Effects of Shared Decision Making on Opioid Prescribing After Hysterectomy.

The institutional review board at the University of Michigan determined that this quality initiative project was exempt as it did not satisfy the definition of research according to the U.S. Department of Health and Human Services or the U.S. Food and Drug Administration 21 , 22 and provided approval for publication. The primary outcome was total opioids prescribed prior to versus after implementation of the decision aid. Secondary outcomes included opioid consumption, patient satisfaction, and refill requests after intervention implementation. We estimated a sample size of 102 patients in each cohort in order to detect a clinically important effect size of a 20% reduction in OME prescribed at discharge. The quality improvement intervention was conducted as follows: each patient was met in the preoperative holding bay on the day of surgery by a resident or fellow-level surgeon. Using a visual decision aid, the patient and any present support members were uniformly educated regarding pain management after hysterectomy. The decision aid was a double-sided 6×4-inch decision aid, with the front side outlining pain expectations after surgery, risks of opioids, and common side effects of opioids ( Figure 1 ). The back side of the card explained our recommended three-tier pain management protocol—starting with scheduled ibuprofen if clinically acceptable, followed by addition of acetaminophen, and finally, the addition of opioids if needed. While acetaminophen was encouraged, patients were counseled to add this medication as needed given our focus on developing an individualized pain plan and minimizing unnecessary medication exposure. The final portion of the education tool included a scale from zero to the maximum number of opioids recommended by institutional guidelines based on the surgical route of hysterectomy (vaginal, laparoscopy, or abdominal). 2 Two arrows were placed on the scale. The first arrow indicated the average number of tablets a cohort of patients previously reported using at our institution from August 2015 through December 2015. 1 The second arrow indicated the maximum number of tablets recommended by hospital and state guidelines. 2 The patient was then asked to choose the number of opioid tablets on this scale deemed to be the appropriate prescription size. Providers were educated to utilized the decision aid with patients scheduled for either inpatient or outpatient surgeries. The training phase of the quality improvement initiative was conducted from February 1, 2018 to February 28, 2018 during which time the decision aid and the workflow for this project were revised in response to direct feedback from patients, families, preoperative staff, and providers. The implementation phase was conducted from March 1, 2018, through July 31, 2018, and included all consecutive patients who underwent hysterectomy for benign indications at our institution (post-decision aid cohort). We excluded patients with a cancer or obstetric indication, as well as any postoperative complication that significantly altered postoperative recovery or resulted in length of stay greater than 7 days. On June 1, 2018, the state of Michigan enacted a law that required standardized counseling regarding opioids for any new opioid prescription. 23 While the initial decision aid included the majority of these required talking points, the tool was revised to include three supplemental items: 1 ) the potential adverse effects to a fetus or neonate with maternal opioid use disorder, 2) the felony charge associated with distributing opioids to anyone other than the person to whom the medication was prescribed, and 3) information regarding proper disposal of leftover opioids. This modification allowed for the front side of the decision aid to comply with all Michigan state law requirements and streamline resident physician preoperative workflow. Quality assurance was ensured via multiple modalities. First, during the training phase, residents and faculty alike were educated regarding the opioid epidemic and the planned shared decision making intervention using an informational video presented at a departmental meeting and then made available online. Next, residents were trained on the use of the decision aid during protected education time. Third, an email was sent the day before each scheduled hysterectomy to the resident or fellow scheduled to assist in the case. The email included a link to a video demonstrating proper use of the decision aid ( Video 1 , available online at http://links.lww.com/xxx ), and each trainee was asked to respond to the email following the case indicating whether the decision aid was used. Periodic unscheduled in-person assessments of the decision aid being used at the bedside were performed, with each trainee being observed using the decision aid and given feedback at least once. The patients included in the post-decision aid cohort were contacted via telephone 14 days postoperatively. If we were unable to reach a patient, two additional attempts were made on subsequent days. Telephone interviews were conducted using a standardized script and inquired about the patient’s postoperative pain experience, use of opioids postoperatively, need for refills, number of leftover tablets, and recollection of and feelings towards the decision aid that was used in the preoperative holding bay. All questions regarding patient experiences with postoperative pain management were modeled on a five-item Likert scale. All opioid data in this study was converted to oral morphine equivalents (OME) using the guidelines published by the Centers for Disease Control and Prevention in order to simplify comparison between different opioid formulations. 