Machine learning can fix how we manage health on a global scale
Democratising patient power will improve people’s ability to manage their own health, regardless of income
Harnessing machine learning to improve health is a major ambition for both medical practitioners and the healthcare industry. If the two can join forces on a global scale in 2019, with the right investment and the right approach, AI could propel a revolution to democratise global health and to leapfrog access to health services in low- and middle-income countries.
A chronic shortage of human resources is one of the major obstacles to better health and healthcare in many resource-poor settings. When it comes to global health, artificial intelligence offers huge opportunities to fill the gap left by critical healthcare worker shortages, particularly if combined with mobile phone technology.
For example, social enterprises such as Peek Vision can use smart-phone based technology to enable healthcare providers to deliver cost-effective and targeted treatment to people with eyesight problems. In addition, through developments in personalised care, wearable devices and image recognition for diagnostics, technology is opening up new opportunities for patients to take better care of themselves. With the rise of chronic conditions in nearly all countries, a growing elderly population and stretched health services, empowering patients to help prevent disease makes a lot of sense.
While there is a wealth of innovation going on in health in the context of AI, there is not enough evaluation and validation as to whether these new technologies actually improve health at all. Such evaluations should be a priority for health research funders.
To make sure new technology and digital health solutions can be accessed by all – not just the wealthy – and address both infectious and non-infectious disease, it is critical that there is a co-ordinated approach to their evaluation and delivery. This is no different to, say, the rolling out of vaccines and new drugs. However, the rules are not fully defined for machine learning and artificial intelligence in healthcare, which risks undermining trust and success. We therefore need to make sure that we are delivering innovation for social impact, but also that we are innovating the ways in which we deliver innovation, with the appropriate governance and regulatory frameworks to ensure equitable access and impact.
As scientists we are reliant on data and the same principles of robust data collection and analysis should apply to machine learning solutions, ensuring that these data are used ethically and that patient privacy is protected. Health apps and medical digital innovation also needs clinical validation and rigorous grass roots development and distribution. This is where the source is also so crucial – as Melinda Gates and others have warned, bias embedded in the model can generate new data that reinforce bias. So we need to make sure that data are collected in a way that reflects our diverse societies by, for example, ensuring that there is gender and ethnic balance among programmers and that local context is built into new technology and innovation.
Looking to the future, we must not repeat mistakes of the past, as happened in Africa in the 1990s, when large-scale interventions for HIV/AIDS relied on western understandings of the epidemic, and subsequently failed. Low and middle-income country partners must lead the way in shaping the technological innovations which could make the greatest difference to health among their own populations.
Home-grown solutions like the mPedigree Network, founded in Ghana by Ghanaian Bright Simons, is an innovative end-to-end mobile tracking system that empowers patients and health professionals to identify counterfeit medicines by strengthening the authenticity of supply chains of medical products, including vaccines. With Google opening an AI lab in Ghana and dedicated education programmes launching in Africa to harness local skills, hopes are rightly high that this trend will continue apace in 2019. And crucially, rigorous evaluation and scrutiny must be applied to AI solutions to ensure the quality and safety standards are in place and that the shift in personalising healthcare for all is an unquestioned force for good.
With that in mind, public-private partnerships will be strengthened in 2019 to create a suite of robust, effective and equitable digital solutions. These will harness the power of AI to democratise patient power and people’s ability to manage their own health, including in low and middle-income countries.
By PETER PIOT, Peter is director of the London School of Hygiene & Tropical Medicine and a microbiologist who helped discover the Ebola virus
Article Source: https://www.wired.co.uk/article/machine-learning-healthcare-ebola