Definition development and prevalence of new persistent opioid use following hysterectomy.

Using data from a single national private health insurer, we propose a data-based definition of new persistent opioid use among opioid naïve women undergoing hysterectomy based on three criteria in the six months after surgery: 1) number of fills ≥ 2; 2) ≥ 1 fill every three months; and 3) ≥ 1,150 mg OME or ≥ 39 days’ supply. Based on this definition, the prevalence of new persistent opioid use in our study was 0.5%. On average, new persistent opioid users received just over two 5-mg hydrocodone tablets per day (345 tablets/165 days) during the 6-month postoperative period. Our study extends the literature by providing a data-driven definition for new persistent opioid use specific to hysterectomy, which was previously lacking. Prior studies have reported the prevalence of new persistent opioid use following surgical procedures; however, the criteria used to define new persistent opioid use were determined a priori , included a variety of surgical procedures, or used definitions derived from a non-surgical cohort. 4 – 6 A study by Brummett et al did include hysterectomy as one of the surgical procedures; they reported the incidence of new persistent opioid use—defined as one or more opioid prescription filled between 90–180 days postoperatively—to be over 5% in this group, which is 10 times higher than what we report. Our definition was purposefully more stringent in an effort to more accurately identify persistence of use by requiring at least one opioid prescription in the early postoperative time period (15–90 days) and also in the later postoperative time period (91–180 days). Another study by Sun et al used a more conservative definition of new “chronic” opioid use—defined as 10 or more fills or > 120 days’ supply in the first postoperative year—for 11 surgical procedures (not including hysterectomy) and reported prevalences more consistent with our results, ranging from 0.119% after Cesarean delivery to 1.41% after total knee arthroplasty. 5 However, total OMEs for the same number of prescriptions or days’ supply can vary widely depending on the specific opioid and dose prescribed. Therefore, our definition accounts for this possible variation by also including criteria for OME quantity. Furthermore, our criteria used for days’ supply and OME quantity were derived from the distribution of use in our cohort and are specific to hysterectomy. We identified two potentially modifiable risk factors for new persistent opioid use following hysterectomy: abdominal route of hysterectomy and preoperative opioid fill. Nearly one-third of hysterectomies in the cohort were done abdominally. While this is an improvement from prior studies showing a national abdominal hysterectomy rate of over 50%, 1 opportunity still exists to decrease utilization of the abdominal approach. Abdominal route of hysterectomy is a known risk factor for increased postoperative pain and opioid use. 9 , 10 Some indications for hysterectomy, like malignancy, may preclude the use of a minimally-invasive route. In our study, hysterectomies done for gynecologic malignancy versus other indications were more often performed abdominally (54.6% vs 27%, p <.0001). Despite this finding, overall, only 1.4% (90/6,637) of abdominal hysterectomies were done for a gynecologic cancer indication. Our results suggest that efforts to increase utilization of minimally-invasive hysterectomy may help decrease the prevalence of new persistent opioid use after hysterectomy. The other potentially modifiable risk factor we identified was a preoperative opioid fill within 30 days of the hysterectomy. We included women with opioid prescriptions in this time period, as some surgeons may provide the prescription for postoperative pain medications during a preoperative visit; however, we are unable to know the exact indication for the preoperative opioid prescription. We know that chronic opioid use prior to surgery increases risk of chronic opioid use after surgery; 11 – 13 however, our finding suggests that even receiving an opioid prescription within a month of surgery increases the odds of persistent use postoperatively. Based on this finding, we recommend against the routine practice of providing preoperative opioid prescriptions, as well as avoiding starting opioid naïve patients on opioids prior to hysterectomy, if possible. The remaining factors we identified to be associated with increased odds of new persistent opioid use were increasing age, gynecologic malignancy, and depression/anxiety—all of which have been shown to have precedence in the literature. 4 , 5 , 14 The main limitations of the study are in those inherent with using a large administrative database. We used pharmacy claims data to estimate opioid use; however, we were unable to assess actual opioid consumption, use of over-the-counter pain medications, or prescription neuropathic pain medications. We were also unable to determine the indication for the opioid prescriptions. We did not have access to potentially relevant clinical data such as uterine size or specimen weight. Our study population was limited to opioid naïve women < 64 years old with a single national private insurer; therefore, our definition for new persistent opioid use may not be valid when applied to women ≥ 64 years old, the uninsured, or those with other types of insurance, or to women with chronic opioid use prior to hysterectomy. We were unable to control for patient-specific pain sensitivity or predisposition to opioid dependence, which may have confounded our results. Finally, as our analysis is limited to the first six months after hysterectomy, we cannot draw any conclusions regarding long-term adverse clinical outcomes related to our definition of new persistent opioid use. Strengths of our study include steps taken to limit the study cohort to opioid naïve women without additional surgical procedures in the study time period. We also used robust analyses to control for potential clinically relevant variables and regional variation in practice patterns. In summary, we propose a data-driven definition for new persistent opioid use after hysterectomy in opioid naïve women. Based on our definition, the prevalence of new persistent opioid use in this population is low; however, two potentially modifiable risk factors are preoperative opioid prescription and abdominal route of hysterectomy.

