AI will improve healthcare and cut costs - if we get these 4 things right

The potential for data-driven medicine and predictive analytics is clear. AI and machine learning are assisting healthcare providers with more informed clinical decision-making and enabling patients to take an active role in their own health, with access to real-time data via wearables and mobile apps.

However, as we’ve seen with many previous social, economic and business challenges that hailed big data as the solution, the key to success will be creating enduring systems for collecting, cleaning and analyzing this wealth of information in thoughtful, efficient, secure and actionable ways. What’s more, systems will need to seamlessly interconnect from doctor to patient to hospital, pharmacy and lab, allowing for information to flow where and when needed. To do this, we must address several key issues:

The proliferation of medical data, from electronic health records to real-time data from connected medical devices, is the foundation for this new era of personalized, precision medicine. The challenge is to create a system of data collection without siloes that is capable of scaling to incorporate new data sources as emerging technologies like nanobots and connected environments – e.g. smart hospital bed, connected cast etc. – begin to contribute valuable data as well.

One example of this is athenahealth, which serves more than 100,000 medical providers with cloud-based services to track performance data, process claims, manage patient records and more. According to the company’s network data, the average doctor spends 40% of their time processing thousands of administrative documents and forms, and chasing down hundreds of missing lab and imaging orders. Athenahealth knew they needed to unlock the data trapped by its existing infrastructure, some of it still using paper-based archives and other content in inflexible databases. A completely new systems architecture was needed, so they built cloud-based services aimed at reducing the administrative burden placed on doctors. This enables AI and analytics tools to capture new insights, process real-time data, and deliver personalized services. This is leading to better doctor-patient experiences through automation and easing operations.

It’s becoming easier to imagine tech-enabled personalized medicine in wealthy countries and hubs of technological innovation, but how will we ensure that those most in need have access to life-saving medical data? One solution will be to tap mobile devices – recent research shows that the developing world has hit over 98.7% mobile phone adoption; creating mobile solutions that allow both doctors and patients to access data will be crucial in bringing this medical revolution around the world. Another solution will be leveraging edge computing, which enables devices to operate without disruption even when they’re offline or internet connectivity is intermittent, allowing for data collection and analysis in remote locations like in a refugee camp or rural village.

There has been a big emphasis recently on the pitfalls and potential for bias in artificial intelligence (AI); when building algorithms and models for predictive medicine, ensuring that they are free of bias could literally be a life or death matter. Currently, healthcare inequalities are systemic and closely intertwined with social inequalities – if developed properly, these AI-infused systems could go a long way towards mitigating these inequalities by removing human bias. However, there is also the risk that these technological advancements could perpetuate existing inequalities. Assessing the data used to train these models and ensuring just representation of different groups will be of the utmost importance for this next phase of medical advancement.

Beyond that, cheap genome sequencing, big data analytics, health sensors, wearables and artificial intelligence will enable medicine to move away from general solutions to personalization and precision. All these technologies are important pieces, but AI holds the key. That’s due to the vast amounts of information that will need to be stored, parsed, put into context, correlated with existing research and procedures and analyzed. Only AI, likely through machine learning and deep learning algorithms, will be able to produce personalized solutions and targeted treatments. Though further advances are needed across several technologies, including AI, there’s little doubt the precision vision will be realized. However, ethical and legal issues will need to be resolved before personalized, precision medicine becomes commonplace.

For truly personalized medicine to work, it needs to be readily available and that requires both data and systems integration. Healthcare information has been notoriously siloed, and this barrier is largely resistant to change. This has been for any number of reasons, among them: legacy systems not architected for information exchange; competitive firewalls; a 'not invented here' mentality; lack of industry-wide application-programming interfaces (APIs); patient privacy issues. To some degree, these have been overcome within the boundaries of an individual healthcare system, but rarely beyond. These barriers have been falling and will continue to do so as the demand for personalized, data-driven medicine necessitates greater information mobility. This integration will allow for commonplace holistic healthcare scenarios.

Here’s once such sequence: A heart monitor embedded in your shirt provides real-time data to a cardiologist, who in response to worrisome changes can then send updated prescriptions to your pharmacist, who can, in turn, send an alert to your virtual assistant – whether smartphone, watch, glasses or some other yet to be invented device – while you are driving to say that your medications are ready to pick up. Then your GPS will automatically update the route to the pharmacy, where you arrive in your self-driving car and pay for the prescription using your phone or retinal scan. All the devices and applications in this example are fully integrated, adjusting seamlessly as you move from one place to another.

To achieve these benefits, whether access to information or applying the latest AI algorithms implies a data infrastructure that connects all the myriad devices. This spans data centers, multiple clouds and vast numbers of edge devices, enables easy and fast provisioning of new systems, rapid deployment of new applications and updates, and ready availability of needed data services such as encryption and security. Such a platform must be scalable to the greatest degree possible. Google built such a platform for its own use – Borg – and similar open-source solutions like Mesos provide this type of platform for everyone else.

These scalable platforms are fundamental to the delivery of AI in healthcare, including data driven-medicine and predictive analytics. This connectivity and integration plus AI-powered algorithms for diagnosis and prescription will create both a platform for rapid determination and delivery of personalized, precision medicine anywhere and an economy of scale to dramatically drive down costs.