Population

Population: The P in PICO and the People a Health Economic Model Represents

Every health technology assessment answers a question about particular people, and the answer can change when the people change. A treatment that offers good value for patients at high risk of an event can offer poor value for those at low risk. This page explains how HTA bodies specify the population, how the term shifts in meaning between trials, models, budget impact analysis and value of information, and why subgroups matter, with a worked example in which the mix of patients moves a cost-effectiveness ratio across a threshold. It covers the decision-problem sense of the term, not the demographic sense used in population health.

The population as the first element of a decision problem

A decision problem in health economics is commonly framed around four elements: the population, the intervention, the comparator or comparators, and the outcomes. The same structure, known as PICO, frames the question in a systematic review, where the Cochrane Handbook makes the population, intervention and comparison the basis of the eligibility criteria. The population comes first because it determines which comparators are relevant and which studies count as evidence.

NICE's manual (PMG36) states that the scope defines the population as precisely as possible, including where, why and how the technology is used in the care pathway, and may highlight subgroups whose clinical or cost effectiveness might differ. In the NICE reference case the decision problem comes from the scope, and the economic evaluation should open with a statement naming the technologies compared and the relevant patient groups, with any departure from the scope justified.

The EU HTA Regulation (2021/2282) defines the scope of a joint clinical assessment in terms of patient population, intervention, comparators and health outcomes. Its recitals state that the joint report takes no position on the target population in which the technology should be used and that pricing and reimbursement remain a national matter, so the choice of population to fund stays with each Member State.

One decision, several populations

The word population is used for several groups within one appraisal, and mixing them up leads to errors. NICE's manual notes that for many technologies there may be multiple populations, and the groups below overlap without being identical.

The narrower concepts map onto these rows: Target Population to the decision population, Trial Population to the trial row and Eligible Population to the budget impact row, with Treated Population the subset who actually receive the technology in a given period. NICE's manual states that, for medicines, the relevant population will normally be informed by the expected marketing authorisation. In other cases, some people who could use a technology may be left out of a scope to keep the evaluation to a reasonable size, which does not mean the technology is inappropriate for them.

How a population enters a model

A model represents the decision population through its starting characteristics and its parameters. In a cohort model, the modelled population is allocated among health states at the start, and everyone in a state shares the same transition probabilities. The ISPOR-SMDM report on state-transition modelling notes that characteristics known at the time of the decision, such as age, sex, comorbidities and disease stage, can be used to define the starting cohorts.

Parameters must also suit the population. NICE's manual asks that data sets be justified by their suitability to the population of interest, and suggests applying relative treatment effects from randomised trials to baseline risk data for the populations or subgroups of interest. Those baseline data can come from observational studies that describe the decision population more closely than the trial does.

Subgroups and heterogeneity within a population

For many technologies, people within the same population gain different amounts. The ISPOR-SMDM report on parameter estimation and uncertainty calls this heterogeneity: the extent to which between-patient variability can be explained by patients' characteristics. Its relevance lies in identifying subgroups for whom separate cost-effectiveness analyses should be carried out. Heterogeneity differs from parameter uncertainty, which concerns how precisely an average value is known, and from random variation between identical patients.

NICE's manual asks for clinical and cost-effectiveness estimates for each relevant subgroup in a cost-utility analysis. Subgroups should preferably rest on an expected differential effect from biologically plausible mechanisms, social characteristics or other clearly justified factors, and post hoc searching for subgroup effects is viewed sceptically. In its section on threshold analysis, the manual adds that threshold analysis should not be used to justify restricting the population to a subgroup on cost-effectiveness grounds. Subgroup analysis and treatment effect heterogeneity cover the methods.

When a population is made up of distinct subgroups, its mean incremental cost and effect are weighted averages of the subgroup means:

$$ \Delta C = \sum_{k} w_k , \Delta C_k, \qquad \Delta E = \sum_{k} w_k , \Delta E_k, \qquad \text{ICER} = \frac{\Delta C}{\Delta E} $$

where $w_k$ is the share of the population in subgroup $k$, with the shares summing to one, $\Delta C_k$ and $\Delta E_k$ are the mean incremental cost and incremental health effect in subgroup $k$, $\Delta C$ and $\Delta E$ are the population means, and $\text{ICER}$ is the incremental cost-effectiveness ratio for the whole population. The population ratio is calculated from the weighted means, not by averaging the subgroup ratios.

Worked example: one treatment, two risk groups

The figures below are illustrative. A new treatment is compared with current care in a population with two subgroups that differ in baseline risk, with discounted costs and QALYs per person.

1. Combine the subgroups. The population mean incremental cost is 0.6 × 4,000 + 0.4 × 3,000 = 3,600, so £3,600 per person. The mean QALY gain is 0.6 × 0.10 + 0.4 × 0.30 = 0.18.

2. Calculate the population ratio.

$$ \text{ICER} = \frac{3600}{0.18} = 20000 $$

where $\text{ICER}$ is the cost in pounds per QALY gained for the population as a whole. Averaging the subgroup ratios instead would give 0.6 × 40,000 + 0.4 × 10,000 = 28,000, which is wrong because it ignores the different QALY gains behind each ratio.