Overall, there were 54,976 hysterectomies performed for women with continuous insurance enrollment during the study period. Of these, 18,707 were excluded for age ≥ 64 or age ≤ 17, 4,321 were excluded for having other surgical procedures during the study period, and 7,617 were excluded for having a preoperative opioid fill between 243 and 31 days prior to surgery. The final cohort included 24,331 women for analysis ( Figure 1 ). Overall, the median total dose of opioids given in the perioperative period (30 days before to 14 days after hysterectomy) was 225 mg OME (IQR 150, 300), which is equivalent to 30 5-mg oxycodone pills or 45 5-mg hydrocodone pills. Of those who met the fill criterion of our definition, median total dose of opioid prescriptions in the 6-month postoperative period was 537.5 mg OME (IQR 300, 1150) and median days’ supplied was 17 (IQR 10, 39). Therefore, based on the top quartile values for both variables, we defined “new persistent opioid use” as women with a total dose of ≥ 1,150 OMEs or days’ supplied ≥ 39 and ≥ 2 opioid fills, with at least one fill each during the 15–90 day and the 91–180 day postoperative time periods. Based on this definition, 0.50% (122/24,331) met criteria for new persistent opioid use following hysterectomy. Among new persistent opioid users, median total opioid dose prescribed during the 6-month postoperative period was 1,725 mg OME (IQR 1200, 3000), which is equivalent to 230 5-mg oxycodone pills or 345 5-mg hydrocodone pills. Table 1 shows bivariate comparisons of demographics, and perioperative variables between new persistent and non-persistent opioid users following hysterectomy. Women with new persistent opioid use were older and a higher proportion were African-American; however, education level and division of the country were similar when compared to the non-persistent users. One in four new persistent users had a preoperative diagnosis of depression/anxiety—a rate over twice that seen in non-persistent opioid users; however, rates of gynecologic pain diagnoses were similar. Over half of hysterectomies in the new persistent users group were performed via the abdominal route—a rate nearly twice that seen in the non-persistent users. Conversely, vaginal hysterectomy was half as common in the new persistent versus non-persistent group. Indications for hysterectomy differed between groups, with the new persistent opioid users having a higher prevalence of gynecologic malignancies and lower prevalence of benign indications ( Table 2 ). Table 3 shows comparisons of number of prescription fills and OMEs at the different time intervals during the study period. Overall, 84.78% of women had an opioid fill during the perioperative period. One in three women in the new persistent group filled an opioid prescription in the 30 days prior to hysterectomy, which was twice the rate seen in the non-persistent group. Of women with a preoperative diagnosis of chronic pelvic pain, cyclic pelvic pain or dyspareunia, 19.63% (1,319/6,721) had a preoperative opioid fill compared to 15.91% (2,801/17,610) of women without these diagnoses (p <.0001). Using variables significant in bivariate analyses, a hierarchical logistic regression model was developed to examine factors independently associated with new persistent opioid use while controlling for regional variation ( Table 4 ). Gynecologic malignancy as the indication for hysterectomy was the factor most strongly associated with new persistent opioid use and increased the odds seven and a half-fold. Abdominal route of hysterectomy was associated with a 360% increased odds of new persistent opioid use compared to the vaginal route, while no significant associations were seen between vaginal and other minimally-invasive routes. Women who filled an opioid prescription within 30 days prior to hysterectomy had nearly a three-fold increased odds of being a new persistent opioid user, and preoperative diagnosis of depression/anxiety increased the odds by 260%. For every increased year of age, odds of new persistent opioid use increased by 4% and African-American women had a 60% increased odds of becoming a new persistent user. C-statistic for the model was 0.74. Analysis of hysterectomy route and gynecologic malignancy indicated that a significantly higher proportion of abdominal hysterectomies were done on patients with gynecologic malignancy. Since route of hysterectomy and gynecologic malignancy were included in the final model, we subsequently assessed if these variables were collinear with each other by analyzing the Spearman Rank correlation, which was found to be not significant ( ρ = −0.05). Furthermore, we fit an alternative model with hysterectomy route and gynecologic malignancy as an interaction term to understand if hysterectomy route is an effect modifier. The interaction term was not significant in multivariable analysis; therefore, both hysterectomy route and gynecologic malignancy were included as independent effects.

