Machine Learning-Based Prediction of Major Perioperative Allogeneic Blood Requirements in Cardiac Surgery (PREMATRICS)

Cardiac surgery is one of the clinical surgical specialties that carries a particularly high risk for patients to suffer from severe bleeding perioperatively and consequent anemia, which subsequently requires transfusion of allogeneic blood.

Although a surprisingly high number of patients in cardiac surgery do not require perioperative transfusions, it is primarily those patients who do require transfusion who are subsequently at risk for a worse outcome.

In recent years many studies have been published discussing measures that can assist physicians in avoiding the triad of anemia, bleeding, and transfusion in cardiac surgery. Within these publications, the implementation of Patient Blood Management (PBM) is advised. PBM is a set of measures aimed at improving patient outcome by reducing perioperative bleeding and thus preventing both anemia and bleeding.

The three pillars of this bundle are the preoperative preparation of anemic patients with iron, erythropoietin, folic acid and vitamin B12, the prevention of intraoperative blood loss and the reasonable indication for allogeneic transfusions.

Nevertheless, it must be mentioned that the implementation of at least part of these measures is laborious, and full implementation of the recommended bundle is therefore rarely achieved. As a consequence, the full potential of Patient Blood Management is not always realized. Unfortunately this means that transfusion of allogeneic blood cannot be prevented in many patients.

A small proportion of patients undergoing cardiac surgery requires a very large amount of allogeneic blood perioperatively. These patients are typically those with a particularly poor outcome. Massive transfusion of allogeneic blood in this situation is an indicator of complications and a cause of increased mortality.

Although cardiac surgeons and anesthesiologists believe they can assess which patients are at high risk for hemorrhage, recent publications indicate that there is an urgent need for adequate predictive methods. A variety of studies exist that attempt to predict perioperative transfusion requirements, but to date have been plagued by several limitations. Either the previous publications do not focus on the prediction of massive transfusion of allogeneic blood, i.e. administration of ten or more packed red blood cell units perioperatively, but on much lower transfusion volumes, have only low predictive strength to predict massive transfusion in daily clinical practice, or are hardly usable for true prediction because they use factors (features) that are not strictly present only in the preoperative phase.

If an accurate prediction model based on a few features could be created and those patients particularly at risk of massive transfusion of allogeneic blood could be identified, it would subsequently be possible to develop an adapted clinical pathway that would allow patient care to be improved and individualized interventions adapted to the situation to be implemented.

In the best case, an adapted care of patients would be possible, which is able to increase the acceptance for the use of even complex measures of patient blood management. This is especially true for measures such as preoperative preparation with iron and/or erythropoietin, the use of a cell saver, and a particularly careful surgical approach.

Even if it is difficult to apply all measures of patient blood management in all patients, it would be possible with an approach as described to identify those patients who would benefit most from individualized approaches.

Study Overview

• Status: Completed
• Conditions: Transfusion-dependent Anemia
• Intervention / Treatment: Biological: Massive Transfusion of Allogeneic Blood

• Study Type: Observational
• Enrollment (Actual): 3782

Contacts and Locations

Study Locations

• Austria
  • Upper Austria
    • Linz, Upper Austria, Austria, 4021
      • Kepler University Hospital

Participation Criteria

Eligibility Criteria

• Ages Eligible for Study: Child; Adult; Older Adult
• Accepts Healthy Volunteers: No
• Genders Eligible for Study: All

• Sampling Method: Non-Probability Sample

Study Population

As described in the inclusion criteria.

Description

Inclusion Criteria:

• All patients that underwent cardiac surgery at the Kepler University Hospital in the period between 2010-01-01 and 2019-12-31.

Exclusion Criteria:

• Patients below 18 years of age.
• Presence of congenital heart disease.
• Revision surgery of the same patient.

Study Plan

How is the study designed?

Design Details

• Observational Models: Cohort
• Time Perspectives: Retrospective

Number of groups / cohorts

2

Cohorts and Interventions

Group / Cohort	Intervention / Treatment
Massive Transfusion Positive	Biological: Massive Transfusion of Allogeneic Blood; Massive Transfusion of Allogeneic Blood, > pRBCs
Massive Transfusion Negative

What is the study measuring?

Primary Outcome Measures

Outcome Measure	Measure Description	Time Frame
AUROC for Classification of Necessity for Massive Transfusion	AUROC for Classification of Necessity for Massive Transfusion	2010-01-01 to 2019-12-31

Secondary Outcome Measures

Outcome Measure	Measure Description	Time Frame
Confusion Matrix Values	Confusion Matrix Values	2010-01-01 to 2019-12-31
Descriptive Statistics	Descriptive Statistics (age, gender, EuroSCORE II, ASA, NYHA, CCS, main surgical procedure, ventricular ejection fraction, endocarditis present, myocardial infarction present, chronic pulmonary disease, immediacy requirement of procedure (elective, emergency), arteriopathy present, diabetes present, previous cardiac surgery done, critical condition present, aortic surgery, pulmonary hypertension present, renal impairment (none, moderate, severe, dialysis)).	2010-01-01 to 2019-12-31

Collaborators and Investigators

• Sponsor: Kepler University Hospital

Publications and helpful links

General Publications

• Tschoellitsch T, Bock C, Mahecic TT, Hofmann A, Meier J. Machine learning-based prediction of massive perioperative allogeneic blood transfusion in cardiac surgery. Eur J Anaesthesiol. 2022 Sep 1;39(9):766-773. doi: 10.1097/EJA.0000000000001721. Epub 2022 Jul 20.

Study record dates

Study Major Dates

• Study Start (Actual): June 16, 2021
• Primary Completion (Actual): June 30, 2021
• Study Completion (Actual): July 20, 2022

Study Registration Dates

• First Submitted: April 16, 2021
• First Submitted That Met QC Criteria: April 20, 2021
• First Posted (Actual): April 23, 2021

Study Record Updates

• Last Update Posted (Actual): November 29, 2022
• Last Update Submitted That Met QC Criteria: November 22, 2022
• Last Verified: November 1, 2022

More Information

Terms related to this study

• Keywords: Artificial Intelligence; Transfusion; Machine Learning
• Other Study ID Numbers: PREMATRICS

Plan for Individual participant data (IPD)

• Plan to Share Individual Participant Data (IPD)?: No

Drug and device information, study documents

• Studies a U.S. FDA-regulated drug product: No
• Studies a U.S. FDA-regulated device product: No