An Algorithmic Approach to Ventilator Withdrawal at the End of Life

The proposed study is an important, under-investigated area of ICU care for terminally ill patients undergoing terminal ventilator withdrawal. The proposed research has relevance to public health because an algorithmic approach to the ventilator withdrawal process will enhance clinicians' ability to conduct the process while assuring patient comfort, using opioids and/or benzodiazepines effectively.

Terminal ventilator withdrawal is a process that entails the cessation of mechanical ventilatory support with patients who are unable to sustain spontaneous breathing and is commonly performed in the ICU. Ventilator withdrawal is undertaken to allow a natural death. Opioids and/or benzodiazepines are administered before, during, and after as an integral component of the ventilator withdrawal process to prevent or relieve respiratory distress, but there are few guidelines to determine how much to administer or when. Insufficient opioid and/or benzodiazepine administration places the patient at risk for unrelieved respiratory distress and preventable suffering. Conversely, excessive medication administration may hasten death, an unintended consequence, and one that concerns clinicians. The effective doses of medications given during ventilator withdrawal are unknown. The investigators hypothesize that an algorithmic approach to ventilator withdrawal, relying on a biobehavioral instrument to measure and trend distress, will ensure patient comfort, and guide effective opioid and/or benzodiazepine administration. The investigators plan to use a stepped wedge cluster randomized controlled trial with all clusters providing unstructured usual care until each cluster is randomized to implement the algorithmic approach (intervention). The proposed study is innovative because there is no standardized, evidence-based approach guided by an objective measure of respiratory distress to this common ICU procedure. The study has broad clinical significance to provide knowledge that can potentially reduce patient suffering.

All study sites begin in usual care and each site is randomly assigned to crossover to the intervention arm until all sites conclude in the intervention arm.

The medical intensive care unit in four hospitals will comprise the clusters. All four clusters begin the study under the control condition. Ventilator withdrawal is conducted by the usual personnel in those units. Data is collected through observation of the process and the respiratory comfort of the enrolled patients. Each cluster is randomly selected to sequentially cross over to the intervention. The remaining clusters continue with usual care (control) until selected for crossover.

Each cluster is randomly selected to sequentially crossover to the intervention. When crossed over to the intervention the assigned intensive care nurse conducts the ventilator withdrawal according to the algorithm. The algorithm is informed by an objective measure of patient respiratory comfort. Data is collected through observation of the process and the respiratory comfort of the enrolled patients.

Steps and decision trees in the algorithm include in descending order: Ascertain patient consciousness, perform cuff-leak test, evaluate for indications for pre-medication, select a withdrawal method, assess for respiratory distress with Respiratory Distress Observation Scale, medicate for respiratory distress with morphine, make an extubation decision, ascertain need for continuous morphine, ascertain need for supplemental oxygen, assess for post-extubation stridor, treat post-extubation stridor

Respiratory comfort will be measured with the Respiratory Distress Observation Scale at baseline, at every ventilator change, after the ventilator is turned off, every 15-minutes for 2 hours after the ventilator is turned off.

Inclusion Criteria: Patients undergoing ventilator withdrawal Exclusion Criteria: Patients who are conscious and cognitively intact Patients who will undergo organ donation after ventilator withdrawal Patients who are brain dead Patients with bulbar amyotrophic lateral sclerosis Patients with C-1 to C-4 quadriplegia Patients with locked-in syndrome

The investigators will provide public access to the de-identified data files through two open repositories: Wayne State University's DigitalCommons (http://digitalcommons.wayne.edu/), which will provide perpetual access to the data, and the Inter-University Consortium for Political and Social Research's openICPSR (https://www.openicpsr.org/), which will provide access to the data for at least 10 years.