Methods:
This is a retrospective cohort analysis of women at a single high-volume academic institution who underwent a cesarean section and were discharged from the hospital. This study included 1,494 women hospitalized for cesarean delivery from July 1, 2016 to December 31, 2016 and then from January 1, 2018 to August 31, 2018, excluding women within the one year surrounding the deadline for the NYSDOH mandated Opioid Prescriber Training Program (January 1, 2017 to December 31, 2017). This tertiary center is in an urban area with a multilingual, racially and ethnically diverse population. Over 80 attending physicians, residents, and physician assistants employed by the hospital system treated these patients.
At our institution, women routinely receive neuraxial anesthesia (epidural and/or spinal) for cesarean delivery. In the post-operative period, the obstetrical team manages pain, and women generally receive multi-modal pain management including long acting narcotics, most commonly oxycodone or Percocet (oxycodone-acetaminophen). Discharge medications are prescribed at the individual provider’s discretion. There are no current guidelines at our hospital for inpatient or outpatient narcotic prescriptions. Though there is a post-Cesarean order set in our electronic medical record system EPIC, there is no discharge navigator or discharge order set that includes prescriptions.
The Institutional Review Board at Montefiore Medical Center, Albert Einstein College of Medicine approved this study. We obtained information via chart review within EPIC. Our center’s electronic medical record system includes all outpatient and inpatient records. We queried demographic information including age, race, ethnicity and primary language. We then obtained all clinical and pharmacologic data, including patient and surgery specific characteristic as well as inpatient medications and outpatient prescriptions, directly from the electronic medical record. The data was transferred to an electronic database and double checked independently by two members of the research team. If there was missing data, it was labeled as “unknown”. We used the same electronic record to see if a narcotic was prescribed at patient discharge, and if so, we included information on the type, strength and number of narcotic pills prescribed. We converted all narcotics into total morphine milligram equivalents (MME) using conversion rates from CDC.gov to more effectively compare amounts among the different opioids.7 We obtained this value by converting each opioid dosage to MME and then multiplying by the number of pills.8 The literature frequently uses this MME conversion to compare amounts between narcotics. Our primary outcome was total MME prescribed for outpatient use. We analyzed the total outpatient MME prescribed at discharge before and after the mandated NYSDOH Opioid Prescriber Training. Secondary outcomes included analyzing outpatient opioid prescription habits by provider level as well as identifying trends in outpatient opioid prescription patterns related to the amount of inpatient narcotic use, and patient, surgical and hospital-specific factors. Since this study was not focused on actual patient use, we did not collect information on whether the prescriptions were filled or not.
We computed descriptive statistics (frequencies, medians and interquartile ranges) to summarize patient, surgical and hospital-specific factors pre and post intervention, as well as across all patients. We assessed the association between cohort (pre vs. post-intervention) and each factor via chi-square test. In-house opioid use was categorized as < 50, 50-100, and > 100 MME. Total amount of opioid prescribed at discharge was categorized as 0, <150, 150, and >150 MME. These categories were chosen for the discharge prescriptions since 150 MME was the median amount of narcotic prescribed both pre and post-intervention. In order to examine the association between opioid prescription and patient, surgical and hospital factors, ordinal logistic regression models were estimated for both the pre and post-intervention periods. We examined univariate and fully adjusted models based on a set of a prioriclinically relevant variables. Age was categorized as ≤25 years, 26-30, 31-35,36-40,41-45,46-50, and > 50.  BMI was categorized as normal (<25), overweight (25-29.99), obese 1 (30-34.99), obese 2 (35-39.99) and obese 3 (40+).  Surgical time was categorized as < 30 minutes to > 180 minutes in 30-minute increments.  Age, surgical time, and BMI were entered into the model as ordinal variables based on the above categorization. Odds ratios (OR) and corresponding 95% confidence limits were estimated. Two-sided p-values less than 0.05 were considered statistically significant. All analyses were performed in SAS version 9.4 (SAS Institute Inc., Cary NC).