The phased roll-out of England’s Clinical Practice Research Datalink (CPRD) has taken a significant step forward with new governance approvals from the Health Research Authority (HRA) that enable the CPRD to make available anonymised and linked data from a wide range of NHS sources. 

CPRD now has permission from the HRA’s National Research Ethics Service and the Confidentiality Advisory Group (CAG) to link patient-level data from primary care and other National Health Service datasets in such a way that qualifying researchers have access to anonymised data throughout the care pathway – an attractive proposition in a health system increasingly geared to long-term outcomes and real-life value.

“This creates a resource that is vital to helping health researchers access a more detailed picture of the journey of care,” CPRD noted.

Launched in April 2012, CPRD is the NHS’s observational-data and interventional-research service. It is jointly funded and provided by the National Institute for Health Research (NIHR) and the Medicines and Healthcare products Regulatory Agency (MHRA), with the latter hosting the service.

In its original form, CPRD combined the resources of the General Practice Research Database (GPRD), which had been collecting anonymised longitudinal medical records from primary care since 1987, and the Research Capability Programme piloted over the previous four years.

There are around 50 datasets in total available from the NHS, with a very high population of primary-care data. Governance permissions from the HRA’s Ethics and Confidentiality Advisory Groups mean these datasets can now be supplied to, and linked for, any approved researcher with an approved research protocol.

Improved linkage

As director John Parkinson points out, very few of the NHS datasets were connected before the CPRD was set up. And researchers used to have to go to National Information Governance Board (NIGB) for clearance each time they wanted to link datasets.

Around seven years ago, for example, the General Practice Research Database provided a gateway to hospital-episode statistics, cancer/myocardial infarction datasets and air-pollution data.

With block approval to link primary-care and other datasets, that gives researchers the benefit of eight more sets of cardiovascular data alone – subject, of course, to each protcol being approved.

The complete library includes medicines prescribed in primary care; drugs administered in hospital; laboratory data; consultations; GP and hospital coded disease data; disease registers; and central death data.

This is all “highly granular” data, Parkinson notes. Linkage also means researchers can cross-validate datasets, enhancing the integrity of their investigations.

Centralised approvals

Another marked benefit of the new governance approvals, in terms of CPRD’s core mission to expedite and enhance health research, is that they reduce the governance burden on research teams by centralising a large portion of the approval requirements for data access.   

CPRD now holds ethics approval for all observational protocols approved by the service – i.e., a centralised system of ethics clearance for CPRD observational studies.

For its part, the CAG has granted CPRD permission to go ahead with any protocols from researchers, subject to an expedited protocol review by the Independent Scientific Advisory Committee (ISAC) for MHRA database research.

ISAC is a non-statutory expert advisory body established in 2006 to provide advice on research-related requests to access data from the Yellow Card Scheme (for adverse drug reactions) or through the Clinical Practice Research Datalink.

The committee will consider various aspects of a research protocol, whether scientific (e.g., medical, statistical, methodological) or degree of linkage (number of datasets), and will then give the protocol a risk rating.

If the protocol is low-risk, it can go through without any further attention. If it is medium- risk, researchers will be told to exercise caution.

If the protocol is high-risk (e.g. rare-disease research with a higher possibility of patient identification), it will have to go back to the CAG for further assessment

Range of datasets

Meanwhile, CPRD is working to access a whole range of NHS datasets.

As Parkinson comments, healthcare research is “like a million-piece jigsaw” where there is always one last piece missing. Higher population coverage of primary-care data plus linkage to other datasets almost completes the puzzle.

In primary care, the aim is to get as close as possible to covering 100% of available data.

So far CPRD has achieved 8-9% coverage, but a joint programme with the NIHR’s Primary Care Research Network could add another 20% of coverage by recruiting some 2,000 general practices to contribute their data directly to CPRD.

A letter to this effect went out to general practices in mid- to late-August and CPRD is now talking to practice managers. 

Data anonymisation

Parkinson also underlines CPRD’s “powerful” data-stewardship model, which already meets the recommendations on information governance in the health and care systems made by the second Caldicott-review report in April 2013.

For example, participating general practices are given a unique embedded patient key and practice ID. The patient identifier and practice number are then encrypted for a second time before the data are made available to researchers via the CPRD.  

Patient consent

CPRD has also trodden carefully with patient-consent issues. Technically and legally, if the data are anonymised there is no need to seek patient consent for their use, Parkinson observes.

Nonetheless, CPRD promotes a patient opt-out system for anonymised practice data through a joint poster and leaflet campaign with GPs.

It has been doing this for seven years in five million patients, and during that time only just over 1,000 patients have chosen to opt out – a number that should not distort any research findings from the datasets, Parkinson comments.

Often, he adds, patients say they are surprised to find that anonymised practice data are not used more for research.

Personalised medicine

As Parkinson notes, dataset linkage through the CPRD also plays strongly to the growing trend for personalised medicine.

With a patient’s consent, CPRD can add 25 years of prior data to a clinical-trial dataset, adding in patient sub-groups and helping to target high-responsers to therapy. Researchers can then continue to track patients once the trial has ended.

The service is also key to designing and implementing adaptive clinical trials, providing a bridge into real-world settings, “from efficacy into effectiveness”, Parkinson says.