Development and Prospective Validation of a Pediatric Heart Failure Risk-Prediction Model for 30-Day Mortality in US Children Listed for Heart Transplantation

Abstract

Introduction: Hyponatremia (HN) is linked to increased heart failure (HF) severity. The frequency of HN in HF is 25-30%. While clinical markers in HF are linked to HN, little data is available describing the impact of HN on total and intensive care unit (ICU) utilization metrics. This study assessed the effect of HN on total and ICU cost ($), total and ICU length of stay (LOS), and ICU admission in hospitalized HF patients (pts). Methods: The Premier hospital database was used to construct a matched sample of inpatients discharged between January 1, 2007 and June 30, 2009. HN pts (n5558,818) were identified using primary or secondary diagnosis of HN (ICD-9-CM 276.1) and were matched to a non-HN control group (n5558,815) using propensity score matching. Matching was based on patient age, gender, race, 3M APR -DRG assignment, comorbidities, hospital size and geographic region. The HF sub-sample (HN5199,803; non-HN5213,562) had a primary diagnosis of HF (ICD-9-CM 428.23, 428.33, 428.43). Bivariate analysis utilized chisquare tests. Cost was analyzed using a multivariate general linear regression (GLM) model with gamma distribution and log-link, LOS was analyzed with GLM with negative binomial distribution and log-link. ICU admission and hospital readmission were analyzed using multivariate logistic regression. Results: Bivariate analysis (see table) showed that HN pts had significantly higher total and ICU $, total and ICU LOS, and percentage of ICU admissions compared to non-HN pts. Multivariate analysis showed that HF pts with HN had significantly higher total $ (19% higher; CI5.15-.24; p!.0001), ICU $ (23% higher; CI5.14-.34; p!.0001), total LOS (22% higher; CI5.19-.25; p!.0001), and ICU LOS (21% higher; CI5.13-.30; p!.0001) than non-HN pts with HF. HN pts with HF were significantly more likely to be admitted to the ICU (OR51.23; CI51.15-1.30; p!.0001).