Predicting Avoidable 30-Day Hospital Readmissions

A common costly component to overall healthcare costs is hospital readmission. The 30-day readmission rate in the United States is 18% according to recent studies, and the cost of readmissions among Medicare beneficiaries is an estimated $17 billion annually.

It has been determined that some hospital readmissions are avoidable, and a facility’s rate of readmission is now being used as a benchmark of quality, with financial penalties in place for hospitals with high risk-adjusted rates.

Because interventions to reduce preventable readmissions are often expensive to implement, a model that would effectively predict 30-day readmission risk in general medical patients would help clinicians target transitional care interventions most efficiently.

Researchers recently derived and internally validated a prediction model for potentially avoidable 30-day hospital readmission in medical patients using readily available administrative and clinical data. They reported results of a retrospective cohort study of the model in JAMA Internal Medicine [2013;173(8):632-638].

The study was conducted at Brigham and Women’s Hospital, a 750-bed academic medical center in Boston, Massachusetts. Study participants were all patient discharges from any medical service between July 1, 2009, and June 30, 2010. Only hospitalizations with a length of stay >24 hours were included in the analysis.

There were 3 possible outcomes among the patients included: (1) admissions not followed by any 30-day readmission; (2) admissions followed by a 30-day potentially avoidable readmission; and (3) admissions followed by a 30-day unavoidable readmission.

There were 12,383 patients discharged from the medical services of the hospital during the study period. Of those, 1652 were excluded because of death before discharge, transfer to another acute healthcare facility, or because the patient left against medical advice.

Of the eligible 10,731 discharges, 22.3% (n=2398) were followed by a 30-day readmission, 879 of which were identified as potentially avoidable (8.5% of all index discharges, 36.7% of readmissions). The researchers randomly divided the 8333 admissions not followed by a 30-day readmission and the 879 potentially avoidable readmissions into a derivation set (two thirds [n=6141]) and a validation set (one third [n=3071]). Overall, 7123 unique patients accounted for all 9212 index discharges.

The prediction score identified 7 independent factors, described as the HOSPITAL score: (1) hemoglobin at discharge, (2) discharge from an oncology unit, (3) sodium level at discharge, (4) procedure during the index admission, (5) index type of admission, (6) number of admissions during the last 12 months, and (7) length of stay.

In the validation set, 26.7% of patients were classified as high risk, with an estimated potentially avoidable readmission risk of 18.0% (observed, 18.0%). The HOSPITAL score had fair discriminatory power (C statistic, 0.71) and had good calibration.

Study limitations cited by the researchers include an inability to identify readmissions that occurred outside the study hospital network, not identifying or including inpatient deaths in the outcome, and the possible exclusion of other predictors of potentially avoidable readmission such as functional status, health literacy, degree of social support, and previous medication adherence.

In summary, the researchers said, “This simple prediction model identifies before discharge the risk of potentially avoidable 30-day readmission in medical patients. This score has potential to easily identify patients who may need more intensive transitional care interventions.”