and any threshold-crossing inputs stand out.

Full lesson → M9 · One-way sensitivity analysis & the tornado

Transition probability

The probability of moving from one health state to another (or staying) in a single cycle. Collected into a transition matrix whose every row sums to 1.

Full lesson → M8 · Markov models

Transitivity

The assumption that the trials being linked are similar enough in their effect modifiers that the shared comparator behaves the same across them; if it fails, the indirect estimate is biased.

Full lesson → M4 · Indirect comparisons and network meta-analysis

Trial-level association

Across many trials, how well the treatment effect on the surrogate predicts the treatment effect on the hard endpoint; usually an R². The evidence that actually validates a surrogate.

Full lesson → M6 · Endpoints: hard vs surrogate

Trim-and-fill

A method that imputes hypothetical 'missing' studies to gauge how sensitive a pooled result is to asymmetry, a sensitivity analysis, not a correction.

Full lesson → M4 · Publication bias and the funnel plot

Tunnel states

A short, ordered chain of temporary states patients pass through in sequence, used to give a Markov model limited memory of time since an event (e.g. months since surgery).

Full lesson → M8 · Markov models

Type I error (false alarm)

Concluding there's an effect when there really isn't, a false positive. How often it happens is set by the significance threshold (0.05 means a 1-in-20 false-alarm rate).

Full lesson → M3 · Significance and power

Type II error (miss)

Failing to detect a real effect that's genuinely there, a false negative. Usually the result of low power, mostly from too small a sample.

Full lesson → M3 · Significance and power

Uncertainty calibration

Matching your tone to how much each claim can bear: state the solid plainly, flag the fragile explicitly. Volunteering your own weaknesses builds credibility; a concealed weakness, once exposed, collapses trust in the whole report.

Full lesson → M14 · Writing the recommendation: the decision-maker's tool

Unit cost

The cost of one unit of a resource (one bed-day, one surgeon-minute), multiplied by quantity used to build a total.

Full lesson → M5 · Costing: top-down vs bottom-up

Unmet need

The extent to which patients lack any satisfactory alternative treatment. A value dimension that cost-per-QALY ignores but that decision-makers weigh heavily, a natural MCDA criterion.

Full lesson → M13 · Multi-criteria decision analysis (MCDA)

Uptake

The share of eligible patients who actually receive a technology. Rarely 100%, and it rises over years toward a plateau rather than arriving at once, so budget impact is a stream, not a snapshot.

Full lesson → M10 · Population, uptake, and horizon

Usual care

The real existing standard of treatment, used as the comparator in a pragmatic trial, so the question becomes "is this better than what we do now?" rather than "is it better than nothing?" (placebo).

Full lesson → M11 · Pragmatic trials

Utility

A preference-based value for a health state on a 0–1 scale (1 = full health, 0 = death), revealed by what people would trade to escape the state.

Full lesson → M6 · Where utility numbers come from

Value set

A pre-built table assigning a utility to every possible EQ-5D health state, produced earlier by valuing states with a choice-based method on a population sample; country-specific.

Full lesson → M6 · Where utility numbers come from

Value-based pricing

Setting a drug's price from the value it delivers: reverse the ICER, set it equal to the threshold and solve for price. The value-based price is a ceiling (the point of indifference), roughly threshold × QALYs gained, adjusted for other cost changes.

Full lesson → M12 · Pricing: reference and value-based

Variance

The average squared distance of values from the mean; its square root is the standard deviation.

Full lesson → M3 · Distributions and the normal curve

Visual Analogue Scale (VAS)

Rating a health state on a 0–100 line with no trade-off involved; not a preference-based method, and systematically lower than TTO or SG.

Full lesson → M6 · Where utility numbers come from

Weight (MCDA)

How much a criterion matters relative to the others. Weights are where the value judgements live, and cost-per-QALY is just an MCDA with all weight on health gain and zero on everything else.

Full lesson → M13 · Multi-criteria decision analysis (MCDA)

Willingness to pay

What people will pay to reduce risk or gain health, estimated from surveys or behaviour; underlies CBA's money valuation of health and is entangled with ability to pay.

Full lesson → M5 · The four types of economic evaluation

Winner's curse

The tendency of small studies that reach significance to overstate the true effect, because only an exaggerated, lucky result could clear the bar with few patients. A big reason early findings shrink on replication.

Full lesson → M3 · Significance and power

Worse than death

Health states rated below 0 on the utility scale, because people would choose death over enduring them.

Full lesson → M6 · The QALY as a construct

Years Lived with Disability (YLD)

The ill-health part of a DALY: years spent with a condition, weighted by its disability weight.

Full lesson → M6 · The DALY and how it mirrors the QALY

Years of Life Lost (YLL)

The premature-death part of a DALY: years lost measured against a standard life expectancy, not the patient's real prognosis.

Full lesson → M6 · The DALY and how it mirrors the QALY

Z-score

How many standard deviations a value sits from the mean, a way to express 'how typical or surprising' any single value is, on a universal scale. (Full treatment later in M3.)

Full lesson: M3

τ² (tau-squared)

The estimated variance of the true effects across studies, in the squared units of the effect measure, the magnitude of heterogeneity.

Full lesson → M4 · Heterogeneity and I²