Outcomes
Table 2 summarizes the prescribing patterns pre and post training. A lower percentage of prescriptions post- training were >150 MME (21.3%) as opposed to prior to the training (45.1%) and a higher percent were < 150 MME post-training (22.5% vs. 13.7%; p-value for association < 0.001). Despite this shift, the median amount prescribed remained 150 MME pre and post-intervention. Table 3 examines prescribing patterns stratified by prescribing provider level. Pre-intervention, there was a statistically significant association between provider level and narcotic prescription category (p-value = 0.03). Residents were 1.63 times more likely to prescribe higher categories of narcotics compared to physician assistants, fellows, and attendings (95% CI: (1.04, 2.54)). In an a priorimodel adjusted for age, race, body mass index (BMI), surgery time, prior cesarean section, in-house infection, in-house narcotic use and whether other procedures were performed, this association remained statistically significant (OR= 3.42; 95% CI: (1.14,10.24); data not shown). Post-training, all provider levels had reduced proportions of narcotic prescriptions in the > 150 MME category. We observed large reductions in narcotic prescriptions > 150 MME after the training period in 2nd through 4thyear residents such that during the pre-intervention phase, 84% and 65% of 2nd and 3-4th year residents prescribed > 150 MME respectively, compared to 22% and 28% post-intervention. In post-intervention univariate and adjusted models, there were no statistically significant associations between provider level and the amount category of narcotics prescribed (p-values 0.30 and 0.65 respectively). Total in-house narcotic use was observed to be associated with higher levels of narcotics prescribed post-intervention in univariate analysis, however this effect was not statistically significant in the a priori adjusted model (OR>100 vs. ≤ 50 MME =1.10, 95% CI: (0.71,1.72)). There were no patient or surgical factors that were associated with the amount of narcotics prescribed in either univariate or multivariable analysis. Therefore, with the exception of in-house narcotic use, multivariable models did not yield qualitatively different results and were thus not included. Table 4 summarizes univariate associations for selected patient and surgical characteristics pre and post intervention.