By Mark Lambrecht, PhD, Director, Health and Life Sciences Global Practice at SAS.
Data analytics in health care isn’t anything new. SAS has been extracting insights from data for the benefit of patients, physicians and health care systems for more than four decades. Today, the ability exists to integrate diverse data sets, and that capability brings with it the potential accelerate health care treatment options.
In the US, the state of Nevada is pioneering new ways diverse data sets can be used to benefit populations. Researchers at the Healthy Nevada Project are applying analytics to clinical, environmental, socioeconomic, and genetic data to advance personalized, precision medicine for individuals while also improving citizens’ overall health statewide – a concept referred to as population health.
The power of correlation
When different sets of data are introduced, a computer can look for correlations and patterns between them. Maybe a certain disease or allergy is prevalent in a certain region, or maybe a certain lifestyle gives rise to illnesses even though the connection between them isn’t readily apparent. Artificial intelligence and machine learning are great at finding patterns and correlations humans cannot readily identify. This opens the door to a new form of health care, where prevention is key.
During the and& festival, I had the chance to discuss the potential impact of AI and analytics on our current health care model with some experts in the field. They all agreed the future of health care should be AI-driven, if only to make it ultimately more human. The power of AI and analytics extends beyond the identification of broad patterns. In theory, an algorithm is far better at diagnosing a patient and deciding possible treatments than a doctor, especially if the underlying data used for the diagnosis is large and complex, and the task at hand can be automated. AI can access a nearly unlimited database of health care records. Those historical datasets can prove invaluable in determining the best course of action in treating any given patient.
Today, we’re already seeing the benefit of AI-powered health care. An algorithm that can identify cancer in CT scans of a lung already exists, and a computer has proven itself to be better than doctors at diagnosing lung infections from X-ray data.
By trusting AI to make diagnoses, doctors can focus once again on the human aspect of health care. They can spend more time actually interacting with their patients and, for instance, gauging their mental and emotional health.
It is obvious that the usefulness of AI- and analytics-driven health care correlates with the amount of available data. At and&, SAS colleague and medical doctor, Joost Huiskens, mentioned the Caesar Project. Researchers contributing to the project use data to try to improve the outcome of cancer patients. By giving algorithms access to, for instance, historical health care records and genetic profiles, AI can indicate the best course of action per patient. Maybe an operation is the optimal choice, because data indicates a high success rate in previous patients with similar genetic makeup, or maybe AI advises against such an operation because the correlation of available data shows the risk is too high.
When an AI program has access to vast amounts of medical records, environmental data and personal information, it can quickly discover the usefulness of certain types of treatment. But that’s not all. There is no reason why it should wait to sound the alarm. If data indicates a person has a high risk of heart failure, it could encourage doctor and patient to take preventive action. Preventive medicine could become very personal. For example, a physician saying, “The data indicates that 90 percent of the people with your weight, diet and genetic makeup develop diabetes,” is a far more potent motivator for someone to take preventive action than merely hearing, “Being obese can cause diabetes.”
The privacy issue
AI can fundamentally turn the health care model from a reactionary to a preventive, but there is one major obstacle to overcome. The data required by health care analytics systems are some of the most personal data in existence. People are understandably very reluctant to allow access to their historical medical records. Dr. Huiskens calls it unethical to not treat patients using data accumulated by treating previous patients, and while I agree, the general public needs to do so as well.
In our daily lives, we readily trade personal data for convenience. Facebook and Google know almost everything about many of us, and what we get in return is hardly life-saving. I think people will be ready to make a similar trade-off in health care, especially if it can be guaranteed relevant data will be made anonymous. Fortunately, emerging technologies such as blockchain allow us to use valuable medical data from personal records without the need to link the data to identifiable personal information.
I believe AI-driven health care is the way forward. If we focus on implementing a robust and secure framework for handling sensitive medical information anonymously, I’m sure we’ll find that preventive health care, with AI inside, is just around the corner.