“Predicting rain doesn’t count. Building arks does.”
— Warren Buffet
THE TRUE BENEFITS of artificial intelligence, machine learning, natural language processing, robotics, and data will be seen when we move away from our current fee-for-service model of healthcare and towards preventative medicine. The idea is simple: instead of waiting for people to get sick and then trying to treat their symptoms, we can head illnesses off at the pass and stop them from becoming a problem in the first place. It might cost a little more up front, but it could save the healthcare industry a huge amount of money in the long run.
Predictive medicine and preventative medicine are similar but different, in that predictive medicine revolves around identifying what’s likely to happen in the future while preventative medicine involves taking active steps based upon these insights. Predictive medicine will naturally lead to preventative medicine, but it’s predictive medicine I want to talk about right now because artificial intelligence essentially relies on predictions derived from massive amounts of data. The more data that’s fed into the algorithm, the better the predictions become.
The interesting thing is that once we get into the business of predicting health, the industry will be forced to radically redefine itself. As new opportunities arise for health maintenance and prevention, so too will the number of services that are available to us.
There are different ways of approaching predictive medicine, from genomics and proteomics to cytomics and genetics. Genetic testing is arguably the most powerful way of identifying potential diseases decades prior to any actual symptoms developing. In some circumstances, it could even be used to identify potential diseases when babies are in the womb so that people who are more susceptible to disease can take preventative measures and modify their behaviors to avoid potential risk factors. This is where predictive medicine becomes preventative medicine.
The overall aim of predictive medicine is to flag risk factors so that physicians and patients can work together to reduce the chances of future problems. For example, patients with a greater risk of heart attacks and irregularities could receive more regular EKGs and cardiologist appointments. It’s intended for both healthy people and for those with existing diseases, but the goal in both cases remains the same: to use predictions and preventative medicine to give people the best possible quality of life.
Predictive Medicine Examples
There’s no shortage of great examples of predictive medicine, and I’ve already shared many of them elsewhere in both my books and my articles. Here are just a few more to help you to wrap your head around some of the potentials that predictive medicine has to offer.
NEWBORN SCREENING: Typically conducted shortly after birth, the goal is to identify potential genetic disorders as early as possible. This is currently one of the most widespread forms of predictive medicine thanks to US state law which mandates taking blood samples of every newborn baby across every state.
RISK TESTING: This approach to predictive medicine looks to see whether patients have risk factors that could exacerbate the likelihood of a disease. For example, a 50-year-old heavy smoker is more likely to suffer from lung cancer, emphysema and other diseases than a 20-year-old non-smoker.
DIAGNOSTIC TESTING: When a doctor has made a tentative diagnosis, diagnostic testing is used to confirm or refute the diagnosis. For example, a celiac disease blood test could be carried out to determine whether a patient is suffering from gluten intolerance or whether some other issue is the root cause.
PRECONCEPTION TESTING: The idea here is to test parents before they start trying to conceive a child to identify whether either (or both!) parents carry a gene mutation that could cause genetic disorders. Parents can then make a more educated decision on whether to try for a child or not based on any potential risk factors.
DIRECT TO CONSUMER TESTING: This relatively new phenomenon is characterized by services like 23AndMe which allow people to test their genes with no need for a physician to act as a go-between. These tests may not be as comprehensive as other types of predictive medicine, but they do have the advantage of increased accessibility and greater privacy. Plus they return power to the hands of the consumer — which in this case is the patient.
While I was carrying out my research for this book, I was fortunate enough to come across Prediction Machines: The Simple Economics of Artificial Intelligence by Ajay Agrawal, Joshua Guns and Avi Goldfarb. I read a lot of books during my travels, often listening to them via audiobook, but this one stands out as the best resource on the subject that I’ve come across. It’s chock full of insights that can apply to any industry — but arguably to healthcare in particular.
Prediction machines will reduce uncertainty, but they may not totally eliminate it. Still, anything that gives physicians an edge could make a huge difference to our struggling healthcare system. For example, the authors talk about what would happen if prediction machines were used to examine tumors. If they could give us a definitive answer of whether they were benign or not with no room for error, the doctor would find it easier to know whether to order an invasive procedure, such as a biopsy, to find out more. “Ordering a biopsy is the less risky decision,” the authors explain. “Yes, it is costly, but it can yield a more certain diagnosis. Seen in this light, the role of the prediction machine is to increase the doctor’s confidence in not conducting a biopsy. Such non-invasive procedures are less costly, especially for the patient, they inform doctors about whether the patient can avoid an invasive exam, like a biopsy, and make them more confident in abstaining from treatments and further analysis.”
