Predictive Analytics for Inpatient Postoperative Opioid Use in Patients Undergoing Mastectomy

Abstract

Introduction: The use of opioids in mastectomy patients is a particular challenge, having to balance the management of acute pain while minimizing risks of continuous opioid use postoperatively. Despite attempts to decrease postmastectomy opioid use, including regional anesthetics, gabapentinoids, topical anesthetics, and nonopioid anesthesia, prolonged opioid use remains clinically significant among these patients. The goal of this study is to identify risk factors and develop machine-learning-based models to predict patients who are at higher risk for postoperative opioid use after mastectomy. Methods: In this retrospective cohort study, we collected data from patients that underwent mastectomy procedures. The primary outcome of interest was defined as oxycodone milligram equivalents (OME) greater than or equal to the 75% of OME use on a postoperative day 1. Model performance (area under the receiveroperating characteristics curve (AUC)) of various machine learning approaches was calculated via 10-fold cross-validation. Odds ratio (OR) and 95% confidence intervals (CI) were reported.Results: There were a total of 148 patients that underwent mastectomy and were included. The medium (quartiles) postoperative day 1 opioid use was 5 mg OME (0.25 mg OME). Using multivariable logistic regression, the most protective factors against higher opioid use was being postmenopausal (OR: 0.13, 95% CI: 0.03-0.61, p = 0.009) and cancer diagnosis (OR: 0.19, 95% CI: 0.05-0.73, p = 0.01). The AUC was 0.725 (95% CI: 0.572-0.876). There was no difference in the performance of other machine-learning-based approaches.Conclusions: The ability to predict patients' postoperative pain could have a significant impact on preoperative counseling and patient satisfaction.

Prediction of Complications and Prognostication in Perioperative Medicine: A Systematic Review and PROBAST Assessment of Machine Learning Tools

“…Development of postoperative complications: Most studies focused on the prediction of postoperative acute complications 26–28,42,44–49,59,99,110,112,113,118,123 such as pain and opioid use, 53–57 postoperative atrial fibrillation (new-onset atrial fibrillation), 82 postoperative risk of stroke or myocardial infarction, 50,71,77 and delirium or cognitive decline. 65–70 Other models focused on the risk of developing pneumonia or respiratory failure, 83,85,125 acute kidney injury, 43,52,58,60–63,120–122 liver failure 117 or development of sepsis or surgical site infection.…”

Predictive models for fentanyl dose requirement and postoperative pain using clinical and genetic factors in patients undergoing major breast surgery

“…We compared our predictive models with prior reports and are presented in Table 5. 9,27,30,35,40 Tsuboi et al 35 developed a multivariable linear regression model for predicting the outcome of 24-hour postoperative fentanyl consumption (R-squared 5 0.302) in patients with inflammatory bowel disease. Yoshida et al 40 developed multivariable linear regression models to predict 24hour postoperative fentanyl requirement (R-squared 5 0.145) and perioperative fentanyl requirement (R-squared 5 0.185) in patients undergoing orthognathic surgery.…”

“…Similar to our study, these 2 studies developed models to predict postoperative fentanyl requirement directly; however, the performance of our predictive model was higher (R-squared 5 0.313) than the abovementioned studies. Dolendo et al 9 and Nair et al 27 developed multiple statistical models to predict and classify patients based on the postoperative opioid requirement in patients undergoing mastectomy and ambulatory surgeries, respectively. Pan et al 30 developed a multivariable linear regression model for predicting perioperative analgesic requirement in patients undergoing Caesarean section using physical and psychological tests.…”

Artificial Intelligence and Machine Learning in Cancer Related Pain: A Systematic Review

“…Generalized linear mixed models (GLMMs), least absolute shrinkage and selection operator (Lasso), Linear regression (LR), Logistic regression (LR), Bayesian regression [42], [32], [43], [33], [34], [35], [27], [44], [45], [46], [39], [41], [31] Model that estimates the relationship between one dependent variable and one or more independent variables using a line.…”

“…The study found that the ANN outperformed the LR model in predicting treatment response. Dolendo et al (2022) [45] analyzed 148 cancer patients who underwent mastectomy, utilizing LR to assess factors affecting postoperative opioid use. The study suggested that these models, if implemented, could significantly impact preoperative counseling and patient satisfaction by aiding in the prediction of postoperative pain.…”