Feature, Report

Big data: Cracking the eyecare code

Big data is being touted as a key trend that will shape Australian eyecare for decades to come. RHIANNON BOWMAN examines how the sector is grasping a data-driven era and how this will impact future eye health models and policy.

Big data – and its inherent ability to reveal information unavailable to traditional research methods – is becoming a dominant driver in the future direction of Australian eyecare.

Whether it be identifying patterns in care, informing eye health policy or for the development of artificial intelligence-based diagnostic systems, it is widely agreed that data holds the key to overcoming the biggest challenges facing the ophthalmic sector today.

Optometry Australia (OA) is among a host of organisations that believe a data-driven approach will help shape the next 20 years of eyecare.

In its report: Optometry 2040 taking control of our future, it states that data driven-decision making can improve efficiency and productivity at the system and practice level. It also offers benefits in terms of performance management and quality improvement, while providing more clarity on cost-benefit analysis in patient management decisions.

While there is excitement around the transformative effect of big data, its emergence also raises a host of legal and ethical questions to consider.

In Australia, OA reports a current “crisis of trust” in data and privacy. As more technology is embedded into practice, it has identified a need to balance the increased capacity for data collection and analysis with trust between the clinician and patient. Insight examines how the sector is carefully embracing a data-driven era.

Big data research

To explain the true power of big data in Australian eyecare, few are as well-credentialed as Professor Claire Vajdic.

Claire Vajdic, UNSW.

An optometry graduate from the University of New South Wales (UNSW), Vajdic later pursued a career in research and a PhD in public health. Since 2015 she has been head of the Cancer Epidemiology Research Unit at the Centre for Big Data Research in Health at UNSW – Australia’s first research centre dedicated to this field.

It aims to maximise the use of all possible sources of health big data, and enhance the health and wellbeing of Australians and the global community.

“Big data, for example, can help identify those at risk of diseases like diabetic retinopathy, and the best referral or management practice. The datasets we develop are about patterns of healthcare and predicting better outcomes,” she says.

“It’s exciting to also think of the future when wearables will come together with administrative and detailed clinical records in order to achieve better patient outcomes.”

In health, the term ‘big data’ includes the millions of records generated routinely by health services, real-time clinical data captured at the point-of-care and genomic data produced in research and clinical settings. It also involves health-related data from the population at large through technology such as wearable devices and social media.

Vajdic says the centre’s objective is to improve population health, working with clinicians, hospitals and data scientists and using existing health records. Its research is facilitated by NSW Health’s Centre for Health Record Linkage (CHeReL), and the Australian Institute of Health and Welfare’s (AIHW) Data Integration Services Centre, which link multiple sources of data and maintain a record linkage system that protects privacy.

“Our linked administrative health data is provided to approved researchers only and provides population coverage so we’re able to minimise bias,” Vajdic says.

“We can receive linked optometric and allied health data, including Pharmaceutical Benefits Scheme and Medicare Benefits Schedule records, so we can look for patterns in that data at a group level, not individually identified practitioners,” she says.

Vajdic says preserving privacy is a key tenet of the centre’s work, and it is serious about a concept termed the ‘separation principle’.

Data can also illuminate the use of prescriptions for conditions such as diabetes and glaucoma.

This principle means no one working with the data can view both the linking (identifying) information about an institution, clinician or patient, such as name, address, or date of birth, together with the merged analysis (content) data such as diagnostic or treatment information.

“As researchers, we are granted access to content data, not identifying variables, for privacy and confidentiality reasons,” she says.

“Specialist linkage centres like CHeReL, which take care of the separation principle, have been a wonderful development over the last 10 years; it has transformed the quantity and quality of data we can access and analyse.”

Vajdic explains that institutes such as the SAX Institute, based at the University of Technology in Sydney, have also transformed the big data field with developments such as SURE, the Secure Unified Research Environment, a computing solution that brings researchers together in order improve healthcare.

“It’s upped the infrastructure to allow for secure, large-scale data analysis,” she says.

“We can access eyecare data captured at a tertiary level, such as cataract surgeries performed in hospitals, or research who is gaining access to the expensive drugs to treat age-related macular disease, whether vulnerable populations have the same access to care, and the risk factors for receiving a later diagnosis. We are looking for patterns in the data.

“Data can also illuminate the use of prescriptions for conditions such as diabetes and glaucoma. Using the separation principle, we can look at prescription data for the entire community while preserving the identity of practitioners writing the prescriptions.”

Vajdic says researchers can also use linked administrative health data to investigate the safety and quality of health care.

“For example, some clinicians may claim they have the ‘most complicated or high-risk patients’ as an explanation for poorer outcomes. We can compare like-with-like and can demonstrate if a hospital or group of practitioners are aligning with guideline-based care. We can compare different models of care and look for any downstream effects.”

