In my previous blog post, I discussed three core disruptive influences that must be considered when designing the architecture of clinical information systems.
The three disruptive influences are:
- Changes in the models of care
- Patients as active participants
- ‘Hospitals without walls’
In this post, I discuss the current landscape with healthcare information technology and how current products have evolved to deal with these disruptions as well as their deficiencies. My third post discusses an open future.
Current solutions and their problems
A number of strategies have already been adopted to deal with these disruptive influences:
- Enterprise-wide clinical information systems hosted by organisations to provide clinical record keeping as well as administrative functionality.
- Research portals and registries
- Patient portals to allow patients to view and contribute to their record - “personal health records”
- Patient-facing mobile device applications to allow recruitment into research.
- Organisation-by-organisation procurement and data sharing agreements with commercial entities.
Monolithic organisation-specific systems
Computer software deployed within healthcare has traditionally been procured and hosted by the organisation in which it is used. Many health organisations in the United Kingdom (UK) have closed networks in which desktop computers are wired into the network and run a mixture of different types of software with file-based network storage. More recent deployments of electronic patient record (EPR) systems such as Cerner’s Millenium and Epic within the UK have repeated this historic approach centred on an organisation.
Such an architecture benefits that organisation in meeting its statutory duty to care for patients, combining administrative and clinical data. In the United States, where these products were designed, that combination permits item-level billing and as a result they are integral to business operations.
However, this type of architecture is less suitable for cross-organisational working in which multiple disparate teams of health professionals and the patient collaborate. Such an architecture means that community teams, for example, might need to be registered directly as users on any systems in which data on their may reside to permit interaction with other professionals involved in a patient’s care. As a result, professionals may need to switch between applications in order to view a complete record. Similarly, single organisation-bound systems do not easily support sharing information between organisations so that information relating to a patient presenting as an emergency when on holiday to a different facility may not be readily available.
Currently, solutions to cross-organisational working adopt a health-information exchange (HIE) model in which information from multiple organisations, typically within a defined geographical region, exchange information. Commercial companies are now increasingly adopting data standards to try to provide semantic interoperability, and it is likely that their products will transform to a ``platform’’ on which other applications can run.
There are several important initiatives encouraging and supporting semantic and technical interoperability:
- Apperta foundation: http://www.apperta.org
- INTEROPen http://www.interopen.org
- HL7 FHIR: CareConnect profiles (UK): https://github.com/HL7-UK/CareConnect-profiles
- openEHR: http://www.openehr.org
- SNOMED-CT: http://www.snomed.org/snomed-ct, browser: http://browser.ihtsdotools.org, my blog posts: http://wardle.org such as http://wardle.org/clinical-informatics/2017/06/02/future-hospital-snomed-ct.html
The current focus is therefore on increasing interoperability at an organisation level, or between organisations within a defined geographic region (such as the Great North Care Record).
The current solution to giving patients access to their records is a secure `patient portal’, or multiple portals providing different functionality.
While `patient portals’ may allow patients to communicate with their teams, see their results and contribute data to their records, such portals are frequently tied a single organisation or service. For example, Cerner offer a proprietary patient portal which is linked with its own EPR. Even when patient portals are not tied to a single health organisation such as with ‘Patients Know Best’ (PKB) an organisation must have signed up to use that provider in order to allow a patient to interact with them via that portal. In Wales, there is a patient portal available for certain GP surgeries providing functionality for booking appointments but this is not universal and there is a separate portal for the recording of patient reported outcome measures, the results of which are not made available in clinicians. The Healthwise Wales programme is a separate portal that allows the population of Wales to register for a research project on wellbeing and health, but it does not integrate with or support the enrolment into any other research.
As a result, the choice of patient portals is by the organisation and not the patient, who may wish to interact with multiple providers of healthcare and flag their interest in multiple research projects.
As such, how can a patient interact with all of their disparate health services across multiple organisations? Do we expect our patients to use multiple portals depending on whether they are visiting the Royal Free or the Royal London Hospital?
