A validated rule for predicting patients who require prolonged ventilation post cardiac surgery

Area of Science:

• Cardiology
• Intensive Care Medicine
• Health Services Research

Background:

• Prolonged mechanical ventilation post-cardiac surgery presents significant logistical challenges for cardiac surgical intensive care units (CSUs).
• Accurate prediction of patients requiring extended ventilation is crucial for optimizing resource allocation and patient management.

Purpose of the Study:

• To derive and validate a clinical decision rule for predicting patients at high risk of prolonged ventilation.
• To enable optimized timing of operations for high-risk patients, considering CSU workload.

Main Methods:

• Analysis of the North Staffordshire Royal Infirmary (NSRI) Open Heart Registry (April 1998-May 2002).
• Definition of prolonged ventilation as >24 hours.
• Assessment of the Parsonnet score and development of a novel decision rule using logistic regression and recursive partitioning.
• Validation of the rule on the Blackpool Victoria Hospital (BVH) Open Heart Registry.

Main Results:

• A total of 3,070 patients were analyzed; 201 required prolonged ventilation.
• The developed rule, incorporating Parsonnet score (>7), ejection fraction, operation status, PA pressure, and age, identified 50% of high-risk patients with >90% specificity.
• The Parsonnet score alone was less accurate, misclassifying a substantial number of patients.

Conclusions:

• The derived clinical decision rule effectively identifies approximately 14% of patients as high-risk for prolonged ventilation.
• This rule facilitates more efficient utilization of limited CSU resources through appropriate surgical scheduling.