P-002 - (Poster #2) Single-center, retrospective study to estimate opioid-induced respiratory depression and/or oversedation risk in hospitalized patients with cancer receiving palliative care

Serious opioid-induced respiratory depression (OIRD), oversedation, and/or overdose continue to remain a global public health concern. There is ample evidence of numerous prediction models estimating OIRD, oversedation, and/or overdose risk. Examples include the PRediction of Opioid-induced respiratory Depression in patients monitored by capnoGraphY (PRODIGY), the Oversedation Risk Criteria (ORC), the Opioid-Related Adverse Drug Events (ORADEs), and two versions of the Risk Index for Overdose or Serious Opioid-induced Respiratory Depression (RIOSORD). A key difference between models are the data utilized for model development. Several prediction models utilized outpatient clinical and administrative claims data (e.g., RIOSORD), while others have utilized data from hospitalized patients (e.g., PRODIGY, ORC, ORADEs). The ORC was developed from all hospitalized patients receiving opioids (n = 163,190) from one of 12 acute care, Texas-based hospitals from July 2016 to June 2018. The ORC is a 24-item scoring system evaluating eleven predictive factors for OIRD/oversedation. The weighted predictive factors are summed to quantify a score from 0 to 31. The ORC score is stratified as low (<9), moderate (10-20), and high risk (>21).

Purpose/Objectives:

Most published prediction models utilized multivariate, stepwise, logistic regression analyzes and provided acceptable measures of discrimination. However, several models had insufficient or lacked model validation. Model validation utilizes external data that is distinct from the data used for model generation. The authors of the ORC recommended that the ORC be evaluated in different clinical settings and clinically important subgroups. Consequently, this study evaluated the ORC to identify the most common predictive factors and to estimate OIRD/oversedation risk in hospitalized patients with cancer receiving palliative care. The ORC prediction model was evaluated given the ORADEs model excluded patients with cancer during model development. The PRODIGY model was not evaluated due insufficient continuous pulse oximetry and capnography measurement data.

Methods:

This was an IRB-exempt study conducted at an academic medical center. Initial consultation requests were by the patient’s medical team to the inpatient palliative care service. Consultation reasons included management of pain and other symptoms, psychosocial support, advance care planning, and/or a combination of reasons. Inclusion criteria included patients at least 18 years of age, diagnosed with cancer, hospitalized at Jacobs Medical Center, and had documentation of the consultation. Palliative care consultations from patients hospitalized more than once during the data collection period were excluded.

 
A patient list was generated from January 1, 2022 to July 31, 2023. The date of the initial palliative care consultation was defined as the index date. ORC predictive factors, opioid and concurrent sedating medication use were obtained from the EHR (EPIC). Concurrent sedating medication use was defined as no more than 24 hours prior to the initial palliative care consultation. International Classification of Disease, 10^th revision (ICD-10) codes were used to confirm any history of renal insufficiency (N00 – N19), hepatic insufficiency (K70–K77), chronic obstructive pulmonary disease (J44.0, J44.1, J44.8, J44.9) and sleep apnea (G47.3). A paired t-test evaluated the ORC on the index date before and after palliative care recommendations.

Results:

There were 650 palliative care consultations included in the data analysis. Excluded from the analysis were patients with no history of cancer (n = 98) and hospitalized patients with no palliative care consult documentation (n = 2). Most patients were Caucasian women with metastatic gastrointestinal cancer. The most common reason for the palliative care consultation was for pain (n = 380, 58%). Observed ORC predictive factors were concurrent sedating medication (n = 419, 64%) and patients who were not opioid naive (n = 481, 74%). Baseline ORC score (mean ± SD) was 10.7 ± 5 and significantly increased after initiation of palliative care recommendations (12 ± 5.7, p < 0.001). Additionally, concomitant use of a sedating medication increased by 9% (n = 473). Based on the baseline ORC score, most patients were moderate risk for an OIRD/oversedation event (n = 358, 55%).

Conclusions/Implications for future research and/or clinical care:

This study evaluated the ORC prediction model to estimate OIRD/oversedation risk in hospitalized cancer patients receiving palliative care. Data utilization from a different patient cohort at a different institution, which the ORC was originally generated from, also served as a means of external validation. Observed ORC predictive factors were concurrent sedating medication and patients who were not opioid naive. Implementation of palliative care recommendations significantly increased the ORC score but did not affect risk class severity as the majority of patients were at moderate OIRD/oversedation risk. Future studies need to evaluate agreement of risk class severity versus observed OIRD/oversedation prior to general use.