The future in big data, Vajdic says, will be focused on areas that can be improved and minimising unwarranted variation in care.

“Our clinical analytics group has expertise in artificial intelligence, including deep-learning and natural-language processing. Some potentially important clinical data is currently only captured in free-text fields. It is unstructured and we are using these analytical approaches to discover new, actionable insights about patient care and patient outcomes.

“Artificial intelligence is aiding pattern recognition and clinical diagnosis using image-based data, such as those generated by various scan-technology in eyecare.”

When it comes to ethical and legal considerations of data use, Vajdic believes these are long-standing strengths of her profession.

“We are guided by NHMRC’s National Statement on Ethical Conduct in Human Research. All actors in the big data pipeline also adhere to the Federal Government’s Five Safes framework. It is a risk assessment framework for data access, ensuring safe people, safe projects, safe settings, safe data and safe outputs.

“There is also pending Commonwealth legislation expected in the next four to six months from the Office of the National Data Commissioner, established last year. When the legislation passes Parliament, it will be known as the Data Availability and Transparency Act (DATA).”

Kate Taylor, Oculo.

Anecdotal to evidence-based

Cloud-based platform Oculo is an Australian company focused on the secure, instant transfer of clinical correspondence between healthcare professionals. Currently, more than 2800 optometrists and 730 ophthalmologists use the system in Australia.

From her front-row seat in clinical communication, founder and CEO Ms Kate Taylor believes the medical sector is in the midst of a revolution, transitioning from anecdotal to evidence-based care. She believes future healthcare lies in an analytics-driven approach.

“Anywhere that moves from the descriptive to the digital may be disrupted. Eyecare clearly uses a lot of digital technologies for diagnosis and management – and increasingly for communication,” she says.

“There’s also significant need to improve quality and efficiency for the health system, as well as to better support patients in their own care. Data-driven care tools will offer opportunities on many levels.”

With troves of big data being collected, stored and interpreted, it stands to reason that the sector faces inherent challenges of responsible data protection and use.

“Healthcare relies on trusting relationships: trust between patients and care providers, trust between colleagues, trust in the system as a whole. Data issues create a whole new element of trust in healthcare interactions, even when we are seeing other sectors breaking trust with consumers,” Taylor says.

“It’s much easier to lose trust than to regain it, so we need to engage in transparent and respectful dialogue about data use and protection and to recognise that all parties are learning as the technological opportunities grow. That includes patients, providers, policymakers and payers.”

Strategies spawn from data

Naomi Barber, Specsavers.

With 400 stores now open across Australia and New Zealand, Specsavers is uniquely placed to share how big data influences its eyecare model.

Last year alone, the company conducted more than 3.5 million consultations – accounting for almost 40% of patients presenting to optometry. Its latest Eye Health Report featured data from 8.5 million patient journeys, making it the largest publicly available data set of its kind in the region, increasing by an additional 3.5 million data points from 2018.

Specsavers professional services manager Ms Naomi Barber says that while record management compliance is a critical component, the company is also interested in using the data to develop large-scale strategies to overcome national eye health challenges – such as improved detection and referral rates for glaucoma and engaging people with diabetes for more regular eye tests.

“We continue to focus on integrating equipment and software to enable optometrists to capture key clinical information efficiently and accurately. This allows us to collate large-scale data, identify what leads to effective detection of eye disease, and report on trends and variability in practice,” Barber says.

“This data is crucial to implementing clinical initiatives that drive enhanced standards of care. We know this because we can measure patient outcomes now.”

Specsavers has integrated its patient management system Socrates with e-referral platform Oculo to facilitate accurate, timely, secure and two-way communication with other providers, including GPs and ophthalmologists.

The system is designed to place the patient at the centre of communication and, as such, the data Specsavers obtains is patient-outcome centric and underpins its mission “to transform eye health”.

These systems have also been key to the development of benchmarks that provide context and a point of reference for optometrists in their practice.

Each week Specsavers optometrists receive a personal report which includes a broad range of data, including their Medicare item number utilisation and personal detection and referral rates for individual eye diseases. This is set alongside whole-of-practice benchmarks and national comparisons, allowing each optometrist to compare their activity alongside that of their colleagues.

“We believe the goal for every optometry practice should be to embed benchmark reporting as a key component of professional development, to ensure it is readily measured in terms of the benefit to patients,” Barber says.

The distribution of benchmark reporting through the Specsavers network has been coupled with key clinical initiatives and is said to have resulted in measurable enhancement to patient outcomes.

For example, it was due to their optometrists’ data on glaucoma referrals that Specsavers has been able to validate the correlation between visual field performance and detection of glaucoma approaching population prevalence.

“This has resulted in the first ever evidenced-based benchmark for visual field performance in primary optometric practice,” Barber says.

