Will machine learning replace medical education in preventing infectious disease pandemics? « Back to Blogs
Machine learning and artificial intelligence have enormous potential in healthcare.
Machine learning algorithms can already read radiological images, make diagnoses, and advise on treatment options. Just a few years ago these were all tasks that required a trained and qualified healthcare professional to perform. In countries that are short of doctors, it is tempting to believe that machine learning and artificial intelligence could come to the rescue and build capacity – but without people.
It is the poorest countries in the world that are short of doctors and it is many of these same countries that are most at risk of infectious disease pandemics. When pandemics do happen in these countries, healthcare professionals are in the frontline and so are often affected themselves. This can have long-term effects on the strength of health systems – which are left even more short of doctors than they were at the beginning of the outbreak. These are all good reasons to think about how machine learning algorithms could play a role in managing infectious disease pandemics. (1)
Healthcare is not simply a purely scientific discipline – it requires empathy and communication skills.
But there are a myriad of healthcare, technical and ethical issues that would need to be overcome before this became a reality. First of all, healthcare is not simply a purely scientific discipline – it requires empathy, and communication skills to convey that empathy. This is true in all disciplines – including that of pandemic infectious diseases. At present machine learning algorithms simply cannot do this. Secondly machine learning algorithms require reliable data and good technology to analyse that data. But poor countries often lack data and frequently don’t have good technology. Once again it is those same poor countries where infectious disease pandemics are most likely to happen. Thirdly the prevention and management of pandemic infectious diseases requires ethical judgement as well as scientific evidence-based algorithms. Can we trust artificial intelligence to always make the right decisions? Let’s look at the following scenario. Imagine that there is an Ebola outbreak in an African country in 2030. There is a shortage of healthcare professionals in the country, so a machine learning algorithm makes some of the decisions about healthcare. There is also a shortage of resources – so the machine learning algorithm makes decisions about who should be prioritised for treatment. The algorithm “learns” as it goes, and it quickly learns that poor people have a poor prognosis and are least likely to respond to treatment. It then makes a decision based on cost effectiveness and health economic outcomes to deny treatment to poorer patients. This would likely be unethical – but there is a chance that it could happen.
At present when we say clinical decision support, we mean knowledge-based resources that are evidence based, continually updated, and practical and accessible to healthcare professionals on all devices (and especially on mobile devices in resource poor countries). (2 3) Our own resource – BMJ Best Practice – is a good example. But the next step might be machine learning. It could support clinical decisions – but it is unlikely to be ready to replace the clinical decision maker for the moment.
- Meskó B, Hetényi G, Győrffy Z. Will artificial intelligence solve the human resource crisis in healthcare? BMC Health Serv Res. 2018; 18: 545.
- Brunner J, Chuang E, Goldzweig C, Cain CL, Sugar C, Yano EM. User-centered design to improve clinical decision support in primary care. Int J Med Inform. 2017 Aug;104:56-64.
- Walsh K. Mobile Learning in Medical Education. Ethiopian journal of health sciences. 2015;25(4):363-6.
Kieran Walsh works for BMJ – which produces clinical decision support resources in infectious diseases.
The purpose of this document is to educate and to inform. The content of this document does not constitute medical advice and it is not intended to function as a substitute for a healthcare practitioner’s judgement, patient care or treatment. The views expressed by contributors are those of the authors. BMJ does not endorse any views or recommendations expressed in this document. Readers should also be aware that professionals in the field may have different opinions. Users of this document hereby agree not to use its content as the basis for their own medical treatment or for the medical treatment of others.