Visual Models Save Lives: How a Decision Tree Revolutionized Heart Attack Diagnosis

ArgonDigital - enterprise automation experts

Share This Post

Heart Attack Decision Tree
Recreation of the model used in the Cook Co. ER

Beyond adding visual flair to a requirements document, visual models enable what Malcolm Gladwell, in his book Blink: The Power of Thinking Without Thinking, calls “rapid cognition.” By streamlining information about any topic, visual models allow us to cut straight to the point without spending time sifting through unnecessary details. It’s hard to find a better example of how visual models can impact an organization than when Brendan Reilly, a Cook County hospital administrator, sanctioned the use of a decision tree to diagnose heart attacks throughout the hospital’s cardiac unit.

When Reilly first arrived at Cook County Hospital in Chicago, he was faced with an immediate problem: the hospital had a critical shortage of beds available to treat patients who were experiencing chest pains. Obviously, chest pains can be symptomatic of heart attack, but not every patient suffering from chest pains is actually having a heart attack. The only way to be 100% sure of a heart attack diagnosis is to admit a patient to the cardiac unit and run a series of expensive tests. However, because of the general lack of health insurance for many of the hospital’s patients, this method of diagnosis dramatically inflated the hospital’s costs. With that in mind, Reilly turned to a decision tree developed years earlier by a U.S. Navy cardiologist Lee Goldman. Goldman spent years developing and testing a single model that would allow submarine doctors to quickly evaluate possible heart attack symptoms and determine if the sub had to resurface and evacuate the chest pain sufferer. Whether saving hospital beds or limiting the unnecessary disclosure of a submarine’s position, the motivation for using the model was simple: make complex decision-making more efficient.

While rising costs inspired Reilly’s decision to institute the decision tree, there was widespread concern that the use of this model would lead to a greater number of heart attack sufferers being denied care. For doctors, the prime focus was ensuring patient safety above all else, but this over-reliance on costly tests was driving up the hospital’s costs and limiting the amount of resources available to other patients who also needed their help. When the model’s diagnostic prescriptions were finally put into place, the conclusion was undeniable: unnecessary hospitalizations were greatly reduced and health outcomes for patients remained consistent with previous levels of care.

The model’s success can be traced to the basic fact that more information does not necessarily lead to better decisions. In fact, psychologists have investigated this issue, testing how different levels of information affected the accuracy of a doctor’s diagnoses. Surprisingly, what they found is that more information did, in fact, make doctors more confident in their diagnosis which, on the surface, seems like a good thing. The only problem was that more information did not help them make better diagnoses. They continued to exhibit only a 30% accuracy level when it comes to determining patient ailments, so more information does not necessarily lead to better outcomes.

Visual models are not just a luxury we should fit into our practice, be it medical, software, or naval strategy; they are essential tools in the quest to spread rapid cognition.

More To Explore