If you have registered for an event, you may log in to manage your registration.
Researchers at BioGrid Australia member and/or collaborator institutions may submit new or amended applications requesting access to data connected to BioGrid using our online application system.
Authorised researchers can explore and visualise data on specified connected data utilising a web interface.
Authorised researchers can utilise this comprehensive data and statistical analysis application to analyse, explore and visualise connected datasets for approved audit and research projects.
Using novel data science to monitor infections in cancer patients
Professor Karin Thursky, the implementation lead for the National Centre for Infections in Cancer, and Associate Director of Health Services Research and Implementation Science
Peter MacCallum Cancer Centre
In the fight against cancer, there’s a lot to be concerned about for clinicians treating vulnerable patients. Infections are a serious and common complication in these patients. Bacterial infections which occur when white cell counts are low are very common, and easy to diagnose and treat. However, there are other infections such as fungal infections , which are difficult to treat and therefore can have serious implications for a patient’s treatment and survival.
Invasive fungal infections in cancer patients are relatively rare, affecting 2-10% of patients, but the impact is significant especially for those being treated for leukemia and requiring bone marrow transplants, or patients requiring stem cell transplants. These procedures lead to the patient’s immune system to be profoundly suppressed, thereby making them at extreme risk of rare and serious infections.
A fungal infection in one of these patients has a very high mortality rate – upwards of 40%. They’re expensive to treat, and if the patient survives the infection there will still be a significant impact on their future outcomes.
To make matters even more complex, there are only a handful of pharmaceutical companies producing antifungal drugs, and the development pipeline is almost non-existent. During COVID-19, the global supplies of these drugs were quickly consumed when mucormycosis (‘black fungus’) broke out in COVID patients, placing stress on the availability of treatment to cancer patients.
Prevention is the key to controlling fungal infections in clinical settings
Given the challenges in treating fungal infections, a critical part of their control is prevention. The Invasive Fungal Infections Surveillance Project aims to arm clinicians with the data they need to recognise the signs of a potential problem.
Professor Thursky is leading an NHMRC $1.5 million project grant : Meeting the challenges of invasive fungal infection: antifungal stewardship and effective surveillance in high risk patient groups. This is a digital health program of research which has established a dedicated auditing program for invasive fungal infections and antimicrobial prescribing called the Antifungal NAPS.
The project is also utilising natural language processing and machine learning to establish an automated surveillance program for invasive fungal infections in high risk patients The system requires access to a wide variety of clinical data from the electronic medical record, microbiology, histology, prescribing data, Computerized Tomography (CT) and Positron Emission Tomography (PET) scan data.
The team are using implementation science to ensure that the system is useful for the clinicians looking after these high risk patients. A major challenge however, lies in pulling all that data together and using it to assess whether an infection is likely, probable, possible, or not likely, which will require training of the data by infection experts.
“We’re building assistive technologies to harness information from multiple sources and pull it together to provide medical information back to the clinicians that are managing programs around prevention of these infections,” says Professor Thursky.
This project utilises the knowledge of infectious diseases experts in infections in cancer patients to train and test machine learning-based interpretable NLP algorithms and machine learning models. The approach had already been tested on CT scan reports from three hospitals with different reporting styles, and was found to work as hoped.
The resulting algorithms will be used to provide indications of possible missed infections to clinicians in their infection surveillance, allowing institutions to monitor and transform their hospital building practices and policies. This will be an assistive technology to support surveillance, rather than a predictive tool – an important distinction.
Most hospitals do not have the resources to undertake this kind of surveillance, which has only ever been done in a research setting and so will be a major contribution to the digital health space.
In fact, there isn’t another platform in the world that is carrying out this sort of work.
How does BioGrid do it?
BioGrid is enabling the data connectivity that empowers IFI surveillance
BioGrid has worked closely with the Victorian Comprehensive Cancer Centre Alliance over the past four years and pioneered patient record linkage of hospital, clinical registry and general practice data, so their experience and unique platform with aggregated, de-identified data was key to demonstrating the project’s feasibility.
Once the data from the BioGrid platform is collated, it will be used in a clinical application which will be built out with dashboards and reporting functionality. The main clinical teams that will be interacting with the system will be the infectious diseases experts who manage these antifungal stewardship programs, looking at elements such as breakthrough rates, and which new drugs seem to be associated with increased (or decreased) risk of infections.
Leveraging BioGrid’s unique and proven data governance framework, linkage to other clinical and administrative datasets will be possible when the project data has been linked to the BioGrid platform. BioGrid’s established working relationship with the Australian Institute of Health and Welfare (AHIW), will assist in streamlining linkage with AIHW to federal datasets such as National Death Index, Medicare Benefits Scheme and Pharmaceutical Benefits Scheme.
“The development of data science machine learning is a continuous improvement; it’s not as simple as building an algorithm and pushing it out to the application, and sitting back and letting it run,” says Professor Thursky. “In the future our data scientists will need to work within the BioGrid platform to harness data as more data becomes available and refine and improve the algorithms."
Because BioGrid has a big footprint in the cancer space, there are already BioGrid servers in place in many of the hospitals that are looking after these high risk patients. Therefore there’s a low barrier to entry for new sites coming on board because of BioGrid’s existing presence.
+61 3 9342 2690
For technical difficulties using SAS to access BioGrid Australia data, please contact our Help Desk during office hours.
© 2023 BioGrid Australia