Another great example is that of medical imaging, which the authors explain will need human oversight for the foreseeable future, although it’s possible that artificial intelligence will take over in the long term. “Imaging is costly, both in terms of time and in the potential health consequences of radiation exposure,” the authors explain. “For some imaging technologies, as the cost of imaging falls, the amount of imaging will increase. So it’s possible that in the short and possibly medium terms, this increase will offset the decline in the human time spent with each image.”
The authors also have some excellent advice on how to choose whether to use a specific AI tool in your business. “Every task has a group of decisions at its heart,” they explain. “And those decisions have some predictive element. We suggest taking those tasks and [breaking them down] into their constituent elements. Separate the parts of a decision into each of its elements. To see how this works, let’s consider the startup Atomwise, which offers a prediction tool that aims to shorten the time involved in discovering promising pharmaceutical drug prospects. Millions of possible drug molecules might become drugs, but purchasing and testing each drug is time-consuming and costly. How do drug companies determine which to test? They make educated guesses — or predictions — based on research that suggests which molecules are most likely to become effective drugs.”
Atomwise CEO Abraham Heifets explains, “For a drug to work, it has to bind the diseased target, and it has to fail to bind proteins in your liver, your kidneys, your heart, your brain, and other things that are going to cause toxic side effects. It comes down to, stick to the things you want to stick to, fail to stick to the things you don’t.”
“If drug companies can predict binding affinity then they can identify which molecules are most likely to work,” the authors explain. “Atomwise provides this prediction by offering an AI tool that makes the task of identifying potential drugs more efficient. The tool uses AI to predict the binding affinity of molecules so Atomwise can recommend to drug companies, in a ranked list, which molecules have the best binding affinity for a disease protein. For example, Atomwise might provide the top 20 molecules that have the highest binding affinity for the Ebola virus. Rather than just testing molecules one at a time, Atomwise’s prediction machine can handle millions of possibilities. While the drug company still needs to test and verify candidates through a combination of human and machine judgements and actions, the Atomwise AI tool dramatically lowers the cost and accelerates the speed of the first task of finding those candidates.”
The Problems with Predictive Medicine
Predictive medicine isn’t perfect, particularly now at this early stage. I prefer to think of myself as a realist instead of an optimist, which is one of the reasons why I acknowledge that there’s a lot of work to be done before we can usher in the new dawn in healthcare that we’re so close to — and yet so far away from.
One of the problems with predictive medicine is the risk of false positives. Even at a 99.9% accuracy rate, that’s 1 in 1,000 people being subjected to the unnecessary stress and strain of thinking they’re at risk of something that will never develop. Then there’s the fact that we may end up giving huge amounts of medication in a bid to prevent illnesses when many people might never have developed the disease in the first place. That would be wasteful, expensive and inefficient, but there’s worse. Many medications have unpleasant side effects which can reduce patients’ quality of life.
There are also the ethical implications of predictive medicine. For example, what if employers start mandating genetic testing for every employee and if they used that data to determine who to invest in and to give promotions to? And what if health insurers start to require people to take genetic tests before they’ll offer them coverage?
These are just a few of the ethical questions that we’ll have to think about, but the good news is that there may just be a precedent. Back on May 21st, 2008, then-president George Bush signed the Genetic Information Nondiscrimination Act into law. According to KaiserNetwork.org, “Under the bill, employers cannot make decisions about whether to hire potential employees or fire or promote employees based on the results of genetic tests. In addition, health insurers cannot deny coverage to potential members or charge higher premiums to members because of genetic test results. Supporters called the bill the ‘first major civil rights act of the 21st century’ and said they hope it will encourage more people to participate in clinical research for treatments of specific genetic sequences.”
Want to learn more?
I talk more about new technologies and their impact on the healthcare industry in my book, The Future of Healthcare: Humans and Machines Partnering for Better Outcomes. Click here to buy yourself a copy.