Machine Learning and AI in Healthcare Coming Up Short (Part 1)

AI in healthcareAn independent study on machine learning and artificial intelligence (AI) was released by the McKinsey Global Institute (MGI) in June 2017, focusing on the following central question: “Is artificial intelligence the next digital frontier, and if so, are businesses ready for it?”

The report examined how companies across different industries have begun adopting these technologies. MGI offered insight into the expected profit margin increases that each sector can expect with AI’s improvements.

An article in Wired on artificial intelligence argued that AI “will quickly become more than an underlying technology capability; it will permeate intelligent enterprises and advance to a fundamental tool for daily engagement.” AI is changing the very way that we interact with (and relate to) technology. The MGI report echoes this sentiment and provides concrete data from surveying more than 3,000 senior executives regarding AI technology, deployment, and potential industry impact.

It is interesting to note that the study identified a gap between AI investment and commercial application. While machine learning has the potential to accelerate revenue (and while the healthcare industry in particular expects to see the greatest increase in profit margins as a result of technology adoption), AI gets below-average attention in the healthcare industry. Why is this sector not pursuing AI, considering they will see the most profit from it?

The McKinsey Global Institute study focused on five aspects of AI: robotics, computer vision, language, virtual agents, and deep machine learning. Its findings have determined that AI investments are booming, thanks in large part to tech giants like Baidu and Google. However, AI adoption is stuck in the early stage of experimentation. Most surveyed firms remained skeptical of ROI and very few deployed AI at scale across their organization, although this is unsurprising as AI depends on unique data and has no adoption shortcuts.

Projected benefits of adopting AI in healthcare

The expected benefits of AI adoption for the healthcare industry are as follows:

  • Quicker diagnoses, better treatment plans, and new approaches to insurance
  • In the US: $300 billion in possible savings for population health forecasting
  • In the UK: £3.3 billion possible savings for preventive care and non-elective hospital admission reductions
  • 30–50 percent improvement in nurse productivity
  • Up to 2 percent GDP savings for operational efficiencies in developed countries
  • 5–9 percent health expenditure reduction by to tailor treatments and keep patients engaged
  • $2 trillion–$10 trillion savings globally by tailoring medications and treatments
  • Adding 0.2–1.3 years onto the average life expectancy

You read that right. Not only will AI save money, increase productivity, and tailor treatment plans to individuals, it is expected to make people live nearly a year longer.

And yet...

Disappointing industry response

The healthcare industry has one of the lowest AI technology adoption rates out of all studied industries, despite promised profits. Of its current adoption, most of the technology is concentrated in operations and customer service, using tools like speech recognition and computer vision.

Why is this adoption rate so low, then, if the reward is high?

Slow to adopt digital technology, slow to adopt AI

The MGI study found that sectors slow to adopt digital technologies tend to be slow in adopting AI, too. Nearly 25 percent of American hospitals (and over 40 percent of physicians) have yet to adopt electronic health record systems, so it’s no surprise that AI adoption rates are low in the healthcare industry. In most hospitals, operations management functions are still manually processed. Those organizations that do use digital records don’t necessarily share their data with patients or out-of-network providers, so patients commonly have to repeat tests and recount their medical history to other doctors. If the healthcare industry hasn’t been able to comprehensively adopt digital medical records, knowing their clinical utility, then healthcare machine learning and AI opportunities are still far behind in gaining acceptance.

Data availability

Currently, the US healthcare industry has realized only 10-20 percent of opportunities for AI and machine learning application. A large barrier to realizing the full potential of AI technology is data availability. Patient level data is notoriously guarded, highly regulated, and protected. Sensitive, high-quality data sharing for AI applications will require an overhaul of established regulations, because as it stands now, patient records are difficult to collect and report.

It’s a bit of a catch-22. In order to have high functioning AI and deep machine learning, the applications need access to high-quality data. However, once AI is implemented, the healthcare industry (particularly pharmaceutical manufacturers) can use the technology to drive complex aspects of business without relying on direct access to data. The pending GDPR compliance in the EU is one example of difficult regulatory practices that limit patient level data ability. Manufacturers can turn to machine learning when, for example, they can’t access accurate patient count. The AI algorithm uses shipping data, demographic data, and prescription data to accurately derive the patient count, so decision-makers can better manage sales compensation, forecasting, and mergers and acquisitions (M&A) for their organization.*

AI limitations

There will most likely be hesitation to place patients’ well-being into the hands of a machine. We’ve seen this reluctance play out with self-driving cars – there was a public push-back due to a distrust of machines, irrespective of demonstrable results. So for AI in healthcare, simply put, people will be wary because bedside manner doesn’t exist in robots. Yet.


The benefits of AI in healthcare are very promising, but there are immediate challenges that need to be addressed before the technology is implemented on a large scale. In addition to regulations, public opinion, and organizational barriers, there will be ethical and legal sensitivities. The solution will require data science and pharma-specific data domain knowledge, a combination that is hard to come by. That’s why, in order to take advantage of this rapidly emerging space, more and more organizations are turning to technology and outsourced data science expertise.

Leaders in the AI healthcare movement

RxDataScience, a leader in data science software solutions for pharmaceutical and healthcare companies, partnered with the Indian Institute of Technology – Mandi in a joint research venture to explore the practical applications of AI and machine learning in the healthcare industry.