30 - Validation of a Costing Algorithm in the Neonatal Intensive Care Unit and Identification of Cost Drivers for Neonates

Poster Number: 30
Publication Number: 30.338

Elias Jabbour, McGill University Faculty of Medicine and Health Sciences, Montreal, PQ, Canada; Sharina Patel, McGill University Health Centre Research Institute, Montreal, PQ, Canada; Petros Pechlivanoglou, The Hospital for Sick Children, Toronto, ON, Canada; Prakesh Shah, University of Toronto Temerty Faculty of Medicine, Toronto, ON, Canada; Marc Beltempo, McGill University, Montreal, PQ, Canada

Background: Neonatal Intensive Care Units (NICUs) account for over 35% of pediatric in-hospital costs, thus implying that improved understanding of expenditures is vital for increasing efficiency of care. The Canadian Neonatal Network (CNN) Algorithm is the first to provide case-specific costs based on resource usage among infants born < 37 weeks gestational age (GA) but has not yet been validated for other NICU populations.

Objective: To validate the CNN costing algorithm in 7 case-mix groups with real-time costs obtained from hospital-specific financial software (CPSS) and assess variations in costs within the NICU. 

Design/Methods: Retrospective cohort study of all patients admitted within 24h of birth to a tertiary-level medico-surgical NICU 2016-2019. Patient characteristics and predicted costs were obtained from CNN databases. Costs were adjusted using Standard Hospitalization Costs from the Canadian Institute for Health Information to account for inter-institutional variations and compared using Spearman correlation coefficients (rho).

Results: Among the 1795 live infants admitted into the NICU, 169 (9%) had major congenital anomalies, 164 (9%) were < 29 weeks GA, 189 (11%) were 29-32 weeks GA, 452 (25%) were 33-36 weeks GA, 86 (9%) were term infants diagnosed with Hypoxic-Ischemic Encephalopathy (HIE) and received therapeutic hypothermia, 194 (11%) were term infants requiring respiratory support and 541 (30%) were term-others. Median NICU costs widely varied from $5,155 for term-others to $173,638 for infants < 29 weeks GA; however, there was little difference in mean daily costs (range: $1,136-$1,920). Overall, the CNN algorithm strongly correlated with real-time costs across all case-mix groups (rho range 0.78-0.98). Conclusion(s): The CNN algorithm robustly predicts NICU total costs for 7 case-mix groups. Unit producing personnel represented the highest cost proportion, and daily costs decreased across tertiles of hospitalization. This suggests that reductions in length of stay through optimal discharge strategies could be the most efficient method to reduce NICU costs.
Unit-perspective number of admissions, total patient days and costs per case-mix groupFigure comparing the proportion of total admission number, patient days and costs for the 7 case-mix groups within the study years.
Correlation heatmap comparing predicted and real-time costs per cost center across case mix groupsSpearman correlation coefficient enrichment plot comparing CNN (predicted) and CPSS (real-time) costs across 8 cost centers for 7 case-mix groups (color-coded from blue to red, for low to high Rho values respectively).