Authors

R W Issitt¹; B Margetts¹; J Booth¹; W Bryant¹; R M Crook¹; A Robertson¹;  ¹ Great Ormond Street Hospital for Children, UK

Objective

Colloid Osmotic or Oncotic Pressure (COP) is a primary force in the movement of solutes between the capillary vessel and interstitial space. In the intensive care setting a COP of 15mmHg is associated with 50% survival and animal studies have shown that a COP <16mmHg is associated with respiratory distress. However, we have previously demonstrated that patients with complex congenital cardiac heart defects have similar oncotic pressures to other critically ill neonates (<15mmHg) which are significantly lower than has been previously reported for other paediatric cardiac patients. We sought to create a model the might use preoperative COP to act as an indicator of disease severity and a predictor of post-operative acute renal failure.

Method

Sixty consecutive patients were enrolled. Data were collected on baseline, on-bypass, rewarm and post-modified ultrafiltration COP, Partial Risk Adjustment in Surgery Score (PRAiS) as well intraoperative data from the heart-lung machine, and postoperative renal injury status; defined as peritoneal dialysis catheter placement, urine output below national standard limits, and requirement for initiation of renal replacement therapy. Bayesian, general linear post-operative risk of death and renal failure models were constructed with the Stan modelling language and fit using Hamiltonian Monte Carlo estimation techniques in R (version 3.5.0).

Results

There was a demonstrable link between baseline COP and risk of death, which is especially prevalent in children under 3 years old. A predictive model was created based upon baseline COP and the type of cardiac lesion (acyanotic vs cyanotic). Patients that developed renal failure postoperatively had significantly lower baseline COP (p < 0.001, T Test) and we created a binomial model based upon baseline COP, age and type of cardiac lesion. This data was then used to construct an interactive application to predict the probability of renal failure following cardiac surgery during the pre-bypass period.

Conclusion

Baseline colloid Osmotic Pressure can be used to predict risk of death and in combination with the patient’s age and type of congenital lesion, can predict the probability of postoperative renal failure. It is hoped that this predictive model and interactive application will allow earlier identification and therefore clinical intervention of those patients at high risk of renal replacement therapy. Further work is planned to use machine learning/artificial intelligence to adjust the model depending upon intraoperative bypass events.