24 A standardized case report form was used to abstract patient and surgical characteristics from the electronic medical record for all patients who met inclusion criteria during the post-decision aid period. Data was abstracted by a gynecologist (A.V. or S.A.) or trained study personnel (M.S. or C.C.) under the frequent supervision of a gynecologist (A.V. or S.A.). Demographic and medical history included age (years); history of major medical comorbidity (hypertension, diabetes mellitus, other); current use of medication for depression, anxiety, or neuropathic pain; chronic opioid use; smoking status; surgical history including history of surgery for pelvic pain or endometriosis; body mass index (BMI), and American Society of Anesthesiologists physical status classification. Indications for hysterectomy were determined by review of the outpatient clinic notes and operative note, and patients could have multiple indications. Route of surgery was categorized as abdominal, vaginal, or laparoscopic. The laparoscopic group included both robotic and traditional laparoscopic approaches. Laparoscopic-assisted vaginal hysterectomies were included in the vaginal group, as the majority of these cases at our institution are performed in conjunction with uterovaginal prolapse repair. Additional procedures at the time of hysterectomy were noted. Finally, the type of opioid, dose, and total number of pills prescribed on the day of discharge (primary outcome) and within 14 days postoperatively were obtained from the electronic medical record. In order to compare postoperative opioid prescribing patterns prior to and after the implementation of the decision aid, we retrospectively identified all patients who underwent hysterectomy for benign, non-obstetric indication at our institution between July 1, 2017 and February 28, 2018 (pre-decision aid cohort). Of note, this cohort did not include the patients from which average opioid consumption after hysterectomy was used to develop the decision aid. For this pre-decision aid cohort, we used a standardized case report form to abstract patient and surgical characteristics, including surgical approach to hysterectomy, additional procedures, OME prescribed at discharge, and opioid refills obtained through our institution within two weeks postoperatively. As 2-week phone calls were not performed in this cohort, refill requests were based on review of the hospital record and did not include self-reported prescriptions obtained outside of the institution. Comparisons between the pre-decision aid cohort and the post-decision aid cohort were made based on data obtained by chart review. We performed a subgroup analysis of the post-decision aid cohort to examine the relationship between the amount of tablets selected and perioperative patient characteristics. Patients were divided into three groups (low, medium, and high) according to the amount of opioid tablets chosen from the presented opioid scale with use of the decision aid. The medium group was defined as those who selected the average number of opioids used for the surgical approach, the high group consisted of those who chose close to or the maximum permitted, and the low group was defined as those who selected well below the average. For example, for patients who were discharged with oxycodone, this equated to the low group including all patients who chose from 0 to 6 tablets, the medium group including all patients who chose from 7 to 13 tablets, and the high group including all patients who chose from 14 to 20 tablets. We then analyzed these groups according to preoperative patient characteristics, indication for surgery, postoperative opioid use, and patient satisfaction. REDCap was utilized for data compilation and data analyses were performed using STATA 14.0. Descriptive data were analyzed using chi-squared or Fisher exact test where applicable for categorical variables, Student t -test or one-way analysis of variance for normally distributed continuous variables, or by the nonparametric Wilcoxon rank-sum test if the distribution was non-normal. Descriptive statistics on continuous variables without significant skewness were reported as mean and standard deviation, while statistics on skewed variables were reported as median and interquartile range. Any questions not answered by patients during the phone calls were excluded from the data analysis for those specific questions. We followed STROBE reporting guidelines for observational studies. 25

During the study period, the post-decision aid cohort included 170 patients who underwent hysterectomy for benign indications with 159 (93.5%) patients participating in the use of the decision aid (93.5%). The medical record of the 11 patients for whom the decision aid was not utilized were reviewed and were eligible, but the resident reported the decision aid was not used. One patient who utilized the decision aid was excluded from the analysis due to significant perioperative complications, including unplanned small bowel resection with postoperative ileus and prolonged hospital stay of eight days. One hundred ten (69.6%) cases were performed laparoscopically, 40 (25.3%) performed vaginally, and 8 (5.3%) performed abdominally. Telephone surveys were completed for 141 (89.2%) participants ( Figure 2 ). While our study included small sample sizes, there were no significant differences in age or surgical characteristics between the pre-decision aid and post-decision aid cohorts ( Table 1 ). The proportion of patients who were discharged on the same day as surgery increased between the two groups, which aligns chronologically with implementation of same-day discharge protocols for minimally invasive hysterectomy in our department. Table 2 illustrates the differences in prescribing patterns between the two cohorts. The patients in the post-decision aid cohort were discharged with significantly fewer OME than the patients in the pre-decision aid cohort (92 ± 35 versus 160 ± 81, P <.01). The proportion of patients who requested refills in the first two weeks after surgery in the post-decision aid cohort was higher than the pre-decision aid cohort, but this did not reach statistical significance (post-decision aid cohort: n=15 (9.5%); pre-decision aid cohort: n=14 (5.7%); P =.15, Figure 3 ). Chronic opioid users (n=6, 40%) were more likely to request refills as compared to non-chronic opioid users [n=7 (6.2%), p<.001]. Tobacco users (n=5, 41.7%) were more likely to request refills as compared to non-tobacco users [n=8, (6.9%), p<.001]. Characteristics of opioid prescribing and opioid use in the post-decision aid cohort only are summarized in Table 3 . Most patients (n=121, 76.6%) chose fewer than the guideline-allotted maximum—an average of 49.4 ± 33.5 OME less than the maximum permitted prescription per patient. This equated to approximately seven fewer oxycodone 5 mg tablets per patient. 