Unlocking the Value of Health Data Reuse
By Damilola Ogungbemi
Former Data Scientist
Leeds Institute for Data Analytics (LIDA)
Health data are often collected for a specific purpose. A clinical trial may compare treatments, an electronic health record may document patient care, and a disease registry may monitor
diagnoses and outcomes across a population. Yet the scientific value of these data does not necessarily end when their original purpose has been fulfilled. Well-characterised datasets may
contain years of clinical observations, laboratory measurements, treatment information and outcomes that can support questions beyond those envisaged when the data were first
collected.
Research increasingly recognises this potential. Reusing existing health data can support validation of previous findings, individual participant data meta-analyses, methodological
research and new analyses addressing questions beyond the original trial objectives [2,3].
I encountered this through my work with data from Myeloma XI, a large Phase III, open-label,
randomised adaptive trial conducted across 110 NHS hospitals in England, Wales and Scotland [1]. The original trial evaluated treatment strategies for patients with newly diagnosed multiple
myeloma. Our project returned to the data with a different objective: to investigate whether longitudinal patterns in biomarkers during treatment were associated with subsequent patient
outcomes. Rather than considering biomarkers at a single time point, repeated measurements were used to characterise patterns of response during early treatment and examine their
relationship with progression-free and overall survival. As part of the project, we developed an interactive dashboard to communicate the resulting patient trajectories and survival analyses.
Explore the Myeloma XI project and dashboard
The analysis did not require recruitment of a new cohort or collection of new biological samples. The data already existed; the research question was new. This is an example of health data
reuse: deriving additional scientific value from data beyond the work for which they were originally generated.
At a time when health research is increasingly data-intensive, an important question is therefore not only how we generate more data, but how effectively we use the data we already have.
Getting More Value from the Health Data We Already Have
Data sharing and data reuse are related but distinct. Making a dataset available does not itself generate new knowledge. Scientific value is realised when researchers can understand the
data, access them appropriately and use them to address meaningful questions.
Recent work on clinical-trial reuse argues that effective reuse can maximise the value of
research and reduce research waste [2]. This is important because generating high-quality clinical research data requires substantial investment in study design, governance, recruitment,
clinical sites, laboratory and data-management infrastructure, and often years of follow-up. It also depends on participants contributing their time, information and, in many studies, biological
samples.
Data reuse does not remove the need for new studies. New data collection remains essential when existing datasets cannot answer an important question. Instead, reuse encourages
researchers to consider what relevant information already exists and whether additional knowledge can responsibly be derived from it before beginning new data collection.
The scientific applications are broad. Individual participant data from multiple trials can be pooled to increase statistical power or investigate treatment effects in specific subgroups.
Existing datasets can support the development and evaluation of new analytical methods. Previous analyses can be reproduced to assess robustness, while linkage to other data sources
may extend follow-up beyond the active study period [2,3].
Data reuse is therefore one way of increasing the scientific value generated from investments already made by participants, researchers, funders and health systems.
What It Takes to Make Health Data Reusable: Data science as an enabler of reuse
The existence of a dataset does not mean that it is ready for reuse. An important contribution of data science and data management is to bridge the gap between data that has been collected
and data that can be reliably understood and analysed for a new purpose.
This begins with understanding how the data were generated. Clinical-trial data are usually collected according to a protocol: eligibility criteria define the study population, measurements may be scheduled at specified intervals, and outcomes are defined in advance. Electronic health record data arise differently because they are primarily generated through patient care. The presence and frequency of measurements may therefore depend on clinical need, healthcare utilisation or professional judgement. NHS England cautions that information collected for direct care requires careful interpretation when subsequently used for research because it may not have been collected for that purpose [4].
Preparing existing data for reuse therefore involves more than conventional data cleaning. Data dictionaries and metadata must be understood; variable defi nitions and units may need harmonisation; duplicate, implausible or inconsistent records require investigation; dates and follow-up periods need to be reconciled; missingness must be characterised; and transformations from source data to analytical datasets should be documented and reproducible.
