South Australian Health Pioneers State-Wide Synthetic Data Initiative for Safe EHR Data Sharing
[ad_1]
Open to anyone with an idea
Microsoft for Startups Founders Hub brings people, knowledge and benefits together to help founders at every stage solve startup challenges. Sign up in minutes with no funding required.
In the era of AI, data has become the new currency on which transformative models are trained. At the same time, the need for privacy and regulatory oversight is critical. This is especially true in healthcare. Hospitals and medical researchers want to use patient information to learn more about diseases and improve treatments, but they also must ensure they’re keeping that information safe and private.
South Australian Health (SA Health) is trying something new to deal with this challenge. Through the Synthetic Data Initiative, SA Health is working with Gretel AI to tackle this challenge. Gretel AI helps to create synthetic data that looks and acts like real patient information, but without revealing anyone’s personal information.
SA Health is exploring innovative methods to navigate the complexities of accessing and sharing Electronic Health Records (EHR) across healthcare entities and systems for research purposes. By using this Gretel AI, SA Health is hoping to find a better way to use patient data for research without risking people’s privacy. It’s a new approach that could make a big difference in how healthcare organizations handle sensitive data.
Read more below.
Problem
South Australia Health (SA Health), the collective body of agencies entrusted with the health and wellbeing of Southern Australia have been challenged for years, like other healthcare industry leaders, to leverage patient data to improve clinical practices, inform data-driven decisions, and power research and discovery—all while safeguarding patient privacy. Managing the health and wellness of a population exceeding 1.8 million individuals across a vast geographic expanse of 400,000 square miles and encompassing over 200 diverse ethnic communities, the group shoulders a substantial mandate. Given these ambitious objectives, ensuring secure access to Electronic Health Records (EHR) across diverse teams and specialties becomes mission critical.
“At South Australia Health, like many in the healthcare sector, we faced the complex challenge of leveraging patient data for advancing medical research and improving our practice while rigorously protecting patient privacy and meeting compliance requirements”- SA Health chief clinical information officer Rhys Parker
Traditionally, obtaining answers to fundamental questions about patient care used to be a time-consuming process, often spanning months or even years for physicians. The delay stemmed from the intricate procedures accompanying each data request, which entail multiple ethics committee reviews due to the heightened sensitivity of EHR records. The prolonged nature of this process has significantly hindered hospitals, public health research entities, and clinical practitioners in leveraging data to enhance patient treatment and care. Addressing this challenge is crucial, necessitating the development of an efficient method that ethically harnesses patient data, ensuring privacy protection while facilitating research and advancements in clinical practice.
Solution
SA Health in Australia is leading the way by creating synthetic data from real EHR data to advance clinical research. This synthetic data, produced by generative AI models, replicates the patterns and insights of real-world data, yet ensures patient confidentiality since no dataset record is tied to any individual patient. Utilizing privacy-enhancing technologies like differential privacy, provides mathematical guarantees of privacy, simplifying compliance processes. Furthermore, although the use of synthetic data generated by generative AI models is a novel concept in Health and Life Sciences (HCLS) research, it is already yielding promising outcomes in SA Health’s testing—including demonstrating the ability to recreate the results of pancreatitis research studies using purely synthetic datasets. SA Health’s collaboration with the Australian Institute of Machine Learning, under the guidance of Chief Medical Information Officer Dr. Santosh Verghese and Chief Clinical Information Officer Rhys Parker, underscores their commitment to both innovation and privacy.
Image 1: PCA distributions of real-world EHR data (left) and Gretel-generated synthetic (right) datasets, illustrating similar data patterns as part of Gretel’s synthetic data quality score report.
SA Health has integrated Gretel’s synthetic data platform and APIs into its real-time data analytics platform to enable the creation of artificial health records that maintain the statistical value and referential integrity of the original data without exposing sensitive patient information. SA Health leverages Gretel’s privacy enhancing technologies to ensure that a synthesized record can never have a direct correlation with a real patient. The solution is built in a secure hybrid-deployment of the Gretel platform in SA Health’s private Azure Blob storage and machine learning compute infrastructure, providing auto-scaling and robust access controls for managed data access. The stack sits aside other generative AI Azure tooling, including the Azure Databricks platform and Azure OpenAI.
