Why Amazon needs AI at the center of its healthcare strategy

This article was originally contributed to Forbes.com

In December 2017, Amazon, JPMorgan Chase and Berkshire Hathaway announced the formation of a new healthcare company which would use technology to provide high-quality healthcare to patients and families more simply, and at a more reasonable cost. This move and rumors of entry into pharmaceutical distribution shook up stock prices for more established healthcare companies and pharmacy chains.

Amazon’s entry into healthcare is intriguing because medicine is ripe for disruption. In 2016, U.S. per-person healthcare expenses were $10,348, more than double that of other first-world countries that offer universal health coverage ($4,752 in Canada, $4,600 in France, $4,708 in Australia, and $4,192 in the UK). Despite these costs, U.S. medical care is not altogether accurate or safe; medical errors kill more Americans annually than AIDS and motor vehicle incidents. Yet somehow modern medicine has escaped large-scale reform from automation and systems engineering.

Amazon has the potential to change this market. Its decision is modeled after tech giants like Alibaba and Tencent, which have been experimenting with employee healthcare software in China for many years and whose initial targets included online medical advice, drug tracking systems and more recently, artificial intelligence.

However, if Amazon intends to succeed where other industry giants have failed, it is essential for it to build infrastructure that can leverage cutting-edge medical technologies. As the saying goes, “If you want something you have never had, you must be willing to do something you have never done.” In 2018, this means Amazon needs to implement AI software for continuous analytics of automatically structured big data and advanced research technology.

Whether Amazon’s latest endeavor hits the mark or not, these are the trends that tech leaders in the healthcare space ought to be familiar with as we move into a new phase of AI, big data and advanced research technology:

Artificial Intelligence

AI is routinely implemented for machine learning applications to enhance clinical decision making and identify trends. According to The New York Times, over 130 Chinese tech companies were applying AI to increase efficiency and accuracy in overburdened Chinese hospitals. An example use case included the use of machine learning to identify diabetic retinopathy, which extends the capacity of Chinese ophthalmologists who are overburdened. Only 20 eye doctors are available for every 1 million persons, half of what is found in the U.S.

Artificial intelligence will likely become capable of rudimentary logic in the near future as its third wave, which could help it make more complex diagnoses and identify novel correlations. But in the early stages, innovators in healthcare tech could benefit from using AI to quickly identify key financial inefficiencies such as insurance fraud or forecast patients’ healthcare needs based on treatment trends.

Big Data And Continuous Analytics

Tough problems require creative solutions. If healthcare and pharmaceutical tech innovators intend a foray into restructuring health plans, it will require significant data collection and analysis.

As behavioral economists know, sometimes humans behave in funny ways. Although most AI programs optimize known solutions to novel problems (first-wave AI) or iterate on them with machine learning (second-wave AI), data mining can identify novel trends that had not been previously correlated.

In modern applications, data mining is done with descriptive, predictive or prescriptive software and, in the case of tech giants, even online controlled experiments.

A modern medical example is a data-mining surveillance system from the University of Alberta laboratory information management systems, which uses data from bacterial cultures and patient care to generate monthly reports on infection control that are then reviewed by human experts.

Healthcare technology disruptors need to plan for continuous data analytics that equalize medical data from disparate contexts in the design and execution of any modern health software to stay competitive.

Advanced Research Technology

Contrary to popular belief, experimental science in medicine is relatively new. Before the introduction of evidence-based medicine in the 1990s, the majority of Western medical advice was based on observational science and expert opinion. Evidence-based medicine has signified a historical shift in the way Western medical doctors view and treat patients.

Within the new hierarchy of medical evidence, expert opinion ranks the lowest. Double-blind, randomized, placebo-controlled clinical trials are mid-level evidence, and systematic reviews (compilations of multiple clinical trials) are the highest level of evidence. However, evidence is not always available, in which case physicians rely on best judgment and collaborative decision making for treatment decisions.

What is concerning to many health experts is that more than half of current treatments may not be evidence-based. Historically, some very common — and invasive — procedures have turned out to have no benefit or even to be harmful. For instance, a 2018 study showed that stent placement for heart disease, a procedure that can cost up to $14,000, works no better than a placebo to increase exercise tolerance on a cardiac stress test. This is disturbing news, but the positive implications are that this evidence reduces costs, saves patients from having to undergo invasive treatments and redirects resources to more valuable interventions. We simply need more of these kinds of studies and providers who can rapidly evaluate data to develop treatments that get results.

If a healthcare technology disruptor that introduces AI with continuous analytics of automatically structured big data can display statistically significant observational evidence that is more likely to be accurate, the same system can be easily adapted to facilitate large-scale experimental validation studies in clinical trials. This could be disruptive, efficient and beneficial for medicine.

Amazon is likely to be welcomed in healthcare because the industry needs the potential advances they can bring with integrated systems using AI, continuous data analytics from big data and advanced research technology. And where Amazon’s work leads, the rest of the world often follows.