Combating COVID-19 with data and ML
Patients using COVID Care remotely
Patients, staff and visitors screened for COVID-19
Patient engagement success rate
For Austin Health, a tertiary and quaternary health service in Melbourne, Australia, COVID-19 ushered in a new set of challenges not previously encountered. They had to revisit their healthcare delivery models to be able to better care for patients, with or without the COVID-19 pandemic. Austin Health had originally engaged Arden Street Labs to design and develop an Azure-based Innovation Program focused on enabling data science research in digital health. When COVID-19 hit, they immediately refocused efforts on developing a series of novel COVID-19 digital solutions, some within days. By tapping on the potential of data with support from Azure Databricks, Austin Health developed data solutions that empowered patient recovery at home, managed hospital risk and enabled frontline health workers to focus on helping more severely ill patients.
Struggles with gathering structured data securely to enable patient communication
As the COVID-19 pandemic continues to rage on, Austin Health must mitigate the stress on their healthcare system to protect the safety of patients and staff while ensuring timely care is provided.
“The sudden change in demand for services, combined with the need to keep hospital staff, ‘non-COVID” patients and people potentially infected with COVID-19 physically separated, led to unique challenges. We required a rapid re-design of how we communicate with patients and how we gather information without bringing large numbers of people together at the hospital,” said Alan Pritchard, Director, EMR and ICT Services at Austin Health.
Care delivery, especially for non-critical conditions, was de-hospitalized which created physical silos that resulted in the challenge of sharing data in a structured and quantifiable manner to support the delivery of insights to patients and staff.
Austin Health needed a unified data analytics platform that could integrate with the hospital’s data environment to simplify data access, and aid in novel digital health solutions development, so it could focus on safe, quality patient care.
Unifying data and ML for enhanced patient communications
Leveraging Azure Databricks as a foundation for the AI Labs and Operational Pipelines, Arden Street Labs, in conjunction with Austin Health’s clinical and IT teams brought the Azure Innovation Program to life. It aims to facilitate the rapid deployment of models, pipelines and code to production for better patient communications and wellbeing.
Azure Databricks utilizes Delta Lake to provide a robust data solution for huge data merges, reducing complexity when integrating with the broader electronic medical record (EMR) system for data reporting, patient contact and data flow.
Within Azure Databricks, AI Labs and experiments provided a separate ML workspace that standardized data development for data scientists to retrain models and rapidly evaluate new ideas co-designed with the medical teams. Shared workspaces and collaborative notebooks further facilitated this, improving data team productivity and accelerating time-to-insight. Role-based access controls integrated with Azure Active Directory and network security controls ensured the safety and confidentiality of patient data.
By doing so, Austin Health was able to translate and deploy a proof of concept to a fully supported digital health operation that allowed it to quickly adapt amid a rapidly changing environment.
Jeff Feldman, CTO of Arden Street Labs, said, “The nearly dozen solutions we have developed are all built on Azure Databricks as a core foundation. This has allowed us to leverage a rapid lab to operations deployment pattern, whilst maintaining data security and computational scalability.”
With the infrastructure in place to support their digital efforts, frontline healthcare workers could then dedicate their attention on caring for patients, while keeping at-home patients safe.
75% success rate in patient engagement through digital solutions
With the support of Databricks, a robust data management infrastructure was laid for the Azure Innovation Program. A set of digital tools that leveraged in-built clinical rules and ML-based models digitized the COVID-19 response, reduced staff overhead and supported pre-appointment compliance.
“These tools have been fundamental to Austin Health’s ability to manage communications rapidly in a dynamic environment. We are able to give patients and visitors accurate and simple advice; provide information to clinicians about at-risk patients; and adjust the communication rapidly in response to new guidelines or changing situations in the hospital,” explained Pritchard.
The COVID Screen self-screening tool assesses the public’s COVID-19 risk and gives advice on the next best action. Over 30,000 people have accessed it, and 85% of them had their risk of infection minimized by being informed to stay home. The appointment portal has also seen over 75% patient engagement success rate, and has offloaded requirements to call numerous patients.
COVID Risk Management is a series of apps that manages COVID risk for patients and visitors coming to the hospital. Over 300,000 people have been screened using this tool.
Austin Health is also able to track the progress of patients in their homes and arrange for urgent medical care through remote tools such as the COVID Triage and COVID-19 Care Monitor, which has supported nearly 30,000 patients and at-risk staff remotely.
Beyond COVID-19, other departments are starting to use the Azure Databricks Data Intelligence Platform for labs, experimentation and research to support other Austin Health patients with respiratory conditions, post-procedure reviews and audits and patient lifecycle and journey insights to improve outcomes.
