Background:
Outcomes within 30 days of a heart failure (HF) hospitalization are being promoted as metrics of hospital-level quality of care for Medicare beneficiaries. However, few clinically oriented models suitable for use at the point of care have been derived and validated in cohorts of patients with HF with reduced ejection fraction (HF-REF) and HF with preserved ejection fraction (HF-PEF).
Methods:
Patients 65 years and older who were hospitalized in the American Heart Association's Get With the Guidelines (GWTG)-HF registry were linked with Medicare claims from January 2005 to December 2009. Multivariable models using the generalized estimating equation method were developed for 30-day mortality after admission, 30-day readmission after discharge, and 30-day readmission or death after discharge. Patients transferred to another facility or with in-hospital deaths were excluded from the models of 30-day readmission or readmission/death after discharge. The predictors were selected from a derivation sample (70% of study cohort) and refit using the whole study sample; then, the refit model was validated in an independent derivation sample (30% of study cohort). C-statistic was used to evaluate model performance on discrimination and plots comparing predicted versus observed probability were used to assess calibration. The models were evaluated among patients either with HF-PEF or HF-REF.
Results:
Among 33,349 patients at 160 hospitals, 3,002 (9.1%) died within 30 days of admission, 7020 (22.8%) were readmitted within 30 days of discharge, and 8,374 (27.2%) either died or were readmitted within 30 days of discharge. The derived risk models (Table shows predictive model for 30-day death or readmission) discriminated well in patients with HF-REF (c-statistic for death: 0.77, readmission: 0.63, death or readmission: 0.67). Similarly, these models maintained their discriminative capacity in patients with HF-PEF (c-statistic for death: 0.75, readmission: 0.61, death or readmission: 0.64).
Conclusions:
In patients hospitalized at GWTG participating institutions with both preserved and reduced EF, the GWTG-HF predictive models stratify risk for 30-day outcomes in patients with HF. Further study is needed to test effectiveness of integrating these clinical risk models at the point of care.
Point-of-care screening for heart failure with reduced ejection fraction using artificial intelligence during ECG-enabled stethoscope examination in London, UK: a prospective, observational, multicentre study
