Comparison of Risk Prediction Models to Estimate Opioid-Induced Respiratory Depression, Oversedation, and Overdose in Patients with Cancer.

Abstract

Numerous opioid-induced respiratory depression (OIRD), oversedation, and overdose prediction models exist to quantify a probability or estimate risk severity for a future event. The primary aim was to determine OIRD, oversedation, and overdose risk severity (i.e., low, moderate, and high) and agreement of risk severity between three previously published prediction models. This single-center, retrospective analysis evaluated 134 patients with cancer. Sixty-five (49%) patients were Caucasian. Forty-three (32%) patients were diagnosed with gastrointestinal cancer. Predictive factors from prediction models were concurrent sedating medication (n = 119, 89%), female sex (n = 85, 63%), a mental health diagnosis (n = 68, 51%), and antidepressant use (n = 55, 41%). For most patients, risk severity varied between moderate to high risk. Risk class severity was significantly different between prediction models (p ≤ 0.05). Frequencies of risk severity agreement between all three prediction models, between two prediction models, and no agreement was 16% (n = 22), 69% (n = 93), and 14% (n = 19), respectively. Additional research is needed to evaluate model calibration to increase OIRD, oversedation, and overdose prediction model validity and generalizability for future clinical implementation.