“By pushing patient files though Gretel’s synthetic data platform patient files are untraceable while the information created is invaluable.” – Chief Medical Information Officer Dr. Santosh Verghese
SA Health is building a future in which it provides state-wide managed access to privacy-protected medical datasets through the establishment of trusted research environments across its services. Leveraging the security benefits of Azure’s secure collaboration zones as well the added privacy protection of Gretel-generated synthetic data, access to EHR dataset will become readily available to specialists, accelerating innovation in medicine at a previously unimaginable pace.
Image 2. SA Health ecosystem-wide architecture showing the full journey from data sources to data consumers, utilizing Azure as the cloud provider and incorporating Gretel as one of the key Data Access and Privacy (DAP) services.
One such example lies in rapid innovation for treating Severe Acute Pancreatitis (SAP), a disease that has overburdened hospitals in most western countries, including Southern Australia, and has a high mortality rate with a 33% mortality rate in the first year of diagnosis. Democratizing access to this data across researchers will greatly accelerate critical research into early detection as well as positively impact how hospitals proactively plan and ensure adequate staffing to handle high patient influx.
Image 3. This figure shows the Synthetic Data Quality Score (SQS) of the synthetic data generated at a high Privacy Protection level. The SQS score is a weighted combination of individual scores that compares the statistical integrity of the synthetic data with the real-world data.
The analytics platform also provides transparency to all data consumers into any variance between the original and synthetic records, allowing users to make informed choices about appropriate data use. The Gretel and SA Health synthetic data solution promises to provide all parties across its vast networks of 27,000 practitioners and researchers with real-time access to the data they need without lengthy delays.
What’s next
The Gretel SA health partnership is off to a promising start. SA Health ethic committees have deemed that since the Gretel solution provides advanced privacy protection with mathematical guarantees, it removes the need for multiple privacy and compliance reviews. The same data requests that took months and years will be achievable in a matter of days through the SA Health analytics platform.
With this foundation in place and an ecosystem-wide deployment of Gretel, SA Health is poised to improve patient care and research through safe access to EHR records to the teams that need it most, including practitioners on the frontlines. Data consumers have also shown early excitement, including Dr. Verghese, an intensive-care specialist in the ICU at Flinders Medical Centre, who predicts that the use of synthetic data will become routine in the medical field in the next few years.
“As part of SA Health’s commitment to public health innovation, we’ve embraced synthetic data as a forward-thinking, privacy-conscious approach to safe EMR data sharing for everyday clinical decision making as well as training machine learning models. Our collaboration with Gretel has been vital in this effort.
With our ecosystem-wide Gretel integration in our secure Azure environment, this advancement in our work is more than a technological leap; it symbolizes a fundamental shift in how we manage and use health data, significantly improving care for our patients with a focus on inclusivity and privacy protection.”— Rhys Parker, Chief Clinical Information Officer, SA Health
Resources:
- SA Health. “About SA Health.” URL: https://www.sahealth.sa.gov.au/wps/wcm/connect/public+content/sa+health+internet/about+us/about+sa+health/about+sa+health#:~:text=SA%20Health%20is%20the%20brand,SA%20Ambulance%20Service.
- The Australian. “Fake patients created from real data helping fast-track research while safeguarding privacy.” URL: https://www.theaustralian.com.au/business/technology/fake-patients-created-from-real-data-helping-fast-track-research-while-safeguarding-privacy/news-story/551f01321807b67483a3c66e8cca6645
- South Australia Government. “Living in South Australia.” URL: https://www.sa.gov.au/topics/about-sa/living-in-sa#:~:text=South%20Australia%20is%20the%20southern,areas%20of%20France%20and%20Germany.
- Australian Nursing and Midwifery Journal. “Mortality and cost of severe acute pancreatitis keeps pressure on hospitals, study finds.” URL: https://anmj.org.au/mortality-and-cost-of-severe-acute-pancreatitis-keeps-pressure-on-hospitals-study-finds/
- ScienceDirect. “Article Title.” URL: https://www.sciencedirect.com/science/article/abs/pii/S142439032300100X
[ad_2]