Machine learning model helps to predict hospital readmissions

A machine learning predictive scoring system is better able to determine the likelihood that discharged patients will be readmitted to the hospital than standard prediction methods.

The ML model, called the Baltimore score, mines hundreds of health data variables in real time and then calculates a score to predict a patient’s chance of returning to the hospital after discharge.

Developed at the University of Maryland Medical System, the algorithm was recently evaluated in a University of Maryland School of Medicine study.

MarylandSchoolofMed-CROP.jpg

In the study, the Baltimore score was matched against standard readmission risk-assessment scores in a cohort of more than 14,000 patients discharged from three hospitals in Maryland, with the goal of assessing their respective abilities at predicting 30-day unplanned readmissions calculated in real time.

“Among three hospitals in different settings, an automated machine learning score better predicted readmissions than commonly used readmission scores,” concludes the study published in the journal JAMA Network Open, which found that the model was most accurate among patients at highest risk. “More efficiently targeting patients at higher risk of readmission may be the first step toward potentially preventing readmissions.”

Also See: UPMC cuts hospital readmission rates with ML algorithm

Specifically, researchers discovered that the Baltimore score was 25.5 percent to 54.9 percent more efficient than comparison scores, “meaning that a similar number of readmissions could be prevented by intervening on 25.5 percent to 54.9 percent fewer patients and thus better targeting resources.”

“If hospitals can better target time and money in planning for discharge to home, then patients may not have to come back to the hospital, with the harm sometimes associated with hospitals, including risks for infection, falls, delirium and other adverse events,” says Daniel Morgan, MD, associate professor of epidemiology and public health at the University of Maryland School of Medicine.

While existing readmission risk-assessment tools analyze a limited set of variables for each patient, the ML model draws from 382 variables including demographics, lab test results, whether the patient required breathing assistance, body mass index, medication usage as well as substance abuse.

“The widespread use of electronic health records has enhanced information flow from all clinicians involved in a patient's treatment,” says E. Albert Reece, MD, dean of the University of Maryland School of Medicine and the John Z. and Akiko K. Bowers Distinguished Professor. “This study underscores how patient data may also help solve the readmission puzzle and, ultimately, improve the quality of patient care.”

According to researchers, the Baltimore score has become automated at three hospitals using Epic EHR in the study—Maryland Midtown Hospital, Saint Joseph Medical Center and University of Maryland Medical Center—and is updated hourly for all patients from admission to discharge.

Going forward, they note that “further work on predicting and targeting readmission prevention efforts needs to account for social determinants of health.”

For reprint and licensing requests for this article, click here.