That Risky Feeling
Photo Source: PxHere
Everyone gets that particular feeling in their stomach when taking a risk. You may place a bigger bet than usual at the casino or on your favorite football team before the game on Sunday. It could be the decision to take a shortcut through a dark alley to make it to the theater on time or cutting it close on a yellow light when driving through an intersection. Risk, in general, refers to the chance of harm, loss, or danger a person faces through exposure to a hazardous situation, person, or substance. A good portion of human mental activity involves deciding what course of action to take and the risks involved with the alternative paths. However, people tend to react to different risks based on perception and not necessarily on facts. In their influential work on the public perception of risk, Paul Slovic, Baruch Fischhoff, and Sarah Lichtenstein discovered in their article titled, “Facts and Fears: Societal Perception of Risk,” that the general public will react differently to a hazard than an expert on that hazard. For example, when asked, 80% people believe death by accident to be more likely than a stroke. When in reality, strokes cause twice as many deaths than accidents. When asked, a group of college students ranked nuclear power number one among thirty perceived threats as the riskiest while experts on nuclear power ranked it near the bottom at 25 out of 30. But experts also get it wrong and sometimes in big ways. For example, an internal memo from Western Union, the telegraph company, in 1876 said, “This telephone has too many shortcomings to be seriously considered as a means of communication. The device is inherently of no value to us.” Determining risk based on scientific studies may help experts to support their risk evaluation of events such as an earthquake happening or the health effects of specific risk factors, but even then, studies are not always sufficient.
Cardiovascular disease causes heart attacks and strokes. According to the World Health Organization in its 2017 fact sheet, cardiovascular disease kills 17.7 million people per year worldwide. That number accounts for 31% of all deaths, which makes cardiovascular disease the number one killer among all the diseases. In light of the lethality of cardiovascular disease, doctors pay very close attention to their patients for signs of potential heart attack and stroke. The American College of Cardiology/American Heart Association (ACC/AHA) developed a Heart Risk Calculator for doctors to predict the risk of heart attack in the next ten years. The calculator takes into account eight risk factors, including age, gender, race, both total and bad cholesterol levels, blood pressure, weight, tobacco use, and presence of diabetes to predict the likelihood of heart attack. The Heart Risk Calculator considers only eight risk factors, but cardiovascular disease results from many factors such as a family history of heart disease or arthritis. Researchers at Nottingham University in the UK led by Stephen F. Weng conducted research to determine if supervised machine learning could make better predictions of Heart attack than the Heart Risk Calculator recommended by the ACC/AHA. In the paper titled, “Can Machine-learning Improve Cardiovascular Risk Prediction Using Routine Clinical Data?” the authors used several types of supervised machine learning to build a prediction system. With 374,000 patient records from the UK Health System, the researchers took a set of records with people who had no heart attacks recorded in 2005 and let the machine learn from the records what types of symptoms and risk factors contributed to heart attacks in the years following 2005. Once the computers were trained on patient records, new records were fed to the machine to see if it could predict heart attacks. Interestingly, the researchers found that the supervised machine learning algorithm produced significantly better predictions than the Heart Risk Calculator. For example, the ACC/AHA accurately predicted a heart attack 72.8% of the time while the supervised machine learning algorithms ranged from 74.5-76.4% accuracy. With millions of people affected by cardiovascular disease, such an improvement in prediction would translate to many lives saved with the application of proper intervention. Moreover, the machine learning found patterns that the doctors did not. The Heart Risk Calculator considered diabetes and high blood pressure to be significant contributors to the risk of heart attack, but the algorithm did not rank either these in the top eight risk factors. Instead, the machine learning algorithm found that body mass index and poverty made for stronger predictors of a heart attack. The sometimes-subtle relationships between different health factors may escape a doctor’s eye, but supervised machine learning can consume vast amounts of information and reveal relationships that even experts may look over.
Risk involves the potential for harm when interacting with a hazardous substance, place, person, animal, or thing. Research has shown that the average person will not always make the same determination of risk than an expert in that area would. Moreover, even recognized experts do not always determine the real danger at hand due to unrecognized connections between risk factors. Machine learning offers in some cases a way to look risk factors and even provide a more accurate prediction of a danger than the experts.
Dr. Smith’s career in scientific and information research spans the areas of bioinformatics, artificial intelligence, toxicology, and chemistry. He has published a number of peer-reviewed scientific papers. He has worked over the past seventeen years developing advanced analytics, machine learning, and knowledge management tools to enable research and support high level decision making. Tim completed his Ph.D. in Toxicology at Cornell University and a Bachelor of Science in chemistry from the University of Washington.