Data quality is therefore central to reuse. NHS England describes high-quality patient records using the CARAT framework and emphasises the importance of how information is structured, described and coded for research and other uses beyond direct care [4].
| Characteristic | Relevance to data reuses |
| Complete | The information required for the intended use is sufficiently represented. |
| Accurate | Recorded information reliably reflects the patient, measurement or clinical event. |
| Relevant | The information is appropriate to the context in which it will be used. |
| Accessible | Appropriately authorised users can retrieve and interpret the information they require. |
| Timely | Information is recorded at an appropriate time in relation to the event it describes. |
Standardisation also facilitates reuse. NHS England highlights structured clinical terminology such as SNOMED CT, while recognising that narrative information can preserve clinical context not easily represented by structured fields alone [4]. Common definitions, metadata and interoperable standards become particularly important when information is combined across datasets or organisations.
Reproducibility is equally important. Version-controlled code, documented cleaning decisions, reproducible data-processing pipelines, clear data dictionaries and transparent derivation of variables allow future researchers to understand how an analytical dataset was produced. Guidance on clinical-trial reuse similarly emphasises early feasibility assessment, multidisciplinary collaboration, interoperability, transparency and reproducible research practices [2].
The role of the data scientist in secondary research is therefore broader than selecting and fitting a statistical model. It includes making complex health information structured, documented,
interpretable and reproducible enough to support reliable research.
The Right Data for the Right Research Question
Promoting health data reuse does not mean that every existing dataset is suitable for every new research question. A dataset may lack the required population or variables; follow-up may be
insufficient; outcomes may not have been captured with adequate precision; missingness may be related to the original data-generating process; or important confounders may never have
been collected.
The Add-Aspirin trial in the UK provides a useful example. Add-Aspirin is a multicentre randomised trial evaluating aspirin following potentially curative cancer treatment. Researchers
investigated whether routinely collected English healthcare-system data could replace some information collected directly from hospitals through trial-specific case-report forms [5].
The suitability of the routine data differed by outcome. Healthcare-system data showed good agreement with trial records for diagnosis and tumour staging and identified all deaths recorded
by the trial, as well as additional deaths not yet captured in the trial database. Cancer recurrence was less reliably captured: cancer-registry data alone identified fewer than half of the
recurrences recorded through trial follow-up. Detailed trial-defined adverse events were also difficult to reconstruct from routine data [5].
The finding illustrates an important principle: the same data source can be suitable for one purpose and insufficient for another.
Health data reuse therefore requires assessment of whether existing data are fit for the proposed research question. This includes understanding the population represented, data
provenance, availability and validity of relevant variables and outcomes, missingness, follow-up and potential sources of bias.
Analytical methods can address some limitations. Missing-data methods, sensitivity analyses, longitudinal models, statistical adjustment and external validation can strengthen analyses when
used appropriately. They cannot recreate an important variable that was never collected, remove confounding from an entirely unmeasured factor or make an inadequately captured
outcome reliable. The research question and characteristics of the data should therefore determine the analytical approach.
Why Health Data Reuse Matters for the UK Now
The case for making better use of existing health data has particular relevance in the UK. NHS data provide comprehensive, longitudinal information for much of the population, with potential
applications ranging from improving patient outcomes and NHS performance to supporting life-sciences research. However, persistent barriers include fragmented systems, complex
access processes and technical incompatibility. Around 80% of UK health data are not organised in a standard format, creating additional work before they can be analysed effectively
[6].
The 2024 Sudlow Review, Uniting the UK's Health Data: A Huge Opportunity for Society, concluded that complexities and inefficiencies across the UK health-data landscape were
delaying the safe use of data for research. It recommended recognising health data as critical national infrastructure, streamlining access processes, establishing a national health-data
service in England, and developing UK-wide standards and accreditation for Secure Data Environments [7].
