Big Data Reveals Factors Leading to Emergency Department Revisits

Researchers at the Rensselaer Polytechnic Institute (RPI; Troy, NY, USA) analyzed Medicare and Medicaid Services (CMS; Baltimore, MD, USA) patient data from 1,149,738 electronic health records (EHRs) to try and understand the characteristics of ED return visits within a 72-hour time frame. The analysis revealed that frequent flyer (FF) patients with multiple revisits account for 47% of Medicaid patient revisits over the time period, and that ED encounters by FF patients with prior 72-hour revisits in the previous six months are thrice more likely to result in a readmit than those of infrequent patients.

From an initial dataset containing over 15,000 potential factors, including primary and secondary treatment codes, demographics, treatment dates and times, and discharge status, the researchers mapped 385 discrete and continuous features using domain knowledge and statistical analyses. The researchers then conducted an analysis to delineate the differences between infrequent and FF revisit patients. The results revealed only three relevant features: alcoholism, living within zip codes with high revisit rates in close proximity to the hospital, and frequent use of the ED and hospital in the past six months.

For the cohort with no prior visits, the models were more complex, involving a combination of diagnosis codes, patient history, and features of the visit such as time, costs, and discharge status. According to the researchers, the marked difference in model features for the two cohorts suggests that distinct opportunities exist among these two patient populations for associated interventions to simultaneously boost the efficacy of care while streamlining ED work flow. The study was published on January 11, 2016, in Big Data.

“Our study demonstrated that there are distinct differences in the determinants for ED revisits among frequent patients and other patients,” concluded lead author James Ryan, PhD, and colleagues. “Further work in the ED revisits should consider the differences in behavior, utilization, and affliction among these groups while exploring how different definitions of readmission events alter the analytic space and the potency of constructed models.”

“This paper shows the impact resulting from the intelligent use of big data for reducing readmissions to emergency departments,” said Prof. Vasant Dhar, PhD, of New York University, editor-in-chief of Big Data. “The authors identify the factors associated with high risk of readmission such as psychiatric and substance abuse and combine them into a model that results in actionable models for decision makers. This type of research will be very useful to healthcare providers as they attempt to align themselves with the requirements of the Affordable Care Act.”

Related Links:
Rensselaer Polytechnic Institute
Medicare and Medicaid Services