Skip to Main Content
Skip to main content
Stories

Say ah and count to 30: Austin Health deploys AI to boost COVID-Care

Holding up the smartphone, the patient counts slowly to 30.  He’s been tested and advised to quarantine himself and recuperate at homeself-reporting symptoms every day, including performing an audio assessment. 

An algorithm assesses the patient’s symptoms including the recording.  Today the patient reports chest pain, an elevated temperature and his count is a little slower– he needs to take a couple more breaths than yesterday to complete the count to 30. A doctor is automatically alerted, reviews the case and follows up with the patient via a telehealth consultation to discuss options and provide advice. It may be that the patient needs to come to hospital or is safe to stay home as long as he continues to do the daily assessment. 

This is the new frontline of health care in Australia – where technology, data and artificial intelligence are combining to optimise patient outcomes by allowing recuperating patients to be monitored each day, supporting clinicians as they triage care remotely. At Austin Health it’s called COVID-Care. 

COVID, care and capacity 

Austin Health is a Melbourne-based tertiary and quaternary health service and the largest Victorian provider of training for specialist physicians and surgeons. In the last financial year, it completed 114,379 inpatient admissions, 89,675 emergency attendances, 28,902 surgical procedures and 279,156 specialist clinic appointmentsAustin Health’s researchers meanwhile are involved in 1,300 clinical trials. 

Preparing for COVID-19 Austin Health focussed on two critical issues – building capacity to cope with anticipated demand, and ensuring patients and staff remain physically separated where possible to reduce the risk of transmission. 

To ensure physical separation the hospital has used Microsoft Teams to allow people to work from home, and also to facilitate virtual multi-disciplinary meetings and ward rounds – optimising patient care without compromising community safety. 

In order to build capacity to meet increased demand, the hospital has also deployed a series of intelligent, innovative solutions that address a range of issues. Called COVID-Care these systems monitor the symptoms of recuperating patients, provide community self-assessment tools, and digitise the patient experience for appointment bookings and rescheduling. 

The smartphone system designed to monitor the symptoms of recuperating patients is just one example of the pioneering work being undertaken at the hospital 

That work began one Wednesday afternoon in early March – a week before the World Health Organisation declared COVID-19 a global pandemic. 

That was when Austin Health’s infectious disease team reached out to Alan Pritchard, the hospital’s Director of EMR and ICT Services to find out if there was a better way of assessing whether people might or might not have COVID-19. 

 Until then the hospital had been managing people who presented at the Emergency Department by having a staff member collect their details on a slip of paper, face to face, and then depending on their responses, either sending them home or for a medical check-up. 

Just digitising the process would reduce the amount of paper handling and associated risks – but what if people could start the process at home and self-screen?

Pritchard had already had discussions with intelligent software specialist Arden Street Labs about how the company could support its transformation plans in the future – COVID-19 meant those discussions needed to be accelerated. 

Two days after being briefed, Arden Street Labs had crafted a triage assessment screening tool to replace the paperbased system, and four days after that transformed it into a self-screening tool that could be used by the general public – one of the first self-assessment tools deployed in Australia. 

On the first day that the self-assessment tool was rolled out more than 25,000 people accessed it, with thousands completing the assessment. The system’s built in logic assessed the likelihood of them being at risk and told 85 per cent of them not to come into the hospital – dramatically reducing the impost on Austin Health’s Emergency Department.  For those that were advised to be tested, the tool used geospatial algorithms to determine the nearest testing clinic and if directed to Austin Health offered to pre-book an appointment. 

The selfassessment tool remains an important first threshold test for the hospital, but the usage rate and deflection rate have dropped significantly since it was first launched – partly because people are more aware of the symptoms of the disease, partly because there is increased testing available. In recent times daily usage has dropped to around 200 people – and only around 4 per cent of them are being told to stay home – the rest are invited to be tested. 

The assessment tool itself is continuously enhanced and the machine learning improved. Jeff Feldman, CTO of Arden Street Labs explains; “We’ve been working day to day with the infectious disease team and others to revise the screening criteria, update the tool and make sure that we’re providing the latest advice to the public. 

Besides being smart, the system has also had security as a priority; essential for a public facing website with a hospital on the backend. Having Microsoft Azure as the digital foundation has elevated the security of the solution, according to Pritchard, injecting confidence about its widespread deployment. 

Streamlined service and symptom surveillance 

Getting the self-assessment tool up and running dramatically improved things in the hospital’s Emergency Department. But Austin Health was also looking for a way to streamline appointments and rescheduling – and not just for COVID-19 patients. 

Tackling more than 110,000 patient admissions a year, and supporting almost 280,000 specialist clinic appointments, Austin Health needed to liaise efficiently with people attending the hospital for other reasons. Given the COVID-19 situation many non-urgent appointments or elective surgeries needed to be rescheduled. 

Pritchard explainsWe had limited tools available to communicate that to patients.” The hospital could phone them, send a letter, or send a text – but there was limited opportunity for efficient, twoay communication with patients to agree changes, confirm or reschedule bookings. 

