UK risk adjustment for 30-day mortality after pediatric cardiac surgery: what it is, where to run it, and audit arithmetic for its outputs
Experimental prototype. Built from published sources but not yet independently verified. Do not use for clinical decisions; check results against the primary references.
Research and audit use only. PRAiS was built to adjust a hospital programme's expected survival for case mix in national audit. It is not a calculator for one child's operative risk, consent or family counselling, and this page does not calculate one.
What PRAiS2 is
Partial Risk Adjustment in Surgery (PRAiS) is the logistic regression model the UK National Congenital Heart Disease Audit (NCHDA) has used since 2013 to report risk-adjusted 30-day survival. PRAiS2 (Rogers 2017) added comorbidity and diagnostic information:
Development data: all pediatric cardiac surgery in UK and Ireland centres, 2009–2014; 21,838 30-day surgical episodes, 539 deaths (2.5%).
Predictors: 15 specific-procedure groups (plus "no specific procedure"), 11 diagnosis groups, 4 comorbidity groups (congenital comorbidity, acquired comorbidity, severity of illness, additional cardiac risk factor), univentricular status, bypass or non-bypass, an era term, and age and weight as fractional polynomials (a·x + b·√x). 36 parameters.
The current version distributed by UCL is PRAiS v4: PRAiS2 recalibrated on NCHDA data from April 2015 to March 2022, extended to all children under 18 years, with prematurity added as a risk factor and the NCHDA specific-procedure algorithm v8.5 (July 2024) (UCL product page).
Why it is not reproduced here
Each episode must first be assigned a specific procedure by the NCHDA specific-procedure algorithm from UK procedure codes, and a diagnosis and comorbidity group from NCHDA code lists. Mapping STS or IPCCC codes to those groups is a large task with room for silent error.
The final coefficients are in the paper's supplemental material (recalibrated on 2009–2015 data), and the model has since been recalibrated again (v4). A copy here would drift from the version the audit uses.
The official software is licensed with the condition that licensees do not redistribute it or run it for third parties.
The model's purpose (programme-level risk adjustment) makes an individual-risk page easy to misuse.
Audit helper: paste predicted risks produced by the official PRAiS software, with outcomes. All fields start blank.
Model, equations and sources
Audit arithmetic (this page)
Expected deaths E = sum of predicted risks p_i (p as a proportion)
Observed deaths O = number of 30-day deaths
O/E = O / E
Cumulative expected - observed after episode k = sum_{i<=k} (p_i - died_i)
(the quantity a variable life-adjusted display, VLAD, plots: above 0 = fewer deaths than expected)
No control limits or funnel-plot bounds are drawn: those need the audit's own method and are produced by the official software and the NCHDA reports.
PRAiS2 risk factors with odds ratios (Rogers 2017, Table 3, development data)
Procedure and diagnosis group odds ratios were not reported because the groups are ordered by age and similarity as well as risk. The ratios below cannot reproduce a prediction on their own.
Risk factor
OR
95% CI
References
Rogers L, Brown KL, Franklin RC, et al. Improving risk adjustment for mortality after pediatric cardiac surgery: the UK PRAiS2 model. Ann Thorac Surg 2017;104(1):211-219. doi:10.1016/j.athoracsur.2016.12.014 (CC BY 4.0; full text retrieved from UCL Discovery)
Brown KL, Rogers L, Barron DJ, et al. Incorporating comorbidity within risk adjustment for UK pediatric cardiac surgery. Ann Thorac Surg 2017;104(1):220-226. doi:10.1016/j.athoracsur.2016.12.013 (abstract retrieved; defines the comorbidity groups)
The proposal for this page cited "Rogers 2023 / Brown 2019"; PubMed shows the PRAiS2 papers are Rogers 2017 and Brown 2017 (above). Later recalibrations are documented on the UCL product page, not in a retrieved paper.
This page never estimates an individual child's risk. For counselling, use the operating team's own outcomes and judgement.
Experimental. Check every result against the official PRAiS software output.