Primary Research

Primary Research in Health Economics and HTA

Primary research produces evidence that did not exist before, by recruiting people or studying settings and recording data to answer a question set in advance. In health economics the main forms are randomised trials that collect costs and quality of life, cohort studies, preference surveys, qualitative interviews and primary costing studies. The reviews and models behind a reimbursement decision draw their inputs from earlier primary studies. HTA bodies also ask for new primary research when the evidence is too uncertain to support a decision, and value of information methods test whether a proposed study is worth its cost.

Primary and secondary research in the evidence chain

Wickham describes the steps of a primary study. An investigator identifies a question in a particular population, designs a project to address it, chooses a quantitative or qualitative method, sets a sample size, recruits subjects, and systematically collects and analyses the data. Secondary research, described in its own article, starts from data or studies that already exist and reanalyses or combines them.

The boundary depends on how data are used as well as how they were collected. The NICE glossary defines primary data as data collected and used for a specific research purpose, and secondary data as data collected routinely, or collected for research, when reused in other studies.

Primary studies sit at the start of the HTA evidence chain. A systematic review identifies and appraises them, a meta-analysis or wider evidence synthesis combines their results, and a health economic model turns the synthesised inputs into costs and QALYs for a decision.

Kinds of primary research used in health economics

Several study types count as primary research when they collect new data for an economic or HTA question. They differ in what they measure.

1. Trials. A randomised controlled trial assigns treatment at random, and economic data can be collected from the same participants as the clinical outcomes. An economic evaluation run this way is usually called a trial-based or piggyback evaluation.
2. Observational studies. An observational study, such as a prospective cohort, follows patients without assigning treatment and can record resource use, costs and outcomes in routine care.
3. Preference studies. Valuation studies generate the utility weights used to calculate a QALY. Devlin and colleagues produced the EQ-5D-5L value set for England from face-to-face interviews in which randomly selected members of the general public completed time trade-off and discrete choice experiment tasks, with valuation data for 996 respondents.
4. Primary costing studies. A micro-costing study measures the quantity of each resource used. Frick lists collection methods that include administrative databases, provider surveys and interviews, chart review, direct observation, programme logs and diaries.
5. Qualitative studies. Interviews and focus groups explore patient experience and the acceptability of a technology, evidence that a model does not capture in costs and QALYs.

Collecting economic data inside a trial

Prospective trials that collect economic data have grown in number, which Ramsey and colleagues link to interest in economic information and to the regulatory and reimbursement requirements of many countries. Their ISPOR good practice report on cost-effectiveness analysis alongside clinical trials recommends that trials evaluate effectiveness rather than efficacy where possible, obtain resource use and health state utilities directly from participants, and integrate economic data collection fully into the study.

Petrou and Gray describe how this works in practice. Use of many resources can be recorded on trial case report forms with little extra burden, while other items need medical records, patient questionnaires or diaries. The total cost for each participant is the quantity of each resource used multiplied by its unit cost, and unit costs may have to be built from accounting data, time and motion studies, interviews with caregivers and case note analysis. A generic questionnaire such as the EQ-5D, completed at several time points and scored with preference weights, is used to estimate QALYs.

Ridyard and Hughes reviewed this work in publicly funded HTA. Of 100 primary research papers funded by the UK HTA programme and published to June 2009, 95 included an economic analysis and 85 recorded patient-level resource use. Common features were extensive use of routine medical records and reliance on patient recall.

Design choices that shape the result

Early design choices fix what an economic evaluation can later say. Three matter most.

1. Perspective. Petrou and Gray note that the perspective of the analysis determines which categories of resource use are recorded, for example healthcare system costs alone or wider societal costs.
2. Recall period. Shorter recall periods reduce forgetting and misdating of episodes, but sampling a short period to represent a long one, such as 12 months, misses much of the use in between.
3. Sample size. Power calculations are almost always based on the primary clinical outcome. Because resource use varies widely, Petrou and Gray explain that economists estimate cost and effect differences and the likelihood of cost-effectiveness rather than testing a hypothesis about it.