Working with Arden Street Labs, Austin Health developed a system that pushed an invitation to patients via SMS, asking them to log into a secure portal where they could agree to an appointment time, ask for a telehealth consultation or reschedule. 

The portal also guides patients attending the hospital through some pre-screening questions to assess their COVID risk. The data collected is then provided through a system developed by Austin Health Application Services Group using Power Apps, to doctors and nurses ahead of any scheduled consultation so that they are fully informed about the patient they are about to see.

We’ve seen about an 8587 per cent success rate in terms of getting patients to use this digital capability for rescheduling or confirming their appointments, which in turn then has offloaded the requirement to call hundreds or eventually thousands of patients, says Pritchard. 

Not only does the process gather more data for clinical staff ahead of a consultation, it reduces the load on Austin Health’s call centre and reduces the risk of no-shows for scheduled appointments. 

Dr Jason Kwong, Austin Health’s Medical Director of Infection Control, says technology allowing patients to be monitored and assessed remotely offers another layer of care to the community.  

Controlling the number of people coming to hospital is a crucial step in the battle against COVID-19 to ensure we can prioritise care to those who are most in need, but there is huge scope to use this technology in a post-COVID world.  

“The care of our patients and staff is always our number one priority. 

“This technology has the potential to help us closely track the progress of patients from afar without breaking quarantine or isolation, keeping patients connected to arrange urgent medical care if required. It’s also incredibly exciting to think about some of the future benefits we may be able to offer to our patients.”

Austin Health is also using the same approach to achieve secure COVID-19 results notification – and for people testing positive or negative, monitoring symptoms daily. This is where the counting to 30 comes in. 

Jeff Feldman explains: “First off, we get the list of swab results on a daily basis. We send out a link to a secure portal for the patient to log into. They log in. They can view their results. They can view a number of different supporting materials depending upon if they’re positive or negative. Then every day following that initial notification, we send them a reminder to log into the portal for 14 days to record their current symptoms. Are they experiencing any shortness of breath? Do they have a temperature? Are they nauseous? There’s a handful of symptoms that we help walk them through. 

All observations are reviewed by a clinician, but if any responses are of concern Austin Health’s medical team is alerted, the patient is phoned for a quick telehealth consult and advised what to do next. Information is also made available to the Department of Health where required, assuring proper compliance and reporting.

Working together, Austin Health and Arden Street Labs are constantly looking to expand the remote patient monitoring capabilities and insights, to further help prioritise followups and care.  For example, where patients have reported shortness of breath, they are asked to count to 30 and the recording is captured and analysed using Azure Cognitive Services. 

If the system identifies a potential problem then clinical personnel are alerted and able to provide further advice or support to the patient. While it’s been developed to support COVID-19 patients, the platform has enduring potential to support other Austin Health patients with respiratory conditions. 

In terms of the underlying technology the solution is a web-based form posted in Web Apps Container in Azure. SQL servers, DataBricks, Azure Machine Learning and Azure Maps operate in the background, with maps used to direct the patient to the most appropriate screening clinic. 

Reports to doctors and nurses are generated using Power BI and there are interactive interfaces allowing secure access to the data by authorised specialists which have been built using PowerApps. 

Security remains a paramount concern to Austin Health says Pritchardbeing able to have access to its own Azure tenant ensures the security of the new applications which are being rolled out to help address the challenges of COVID-19. 

Feldman adds; “Whilst we are using Azure Databricks and Azure ML to support the key analytics capabilities built into the solutions, we’ve taken a very risk-averse approach putting client confidentiality at the forefront.  Therefore, everything employed from Azure has been built leveraging Microsoft’s security capabilities, including highly restricted access and compartmentalised approach to data.” 

Two additional intelligent solutions are currently in final testing; an in-clinic symptom assessment, helping to standardise risk of COVID for in-hospital patients to help alleviate the strain on the infectious disease team; and an all of hospital visitor registration capability, helping to digitally capture who is at the hospital, their COVID risk and supported by QR code functionality and PowerApps for check-in services.. 

While the technology smarts are clearly fundamental to the success of the solution, it’s the one-team approach with Austin Health’s clinical personnel that has been crucial. 

Using an agile approach Arden Street Labs and Austin’s own IT staff have discussed with doctors and nurses what the systems should look like and how they should perform; they’ve then coded a prototype – often overnight – that can then be refined after further checks with the clinicians. 

According to Pritchard that’s critical to ensure that whatever is being developed has real purpose and impact, and is not developed just because it is possible. “Recording the audio and analysing it using some sort of artificial intelligence, was actually a conversation Jeff and I had the very first day we spoke, at the beginning of March, about wouldn’t it be great if? But it doesn’t matter how great a couple of blokes in IT think it is, if nobody else cares. 

With COVID-19 on parade – it turned out the clinicians cared very much indeed. 

We’re focused on supporting our customers during these difficult times. For more information on how we can help your business, please see our dedicated webpage and report on Supporting Resilient Operations. We’ll be updating these resources regularly as the situation evolves.