TRACKing or TRUSTing transfusion prediction: Validation of Red blood cell transfusion prediction models for low transfusion rate cardiac surgery and high transfusion rate post-cardiotomy veno-arterial extracorporeal life support

Renard G. Haumann, Romana Alp, Tomasz P. Płonek, Ed D. Niesten, Jolanda M. Maaskant, Silvia Mariani, Bettina Wiegmann, Jutta Arens, Job van der Palen, Frank R. Halfwerk

Research output: Contribution to conferencePosterAcademic

16 Downloads (Pure)

Abstract

Abstract body
Preoperative identification of patients at risk of red blood cell (RBC) transfusion is necessary to prevent adverse outcomes. Several models can determine this risk. Models like TRACK, TRUST and ACTA-PORT differ in complexity and performance. Some models outperform TRACK, but their complexity limits clinical application. In 2009, the TRACK model was developed with criteria for everyday practice, simplicity and easy clinical implementation. Advances in hemodilution management in Europe has reduced transfusion rates in adult cardiac surgery, necessitating re-evaluation of the TRACK model in low transfusion rate populations.

Methods
The TRACK model was validated using 4053 adult patients who underwent cardiac surgery between 2015 and 2022. Subsequently, the database was divided at random into a derivation and validation data set. Original coefficients of the TRACK model were updated in the derivation data set and validated in a validation data set on accuracy and discriminative ability. Model calibration and discriminative ability were assessed as measures of model performance. Further, the TRACK model will be validated and updated in the same way for predicting blood transfusion in post-cardiotomy ECLS patients.

Results
All variables but age remained significant in the external validation of the TRACK model. The odds ratio of female sex on blood transfusion increased from 1.42 to 2.42 (95% CI, 1.94 – 3.02). The original TRACK model demonstrated an area under the curve (AUC) of 0.76 (95% CI, 0.74 – 0.78) while showing poor calibration indicating overoptimistic estimation of RBC transfusion risk (p < 0.05). The updated TRACK model demonstrated a slightly higher AUC of 0.78 (95% CI,0.75 – 0.81) and showed good calibration over all risk strata (p = 0.19).

Conclusions
Refining the TRACK coefficients improved preoperative at-risk identification. The updated TRACK model improves predicted accuracy and may help clinicians make better discissions, especially in low-transfusion adult cardiac surgery. This study demonstrates the feasibility of RBC transfusion prediction models for adult cardiac surgery. Our ongoing study is evaluating RBC transfusion prediction models for post-cardiotomy ECLS. These results will also be presented at the conference.
Original languageEnglish
Number of pages1
Publication statusPublished - 26 Apr 2024
Event12th EuroELSO 2024 - ICE Krakow Congress Centre, Krakow, Poland
Duration: 24 Apr 202427 Apr 2024
https://euroelso-congress.com

Conference

Conference12th EuroELSO 2024
Country/TerritoryPoland
CityKrakow
Period24/04/2427/04/24
Internet address

Keywords

  • Red blood cell transfusion,
  • Red blood cell transfusion, Cardiopulmonary bypass
  • Cardiac surgery
  • Risk factors
  • Risk models

Fingerprint

Dive into the research topics of 'TRACKing or TRUSTing transfusion prediction: Validation of Red blood cell transfusion prediction models for low transfusion rate cardiac surgery and high transfusion rate post-cardiotomy veno-arterial extracorporeal life support'. Together they form a unique fingerprint.

Cite this