When HTA bodies ask for new primary research

NICE's manual for technology appraisals and highly specialised technologies (PMG36) sets out recommendation categories that call for new evidence. A committee can recommend a medicine with managed access when it has the plausible potential to be cost effective at the currently agreed price but the evidence is too uncertain, new evidence is expected from ongoing or planned clinical trials or could be collected from patients having the medicine in clinical practice, and the data could feasibly be collected within up to 5 years without undue burden. The funding route is covered under managed access agreement.

PMG36 also allows a recommendation only in a research context when evidence of clinical effectiveness is absent, weak or uncertain. The committee first weighs the value of more evidence, the uncertainty in the analysis and what could be gained by reconsidering the decision, the feasibility of the research under the recommendation, the likelihood that it will be commissioned and reported, the time it will take, and its ethical and practical aspects. The manual states that managed access allows further evidence generation in the NHS, while a recommendation only in a research context comes without a requirement for NHS funding.

Claxton and colleagues, in research funded by the NIHR HTA programme to inform NICE, concluded that the right category depends only partly on expected cost-effectiveness. A technology expected to be cost effective may still warrant an only in research recommendation when the research could not be done once the technology is approved, or when significant irrecoverable costs are involved.

Deciding whether new primary research is worth doing

Claxton argued in 1999 that adoption should rest on expected net benefit alone, and that the uncertainty around it matters only for deciding whether more information is needed. Value of information analysis puts a monetary value on that second decision.

Rothery and colleagues set out the measures for an ISPOR task force. The EVPI values resolving all uncertainty, the EVPPI values perfect information on chosen parameters, and the EVSI values a specific data collection exercise. Per-person values are scaled to the population whose treatment the research will inform, discounted over the period in which the findings stay useful. The expected net benefit of sampling is population EVSI minus the cost of the study. A population EVPI below the cost of any study shows that research is not of value, and an ENBS of zero or more is a necessary condition for running a study with no wider value.

$$\text{Population EVSI} = \text{EVSI}_{\text{pp}} \times \sum_{t=0}^{T} \frac{I_t}{(1+d)^t}$$

where $\text{EVSI}_{\text{pp}}$ is the per-person expected value of sample information in GBP of net monetary benefit, $I_t$ is the number of patients facing the decision in period $t$, $T$ is the last period in which the findings inform decisions and $d$ is the discount rate per period.

$$\text{ENBS} = \text{Population EVSI} - C$$

where $C$ is the total cost of the proposed study in GBP.

Worked example comparing two trial designs

All numbers in this example are illustrative. An HTA committee faces a decision about a new treatment, with 4,000 eligible patients a year, findings expected to inform five annual cohorts ($t = 0$ to $T = 4$) and a discount rate of 3.5%. The per-person EVPI from the probabilistic model is 400 GBP, and two trials are proposed.

Step 1 (effective population). The discount factors for five years at 3.5% sum to about 4.6731, so the discounted population is $N = 4{,}000 \times 4.6731 \approx 18{,}692$ patients.

Step 2 (upper bound). $\text{Population EVPI} = 400 \times 18{,}692 = 7{,}476{,}800$ GBP. This exceeds the cost of both designs, so research may be worthwhile, though this alone does not show that either study is.

Step 3 (Design A). $\text{Population EVSI} = 150 \times 18{,}692 = 2{,}803{,}800$ GBP and $\text{ENBS} = 2{,}803{,}800 - 1{,}800{,}000 = 1{,}003{,}800$ GBP.

Step 4 (Design B). $\text{Population EVSI} = 210 \times 18{,}692 = 3{,}925{,}320$ GBP and $\text{ENBS} = 3{,}925{,}320 - 4{,}500{,}000 = -574{,}680$ GBP.

Step 5 (interpretation). The larger trial adds 1,121,520 GBP of population EVSI but costs 2,700,000 GBP more, so the smaller design is the better use of research funds. A full analysis would also count the opportunity cost of delay and benefits to trial participants.

