Healthcare systems worldwide struggle with the problem of patients who return shortly after discharge from the hospital for the same or a related condition.

Prior studies have examined the recurrent readmission problem by focusing on patients with one chronic condition, such as congestive heart failure. A new study carried out in Israel sought to predict the risk of readmission among patients with more than one chronic condition.

Researchers applied statistical and machine learning methods to information from the electronic health records of 16,117 Israeli patients with multiple chronic conditions and frequent readmissions that began with a visit to the emergency department (ED) and continued from 2005 to 2008.

“Our goal was to see if we could predict each readmission to prevent future readmissions. We hope that if doctors had additional information early in the process, they could intervene earlier to avoid having patients return,” said coauthor Rema Padman, professor of management science and healthcare informatics at Carnegie Mellon University’s Heinz College in Pennsylvania.

The researchers indeed found that tracking multiple readmissions over time can help to identify specific groups of patients early based on their patterns of seeking emergency care and being readmitted.

They found that 30-day readmission rates differed significantly by age, gender, levels of creatinine (a waste product from wear and tear of muscles), chronic conditions, and length of stay, measured at each visit to the ED.

“Based on the results of our study, high-risk patients can be identified earlier, which will result in better care in the ED, appropriate and timely interventions for coordinated and personalized care in inpatient and ambulatory settings, and better allocation of resources,” suggested coauthor Ofir Ben-Assuli, a senior lecturer at Ono Academic College in Israel. His fields of expertise are bio-informatics, data mining, health analytics, information systems and information systems policy.

The study, funded in part by a grant from the United States-Israel Binational Science Foundation, was published in MIS Quarterly of the Management Information Systems Research Center at the University of Minnesota.