Author + information
- Antonio Perez,
- Kevin Chagin,
- Alexander Milinovich,
- Xinge Ji,
- James Pavlescak,
- Michael Kattan,
- Wai Hong Tang and
- Randall Starling
Background: Accurately predicting risk for 30-day readmission after hospital discharge remains challenging in heart failure (HF) patients. The HOSPITAL readmission model, which predicts readmission risk based upon clinical variables, has been derived and validated in general and subspecialty medical admissions from multiple centers, with a reported c-statistic of 0.71 (Donze et al. JAMA Intern Med. 2013 and 2016). This model has never been validated in a cohort of HF hospitalizations.
Methods: We validated the HOSPITAL model using clinical data from inpatient hospitalizations part of the Cleveland Clinic Heart Failure Registry (n=28,939), which contains data from all admissions with a primary diagnosis of HF to a Cleveland Clinic Health System hospital from 2010 to present. The HOSPITAL model variables include discharge hemoglobin (Hgb), discharge from an oncology service, discharge sodium (Na), any procedure during admission, nonelective vs elective index admission, number of admissions during the past 12 months, and length of stay (LOS).
Results: Our cohort had a median (IQR) discharge Hgb of 10.9 (9.6-12.4), discharge Na of 138 (135-140), LOS of 5 (3-7) days, and number of prior hospitalizations in the prior year of 1 (0-3). Seventeen percent of admissions were followed by a 30-day readmission. Forty-two percent had at least one procedure during hospitalization; 58% were admitted via emergency department. In univariable analysis, there was a statistically significant difference in Hgb, Na, procedures, LOS and number of prior hospitalizations between admissions with and without a subsequent 30-day readmission (p<0.0001 for all). The validated c-statistic for the HOSPITAL model in our cohort was 0.52. In secondary analysis, we rebuilt the model using beta coefficients derived from our cohort. This re-derived model had a c-statistic of 0.57.
Conclusions: The HOSPITAL model poorly predicts 30-day readmission after hospitalization due to HF, with a validated c-statistic of 0.52. The applicability of the HOSPITAL model to population management efforts in HF is limited. Predicting risk of 30-day readmission after HF hospitalization remains challenging.
Moderated Poster Contributions
Heart Failure and Cardiomyopathies Moderated Poster Theater, Poster Hall, Hall C
Saturday, March 18, 2017, 3:45 p.m.-3:55 p.m.
Session Title: Risky Business: Predicting Outcomes in Heart Failure
Abstract Category: 13. Heart Failure and Cardiomyopathies: Clinical
Presentation Number: 1265M-03
- 2017 American College of Cardiology Foundation