Boundaries with nearby concepts

Primary data collection is the act of gathering new data, while primary research is the whole study built around it, from question to analysis. Secondary research and secondary data analysis work with data or results that already exist. Real-world data from routine records or claims are secondary when reused, following the NICE glossary, although a prospective registry or cohort set up to answer a defined question is primary research. A primary study can still take some items from routine sources, and Ridyard and Hughes found extensive use of routine medical records in HTA-funded trials. A model-based cost-effectiveness analysis draws most inputs from published studies, so it sits mainly on the secondary side.

Limitations and common misreadings

Primary research is slow and expensive. Wickham notes that collecting primary data is usually the most time-consuming and costly part of research.

1. One trial is the economic evaluation. Sculpher and colleagues argue that a single trial is, in most circumstances, an inadequate basis for decisions and should be treated as one source of evidence within a wider synthesis and decision model. Petrou and Gray list truncated time horizons, limited comparators, restricted generalisability and the failure to include all relevant evidence as limits of trial-based evaluation.
2. Primary means more reliable. A badly designed study can be less useful than a careful synthesis.
3. More research is always worth funding. Value of information analysis can show that the expected gain from a study is smaller than its cost, in which case the decision is made on current evidence.

Sources

• Wickham RJ. Secondary analysis research. Journal of the Advanced Practitioner in Oncology. 2019;10(4):395-400. https://doi.org/10.6004/jadpro.2019.10.4.7
• National Institute for Health and Care Excellence. Glossary (primary data; secondary data). Accessed 14 October 2026. https://www.nice.org.uk/glossary
• Petrou S, Gray A. Economic evaluation alongside randomised controlled trials: design, conduct, analysis, and reporting. BMJ. 2011;342:d1548. https://doi.org/10.1136/bmj.d1548
• Ramsey SD, Willke RJ, Glick H, et al. Cost-effectiveness analysis alongside clinical trials II: an ISPOR Good Research Practices Task Force report. Value in Health. 2015;18(2):161-172. https://doi.org/10.1016/j.jval.2015.02.001
• Ridyard CH, Hughes DA. Methods for the collection of resource use data within clinical trials: a systematic review of studies funded by the UK Health Technology Assessment program. Value in Health. 2010;13(8):867-872. https://doi.org/10.1111/j.1524-4733.2010.00788.x
• Devlin NJ, Shah KK, Feng Y, et al. Valuing health-related quality of life: an EQ-5D-5L value set for England. Health Economics. 2018;27(1):7-22. https://doi.org/10.1002/hec.3564
• Frick KD. Microcosting quantity data collection methods. Medical Care. 2009;47(7 Suppl 1):S76-S81. https://doi.org/10.1097/MLR.0b013e31819bc064
• National Institute for Health and Care Excellence. NICE technology appraisal and highly specialised technologies guidance: the manual (PMG36). Published 31 January 2022; last updated 13 October 2026. https://www.nice.org.uk/process/pmg36
• Claxton K, Palmer S, Longworth L, et al. Informing a decision framework for when NICE should recommend the use of health technologies only in the context of an appropriately designed programme of evidence development. Health Technology Assessment. 2012;16(46):1-323. https://doi.org/10.3310/hta16460
• Claxton K. The irrelevance of inference: a decision-making approach to the stochastic evaluation of health care technologies. Journal of Health Economics. 1999;18(3):341-364. https://doi.org/10.1016/S0167-6296(98)00039-3
• Rothery C, Strong M, Koffijberg HE, et al. Value of information analytical methods: report 2 of the ISPOR Value of Information Analysis Emerging Good Practices Task Force. Value in Health. 2020;23(3):277-286. https://doi.org/10.1016/j.jval.2020.01.004
• Sculpher MJ, Claxton K, Drummond M, McCabe C. Whither trial-based economic evaluation for health care decision making? Health Economics. 2006;15(7):677-687. https://doi.org/10.1002/hec.1093