Patients who leave a hospital do not want to have to go back soon. "It is important to reduce rehospitalizations, as they represent a great burden for patients and their relatives," says Tristan Struja. The physician has investigated - as part of his research project funded by Swiss Cancer Research - how well the rehospitalization risk of cancer patients can be predicted.
If you know in advance who will be readmitted to the hospital within 30 days, you can look specifically for follow-up solutions, Struja says. "That could be a stay in a rehab clinic, organizing hospital-external care or transitional care." Sometimes it's also worth keeping someone in the hospital a few days longer if it means they can avoid readmission.
Scientifically based risk calculation
With his colleagues at Aarau Cantonal Hospital, Struja combed through the anonymized electronic medical records of more than 10,000 patients who were treated in the hospital from 2016 to 2018. And calculated a risk score based on the medical history - i.e., the diagnosis and any secondary diseases.
"We compared two calculation methods: a decades-old statistical method, logistic regression, with a modern method that relies on machine learning," Struja says. They found comparably good results - and (based on the large data set from Aarau) were able to correctly predict a re-entry in 70 to 80 percent of cases using both types of risk calculation.
Until now, physicians have relied primarily on gut feeling when assessing the risk of readmission. "Sometimes patients and relatives also don't want to believe the increased risk," Struja says. "It's then easier for physicians to inform if the risk assessment has been done scientifically."