What if we could harness data captured every day in the provision of care and provide valuable insights that support clinical decisions, for example, a suggested treatment and care plan tailored to optimise outcomes for a specific patient?
Machine Learning (ML) is used for processing large and complex data to build useful predictive models that have the potential to revolutionise healthcare through personalised and precision medicine, augmenting decision making and removing cognitive load for clinicians.
Working on a range of clinician led projects with partners from across the health system, we are creating and/or testing ML models to improve patient care with the aim of reduce ramping in our hospital’s Emergency Departments.
- Estimating a patient’s Length of Stay based on a clinical speciality, reducing administrative burden and supporting reduction in length of stay.
- Providing diagnosis tools that use standard tests and observations to help augment decision making in general medicine wards with a view to moving patients to appropriate care settings faster.
- Identifying patients who are at risk of a new admission within a month of discharge from hospital to provide opportunities for interventions.
- Supporting efficiencies in the generation of discharge summaries, to improve continuity of care and handover to General Practice (reducing admissions).
However, there are a number of challenges that need to be addressed before ML can be widely implemented in hospital systems.
One of the biggest challenges is the lack of standardised data. Health data is complex and variable, and it can be difficult to combine data from different sources and from different health systems in a way that is compatible with ML algorithms.
Important diagnostic information, such as notes describing the treatment of a patient are stored in unstructured text documents which makes accessing the information even more challenging.
It can be difficult to answer or even pose new questions if it requires the creation of a new bespoke dataset.
To address these challenges, we are working to combine data from core systems, such as surgical and ambulance data, to create well documented, curated and reusable data sets that are organised, cleansed and validated in collaboration with SA Health’s clinicians and operational staff. These data sets, once anonymised, can be used to create ML models more quickly and easily that will help facilitate a data driven approach to improving health care across the state.
The systematic adoption of ML in healthcare is still in its early stages, but the potential benefits are enormous. By addressing the challenges of applying ML in healthcare, we can improve the quality, efficiency, and affordability of healthcare for everyone.
To know more, reach out to our Clinical Informatics team at firstname.lastname@example.org.