We performed a retrospective analysis of data from Optum Clinformatics Data Mart, a national database that includes both medical and outpatient pharmacy data for over 73 million individuals from a single private health insurer in the United States. The study received exempt status from the University of Michigan IRB (HUM00132182). Hysterectomies performed from January 1, 2011 to December 31, 2014 were identified using Current Procedural Terminology (CPT) and International Classification of Diseases, Ninth Revision (ICD-9) codes. We included women 18–63 years old with continuous insurance enrollment from eight months prior to hysterectomy through the 6-month post-hysterectomy time period. Women 64 years and older were excluded as they could have become eligible for Medicare in the post-hysterectomy time period and in order to avoid confounding due to multiple plans, we wanted to restrict our population to women with a single insurance plan during the study period. Women with obstetrical indications for hysterectomy were also excluded. To identify an “opioid naïve” cohort, we excluded women with any opioid fills from 243 days to 31 days prior to hysterectomy. We did not exclude women with an opioid fill within 30 days prior to hysterectomy, as some surgeons may provide opioid prescriptions for postoperative pain management during this time. Women who underwent additional surgical procedures in the six months following hysterectomy, identified using anesthesia CPT codes, were also excluded ( Supplemental Table 1 ). The “perioperative period” was defined as 30 days prior to 14 days post-hysterectomy and was considered the time period when opioid prescriptions would be expected as part of routine management of postoperative pain. 4 Thirty days prior to hysterectomy was defined as the “preoperative period”, and the time from 15–180 days following the hysterectomy was considered the “6-month postoperative period”. Opioids were converted to oral morphine equivalents (OMEs). Number of opioid prescription fills, days supplied, and total OMEs were analyzed during the preoperative, perioperative, and 6-month postoperative time periods to determine the distribution of these data. Opioids were identified as medications with a therapeutic class of either “Opioid Agonist” or “Opioid Partial Agonist.” OME conversions were calculated using an opioid equivalent look-up. 7 The definition of “persistent opioid use” was based on data from the 6-month postoperative period. We analyzed patterns of opioid fills by month and by quarter, and corresponding distributions of total OMEs and days’ supply. These data were then used to determine cut-offs for our definition. Three criteria were used to develop the definition: 1) number of opioid fills, 2) persistence of use, and 3) quantity of opioids filled (total OMEs and days supplied). To meet the first two criteria, a minimum of two opioid prescriptions had to be filled with at least one in the 15–90 day postoperative time period (first quarter) and at least one in the 91–180 day postoperative time period (second quarter). Women who met this criterion had higher average number of days’ supply in the postoperative period compared to those with ≥1 fill in either quarter or ≥1 fill in any postoperative month. This criterion also establishes a definition for persistence and therefore avoids classifying those with two opioid fills in the first quarter but none in the second quarter as persistent users. Determination of the third criterion—quantity of opioids—was based on the top quartile values of total OMEs and days’ supply for those who met the first two criteria. We chose the top quartile for this criterion in order to identify women whose opioid use was disproportionately higher than average, which subsequently avoids classifying someone as a new persistent user who had one fill per quarter but only received five tablets of hydrocodone with each fill. We obtained demographics including age, race, educational level, and division of the country according to the U.S. Census Bureau and used ICD-9 diagnosis codes to identify hysterectomy indications and pre-hysterectomy comorbidities present within the eight-month pre-hysterectomy time period. Gynecologic and non-gynecologic chronic pain conditions were identified as well as potentially relevant psychiatric diagnoses. The comprehensive list of codes we used for the analyses is shown in Supplemental Table 2 . Bivariate analyses were used to compare new persistent to non-persistent opioid users. For categorical variables, either Chi-Square tests, or Fisher’s Exact Test in the case of small numbers, were performed. Variables that were considered both statistically and clinically relevant as preoperative or intraoperative predictors of new persistent use were considered in a stepwise logistic regression model. As published data show regional variation in opioid prescribing practices after hysterectomy, 8 a hierarchical logistic regression model with U.S. Census Division random effects to control for regional variation was developed, with independent factors associated with new persistent opioid use following hysterectomy. Collinearity between candidate covariates was assessed using Pearson’s or Spearman Rank correlation matrices. Confounding between variables was tested with an interaction product term to assess potential significance. All statistical analyses were performed using SAS Version 9.4 (Copyright 2012, SAS Institute, Inc.).

Hysterectomy is the most common major gynecologic procedure in the United States, with more than 400,000 women per year undergoing this surgery. 1 While opioids are routinely used for postoperative pain control after hysterectomy, increasing attention is being paid to prescribing practices due to the epidemic of opioid abuse and overdose in the United States. 2 Data suggest that opioids prescribed after surgery can lead to chronic use or abuse, even among opioid naïve individuals. 3 The prevalence of new persistent opioid use among opioid naïve patients following surgery ranges from 0.12% to 6.5% and varies depending on the definition used and also by surgical procedure. 4 – 6 The criteria used to define “persistent” or “chronic” opioid use are often chosen arbitrarily and studies looking at chronic opioid use after surgery often include multiple different procedures. Despite the urgent need to better understand the impact of opioid prescribing, we currently lack an evidence-based definition of new persistent opioid use specific to hysterectomy. Therefore, the primary aim of this study was to develop a hysterectomy-specific definition of new persistent opioid use among opioid naïve women. Secondarily, we sought to determine the prevalence of and risk factors for new persistent opioid use in women undergoing hysterectomy.

Supplemental Table 1. Anesthesia Procedures Codes Used to Exclude Women With Additional Surgical Procedures Within Six Months Following Hysterectomy Supplemental Table 2. Current Procedural Terminology (CPT) and International Classification of Diseases, Ninth Revision (ICD-9) Codes Used for Analyses