36: DERIVATION AND VALIDATION OF A NOVEL OPIOID-INDUCED RESPIRATORY DEPRESSION RISK PREDICTION TOOL

Learning Objectives: Nearly half of all in-hospital cardiorespiratory arrests occur on general care floors (GCF), with catastrophic outcomes. Opioid-induced respiratory depression (OIRD) is one potential cause of these events. The PRediction of Opioid-induced respiratory Depression In patients monitored by capnoGraphY (PRODIGY) trial (NCT02811302) investigated the incidence of OIRD using continuous cardiorespiratory monitoring (heart rate (HR), oxygen saturation (SpO2), end-tidal carbon dioxide (etCO2) and respiratory rate (RR)), in patients receiving opioid therapy on the GCF. Methods: The PRODIGY trial was an international, prospective, observational study conducted from April 2017-April 2018 across 16 sites in USA,Europe and Asia. Continuous capnography & pulse oximetry data was collected using the CapnostreamTM respiratory monitor (Medtronic), with screens blinded and alarms silenced throughout the monitoring period (≥24 hr).Standard (spot-check) monitoring continued unblinded.Respiratory depression (RD) events, defined by one of: etCO2 ≤15 or ≥60mmHg for ≥3min, RR ≤5 breaths for ≥3min, SpO2 ≤85% for ≥3min, apnea >30sec, or a respiratory opioid-related adverse event, were adjudicated by an independent clinical event committee. A novel risk prediction tool was derived and validated by bootstrapping. Results: RD occurred in 41.4% of 1,496 patients. In a modified full analysis set (1,384 monitored patients taking opioids), positive predictors of OIRD (p≤0.05) by univariate analysis included age ≥70 to <80 or ≥80 yr, male sex, major organ failure, chronic heart failure or cardiac disease, coronary artery disease, COPD or pulmonary disease, pneumonia, Type II diabetes, hypertension, kidney failure, and opioid naivety; negative predictors included BMI ≥35 kg⁄m and asthma. The multivariate regression model, including age ≥70 to <80 or ≥80 yr, male sex, sleep disorders, hypertension, and opioid naivety, was used to develop a novel OIRD risk prediction tool (PRODIGY score). Derivation model assessment included AUC (0.7676), Brier score (0.1498), R (0.2044), and Hosmer-Lemeshow p-value (0.3649), with low optimism (0.001) between the derivation and validation models. Conclusions: A novel high accuracy multivariate risk prediction tool of OIRD was derived and validated. Implementation of improved monitoring strategies using this tool could prevent respiratory compromise, improve patient safety on the GCF, decrease the burden of rapid response calls & unplanned ICU admissions. CCMCritical Care MedicineCrit Care Med0090-3493Lippincott Williams & WilkinsHagerstown, MDCCM