This two part blog series will delve into arguably the two most important vowels in healthcare: A-I. This first part will discuss what artificial intelligence (AI) is and how it is currently being used to improve clinical care, while part 2 of this series will take a peek into what the future holds for AI.
What is AI?
AI has been around since the mid-1950s, however its role in medical practice has been limited. So the question now is: why all the recent hype? To answer that question let us clear up some confusion about what artificial intelligence is. In short, AI is any computer program that can achieve goals at a level comparable to or better than a human. When considering the role of AI in medicine, it’s important to note that there are different forms of intelligence. In healthcare, most of the computer generated solutions rely on expert knowledge and hard-coded algorithms rather than true computer based intelligence. This should be contrasted with “machine learning”, where computers can learn from the data allowing them to find patterns that a human expert may have missed. In doing so, programmers of this technology can’t always be sure how the programs will derive solutions, or what solutions they will propose. The most recent way of getting computers to learn from the data is known as “deep learning”, a form of machine learning in which patterns in the data are recognized in multiple distinct layers with different levels of abstraction. This approach has the most potential for the medical community because of its promise in diagnostic medicine and cancer research.
The Hype and the Concern
People are excited about AI because of its potential to disrupt healthcare in areas ranging from data analytics and medical diagnostics to lifestyle management and wearables. The market for artificial intelligence in healthcare is expected to hit $6.6 billion in revenues by 2021.1 There has also been an explosion in AI-related healthcare deals with fewer than 20 in 2012 and more than 70 in 2016.2 Many startups are going all in when it comes to offering AI-based solutions. A few examples include:
- Increasing revenues for healthcare providers by reducing insurance claim denials (e.g. Ayasdi),
- Reducing re-hospitalisation rates of the disabled and elderly with real-time data collected from wearables (e.g. Sentrian), and
- Reducing the diagnostic and therapeutic errors of human physicians (e.g. MaxQ).
Unfortunately, most of these new healthcare technologies and devices are unable to significantly move the needle in terms of improving outcomes and lowering mortality rates. For all of those excited about AI and its potential, there are an equal number of skeptics. A recurring fear in popular culture is that machines will wake up, start to self-replicate, and go on the offensive in order to supplant humans. These fears are largely rooted in fiction as people fear the ‘Skynet’ scenario from the Terminator movies. In my eyes, a more realistic concern is that AI will displace jobs in the healthcare sector. In an article from the Oxford Martin School at the University of Oxford, it was estimated that almost 50% of U.S. employment will be taken over by artificial intelligence within the next 20 years.3 This fear continues to permeate throughout the U.S. as blue-collar jobs remain in the cross hairs of technological advancements, but now you are seeing healthcare professionals get nervous as well.
What Role is AI Currently Playing?
Today, most AI systems use pre-programmed algorithms that process and analyse data provided by healthcare providers. Before this type of AI system can be used, it must be ‘taught’ how to synthesize clinical data such as screenings, diagnoses, and treatment plans so it can identify features and outcomes of interest. Healthcare data stored in electronic medical records is drawn from various sources including demographic surveys, electronic devices and wearables, physical exams, and laboratory results. An AI system might be used, for example, when treating cancer. By applying algorithms designed by field experts to the collected medical information, a computer is able to analyze hundreds of treatment plans and provide a tailored recommendation for a given patient.
AI systems are also used to assist with visualization and pattern recognition. This is something that can help radiologists, pathologists, dermatologists, and even ophthalmologists to reduce the number of false positive diagnoses. Unfortunately, it is not uncommon to have two radiologists examine the same image and disagree on how to interpret the results. Research has shown that, in screening for breast cancer, 50-60% of women will get at least one false-positive diagnosis over a 10 year period.4 AI software designed to recognize visual patterns is estimated to be around 5-10% more accurate in analyzing images compared to physicians.5
Funding for AI in healthcare has mostly focused on three areas: oncology, neurology, and cardiology. Interest in these areas has been driven by the fact that they are amongst the leading causes of death and would greatly benefit from early diagnoses in order to limit comorbidities and complications. Luckily, this is an area in which AI can undoubtedly make a difference.
With that being said, AI systems have a much broader scope of potential applications. Due to the potential financial benefits and expected growth of AI in healthcare, a number of companies are pouring significant resources into this technology. In 2017, Accenture Health reported that the top 10 AI applications will produce an estimated annual benefit of roughly $150 billion by 20266, with the following applications leading the way:
- Robot-Assisted Surgery ($40B)
- Virtual Nursing Assistants ($20B)
- Administrative Workflow Assistance ($18B)
- Fraud Detection ($17B)
- Dosage Error Reduction ($16B)
These applications aim to address areas that separate outstanding physicians from ordinary ones, and no it’s not their level of intelligence, but rather it’s how they approach a patients’ problem and how the healthcare system supports them. You can often find some combination of these factors playing a role in the disparity of outcomes across patient populations throughout America.
Conclusion
AI will undoubtedly play a more significant role in healthcare going forwards as physicians continue to find ways to treat more patients, achieve better outcomes, and do so at lower cost. The possibilities seem to be truly endless. The next part of this series will delve into what the future may look like as AI’s capabilities continue to expand.
Kevin Anderson is a graduating medical student at Duke University School of Medicine and will be starting at LEK Consulting later this year. He’s most passionate about healthcare redesign, patient engagement, and the life sciences. His free moments are spent traveling and enjoying sporting events with his wife and daughter.
Image: Pixabay
References:
- Frost & Sullivan, https://ww2.frost.com/news/press-releases/600-m-6-billion-artificial-intelligence-systems-poised-dramatic-market-expansion-healthcare/
- CB Insights; “From Virtual Nurses To Drug Discovery: 106 Artificial Intelligence Startups In Healthcare;” posted February 3, 2017 at https://www.cbinsights.com/blog/artificial-intelligence-startups-healthcare/
- Frey, Carl Benedikt, and Michael A. Osborne. “The Future of Employment: How Susceptible Are Jobs to Computerisation?” Technological Forecasting and Social Change, vol. 114, 2017, pp. 254–280., doi:10.1016/j.techfore.2016.08.019. Available at: https://www.oxfordmartin.ox.ac.uk/downloads/academic/The_Future_of_Employment.pdf
- Defrank, Jessica T, et al. “Influence of False-Positive Mammography Results on Subsequent Screening: Do Physician Recommendations Buffer Negative Effects?” Journal of Medical Screening, vol. 19, no. 1, 2012, pp. 35–41., doi:10.1258/jms.2012.011123. Available from: https://www.researchgate.net/publication/221967401_Influence_of_false-positive_mammography_results_on_subsequent_screening_Do_physician_recommendations_buffer_negative_effects
- Aletan, Samuel O. “Artificial Intelligence Languages and Architectures: Past, Present, and Future.” Artificial Intelligence Methods and Applications, 1992, pp. 431–484., doi:10.1142/9789814354707_0014. Available from: https://www.worldscientific.com/worldscibooks/10.1142/1734
- Accenture, “Artifical Intelligence: Healthcare’s New Nervous System”, https://www.accenture.com/t20171215T032059Z__w__/us-en/_acnmedia/PDF-49/Accenture-Health-Artificial-Intelligence.pdf#zoom=50