A fanfare for software feats

In research, sometimes the process deserves as much accolades as the results. Take two recent studies where IT innovators not only built the foundation for the analyses but they also devised much needed solutions for tackling administrative data.

First, consider the challenge of extracting data about adverse drug events from physicians’ narrative electronic notes. The notes offer a treasure trove of information about reasons for discontinuing a medication, but how to make sense of them? In a paper published in the April 2 issue of the Annals of Internal Medicine, Huabing Zhang, MD, of Brigham and Women's Hospital in Boston, and colleagues described natural language processing software that, combined with an EHR, allowed them to explore statin-related clinical events or symptoms that led to statin discontinuation.

The software could process up to 40 notes a second and could recognize documentation that linked any medication to clinical events. It was specifically validated for statins with a sensitivity of at least 86.5 percent and had a specificity of at least 91.9 percent.

Among other findings, the researchers revealed that structured formats captured only 30 percent of patients who had a statin-related event. While placebo-controlled clinical trials reported an event rate of 5 percent to 10 percent, Zhang et al had a rate of 18 percent, which they attributed to the use of statins in everyday clinical practice. They proposed that their algorithm might be a useful tool for tracking adverse drug events retrospectively or prospectively.

The second study, published in the March issue of the American Heart Journal, provided a cost analysis that compared transradial and transfemoral PCI. The study relied on ICD-9 codes to identify patients who underwent PCI between 2004 and 2009. The catch, the senior author told Cardiovascular Business, was that the codes don’t discriminate by access site.

To tease out which cases used one or the other access site, Steven P. Marso, MD, an interventional cardiologist at St. Luke’s Mid America Heart Institute in Kansas City, Mo., and colleagues designed and validated an algorithm. They searched the Premier research database, which holds data on about 20 percent of all acute care hospitalizations at more than 450 facilities. They validated the algorithm using data from St. Luke’s Mid America, which is not a Premier center and includes access site in its PCI data.

Marso et al found that using the transradial approach saved hospitals on average more than $500, a figure that could double if the case involved a patient at high risk of bleeding complications. Much of the savings derived from shorter length of stay for transradial patients. The study looked at only inhospital PCIs.

“Where I think you can realize greater savings is if you can convert many more patients to transradial [with] same-day discharge,” Marso said.   

Here is a tip of the hat to the algorithm architects in both studies. Have you or a colleague played a key role in designing software that made research happen? Let us know.

Candace Stuart

Cardiovascular Business, editor