From early in the COVID-19 pandemic (July 2020), Digital Health CRC (DHCRC) along with Macquarie University have partnered with a number of primary health networks (PHNs) in Victoria and NSW among other collaborators including Outcome Health in successive research projects investigating how near real-time de-identified data from general practices could inform and help optimise care, first for COVID-19 and then extending to long COVID.
“It was a very difficult situation. No one knew what was happening,” explained Professor Andrew Georgiou, from the Australian Institute of Health Innovation at Macquarie University, of the opportunity that arose from those very early days of the pandemic.
General practice was a key area where COVID-19 care was and continues to be required and happening to this day. “A lot of the insights we’re looking at were real time, but also first time…because we’d never been through it before,” Dr Stephen Weeding, then Strategy and Commissioning Manager Digital Health at Eastern Melbourne PHN (EMPHN), pointed out.
The starting point
Professor Georgiou and team proposed interrogating near real-time PHN-owned de-identified electronic data, collected with permission from general practices via Outcome Health’s POLAR platform, involving general practitioners (GPs) and PHNs to help make sense of the data and produce meaningful, practical insights that would guide appropriate care for patients in response.
“We felt we had a mission, we felt obliged to put out as much information as we could. It was a way of producing and getting messages out there about what was going on and getting a conversation going,” Professor Georgiou said of the aim of the research, and its translation, at this turbulent time.
Bringing together a successful action-research collaboration
With permission from participating PHNs, de-identified medication and patient data from general practice electronic health records was extracted by the data custodians via the POLAR platform and provided to the research team. This involved over 7 million patient records covering approximately 60% and 43% of the Victorian and New South Wales populations, respectively, across 869 general practice clinics from four primary health networks — two in NSW and two in Victoria. [1]
The researchers took an established approach called action research, which involves participation from key stakeholders throughout the entire process, and for these projects the researchers met every 2-3 months with PHN representatives and general practitioners (GPs). [2] For Dr Precious McGuire, GP and Clinical Editor at HealthPathways Melbourne for EMPHN, she was interested in taking part to “see where the data from general practice went and what it was being used for”, but also appreciated GPs were invited to be involved and could directly influence the research process.
The collaboration built in from the beginning to the end and beyond the project was key as it provided input into what to research, whether the findings were meaningful, and putting the findings into context, which really improved the entire process. Researchers have the ability to get the answer, but need the clinicians’ context to ensure they were looking at the right inputs, to produce outputs that are meaningful.
Overall, the collaboration succeeded as everyone involved – the researchers, the GPs, PHNs and Outcome Health – all wanted to provide back into general practice at a time of great pressure and demand, united in making sure the data coming from general practices was being returned in a format that supported GPs to optimise care in constrained and uncertain circumstances.
Producing outcomes to benefit clinicians and consumers
At the start of the process, there was a lot of data available, but “we didn’t know what we didn’t know”, Dr Weeding said. The benefit of involving PHN representatives and GPs could be seen particularly in two ways:
- Helping to identify whether relevant data existed and if so where in the system researchers could find it to answer the questions posed;
- Providing context to determine if the data available was meaningful and could produce useful insights towards care.
One of the initial challenges presented by the data was the variability in how patient information was entered into practice software and coded. Dr McGuire gave an example: “if it’s not entered in a certain way, it’s not coded in a certain way, it doesn’t mean that it’s not there…if they’ve got COVID you actually need to document they’ve had COVID, because you can’t diagnose Long COVID unless you’ve documented they’ve had COVID.”
While there’s still no clear consensus on the number of individuals that will develop long COVID following an acute case of COVID‐19, Australian estimates suggest between 5% and 10% of people who have had COVID‐19 will develop long COVID. [3] However, the project team’s research found only 0.03% in the PHN GP dataset analysed had a diagnosis of long COVID, with the study calling out that “long COVID was almost certainly under‐represented in our results” because of limitations in the data. [4]
For example, they found that only 41.3% of the general practitioner‐led long COVID diagnosis cohort had a previously documented COVID‐19 diagnosis – that is, over half of those diagnosed with long COVID did not have a previously documented COVID-19 diagnosis. They also discovered almost 49,000 records of prescriptions for antivirals to treat COVID‐19 without corresponding documentation of a confirmed COVID‐19 infection; again, indicating the possibility of additional unrecorded acute COVID‐19 cases in the comparison cohort. [4]
Dr Geoff Campbell, GP in the South Eastern Melbourne PHN (SEMPHN) who also contributed to the research, added “You’ll probably find some GPs have never coded long COVID in their life, and some have probably coded just about more cases of long COVID than existing in Australia. Somewhere in there is the truth.”
By having a better understanding of where the data in the practice management system goes and how it could be used, and then having conversations about what the data highlighted, it improved how the doctors and their practices approached clinical notes and coding in turn.
