Abstract
Background
Risk prediction models in heart failure (HF) are typically complex, derived retrospectively from administrative databases, and modest in their ability to discriminate between high, medium, and low risk categories. The complexity of these models makes them difficult to use at the point of care.
Purpose
To determine if a simple risk index using Length of hospital stay (L), number of Emergency department visits in the preceding 6 months (E), and either admission or discharge N-Terminal (NT) prohormone of Brain Natriuretic Peptide (pro-BNP) at the point of care can predict 30-day readmissions in patients hospitalized for HF.
Methods
This is a sub-study of the Patient-Centered Care Transitions in HF (PACT-HF) stepped-wedge cluster randomized trial. We included 772 patients hospitalized for HF at 10 Canadian hospitals. We used log-binomial regression models with Length of stay, Emergency department visits in the preceding 6 months, and either admission or discharge N-Terminal prohormone of Brain Natriuretic Peptide (NT-pro-BNP) as the predictor variables and 30-day all-cause readmission as the outcome. We derived the LENT risk score from the β-coefficients of the regression model (Fig. 1). All the models were adjusted for post-discharge services. We assessed model discrimination with C-statistics and model calibration with the net reclassification index (NRI). We used the bootstrapping approach with 100 runs for internal validation.
Results
The LENT index had a possible score ranging from 1 to 13 (Fig 1). Increments in the LENT risk score were associated with an increased risk of 30-day readmission; a 1-point increase in the LENT index using the admission and discharge NT-pro-BNP predicted a 23% and 19% increase in 30-day readmission risk, respectively. The internal validation produced similar results. Compared to a null model, the LE index had an NRI of 0.35 [95% CI 0.18, 0.53], and admission and discharge NT-pro-BNP further improved calibration of the LE index (NRI 0.15 [95% CI 0, 0.32] and 0.20 [95% CI 0.03, 0.37], respectively). The LENT index offered modest discrimination for 30-day readmission (C-statistic 0.64 [95% CI 0.59, 0.69]), similar to more complex risk models.
Figure 1. The LENT index scoring system
Conclusion
A simple risk index based on Length of stay, Emergent visits, and NT-pro-BNP at the point of care can reliably predict 30-day readmissions. The LENT index offers ease of use over traditional risk prediction models.
Acknowledgement/Funding
Canadian Institutes of Health Research, Ontario MOHLTC, Roche Diagnostics
Patient-Reported Outcomes Improves the Prediction of In-patient and Emergency Department Readmission Risks in Coronary Artery Disease
