Becker's Clinical Quality & Infection Control

37 DATA INFORMATICS & ANALYTICS AI tools to detect infection fall short when analyzing data across health systems By Megan Knowles A rtificial intelligence tools trained to detect pneumo- nia on chest X-rays significantly decreased in performance when tested on data at multiple health systems, a study published in PLOS Medicine found. The study findings suggest AI in medicine must be carefully tested for performance among various popula- tions, or the deep learning models might not perform as accurately as anticipated. The study, conducted at New York City-based Icahn School of Medicine at Mount Sinai, evaluated how AI models identified pneumonia in 158,000 chest X-rays in three medical institutions. The researchers looked at computer system frameworks called convolutional neural networks, which analyze medical imaging and give a computer-aided diagnosis. In three out of five comparisons between medical institutions, the convolutional neural networks' perfor- mance in diagnosing diseases on X-rays from hospitals outside its own network was significantly lower than on X-rays from the original health system. However, the study found convolutional neural networks were highly accurate at detecting the hospital system where an X-ray was acquired. The challenge of using AI in medicine must be carefully tested for performance among various populations, or the deep learning models might not per- form as accurately as anticipated. deep learning models in medicine is that they use a huge number of parameters, making it difficult to pinpoint the variables that drive predictions, such as the types of CT scanners used at a hospital and the resolution quality of imaging, the researchers said. "Our findings should give pause to those considering rapid deployment of artificial intelligence platforms without rigorously assessing their performance in re- al-world clinical settings reflective of where they are being deployed," said senior author Eric Oermann, MD, an instructor in neurosurgery at the Icahn School of Medicine at Mount Sinai. n Johns Hopkins researchers develop tool to predict patient no-shows By Megan Knowles R esearchers from Baltimore-based Johns Hopkins University's Malone Center for Engineering in Healthcare have cre- ated an algorithm to reduce patient no-shows and increase appointment availability. "e new approach developed with our partners at Johns Hopkins Community Phy- sicians has allowed the clinic to add over 70 pediatric appointments to their schedule per week, improving outpatient access for more children while also reducing the no-show rate 16 percent for patients who are highly likely to miss scheduled appointments," said Scott Levin, PhD, associate professor of emergency medicine at Johns Hopkins University School of Medicine. e research team analyzed operations at two Johns Hopkins clinics. On average, 20 percent of patients did not show up for their scheduled appointments. To address this issue, the researchers created a machine learning algorithm that predicts the likelihood a patient will show up for an appointment and considers demographics, economic status and medical history. e al- gorithm calculates a no-show score for each patient. Providers can use these scores to look at their weekly schedule and find which patients are at high risk for not showing up for their appointments. The model, in use at two clinics since September 2017, found patients who visit emergency departments more often are more likely to not show up for scheduled appointments. Patients who use online patient portals to schedule their own ap- pointments are more likely to show up for their appointments. Individual departments are best equipped to use information the model gives to identify the best way to use their appointment slots, the researchers said. Some departments increased outreach strategies, such as having staff make more reminder calls to patients with a high risk of not showing up, while other departments give these high-risk appointment slots to patients who urgently need care. e next step is to integrate the model into Johns Hopkins Medicine's EHR before launching it in the Johns Hopkins Health System and hospitals nationwide, the re- searchers said. n

• Cover