The policy landscape is beginning to respond. The UK Government and Wellcome are investing up to £600 million in the Health Data Research Service (HDRS), intended to provide
approved researchers with more integrated and secure access to regional and national health datasets. Delivery is expected to begin by the end of 2026 [8].
Improved access, however, is only one component of effective reuse. Data must also be understandable, appropriately standardised, sufficiently documented and interoperable where
linkage is required. Researchers must be able to determine what the data can reliably support.
Responsible health data reuse also depends on maintaining public trust and respecting patients' choices about how their information is used. In England, the National Data Opt-out allows individuals to choose whether their confidential patient information is used for certain research and planning purposes beyond their direct care. Where the policy applies, organisations must respect these choices, alongside relevant data protection and governance requirements. The opt-out does not apply to all forms of health research, so patients must understand how their data may be used, what safeguards are in place and what choices are available to them.[9].
The opportunity for the UK is therefore not simply to make more datasets available, but to create an environment in which health data can function as reusable research assets while
maintaining appropriate governance, security and public trust. Achieving this requires infrastructure alongside skilled data managers, clinicians, statisticians and data scientists capable of making health data scientifically useful.
Building Health Data for Use and Reuse
Health data reuse offers an opportunity to generate greater scientific value from data that patients, researchers, funders and health systems have already invested in creating. It should
not replace new data collection where new evidence is needed, but it should encourage us to ask whether existing data can responsibly answer a research question before collecting more.
For researchers, this means considering existing datasets and understanding their limitations before initiating new data collection. For data managers, it means recognising that good
documentation, metadata, coding and data structure determine whether data remain understandable and reusable. For data scientists, data preparation, harmonisation, quality assessment and reproducibility are fundamental to turning existing data into reliable research assets. For trialists and those developing health-data infrastructure, it means considering future
reuse throughout the data lifecycle.
As the UK improves access to health data, the opportunity is not simply to make more data available, but to make better use of what already exists. Responsible reuse depends on data
being well prepared, well documented, appropriately governed and used to answer questions they can reliably support.
References
[1] Jackson GH, Davies FE, Pawlyn C, et al. Response-adapted intensification with cyclophosphamide, bortezomib, and dexamethasone versus no intensification in patients with
newly diagnosed multiple myeloma (Myeloma XI): a multicentre, open-label, randomised, phase 3 trial. Lancet Haematology. 2019.
[2] Varvarà G, Burgwinkel C, Locher C, Ross JS, Mansmann U, Naudet F. Reusing clinical trial data to consolidate and advance medical knowledge. Journal of Clinical Epidemiology.
2025;188:111984. doi:10.1016/j.jclinepi.2025.111984.
[3] Shang Q, Yao S, Ouyang M, Wang X, Luo S. Secondary Analysis of Randomized Controlled Trials: Methodological Considerations and Best Practices. Annals of Clinical Epidemiology.
2025;7(4):137–147. doi:10.37737/ace.25017.
[4] NHS England. High quality patient records: good practice guidelines for GP electronic patient records. Version 1.1. NHS England; 2025.
[5] Macnair A, Gilbert DC, Langley RE, Cafferty FH. Healthcare Systems Data to Improve the Efficiency of Clinical Trials Within the UK: A Prospectively Planned Comparison of Healthcare
Systems Data and Trial-Reported Outcomes Within the Add-Aspirin Cancer Clinical Trial. Clinical Oncology. 2026;53:104059. doi:10.1016/j.clon.2026.104059.
[6] Jones M, Lally C. Accessing national health data for research. POSTnote 774. Parliamentary Office of Science and Technology; 2026. doi:10.58248/PN774.
[7] Sudlow C. Uniting the UK's Health Data: A Huge Opportunity for Society. Independent review of the UK health-data landscape; 2024.
[8] Department of Health and Social Care. Healthcare innovator appointed to Health Data Research Service. GOV.UK; 12 January 2026.
[9] NHS England. National Data Opt-Out. NHS England Digital. Available at: https://digital.nhs.uk/services/national-data-opt-out (Accessed: 11 October 2026).