Outcomes of Donor Hearts Offered for Transplantation, Declined by a Single Center and Transplanted at a Secondary Institution

Abstract

Background: Many predictors to assess risk of heart failure (HF) readmissions have been suggested, yet a broad-based approach to using models is lacking. The purpose of this analysis was to use clinical predictors for readmissions found in a univariate analysis in an urban, tertiary care center to create and propose a clinically useful model for HF readmissions derived from a multivariate analysis. Methods: The enterprise data warehouse and the database registry of the Intermountain Heart Collaborative Study were queried for patients discharged from Intermountain Medical Center with a primary diagnosis of HF between January 1, 2012 and December 31, 2012. Assessment of 30-day all-cause (AC) readmissions extended through January 31, 2013. Patients were grouped into those who were, or were not, readmitted within 30 days of discharge. In a previous parallel study, we considered 71 variables in a univariate analysis, 25 of which reached statistical significance for readmissions. Themes from these variables included demographic and clinical data, clinical treatments, comorbidities, and measures of acuity. These variables were then used to perform a multivariate logistic regression analysis to create a risk prediction model for 30-day AC readmissions. Results: Of 586 eligible HF patients identified, 103(17.6%) were readmitted within 30-days; mean time to readmission 13.5 days ( 8.7). The multivariate analysis resulted in 5 variables that remained significant and were predictive of 30-day AC readmissions: longer length of stay, other admissions within the prior year, lower systolic blood pressure on admission, use of pressors or inotropes, and ever having a diagnosis of pneumonia. The final model showed good discrimination for 30-day all-cause readmission with an area under the curve c-statistic of 0.75, see Figure. Conclusions: Given national pressures to improve HF care and reduce readmissions, assessing variables associated with an increased risk in a meaningful manner is not only timely, but of clinical importance. Defining areas of higher acuityseen in longer lengths of stay and previous hospital encounters -should trigger heightened awareness of risk. Creating useful predictive models that may be tested and validated in larger populations will be the next step in making a difference in HF care.