Intended for healthcare professionals

Analysis

Putting patients in control of data from electronic health records

BMJ 2018; 360 doi: https://doi.org/10.1136/bmj.j5554 (Published 02 January 2018) Cite this as: BMJ 2018;360:j5554
  1. John P New, clinical professor of diabetes and endocrinology1 2,
  2. David Leather, global medical affairs and study lead33,
  3. Nawar Diar Bakerly, respiratory physician1 2,
  4. John McCrae, chief technology officer1,
  5. J Martin Gibson, clinical professor of diabetes and endocrinology1 2
  1. 1Salford Royal NHS Foundation Trust, Salford, UK
  2. 2Manchester Academic Health Science Centre, University of Manchester, Manchester, UK
  3. 3GSK Respiratory Centre of Excellence, GlaxoSmithKline UK, Uxbridge, UK
  1. Correspondence to: J P New john.new{at}manchester.ac.uk

Research using data from electronic health records offers great potential in healthcare, but patients must be informed about how their data are to be used and for what purpose, argue John New and colleagues

Many areas of healthcare use electronic health records to document and store clinical data,1 and there is much interest in using this information for clinical research. However, such use is outside clinical care and requires patient consent if the data are identifiable.

Several approaches can be used to get consent for identifiable data being shared for purposes other than personal care. Organisations have tended to favour an opt-out model, believing that an opt-in approach based on active consent would be unfeasible or lead to a low participation rate. In both models, if not enough patients opt in or too many patients opt out, it could be argued that the data are no longer useful or generalisable.

NHS England’s care.data programme, designed to provide access to health and social care information from different settings, was abandoned, partly because it failed to provide clear information about how data would be used and win the confidence of patients and healthcare professionals.23 Patients must be informed of the intended uses and benefits of sharing their data for research to raise public confidence in data sharing and to avoid patients opting out. Here, we describe our experience of an opt-in approach for sharing electronic health record data and discuss it in the context of national and global approaches to consent for sharing health data in research.

Using electronic health records

Electronic health records can provide comprehensive data on medical history, prescriptions, and service use for large populations. They have an advantage over research databases in that there is no interviewer or patient recall bias (eg, adherence to medication can be estimated by the number of prescription refills). Linking electronic primary care data to secondary care service records allows researchers to assess disease progression and use of healthcare resources. By recording clinically important outcomes in a wide range of participants,456 observational research studies of routine clinical care allow comparative effectiveness research to be integrated into clinical care.7

In England, primary care has used computerised record keeping since 2000 and general practices use one of several, nationally available electronic record systems. However, they are not generally used for research and there is no standardisation between systems. Across the UK, national datasets such as the Clinical Practice Research Datalink use primary care records for clinical research, but these are not linked to other electronic health records and are not suitable for use to monitor a clinical study because they are not updated in real time.

Other countries, including the US, use electronic health records for clinical and other purposes, such as processing insurance claims.8 However, these records are not representative of the general population, as some only contain information on individuals who have healthcare insurance. These records may also be limited to data from hospitals, rather than information from both primary and secondary care. The US has recently taken steps towards greater standardisation of electronic records.9 Adoption of new standards, such as the Fast Healthcare Interoperability Resources (FHIR), is likely to improve the ability to use linked datasets in clinical trials, although it may be some time before these technologies are widely available.

In Europe, Sweden and the Netherlands are leading the way. Sweden has a national health record that can be accessed by healthcare providers across the country as well as by patients and their nominated family or carers.10 Information from different healthcare providers is harmonised by use of standard interdisciplinary terminology (Systematized Nomenclature of Medicine Clinical Terms, SNOMED international). Data from the health record database are copied into national registries that may release data to research organisations, subject to ethics committee approval. Consent for data sharing is on an opt-out basis, and patients can request information from their provider on how their details are used.11

In the Netherlands, most medical records are updated electronically and several systems are in place for the electronic exchange of patient data between primary and secondary care providers.12 The two University Medical Centres in Amsterdam (the VU University Medical Center and the Academic Medical Center) have adopted a shared electronic patient record system and are currently asking patients for consent to merge their data and create a single shared record for both sites.13

Inconsistencies between different sources of electronic health record data can make sharing difficult. In the UK, primary and secondary care providers use different coding schemes and terminology for classifying diseases and these must be cross mapped, although coding for long term conditions has become more robust since the introduction of the Quality and Outcomes Framework in 2004.14 In addition, certain data important for clinical research, such as disease severity, are not well coded in electronic health records.