75.6% of patients (n=102) reported having leftover tablets, and 28.3% of patients (n=39) reported using no opioids postoperatively. Over half (54.6%) of patients reported discontinuing the medication due to manageable or no pain (n=77). 97.6% of patients reported their postoperative pain control as adequate or good (n=121), and 63% of patients reported that their postoperative pain was better or much better than they expected (n=78). The median days of opioid use postoperatively varied based on surgical route, with five days for laparoscopic cases, two days for vaginal cases, and 10 days for abdominal cases ( P <.01). 73% (n=81) of patients reported knowledge of proper disposal of leftover opioid tablets. Among patients who recalled the use of the decision aid, 73% of them found it to be helpful or extremely helpful. In the post-decision aid cohort, we examined clinical characteristics stratified by number of opioid tablets chosen (low, medium, high) with use of the decision aid. Of the 17 patients identified with chronic opioid use prior to hysterectomy, one patient chose a low number (5.9%), six patients chose a medium number (35.3%), and 10 patients chose a high number (58.2%), P =.12. Excluding patients with chronic opioid use, current use of a neuromodulating medication, antidepressant, or benzodiazepine was not associated with the amount of opioids chosen (low, medium, or high), nor was smoking status, obesity (defined as BMI ≥35), or American Society of Anesthesiologists class. Neither the indication for hysterectomy nor a patient’s surgical history was correlated with number of tablets chosen ( Appendix 1 , available online at http://links.lww.com/xxx ). Table 4 illustrates postoperative characteristics within the post-decision aid cohort stratified by number of opioid tablets chosen (low, medium, high). There was no significant difference in refill requests among the three groups. Patients who chose a high number of opioids with which to be discharged were significantly more likely to have used more opioids at the time of the 2-week phone call as compared to those who chose a low or medium number (low: 16.7 ± 17.1 OME, medium: 30.5 ± 35 OME, high: 73 ± 77.8 OME, P <.01). The difference remained significant when stratified by laparoscopic and abdominal route of surgery, but not vaginal route of surgery. For a laparoscopic hysterectomy, this equates to an average difference of six more tablets of oxycodone 5 mg used per patient choosing a high number of tablets as compared to a patient choosing a medium number of tablets. Regardless of number of opioids chosen, patients were similarly satisfied with their postoperative pain experience, and over 96% of patients in each group reported adequate or good pain control after surgery. The majority of patients across all three groups characterized the use of the decision aid as helpful or extremely helpful and endorsed knowledge of proper disposal of leftover medications. Results were similar when subgroup analysis was performed excluding those with chronic opioid use.

In this prospective quality initiative study, shared decision making with utilization of a simple decision aid was associated with a significant decrease in the number of opioids prescribed after hysterectomy for benign indications without a significant increase in refill requests. Despite fewer opioids dispensed, the majority of patients characterized their postoperative pain control as better or much better than expected, as well as the use of the decision aid as helpful or extremely helpful. There is ample evidence that excess prescribing of postoperative opioids has contributed to the opioid crisis, 8 , 26 and in response to this, the development of evidence-based guidelines has effectively decreased the amount of excess opioids prescribed in this clinical scenario. 2 , 9 , 27 Promising attention has recently been paid to the potential that shared decision making may have in decreasing this excess even further. A pilot study illustrated that the use of a decision aid after cesarean section resulted in an approximately 50% decrease in opioid prescribing, 19 which was sustained at a 35% decrease when modified to the department-wide level. 18 Our current study suggests that this positive impact can translate to gynecologic surgery as well with an important difference being that the majority of hysterectomies performed in the current surgical setting are performed on an outpatient basis. Hence, our model of shared decision making illustrates that, despite not having the opportunity to assess a patient’s postoperative pain requirements while inpatient, engaging patients in the conversation preoperatively regarding their postoperative pain management plan is an effective way to decrease excess opioid prescribing. In our study, patients who chose a high number of opioids used significantly more medication in the two weeks postoperatively as compared to patients who chose a medium or low number. This is consistent with existing data from other surgical procedures showing that opioid prescription size correlates with postoperative opioid consumption. 5 , 28 This may be a demonstration of the anchoring and adjustment heuristic, which is frequently associated with portion size in the food industry. 