Solving healthcare challenges through interoperability and artificial intelligence

In December 2017, Google announced that its Launchpad Studio start-up accelerator has commenced working with companies interested in utilising Google’s machine learning and artificial intelligence infrastructure to address healthcare issues1.

Google’s announcement makes real what many in the industry have known for some time: Artificial intelligence in health care is not science-fiction. It’s happening now.

Machine learning has made significant strides in recent years, moving from naïve Bayes algorithms into complex modelling tools that can be used across a range of industries. While progress has been made in assisting the work of pathologists and radiologists2, machine learning is coming closer to the clinical bed-side with a recent multicentre trial demonstrating the effectiveness of machine learning in assisting the detection of deteriorating patients3.

Within Australia, we have seen groundbreaking research output in this domain from the Australian Institute of Health Innovation, under the leadership of Professor Enrico Coiera, including the development of models that use the lived experience of previous patients to help plan care for complex patients, where the traditional experimental and observational evidence for intervention is limited4. Not only can this improve health outcomes for patients, it can also improve health service delivery and the experience of people interacting with the health sector.

Within Australia, there remains a persistent inequality in health outcomes with mortality in areas of lowest socioeconomic status 30 per cent higher than in the highest, and with Indigenous mortality nearly double that of non-Indigenous Australians5. The challenge for the future is to ensure that new, digital, approaches to health care are actively designed to reduce these inequalities. The development of machine learning systems within Queensland’s Inala Aboriginal and Torres Strait Islander Health Care Centre to detect retinopathy6 and overseas research into the possible early detection of suicide ideation through social media content7 provide examples of such approaches.

However, all of these and future advances need data. Machine learning systems need data to learn from, and for a consumer to benefit from these approaches, their own information must be available in a similar form. In Australia, we have a long road to walk to bring our health information into these computable forms. While we have seen improved use of computerised record systems across much of the health care sector, much of our information remains locked in documents, or lacks the consistent use of terminology needed to identify data.

Moving towards an interoperable, atomic data approach to recording care will be a long and sometimes painful process, but it’s one that will bear significant fruit for generations of health care consumers to come.


1 Ricci, M. Google’s new accelerator aims to foster AI startups. pharmaphorum, 2017.

2 Jha, S. and E.J. Topol, Adapting to artificial intelligence: radiologists and pathologists as information specialists. JAMA, 2016. 316(22): p. 2353-2354.

3 Churpek, M.M., et al., Multicenter comparison of machine learning methods and conventional regression for predicting clinical deterioration on the wards. Critical care medicine, 2016. 44(2): p. 368.

4 Gallego, B., et al., Bringing cohort studies to the bedside: framework for a ‘green button’to support clinical decision-making. Journal of comparative effectiveness research, 2015. 4(3): p. 191-197.

5 Australian Institute of Health and Welfare, Mortality inequalities in Australia, A.I.o.H.a. Welfare, Editor. 2014, Australian Institute of Health and Welfare: Canberra.

6 Pires, R., et al., Automated multi-lesion detection for referable diabetic retinopathy in indigenous health care. PloS one, 2015. 10(6): p. e0127664.

7 Larsen, M.E., et al. The use of technology in suicide prevention. in Engineering in Medicine and Biology Society (EMBC), 2015 37th Annual International Conference of the IEEE. 2015. IEEE.