Rogers, L;
Brown, KL;
Franklin, RC;
Ambler, G;
Anderson, D;
Barron, DJ;
Crowe, S;
... Pagel, C; + view all
(2017)
Improving Risk Adjustment for Mortality After Pediatric Cardiac Surgery: The UK PRAiS2 Model.
The Annals of Thoracic Surgery
, 104
(1)
pp. 211-219.
10.1016/j.athoracsur.2016.12.014.
Preview |
Text
Ambler VoR 1-s2.0-S0003497516318288-main.pdf Download (495kB) | Preview |
Abstract
BACKGROUND: Partial Risk Adjustment in Surgery (PRAiS), a risk model for 30-day mortality after children's heart surgery, has been used by the UK National Congenital Heart Disease Audit to report expected risk-adjusted survival since 2013. This study aimed to improve the model by incorporating additional comorbidity and diagnostic information. METHODS: The model development dataset was all procedures performed between 2009 and 2014 in all UK and Ireland congenital cardiac centers. The outcome measure was death within each 30-day surgical episode. Model development followed an iterative process of clinical discussion and development and assessment of models using logistic regression under 25 × 5 cross-validation. Performance was measured using Akaike information criterion, the area under the receiver-operating characteristic curve (AUC), and calibration. The final model was assessed in an external 2014 to 2015 validation dataset. RESULTS: The development dataset comprised 21,838 30-day surgical episodes, with 539 deaths (mortality, 2.5%). The validation dataset comprised 4,207 episodes, with 97 deaths (mortality, 2.3%). The updated risk model included 15 procedural, 11 diagnostic, and 4 comorbidity groupings, and nonlinear functions of age and weight. Performance under cross-validation was: median AUC of 0.83 (range, 0.82 to 0.83), median calibration slope and intercept of 0.92 (range, 0.64 to 1.25) and -0.23 (range, -1.08 to 0.85) respectively. In the validation dataset, the AUC was 0.86 (95% confidence interval [CI], 0.82 to 0.89), and the calibration slope and intercept were 1.01 (95% CI, 0.83 to 1.18) and 0.11 (95% CI, -0.45 to 0.67), respectively, showing excellent performance. CONCLUSIONS: A more sophisticated PRAiS2 risk model for UK use was developed with additional comorbidity and diagnostic information, alongside age and weight as nonlinear variables.
Archive Staff Only
 |
View Item |