Similarly, patients with specific needs, such as those who do not have a fixed address, or those requiring communication via braille or spoken text, must hope that a single provider supports those needs. What if a patient cannot manage appointments themselves, but has a proxy who needs to be involved in their care, such as a nursing home resident?
Research portals and registries
Clinical research has traditionally focused on recording longitudinal patient outcomes in well-defined cohorts of patients recruited as part of observation or interventional trials. As a result, there are now many different research registries which are usually disease or specialty-specific. I have written about the disadvantages of registries before, but essentially they don’t really support clinical practice. The main problems are:
- there are lots of them, so potentially lots of disparate systems (they are not joined up)
- they are not embedded in day-to-day workflow
- there is sometimes little feedback, result or reward, particularly in any timely manner
The current mainstream solution for clinical research is direct recruitment by clinicians in their clinical practice or via advertising in newsletters of third-party charitable organisations. However, technology such as Apple’s ResearchKit provides a simple framework for research teams to build trivial mobile device applications to allow patients to sign up themselves, consent and then take part in any necessary data capture for clinical research. As a result, patients can be recruited in large numbers for a single project. As an example, see the mPower study on Parkinson’s disease published in Nature.
A limitation of these approaches is that research data are typically collected separately from routinely collected clinical data. Historically, this has not had any significant drawback due to a lack of routinely collected clinical data that would actually be usable for clinical research. However, given the other disruptive changes in healthcare, it is likely that structured longitudinal data will be increasingly available from systematic collection in clinical practice. Much like the failed ‘care.data’ project, such data would need to be combined with multiple other disparate sources of information and as a result, patients must be given appropriate tools to understand and control the use of their information. Importantly however, a record of consent must take care to not divulge personal information as a result of its existence. Just as in email communications, if information relating to a withdrawal of consent of a patient from a HIV project is made available, then confidential information is leaked. Thus, a patient’s current and historic opt-in or opt-out choices must be made available so those choices can be enforced and yet remain private.
Organisation-by-organisation procurement and data sharing agreements
Most innovative healthcare information technology solutions must still be procured by organisations and then deployed. As a result, a commercial entity acts on behalf of that health organisation and, if data are transferred, formal data sharing agreements signed. For small and medium sized enterprises, procurement processes are long, complex and expensive and traditionally have favoured larger, more established companies over smaller companies.
Likewise, procuring an established, ‘proven’ solution is perceived as lower in risk than collaborating with a small or medium-sized enterprise in a development partnership. Procurement rules frequently require levels of insurance, financial and regulatory backing that make it more difficult for new enterprises to deploy and scale innovation with the health system.
How then to balance innovation and governance? Unfortunately, there are many examples of innovative projects that are closed down as a result of a loss of trust in the governance of private and confidential medical information. For example, the `care.data’ project aimed to combine multiple large scale datasets together across multiple organisations in order to support population health research within England, but controversy resulted in its closure. Similarly, Google DeepMind’s project to provide a smartphone clinical application (Streams) to clinicians at the Royal Free has been mired in controversy with negative press reports and a referral from the National Data Guardian to the Information Commissioner’s Office (ICO) and a formal ruling pending.
Such debacles result in a perception of wrongdoing even if no information was shared inappropriately and risk the deployment of innovative technologies in the future Unfortunately, few commentators note the consequences of not deploying modern technology in order to support healthcare or indeed note the gap between where we should be and where we current stand.
In summary, existing technology focuses on an organisation, a specialty or a specific project. In essence, the digitisation has followed existing paper-based processes instead of using technology to underpin transforming our healthcare services around the patient.
If you have read my other blog posts, you will already know the answers to the problems posed in this and the previous post; we need to focus on a data-driven standards-based “platform” consisting of multiple well-defined services working together to create vendor-neutral archives of immutable clinical information in structured formats. My next post covers the components of this platform in more detail.