3. Compare with a threshold. At an illustrative cost-effectiveness threshold of £25,000 per QALY, the lower end of the £25,000 to £35,000 range in NICE's manual (PMG36, updated March 2026), the net monetary benefit gain per person is 25,000 × 0.10 − 4,000 = −1,500 in the lower-risk group and 25,000 × 0.30 − 3,000 = 4,500 in the higher-risk group. For the population it is 25,000 × 0.18 − 3,600 = 900. The average hides a net loss in the lower-risk group. Treating only the higher-risk group gives 0.4 × 4,500 = £1,800 per person in the population, double the £900 from treating everyone, which is why NICE asks for estimates for each relevant subgroup.

4. Change the mix. If the population were 80% lower risk and 20% higher risk, the mean incremental cost would be 0.8 × 4,000 + 0.2 × 3,000 = 3,800 and the mean QALY gain 0.8 × 0.10 + 0.2 × 0.30 = 0.14, giving a ratio of about £27,143 per QALY and a net monetary benefit of 25,000 × 0.14 − 3,800 = −300 per person at £25,000. At £35,000 per QALY, both mixes would be cost-effective on average. Only the population has changed, so the case mix in a model should match the case mix expected in practice.

Population size: budget impact and value of information

Cost-effectiveness results are usually expressed per person, so they do not normally depend on how many people are in the population. Two other analyses do. A budget impact analysis needs the size of the eligible population, which the ISPOR good practice report on budget impact analysis lists among its key elements. The report also asks for data that reflect the decision maker's own population and, where a model is used, for people entering and leaving the eligible population over time. Size estimates draw on measures such as local prevalence and incidence.

Value of information analysis scales a per-person result to the population that will face the decision. Briggs, Claxton and Sculpher give a standard form in their textbook on decision modelling:

$$ EVPI_{pop} = EVPI_{ind} \times \sum_{t=1}^{T} \frac{N_t}{(1+r)^t} $$

where $EVPI_{pop}$ is the expected value of perfect information for the population, $EVPI_{ind}$ is the per-person value, $N_t$ is the number of people expected to face the decision in period $t$, $T$ is the period over which the decision is expected to apply, and $r$ is the discount rate. The answer is sensitive to assumptions about $N_t$ and $T$, which are themselves judgements about the population.

Where population definitions go wrong

One error is applying trial results to a decision population that is older, more comorbid or at a different stage of disease than those studied. Another is to treat heterogeneity as uncertainty, folding differences between identifiable groups into a probabilistic analysis instead of reporting results for each subgroup.

Reports can also be unclear about which population a result describes. The CHEERS 2022 statement asks authors to describe the characteristics of the study population and any methods used to estimate how results vary for subgroups. Without that, a reader cannot judge whether a published ratio applies to a local decision.

Sources

• Briggs A, Claxton K, Sculpher M. Decision Modelling for Health Economic Evaluation. Oxford: Oxford University Press; 2006.
• Briggs AH, Weinstein MC, Fenwick EAL, Karnon J, Sculpher MJ, Paltiel AD. Model parameter estimation and uncertainty analysis: a report of the ISPOR-SMDM Modeling Good Research Practices Task Force Working Group-6. Medical Decision Making. 2012;32(5):722-732.
• European Parliament and Council of the European Union. Regulation (EU) 2021/2282 on health technology assessment and amending Directive 2011/24/EU. Official Journal of the European Union. 2021;L 458:1. Recitals 14 and 28, Article 2(9).
• Husereau D, Drummond M, Augustovski F, et al. Consolidated Health Economic Evaluation Reporting Standards 2022 (CHEERS 2022) statement: updated reporting guidance for health economic evaluations. BMJ. 2022;376:e067975.
• McKenzie JE, Brennan SE, Ryan RE, Thomson HJ, Johnston RV, Thomas J. Chapter 3: Defining the criteria for including studies and how they will be grouped for the synthesis. In: Higgins JPT, Thomas J, Chandler J, et al, editors. Cochrane Handbook for Systematic Reviews of Interventions. Version 6.5. Cochrane; 2024.
• National Institute for Health and Care Excellence. NICE technology appraisal and highly specialised technologies guidance: the manual (PMG36). London: NICE; 2022, updated March 2026. Sections 2.2.9 to 2.2.11, 4.2.4, 4.6.15, 4.6.16, 4.7.23, 4.9.1, 4.9.3, 4.9.4, 6.3.4 and 6.3.7.
• Siebert U, Alagoz O, Bayoumi AM, Jahn B, Owens DK, Cohen DJ, Kuntz KM. State-transition modeling: a report of the ISPOR-SMDM Modeling Good Research Practices Task Force-3. Value in Health. 2012;15(6):812-820.
• Sullivan SD, Mauskopf JA, Augustovski F, et al. Budget impact analysis: principles of good practice. Report of the ISPOR 2012 Budget Impact Analysis Good Practice II Task Force. Value in Health. 2014;17(1):5-14.