Predictive analytics as decision support tool

Twitter icon
Facebook icon
LinkedIn icon
e-mail icon
Google icon
 - Computer Graph

In an era that will emphasize value over volume, a predictive diagnostic imaging calculator could serve as a clinical decision support tool, according to an article published online April 18 in the Journal of the American College of Radiology.

José M. Morey, MD, of the University of Virginia, and colleagues outlined the experiences of Augusta Health, a 255-bed community hospital located in Virginia's Shenandoah Valley that is transitioning into an accountable care organization this year. As part of this transition, a quality improvement group primarily consisting of radiologists was tasked with monitoring the hospital’s imaging utilization. To this end, a predictive imaging calculator was developed on the basis of patient, physician and department imaging averages.

An analysis of imaging trends was conducted for all patients from Jan. 1, 2009, through 2013. This included both inpatient and outpatient care settings, as well as patients seen in the emergency department and urgent care centers. The goal was to use the data to work with individual physicians and departments to improve cost-effectiveness.

Morey and colleagues noted that the volume of patients seen annually at Augusta Health did not vary over the study period, though the image count dropped overall, including an 8 percent drop between 2009 and 2010, the largest year-to-year decline.

Not all modalities exhibited the same pattern. Diagnostic and nuclear medicine decreased linearly over the study period, while ultrasound increased linearly. CT, MRI and women’s imaging decreased, but this drop was not linear. While imaging increased in a nonlinear trend in the emergency department—bolstered by an expansion of services at associated urgent care centers—the data showed a linear decrease in the volume of imaging charges in the clinical outpatient, inpatient, referred and surgical day care settings.

All of this information could lead to the creation of a model to calculate appropriate levels of imaging, wrote Morey and colleagues. This tool could help with image rationing and diagnosis on the basis of a variety of predictive indices, including age and sex.

The calculator could also bolster the efforts of quality improvement teams, with physicians who deviate from expected ordering patterns being assisted in the search for alternative diagnostic strategies.

“Predictive analytics based on patient demographics, such as the imaging calculator proposed, may also help hospitals streamline processes for more efficient patient care, as well as create bundled payment models,” wrote the authors. “Imaging information could effortlessly be monitored within an individual institution, as well as shared across institutions for more complex or atypical patient cases in need of imaging. Monitoring trends in imaging utilization is a pivotal component in efforts to enhance efficiency, decrease unnecessary radiologic services, reduce patient radiation exposure, and improve the quality of patient care.”