The impact of opioid safety reports on racial and ethnic differences in post- operative opioid prescribing practices at hospital discharge: exploratory analysis of a clustered randomized controlled trial

doi:10.21203/rs.3.rs-4613746/v1

Download PDF

Research Article

The impact of opioid safety reports on racial and ethnic differences in post- operative opioid prescribing practices at hospital discharge: exploratory analysis of a clustered randomized controlled trial

https://doi.org/10.21203/rs.3.rs-4613746/v1

This work is licensed under a CC BY 4.0 License

You are reading this latest preprint version

Background:

The objective of this study was to examine differences in postoperative opioid prescribing by race and ethnicity before and after the implementation of opioid safety reports, via a randomized control trial (RCT), to reduce guideline-discordant opioid prescribing.

Methods:

This exploratory analysis of an RCT used three years of data (October 2019-October 2022), two before the intervention and one during the intervention, across 19 hospitals and three surgical specialties: general surgery, obstetrics/gynecology, and orthopedic surgery. The analysis included patients ≥ 18 years of age, with known race/ethnicity (Hispanic [any race], or non-Hispanic, Asian, Black, Other or white [NHA, NHB, NHO, or NHW]), who underwent procedures by surgeons who were randomized to receive monthly opioid safety reports (treatment arm) or not (control arm). Safety reports were based on social norms, comparing surgeons’ prescribing to guideline recommendations s or their peers. The primary outcome was total morphine milligram equivalence (MME) prescribed per patient at hospital discharge. Linear regression models were used to examine differences in prescribed total MME by race/ethnicity between control and treatment arms, before and after intervention implementation, with statistical adjustment for numerous patient, procedure, and clinician characteristics.

Results:

Among 55,637 patients meeting study eligibility criteria, overall trends showed decreases in opioid prescribing over time, which were more pronounced among racial/ethnic minority patients compared to NHW patients. Independent of randomized assignment, NHA and Hispanic patients received on average lower opioid quantities than NHW patients within each year, whereas NHB patients received higher opioid quantities in the first and second year and lower opioid quantities in the third year. The intervention, itself, did not impact prescribing differences by race/ethnicity across surgical specialty combined. Stratified by speciality, the intervention appeared to reduce differences in opioid prescribing for NHB vs. NHW patients undergoing obstetrics/gynecological procedures by an average of -20.8 total MME (95% Confidence Interval: -40.1, -1.5).

Conclusions:

Opioid safety reports, based on social norms, showed little impact on reducing gaps in opioid prescribing by race/ethnicity. Future research is needed to identify strategies to mitigate racial/ethnic postoperative opioid prescribing differences at hospital discharge.

Our study found that opioid safety reports designed to promote guideline-based postoperative prescribing at hospital discharge after common surgical procedures did not materially impact differences in opioid prescribing by race/ethnicity.
These findings were potentially due to temporal decreases in postoperative opioid prescribing quantities, which were more pronounced for patients belonging to racial/ethnic minority groups than non-Hispanic white patients.
As a result, differences in opioid prescribing by race/ethnicity were perpetuated, whereby Asian and Hispanic patients receive on average lower opioid quantities at hospital discharge than non-Hispanic white patients.
Future research is needed to promote more equitable opioid prescribing practices.

Postoperative opioid prescribing for at-home pain management has been a contributor to the opioid crisis in the United States (US), as excess prescribed pills can lead to prolonged opioid use or drug diversion, increasing the risks for misuse, abuse, and addiction. Prior work has described these risks in detail, showing that postoperative prescription quantities are strongly associated with patient consumption [1] that patient consumption in the postoperative period successively predicts persistent opioid use [2] And that persistent use is, in turn, associated with increased hazards of developing opioid use disorder and overdose [3, 4].

These risks have not been evenly distributed across racial/ethnic groups. In earlier years of the opioid crisis, risks were concentrated among white communities, in part because of false notions about pain tolerance and illicit drug use among people of color, which prevented them from receiving similar quantities of opioids as white individuals [5–7]. However, opioid related mortality rates have been increasing among Black people and in some cases now exceed mortality rates among white people [8, 9].

In a retrospective analysis of electronic health records (EHR) data from a large healthcare system in northern California between January 2015 and February 2020, we found that non-Hispanic Black (NHB) patients received significantly larger postoperative opioid quantities at hospital discharge among high-volume surgeries examined than non-Hispanic white (NHW) patients after controlling for relevant factors [10] At the same time, we found that Hispanic and non-Hispanic Asian (NHA) patients received significantly smaller opioid quantities than non-Hispanic white patients, suggesting that racial/ethnic differences in prescribing are a complex issue.

