A causal model for predicting the impact of pharmacotherapy on colorectal surgery outcomes

Evidence-based principles in enhanced recovery programs (ERPs) demonstrate substantial improvement in patient outcomes. Determining which latent variables predict composite outcomes could refine ERP pharmacotherapy recommendations.

Using R, pharmacotherapy data were modeled from an existing dataset of adult elective colorectal surgery patients. Primary composite outcome was absence of surgical site infection, venous thromboembolism, postoperative nausea and vomiting, and other in-hospital postoperative complications (POCs). Secondary composite outcome included no postdischarge POCs, hospital length of stay ≤3 days, and no readmission at 7- or 30-days.

Variables with greater odds of predicting both positive primary and secondary composite outcomes included prehospital oral iron and oral antibiotic use, postoperative sugammadex and neostigmine use, postoperative morphine milligram equivalents (MME) ≤ 50, and IV fluid stop by postoperative day 2. Preoperative scopolamine patch (OR = 0.29 and CI = -0.19-0.77) and perioperative gabapentin (OR = 0.46 and CI = 0.06-0.83) had lesser odds for both primary and secondary composite outcomes. Ketamine nonanesthetic bolus, ondansetron IV use, and in-hospital enoxaparin use had paradoxical lesser primary but greater odds for secondary composite outcomes. Prehospital oral laxative use (OR = 0.61 and CI = 0.18-1.04) and postoperative dual IV antibiotics (OR = 0.52 and CI = 0.10-0.94) had lesser odds for primary, but not secondary, outcome.

To improve the odds for positive composite outcomes, oral iron and antibiotics, sugammadex and neostigmine, lower MME, and early IV fluid cessation could be considered essential core items, whereas postoperative dual IV antibiotics and epidural anesthesia might be avoided. Additional research needs to clarify the impacts of in-hospital enoxaparin, ketamine nonanesthetic bolus, and ondansetron use on composite patient outcomes.