Salford integrated record

Ten years ago, Salford, a borough of Greater Manchester, UK, relied on many disease specific registers to identify and manage patients across the district, resulting in much duplication of effort for those with comorbidity. In 2007, the Salford Integrated Record (SIR) was established with real time electronic links between primary and secondary care. The SIR linked these sources with patients’ prescription information, capturing data on everyone accessing health services in Salford.

Key to the success of the project was informing patients about how their data would be used—that is, predominantly to allow local clinicians access to their information for better healthcare. All patients were sent an information letter and asked if they wanted to opt out. Fewer than 500 of 230 000 patients chose to opt out (<0.2%). SIR then became an integral part of primary and secondary care. SIR was developed before the widespread movement to allow patients access to their records, and currently only allows healthcare professionals to access patients’ records. If presented with the same scenario today, an active opt-in approach may have been considered, offering patients different levels of data sharing—such as solely for clinical care or for both clinical care and research.

In 2008, the NorthWest EHealth (NWEH) group was established (https.nweh.co.uk) to use SIR data for research. The group brought together computer scientists, clinicians, researchers, statisticians, epidemiologists, and project managers to develop methods and a governance framework to extract patient healthcare data for use in real world research studies. The Salford lung studies (SLS) are one example of the integrated record being used for research.

Salford lung studies: an example of opt-in approach

SIR helped to quantify the burden of asthma and chronic obstructive pulmonary disease (COPD), identified exacerbations, and evaluated differences in disease outcomes and healthcare resource use in Salford.1516 Building on this, the Salford lung studies (SLS) programme evaluated the effectiveness and safety of an unlicensed inhaled medication for the two conditions in everyday clinical practice in Salford and South Manchester. The studies are the first prospective, randomised controlled trials of an unlicensed drug using electronic health records to capture data on clinical endpoints.171819 The linked database system captured data in near to real-time from trial participants accessing health services at local hospitals and general practices in these areas.

Although SIR’s opt-out approach is adequate for sharing data for clinical care, a randomised clinical trial requires explicit informed consent. The Medicines and Healthcare Products Regulatory Agency required real time patient safety monitoring and information governance for the trial, and the local ethics committee specified that the use of patient data necessitated written consent. NWEH therefore set up data sharing agreements with every general practice and hospital, the Office for National Statistics, and the NHS Secondary Uses Service.

Participating patients signed a consent form which permitted their GPs to share their data in entirety and link to other databases. Explicit information was provided to patients about how their data were to be used and for what purpose. This was achieved by extensive training of primary care nurses, pharmacists, and GPs, who were the main recruiters to the study. Patients discussed the data sharing terms with a GP specially trained in the study, supported by a nurse qualified in good clinical practice. Patients were given an ethics committee approved patient information sheet before opting in, which described the terms of data use. We also established a public education campaign and a website for prospective recruits to further clarify the requirements of the study and detail how their data would be collected and used.17

By consenting to join the trial, participants confirmed that they were comfortable for the NWEH group to hold their electronic records and for researchers (study sponsor) to receive an anonymised subset of these data for analysis of the study endpoints. All patient data shared with the sponsor were predefined, specific to the study, and approved by the appropriate ethical bodies. As the COPD study collected all safety data and most efficacy endpoints through the electronic record, patients who had opted out of SIR were not eligible to participate. However, we invited those who met the clinical criteria for the study and explained that they must opt in to SIR and data sharing agreements to be eligible. Fifty eight patients opted to join SIR so that they could participate in the study. In total 7039 of eligible patients have been recruited (2802 with COPD; 4232 with asthma).

Acceptability

Experience from the SIR project suggests that the opt-out approach is acceptable across Salford for facilitating clinical care. The transparent opt-in approach used in the lung studies informs patients about the purpose for which their data are to be used and who will have access to their data. Trial participants did not object to their information being held by a trusted third party for research purposes and supported anonymised data extracts being sent to the study sponsor. This suggests that patients throughout the UK might be willing to provide informed consent to data sharing and give consent for specific use of their medical information if the government restricts access to health related organisations.20

The Salford experience also provides evidence that local data for local use is acceptable to patients and health providers. This approach is consistent with the recently revised Caldicott principles21 and may be more acceptable to patients than NHS Digital holding all data centrally. Patients’ views were considered when designing SLS, and many expressed their pleasure at being able to take part in clinical research.

