Workforce management and the rising cost of care are just two of the pressing issues healthcare needs to address today. Health systems will need to adopt big data approaches to move toward new patient-centric and technology-driven models of care in an effort to resolve such issues, said Dr Chris Bates, Director of Research and Analytics at UK-based clinical software company TPP.

In a TPP-sponsored presentation at HIMSS22 APAC, Dr Bates shared some of the latest applications of machine learning and data analytics in driving better care delivery and outcomes during the pandemic.

Using big data

On the clinical side, TPP was able to utilise millions of patient EMRs to develop algorithms that can predict if a patient has an ovarian cancer tumour and another that provides a streamlined view of patients who are most at risk of developing cancer. 

Meanwhile, on the operational side, TPP helped optimise nurses' work in rural communities by developing ML algorithms. The technology was able to increase the contact time between nurses and patients by 40% and raised both patient and nurse satisfaction.

During the pandemic, TPP was tasked by the UK government to work on an analytics platform for COVID-19 research. It delivered the open-source OpenSAFELY platform in six weeks, which was then used by epidemiologists and public health analysts to run analytics on "tens of millions of records," including primary and secondary care data, immunisation records, and testing data. Dr Bates said the researchers were able to do this safely at home because the platform was built with tight security layers. 

"What we built there was a secure, trusted research environment that works on massive data and scale. It gives you all the power you need but doesn't compromise that privacy balance," he said. 

In developing the algorithms, Dr Bates emphasised the importance of closely engaging clinicians in the process.

"We worked very closely with clinicians – and I think that's important. We weren't trying to work against clinicians. We were working with them in collaboration to try and better reflect clinical practice and to use all those data items to try and reflect how clinicians actually work using multiple sources of information to make that decision," he said.

Moreover, while patients are willing to let the health system use their data to improve healthcare, they also wanted their data to be handled "exceptionally responsibly," Dr Bates stressed. "We need to anonymise and aggregate data. We need to have levels of security in place."

Novel data sources

From all that work, TPP was able to identify novel data sources which could potentially "impact the future of analytics." These are textual information, citizen data, and genomics information. 

According to Dr Bates, the same AI algorithms they used in detecting ovarian cancer were also applied to support the government's pandemic response.  For example, textual symptom data were run through these algorithms to spot exactly where outbreaks are happening. The AI was also used on text-based data to detect patterns of emerging COVID-19 symptoms.

Measuring data capabilities

Thinking about the future of analytics, Dr Bates implored the audience to "keep on building" capabilities around analytics and machine learning. They should also adopt metrics to measure their improvement in bringing analytics to both clinicians and patients. "The HIMSS [Adoption Model for Analytics Maturity] here, I think, is absolutely fantastic and takes organisations through a great journey from basic analytics up to predictive and prescriptive analytics," he suggested.

"If this [analytics] isn't empowering the workforce if it's not delivering better, safer care, then all of this work is for absolutely nothing," he cautioned.