Dive Brief:
- A predictive software tool developed by Dr. Thomas McGinn, chairman of medicine at Hofstra Northwell School of Medicine, that appears on electronic health records (EHRs), asks physicians about patients' conditions, predicts whether they have an ailment, and may include a recommended treatment course, The Wall Street Journal reports.
- The tool proved effective in a clinical trial for diagnosing strep and another for diagnosing pneumonia. Physicians who used the software wrote 25% less antibiotic prescriptions.
- An emergency room pilot for determining the probability of a pulmonary embolism is still ongoing and a study by Dr. McGinn's research team showed the tool can be used to predict whether someone will develop a dangerous infection acquired at hospitals called Clostridium difficile.
Dive Insight:
Despite some success with the software, a 2015 study in JAMA Internal Medicine found a tool to predict deep vein thrombosis overestimated the likelihood of a clot and many patients received high likelihood scores. However, Dr. McGinn told the WSJ that the tool is flawed for inpatients adding, "you have to know where and when these [predictive models] should be used."
Yet Dr. David Feldstein, an associate professor at the University of Wisconsin School of Medicine and Public Health, who oversees the trial, said there is currently a backlash against clinical decision support. He added that resistance to using these predictive models arises from "click fatigue" with the overload of EHR requirements and best-practice recommendations that pop-up on the computer screen.
A recent article in Harvard Business Review takes a look at different predictive analytical tools being used in various clinical settings and concluded that if these are to help reduce costs and improve healthcare outcomes, healthcare systems have to develop ways to evaluate their clinical impact.