Importantly, our previous retrospective analysis also showed that differences in opioid prescribing among Hispanic and NHB patients compared to NHW patients were not observed for prescriptions that were concordant with surgical procedure-specific guideline recommendations. These findings are consistent with another study of general surgical procedures from an academic medical center [11]. Our findings, and those of Herb and colleagues, are likely attributable to the fact that prescribing guidelines offer a narrower range of opioid pill quantities. It follows, then, that interventions to reduce opioid guideline-discordant prescribing could mitigate racial/ethnic prescribing differences as an unintended but secondary benefit.

In the present study, we examined differences in postoperative opioid prescribing by race and ethnicity before and after the implementation of an intervention to reduce guideline-discordant opioid prescribing. The intervention was implemented via a clustered randomized controlled trial (RCT) from October 2021 to October 2022 and comprised email-based monthly opioid safety reports based on social norms that provided surgeons with feedback on their recent opioid prescribing practices in comparison to published guidelines or their peers. Both safety reports were effective in reducing the percentage of clinicians who prescribed above guideline recommendations and the average total morphine milligram equivalence (MME) of opioids prescribed [12]. We hypothesize that these effects, in turn, may reduce racial and ethnic-based prescribing differences.

Study design

This was a secondary analysis of an investigator initiated RCT conducted at 19 hospitals, across three surgical specialties (general surgery, obstetrics/gynecology [Ob/Gyn], and orthopedics) in a large healthcare delivery system in Northern California from October 19, 2021 to October 18, 2022. Details of the design and analysis plan for the RCT can be found in the published protocol [13]. This secondary analysis used data that were part of the primary analysis of the RCT, including two years of pre-intervention data from October 1, 2019 to October 18, 2021.

Data collection

All data were obtained from the Sutter Health EHR for the three-year period from October 1, 2019 to October 18, 2022. Data collected included hospital discharge prescriptions, patient demographics and comorbidities, procedure characteristics, and prescriber characteristics.

Intervention

Surgeons assigned to the treatment arms received one of two different types of opioid safety reports based on social norms, designed to provide feedback to surgeons on their post-operative prescribing performance at hospital discharge. One report informed surgeons how they were performing relative to published opioid prescribing guidelines (guideline-comparison report). The other report informed surgeons how they were performing relative to their high-performing peers (peer-comparison report), where "high-performing" was defined as those who conformed to guideline prescribing. Thus, both safety reports used the same performance standard, but they differed in how the standard was communicated. The difference between the two treatment arms is subtle and the effectiveness was similar [12] so we combined the two for the purpose of this analysis.

Randomization occurred at the level of the surgical specialty (general surgery, Ob/Gyn, and orthopedics) for the 19 hospitals. For each hospital, a specialty was randomized to one of the following arms, stratified by hospital size: monthly peer-comparison report, monthly guideline comparison report, or usual care (control arm), composing 57 randomized units in total. Surgeons were thus assigned to one of the three arms based on their specialty and location where they performed surgeries. For surgeons assigned to the treatment arms, monthly reports were delivered by email to surgeons who had at least two patients with guideline-discordant prescriptions in the previous month. Surgeons were not notified that they were part of a study to avoid a Hawthorne Effect. The intervention procedures are described in detail elsewhere [12].

Cohort eligibility

In this analysis, we included patients at least 18 years of age, who underwent a surgical procedure during the study time period and were discharged to home with an opioid prescription. We excluded patients for whom both race and ethnicity were missing and those who had a prescription with a total MME > 500 (as we considered this to likely be erroneous).

Outcome variables

The main outcome measure for this study was total MME prescribed at hospital discharge. MME is the product of the milligram-strength, pill quantity, and opioid analgesic morphine conversion factor [14]. A secondary outcome measure was guideline-concordant prescribing. Because opioid prescribing guidelines provide a number (or range) of 5-mg oxycodone pills, we converted these to total MME to compare across prescriptions of other opioids and/or mg-strengths.

Explanatory variables

The main explanatory variable was patient race/ethnicity. Race/ethnicity was operationalized, based on the U.S. Census categories, as Hispanic ethnicity, regardless of race; NHW; NHB; NHA; and non-Hispanic Other (NHO; which included Native Hawaiian, Pacific Islander, Native Alaskan, Native American, mixed race or self-reported ‘other’).

We also used time period and randomized assignment of the surgeon to assess how the racial/ethnic differences in opioid prescribed changed from before to after the start of the intervention. Time was binned into annual periods from October 1, 2019 to October 18, 2022 (three periods in total). The first two periods represented the pre-intervention years (each period was composed of 12 months), and the last period represented the 13 months that the intervention was active.

