Predictive Models of Prolonged Mechanical Ventilation and Difficult Weaning

Abstract

Clinical predictions about individual patients’ total duration of mechanical ventilation or their duration of weaning are routinely made by intensivists as part of care of ventilated patients. Although these predictions may not be systematically elaborated or formally expressed, they influence important clinical decisions. The decision about whether and when to perform a tracheostomy is one of particular importance. Weaning and other aspects of care might be facilitated by an early tracheostomy in patients who will need a lengthy course of invasive ventilation. Other decisions that may also be influenced by such predictions include the initiation of enteral nutrition, the use of intensive glycemic control, the inclusion of patients in clinical trials, and the possible transfer of patients to referral centers for mechanical ventilation or weaning. However, the accuracy of these clinical predictions by intensivists, either in the setting of clinical research [1] or practice [2], has been shown to be quite limited. Therefore, objective tools that allow accurate predictions of these outcomes, prolonged ventilation or difficult weaning, have been sought to assist physicians with these decisions. These tools include the identification of risk factors and the development of predictive models. In this case, predictive models are mathematical tools that combine results of several variables assessed at an early point in the course of mechanical ventilation to estimate either the probability that a patient will require “prolonged” ventilation (or weaning), or its actual duration. This chapter reviews studies aimed at identifying risk factors for and developing predictive models of prolonged mechanical ventilation and weaning.