An automated method of identifying potentially preventable readmissions failed to prove accurate enough for it to replace manual review in a recent Kaiser Permanente study.
Researchers reviewed 459 readmission cases from18 Kaiser Permanente hospitals in Northern California and compared the results using both methods on the same data. Their results are published in BMC Medical Informatics and Decision Making.
Manual review consisted of a chart review tool; interviews with patients, their families and treating providers; and nurse and physician team determination of preventability on a five-point scale. Those findings were compared with those generated from 3M's Potentially Preventable Readmission (PPR) software.
As expected, the automated system identified more cases as potentially preventable (78 percent or 358 cases) vs. manual review (47 percent, 227 cases). The two methods agreed on the preventability of 56 percent (258) of readmissions.
Using manual review as the reference, however, the sensitivity--the percentage of potentially preventable readmissions identified by manual review that also were identified as such by PPR--was 85 percent. Meanwhile, specificity--the percentage of non-potentially preventable readmissions identified by manual review that also were identified by PPR--was 28 percent.
With hospitals facing increased penalties for excess readmissions, they're increasingly looking for technical solutions to better flag those who need better care coordination. Penalties estimated at $227 million are expected to be levied against 2,225 hospitals this fiscal year.
Vanderbilt University Medical Center, for one, is testing a statistical modeling program called Cornelius to predict patients' risk of readmission.
Parkland Health and Hospital System in Dallas found success with heart patents, using an electronic health record-based risk stratification system.
Meanwhile, a collaborative initiative involving 83 hospitals and 93 community partners in Minnesota reduced readmissions through a combination of comprehensive discharge planning, medication management, patient engagement, transition care support and transition communications.
To learn more:
- find the research (.pdf)