29 , 30 In this phenomenon, the initial amount provided serves as the anchor to which the corresponding intake adjusts. This concept presents both a challenge and an opportunity relative to opioid abuse. In surgical patients, the total duration of opioid use has been shown to be the strongest predictor of misuse, 31 and in nonsurgical patients, higher doses are correlated with an increased risk of overdose. 32 Thus, efforts to decrease prescription dose and duration may be an important tool in the larger strategy to combat the opioid epidemic, and additional research is needed regarding impact of prescription size directly on opioid consumption. Increasing disposal of leftover opioid tablets are primary focuses of multiple national health initiatives, 2 , 33 as one of the most common sources of misused opioids remains prescriptions originally intended for family or friends. 7 A large systematic review of six studies encompassing over 800 patients following seven different types of surgical procedures found that 67–92% of postoperative patients reported unused opioids, but only 4–30% had a plan to dispose of them. 8 While our study did not explicitly elicit patient plans for disposal, it is encouraging that 73% reported knowledge of proper disposal of leftover opioids. Future research should explore whether patient knowledge of proper disposal translates into actual disposal of excess opioids. Strengths of the study include prospective data collection, successful implementation of an individualized plan for pain control after hysterectomy for benign indication, and the simple design of our decision aid with minimal impact on preoperative workflow—increasing its practicality as a sustainable effort in reducing excess perioperative opioid prescribing. Despite these strengths, our study has several limitations. Data from our large, academic hospital may not be generalizable to other populations and hospital settings. In particular, our institution allows providers to refill opioids electronically without an in-person visit when deemed appropriate. Hence, patients are reassured that if they choose fewer opioids than they may need, the option for a refill is likely easily accessible. Our study also included few abdominal hysterectomies, which limits interpretation in this cohort. Data pertaining to the pre-decision aid cohort was gathered retrospectively through chart review alone. Thus, secondary outcomes related to opioid consumption and patient satisfaction were not available for comparisons. While our sample size of patients in the post-decision aid cohort who requested refills is extremely small (n=15), our data suggests there may be a correlation between certain preoperative factors and request for refills; future research should prioritize investigation of this in an effort to further personalize postoperative pain management. There is also the potential that a level of response bias exists with patients desiring to select a number of tablets deemed acceptable to the surgical team. The majority of patients, however, were satisfied with their postoperative pain experience and characterized the use of the decision aid as helpful; hence, it does not appear that this greatly impacted the results of the study. Finally, opioid use after discharge was based on self-report as we do not routinely have patients return to clinic for a pill count. We have a responsibility to develop effective strategies to ensure adequate management of postoperative pain without contributing to the epidemic of opioid abuse, and appropriate pain management is a crucial aspect of improving patient outcomes through enhanced recovery after surgery protocols. 34 As we continue to understand the factors that are associated with increased opioid use or misuse after surgery, perhaps one of the most important ways we can address the opioid epidemic is by directly engaging our patients in shared decision making in order to develop personalized postoperative pain management plans. 35 This quality improvement initiative lends further support to the fact that the use of simple decision aids can greatly reduce opioid prescribing without compromising patient satisfaction or postoperative pain.

Opioid abuse has reached epidemic proportions in the United States with more than 11 million people abusing prescription opioids in 2016. 3 42% to 71% of opioids prescribed after surgery are not used, leaving leftover opioids at risk for community distribution. 1 , 4 – 8 Across surgical specialties, this has prompted the development of prescribing guidelines in an effort to decrease the flow of excess prescription opioids, 2 , 9 , 10 but there are few guidelines available for benign gynecology and hysterectomy in particular. The Michigan Opioid Prescribing Engagement Network (M-OPEN) has developed evidence-based guidelines for this indication; however, these guidelines are based on self-reported opioid consumption of the 75th percentile of patients. 2 , 11 Hence, more opioids are prescribed than are used by the average patient even when utilizing these guidelines. Furthermore, this strategy is not personalized according to patient preference or prior experience with opioids. Shared decision making is a process by which both patients and physicians share information, express treatment preferences, and agree on a treatment plan. 12 Decision aids are tools designed to support this. 13 – 15 The positive effect of these tools has been demonstrated in obstetrics and gynecology, most recently in opioid prescribing after cesarean section. 16 – 19 We sought to apply shared decision making to decrease excess opioid prescribing at the time of hysterectomy. As hysterectomy is transitioning to the outpatient setting, 20 we developed a patient decision aid that can be implemented in the preoperative holding area. Our primary objective was to determine if this process reduced opioid prescribing after hysterectomy as compared to prior to decision aid implementation.