“We still were living COVID, but we still needed that frontline context of how [the data is] being collected, how it’s stored, how it’s then retrieved and has been interpreted was really invaluable from that point of view,” Dr Weeding provided as perspective on PHN participation.
Once the relevant data was identified, the researchers looked again to the GPs and PHNs to provide the clinical view that would allow them to look in the right places and draw conclusions that made sense in context, using design-thinking to identify priority areas for research and then rapidly generating knowledge in those areas to support care activities.
Understanding the value and impact of this research
Rapid evidence generation and dissemination
An early valued output of this project was a series of seven snapshots [5] on emerging trends related to COVID-19 diagnosis and treatment, and the impact on patients and medication and testing trends, to help inform GPs keep their fingers on the pulse of the ever-changing trajectory of the pandemic.
With near real-time data and on-the-ground insights from the GPs and PHNs, the research team were able to rapidly synthesise emerging information and evidence into reports, released between December 2020 and June 2021, that were distributed as quickly and as widely as possible to keep clinicians updated.
Since publication, the snapshot series has been downloaded over 1,500 times, and the researchers have also produced a suite of impact measures on the impact of COVID-19 on Australian general practice.
There are also nearly 20 peer-reviewed publications as a result of these projects, with recent examples including a paper discussing key challenges of diagnosing and managing long COVID [6] in the Australian Journal of General Practice, as well another on oral COVID-19 antiviral prescribing in Australian general practice in the Australian Journal of Primary Health. [7]
The value of clinician input in health research and policy in healthcare
Many GPs may not be aware of research that’s occurring using general practice data and how it relates to their day-to-day work, but Drs Campbell and McGuire both stressed the importance of involving GPs in the process of deriving insights from the PHN GP data.
“We’re trying to talk about the same thing, but we don’t often get together and actually out these problems,” Dr Campbell noted. In his years of experience, bringing practices together to talk about what they’re seeing and learning, in near real-time, helped with the spread of evidence-based knowledge and quality improvement activities.
Similarly, Dr McGuire mentioned the reducing influence of general practitioners on policy making over the years. In contrast, the action-research approach of the project allowed more GPs to be involved, and from a personal and professional point of view, she found being “listened to, acknowledged and changing what was being done…was brilliant”.
There’s also interest in using these projects as an exemplar of an industry-led collaborative approach that will enable greater uptake of evidence-based knowledge to practice in health.
“It might help GPs to understand that there is data available, and if they cooperate a bit more, we could actually get more meaningful stuff done for general practice, without major heartache, apart from what they’re doing in normal day to day work,” Dr Campbell mused.
Dr Campbell added that if GPs want to be involved in trying to help with government policy and implementation of health initiatives, then there’s a real need for doctors to code better in the systems for data that will allow for an accurate picture and meaningful insights to be drawn.
Updates to the POLAR platform enabling more accurate COVID-19 and long COVID clinical classification in the PHN GP data
Changes were also rapidly incorporated into the POLAR platform to ensure PHN GP data about COVID-19 and long COVID could be more accurately coded, reflecting the evolving picture of long COVID in general practice.
Previously, any diagnosis entered into the relevant field with ‘COVID’ as part of the entry would be mapped and captured to the group ‘COVID-19’ (SNOMED Code: 840539006). This meant, for example, if a doctor entered free-text such as ‘anxiety post COVID’ or ‘loss of appetite post COVID’ or ‘post COVID cardiac’, these would all be grouped under COVID-19 with no differentiation to indicate these were issues of a longer-term experience after an acute COVID-19 infection.
The work conducted as part of this research project allowed the team at Outcome Health to map contents of diagnosis fields mentioning ‘COVID’ to a new subgroup called ‘Chronic post-COVID-19 syndrome’ (SNOMED Code: 1119304009) where relevant, to indicate likelihood of long COVID. Over 2,000 diagnoses were investigated, regrouped and then remapped to this subgroup as part of the collaboration with clinician experts, validated by the Outcome Health Clinical Review Expert Group.
These changes to the mapping algorithm have now been deployed across the POLAR platform involving almost 2,000 practices across the eastern seaboard of Australia, ensuring GPs have access to an updated and more accurate clinical classification when managing people with long COVID.
Insights and learnings to take forward
“I don’t want another COVID again,” Professor Georgiou said, but “how do you have that environment of collaboration without the giant crisis?”
Professor Georgiou highlighted the advantages of this collaboration in the connection and regular interaction between the different partners, with their perspectives, resources and skills. “I’ve never been able to speak to GPs and PHNs and IT companies all in one room every two months, you know, and I didn’t think it was possible actually.”
Dr Weeding gave credit to the power of the DHCRC as the connector between “all these different groups that probably may sometimes find each other in the wild, but sometimes they don’t. And so [for the Digital Health CRC] to have that coverage, I think was really important to bringing the groups together.”
