Here in downtown Chicago, Cook County Hospital stands as a monument to credit scoring. Aside from inspiring the television show "ER," the hospital is notable because it developed an efficient way to treat people complaining of heart problems. Send potential heart patients home early and you expose the hospital to tremendous liability. Give the patient a full battery of tests and spend $13,800 to find that nothing is wrong. Since this happened more than 30+ times a night, a better treatment method was needed.
The hospital turned to Dr. Lee Goldman, who developed a mathematical model for treating heart patients. He took the output from the EKG and combined it with 3 simple questions - 1) How and where the pain is felt; 2) Is there fluid in the patient's lungs; and, 3) Is the patient's systolic blood pressure below 100. Depending on the answers, a treatment decision is reached. An accelerating pain below the heart and yeses to the last two questions would be a quick ticket to the cardiac care unit.
Now, most of the medical community scoffed, as doctors were used to asking a battery of questions about family history, looking at blood work and observing the patient for a period of time before coming to a conclusion. However, Cook County broke the ER staff into 2 teams, one using the cardiac algorithm and one using traditional methodologies. After 2 years, doctors using the algorithm were found to be 95% accurate at predicting serious coronary problems, which was 70% better than doctors left to their own devices. As hospital's management reviewed the findings, a startling conclusion was reached - extra information is sometimes not an advantage and can be a disadvantage.
Having doctors find out that the patient is a hard-charging CEO of bank, has heartburn and just drank a bottle of reserve cabernet not only takes valuable time, but also may lead to the conclusion that the patient has acid reflux instead of the onset of a heart attack. Many factors regarding heart problems matter, but their importance is overwhelmed by key data points that matter more significantly.
When it comes to commercial loan underwriting, certain attributes like debt service coverage, the project area's economic growth and future absorption rates dwarf traditional items like guarantor's net worth, LTV and development experience. Several commercial scoring models have been developed in the last 3 years that take these factors into account, but do not have a large enough data set of problem bank loans to accurately back-test. Unfortunately, this fact is rapidly changing and we continue to test bank scoring models for accuracy. Since many community banks originate loans under $200k, a scoring model is one of the only ways to drop the cost of processing this small sized loan so that it can be profitable. We are hopeful that we can recommend a model within the next several months that will save community banks countless hours and increase the accuracy of small loan credit decisions.
While many loan officers will no doubt continue to feel more confident the more they know about the borrower, over time there will be an evolution towards the quality of credit information collected. Although grizzled veterans may scoff, it is that very desire for greater information that precisely undermines the accuracy of the lending decision.
To this day, the notion that doctors would do better if they knew less about a patient is a tough one to swallow. However, anybody with heart pain admitted to a hospital certainly is thankful they utilize the cardiac heart algorithm.