In the Greater Manchester area, as part of the devolution of the health and social care budget to the city, the devolved Health and Social Care Partnership is already developing and rolling out DataWell, a platform that will be used to enable health data to be shared between healthcare providers. This could enable the SIR/SLS technologies to be extended to the entire local Greater Manchester population (2.9 million).

The main barriers to the opt-in approach were the time and staff required to explain the terms of data usage to patients. Although it took time to help patients understand how their data would be used, in general they were reassured by these explanations. The most common questions asked by patients were who would be able to see their data and if they would be used for their clinical care; in this case, the data were only used for the study.

Rationalising an opt-in approach

An opt-in approach to data sharing aims to provide an ethical resource to support clinical research that ultimately benefits patients, clinicians, and payers. Patient consent for data sharing allows the linkage of datasets, which is more difficult to achieve with anonymised data. Under an opt-in approach patients should be informed before their data are accessed. The system should also allow for varying levels of data sharing with different stakeholders (eg, for clinical care, for research, for clinical commissioning, or with pharmaceutical or healthcare providers), ideally with patients being informed of who has been provided with their data, for what purpose, and what the outcomes of the research were, to alleviate concerns over misuse of personal data. Patients should be able to amend their opt-in choices at any time, allowing them control of how their data are used.

In light of the NHS strategy to allow patients access to their medical records online, the NHS has been in discussions about developing a simple, secure portal.2223 Parallel incorporation of an opt-in choice for different levels of data sharing, and enabling users to update their permission settings (akin to updating cookie preferences for website access), would add extra value for patients, clinicians, and payers. Development of individual “health accounts” could allow patients to access their data and approve or amend their use by certain organisations at any time. Such developments must be underpinned throughout by effective and transparent communication with patients to build trust.

The static, one-off approach used in SLS is not a perfect model. Ideally, patients would be able to review their consent and opt-out at any time. Additionally, the single level of consent in SLS made all sections of patients’ electronic records accessible. A more refined system that allows patients to select who can access their records, and the level of access for each person, would be an improvement.

Although the costs of implementing an opt-in versus opt-out approach are considerable, the benefits may recoup this outlay. For example, medical tests are often repeated because healthcare providers do not have access to previous results. Linked electronic records could prevent the additional time and costs of repeating these tests. Greater linkage of records could also facilitate future clinical studies, reducing the initial costs.

Key messages

  • Data from electronic health records cannot be used for research without patient consent

  • Opt-in methods for consent have been avoided because of fears about low participation

  • Experience in Salford suggests patients will opt in to data sharing if properly informed

Footnotes

  • Contributors and sources: JPN, JMG, and NDB were involved with setting up the Salford Lung Study, and NDB leads the clinical safety team for the study. JMcC was involved in implementing the IT infrastructure in the Salford Lung Study; DL conceived the concept for and was involved in setting up the Salford Lung Study. All authors contributed to writing this article and approved the final version. Angela Rogers, from Gardiner-Caldwell Communications, provided editorial support in the form of development of a draft outline in consultation with the authors, editorial suggestions to draft versions, collating author comments, copyediting, and fact checking. She was funded by GSK. JPN is guarantor.

  • Competing interests: We have read and understood BMJ policy on declaration of interests and declare the following interests: JPN has received consulting and speaker’s fees, an educational grant from GSK, and is employed by NorthWest EHealth on a grant from GSK for the Salford Lung Study; JMcC is employed by NorthWest EHealth on a grant from GSK for the Salford Lung Study; DL is an employee of, and holds shares/stock options in, GSK; JMG’s institution has received funding from GSK as the Salford Lung Study sponsor; NDB is employed by Salford Royal NHS Foundation Trust and NWEH, and has received travel support and an honorarium to speak for GSK and Novartis.

  • Provenance and peer review: Not commissioned; externally peer reviewed.

References

View Abstract