A systematic review of cardiac surgery clinical prediction models that include intra-operative variables

Area of Science:

• Cardiology
• Medical Informatics
• Surgical Outcomes Research

Background:

• Clinical prediction models (CPMs) for cardiac surgery typically rely on pre-operative data.
• Intra-operative variables are less commonly incorporated into CPMs, limiting their widespread adoption.
• This review systematically identifies CPMs using intra-operative variables for short-term outcomes in adult cardiac surgery.

Purpose of the Study:

• To identify and evaluate clinical prediction models (CPMs) that incorporate intra-operative variables.
• To assess the performance and limitations of existing CPMs for predicting short-term outcomes in adult cardiac surgery.
• To understand the potential added value of intra-operative data in cardiac surgery CPMs.

Main Methods:

• Systematic literature search of Ovid MEDLINE and EMBASE databases (inception to December 2022).
• Inclusion of studies developing CPMs with at least one intra-operative variable.
• Data extraction using a critical appraisal framework and bias assessment tool, with analysis of model performance (discrimination and calibration).

Main Results:

• Twenty-four models were identified, frequently predicting acute kidney injury and peri-operative mortality.
• Common intra-operative variables included cardiopulmonary bypass time and transfusion.
• While model discrimination was acceptable, calibration was often poor or unreported, and most models had high risk of bias.

Conclusions:

• Weak reporting and methodological limitations hinder the adoption of intra-operative CPMs.
• Evidence suggests intra-operative variables can improve CPM discrimination.
• Further research is needed to establish the role of intra-operative CPMs in managing cardiac surgery patients.