Investment to improve integration

During the past five years Specsavers has also been developing a framework to harness integrated equipment, technology, IT, professional development and collaboration strategies.

It plans to use the data obtained from this to draw attention to the challenges the industry faces in eye disease detection. In doing so, it is also able to develop new strategies that are designed to improve eye health outcomes for Australians.

“For example, we are now completing a Socrates update that will fully automate the registration of patients with diabetes onto the KeepSight portal, as part of the standard eye test routine,” Barber explains.

“When the optometrist sets the recall period, in agreement with the patient, their information will be updated – or registered for the first time – on the KeepSight portal.

“This supports the government’s goal of developing a comprehensive record of patients with diabetes engagement with regular eye testing.

“This is critical because we know that regular eye testing of patients with diabetes leads to earlier detection of diabetic eye disease which, in turn, helps save sight.”

Sharing data for common goal

Specsavers has made no secret of its belief that the collection and analysis of patient outcome data is integral for a sustainable model of preventative care.

For patients to truly benefit from future advancements in primary practice, Barber notes that information needs to be shared between all stakeholders to ensure eyecare continues to progress at a consistent standard.

“Aside from the significant role that data plays in demonstrating measurable impact of clinical interventions, we believe that all patient outcome data collected should be used at a national, regional and individual level to keep optometrists abreast of their practice and the impact they are having on patient outcomes.”

To that end, Specsavers will continue to share its data with the Australian Federal Government for a range of purposes and for initiatives such as the Medicare Benefits schedule review process. It also publicly publishes its data through its Eye Health Reports for the wider industry to access.

“We do this because we know the data we collect provides vital evidence to substantiate the importance of subsidised eyecare services and the intrinsic need for affordability and accessibility to drive preventative eyecare,” Barber says.

Peter van Wijngaarden, CERA.

AI solution to the big data challenge

Associate Professor Peter van Wijngaarden, from the Centre for Eye Research Australia, says advances in imaging technologies have transformed eyecare in the past 30 years, with the typical eye health consultation today involving the use of several imaging devices.

This means that a significant amount of data is being generated in each consultation.

“Though these data are powerful aides to clinicians in arriving at a diagnosis and deciding on the best treatment plan, it is becoming increasingly challenging for clinicians to assimilate and interpret all of the data that is being generated,” he says.

“A rather apt phrase to describe the situation is that each clinical interaction is becoming a ‘big data challenge’ and big challenges warrant innovative solutions.”

Accordingly, van Wijngaarden says this challenge has become a major driver for the development of artificial intelligence (AI) solutions. These are being developed for a variety of indications in eye health, but chief amongst these are tools to support image analysis – such as algorithms that can detect and triage retinal diseases from OCT images.

These tools, he says, promise to streamline the analysis and interpretation of complex imaging data. Some systems have the ability highlight key image features that may be of interest to the clinician.

“Analysis is, however, only one part of the big data challenge. Data collection, curation and storage are important considerations, as are data security and privacy. The capacity to link diverse sources of data for a given patient, such as imaging, demographic and clinical data are also important challenges,” he says.

In terms of data storage, van Wijngaarden says each image file is large and many images may be acquired in a given consultation, meaning the challenge of scale becomes significant in a short space of time. There is also the problem of bandwidth for upload to cloud storage – an issue in many clinical sites, especially in rural and remote settings.

Privacy and data security are also considered important factors. International media have reported major breaches, with the National Health Service illegally handing Google-owned artificial intelligence company DeepMind 1.6 million patient records for a trial – without patient consent.

Possibly even larger breaches have also been reported in the US.

Clinical data is also integral for the training of AI systems, and that data has value attached to it. Companies are amassing large clinical databases of eyecare data, which will undoubtedly be monetised in future.

van Wijngaarden says the European General Data Protection Regulations (GDPR) is setting a new bar for privacy and data security.

Key principles enshrined in the GDPR are that personal data can only be used under six lawful bases (consent, contract, public task, vital interest, legitimate interest, legal requirement) and a key aspect of the provision of data under consent is the right of the data subject to revoke that consent at any time.

This also extends to the right to erasure – the ability for the data subject to request that their data is erased at any time. This poses a raft of technical challenges for companies using large amounts of health data. Similar controls are being applied in other jurisdictions.

“Technology may also constitute part of the solution – blockchain, for instance may be used to embed consent with the data itself.

“This may help to ensure that data is only used for the purposes for which consent has been granted and that consent remains linked with the data,” van Wijngaarden explains.

Finally, van Wijngaarden says it is vitally important that healthcare providers are given access to these data in a convenient, comprehensible and secure manner to translate its potential into improved healthcare for patients. This has implications for technology developers, but also for the training and education of healthcare providers who are increasingly required to be expert informaticians.

“In the torrents of data and technology healthcare providers must not lose sight of the empathy and compassion that is the foundation of the care in health.”