The “science of prediction” may sound like a contradiction in terms. After all, science is definitive, prediction is merely a guess.
But not all predictions are created equal. The truth is, there is a science to predicting results, and accuracy depends on the quality of the data that helps determine outcomes.
We’re not talking about predicting NFL scores or the odds of winning at blackjack(though there is a reason the house sticks on 17 and casinos are allowed to print their own money). Rather, the science of prediction concerns determining the right diagnosis and prescribing the correct treatment when patients visit their doctor. And when it comes to something as important as health, is there really any room for guess work?
Not according to Thomas McGinn, MD,chair of medicine at Northwell Health. He says that when it comes to health care, there’s a way to take the “guess work” out of forecasting outcomes and instead rely on the science of prediction. His work in the field of prediction models uses data and evidence-based medicine at the point of patient contact to improve the accuracy of diagnosis and treatment.
Dr. McGinn described his work as “developing clinical prediction models based on patient symptoms and basic labs. After creating a model, we integrate it into the electronic health record [EHR] in an easy-to-use format.” He explained that the EHR, which has become as standard in the doctor’s toolkit as a stethoscope, will “at the point of patient contact, help more accurately diagnose a patient’s condition, provide automated doctor’s notes and recommend prescriptions in many areas of routine medicine.”
It is a unique approach — still in the trial phase — but one that raises eyebrows throughout the medical field.
Dr. McGinn began studying and writing about the accuracy of prediction models years ago, then sought to turn those findings into a practical application by integrating them into the EHR. Other physician researchers from institutions like Yale University, the University of Pittsburgh Medical Center and the University of Wisconsin are either studying the concept with Dr. McGinn or are learning from his innovative work.
“Most major medical centers and health care systems are not doing this sort of prediction model work yet,” Dr. McGinn said. “At Northwell, we’re creating the prediction models and then studying the best way to integrate them into the EHR in an easy-to-use way, so that physicians will not only be able to use it, but also will want to use it. By making prediction models part of the work flow and creating the tools and algorithms that doctors want to use, we are making it a valuable addition to patient care — improving accuracy and treatment, eliminating unnecessary tests and reducing costs as a result.”
Science vs. Myth
The irony is that the prediction model is far more scientific and definitive than much of current medical practice.
“So much of medicine is folklore,” Dr. McGinn said, offering as an example the common misconception that the color of sputum indicates if a patient needs antibiotics for infection. “The color of sputum does not indicate whether an antibiotic is required,” he said emphatically. “That is a myth and 100 percent inaccurate.”
Dr. McGinn said doctors are notoriously inaccurate when providing antibiotics and usually overestimate risk. “Prediction models calculate the probability based on a number of indicators to get more accurate diagnosis and treatments.”
The emergency departments at North Shore University Hospital and LIJ Medical Center are piloting Dr. McGinn’s prediction models, as are numerous primary care centers in Wisconsin and Utah. Furthermore, Dr. McGinn is testing the EHR models with the Patient Simulation Center at Northwell’s Center for Leaning and Innovation, anticipating a rollout throughout the health system’s hospitals and facilities.
“Reviewing the data from ongoing studies at several sites, we are seeing significant benefit, mostly reducing waste and increasing quality,” Dr. McGinn said. “Thousands of patients have participated in the studies. In almost every case, prediction models are found to be accurate and reduce unnecessary and potentially harmful testing and treatments.”
Looking into his crystal ball, Dr. McGinn believes that prediction models will become the standard for delivering care. “It will be a subtle change, but one that will have broad impact on the way we will think about and practice medicine. It will improve patient outcomes throughout health care and reduce waste significantly,” he said.
As a physician scientist who leaves little to chance, Dr. McGinn is compiling all the evidence he needs to back that prediction up. Dr. McGinn’s models use data and evidence-based medicine at the point of patient contact.