Other variables that could potentially influence differences in opioid prescribing and confound the relationship between total MMEs prescribed and race/ethnicity were included in the analysis. These include patient age (years), sex (male, female or other/unknown) and body mass index (kg/m²). Other potential confounding variables for patient, procedure and prescriber characteristics were also created as described below.

Indicator variables (y/n) were created for whether patients had evidence of diabetes, a pain disorder, an opioid prescription in the past 30 days, an opioid prescription in the past 1–12 months, a non-opioid analgesic in the past 30 days, chronic opioid use, an emergency department (ED) visit the past 30 days, and an inpatient visit in the past 30 days. Patients’ physical status, per American Society of Anesthesiologists classification, was categorized as healthy, mild systemic disease, severe systemic disease, incapacitating disease, or moribund [15]. Patient’s primary insurance type was classified as: commercial, Medicare, Medicaid, Worker’s Compensation, other, or missing. Variables were created to indicate whether patients had Medicaid (y/n) or worker’s compensation insurance (y/n) as either primary or secondary insurance.

Surgical procedures were classified based on the surgical specialty (general surgery, Ob/Gyn, orthopedics) and whether the case was elective, emergent, or urgent. In total, there were 35 surgical procedures across the three specialties that had opioid prescribing guidelines and were included in the analysis; a dummy variable was created to control for between-group differences in the distribution of these procedures across racial/ethnic groups. Length of hospital stay was calculated based on admission and discharge dates and was categorized as same-day discharge, 1–2 days, 3–4 days, or 5 or more days.

Opioid prescriber characteristics were also included as they have previously been shown to influence opioid prescribing practices. [16]. These characteristics include prescriber sex (male, female or unknown), years in clinical practice. sex-concordance between the patient and the prescriber (y/n), and whether the prescribing provider was different from the surgeon who performed the procedure (y/n).

Statistical analysis

Descriptive statistics were performed on all study variables. Continuous variables were summarized as means with standard deviations (SDs). Categorical variables were summarized as percentages.

We first examined prescribed total MME by race/ethnicity, for each year across the study period, using Ordinary Least Squares (OLS) regression. Specifically, we regressed an interaction term between race/ethnicity and year on total MME, with statistical adjustment for the abovementioned control variables, as well as a variable for time (months) from the first to the last month of the study period to account for overall temporal trends and a variable to indicate the month-of-year to account for potential seasonality in prescribing. This model can be expressed as follows:

$$\:{MME}_{it}=\:{\beta\:}_{0}+\sum\:_{k}{\beta\:}_{k}RaceEt{h}_{i}^{k}+\sum\:_{t}{\lambda\:}_{t}Yea{r}_{t}+\sum\:_{k}\sum\:_{t}{\tau\:}_{kt}\left(RaceEt{h}_{i}^{k}\times\:Yea{r}_{t}\right)+\varvec{X}\varvec{\rho\:}+{ϵ}_{it}$$

(1)

where MME is the total MME quantity prescribed at hospital discharge for patient $\:i$ in time $\:t$; $\:RaceEth$ represent race/ethnicity category $\:k$ (5-level variable, with NHW as the reference category); $\:Year$ represents the year in which the prescription was written (3-level variable, with Year 1 as the reference category); and X represents a set of control variables. The $\:{\tau\:}_{kt}$ coefficients estimate the difference in opioid prescribing between each race/ethnic minority group $\:k$ and the NHW group in each year $\:t$ relative to this difference in the first year. Standard errors were clustered by the units used for randomized assignment of the surgeon (57 clusters in total). From this model we estimated mean total MME prescribed for each race/ethnicity by year (i.e., marginal adjusted means), as well as the mean differences in total MME prescribed for each race/ethnic minority group versus the NHW group by year.

A logit model was used to estimate the odds that patients’ prescriptions were guideline concordant (yes/no), using the same modelling framework described above. From this model we estimated the percentage of guideline concordant prescription by race/ethnicity and year (i.e., predicted percentages).

We next examined the impact of the intervention on racial/ethnic differences in opioid prescribing by regressing an interaction term for race/ethnicity, randomized assignment of the surgeon, and time period on total MME prescribed, with statistical adjustment for the aforementioned control variables. This OLS regression model can be expressed as follows,

where $\:RaceEth$ represents race/ethnicity (as described above), $\:Post$ represents whether the time period was before or after the intervention (binary variable), and $\:Treat$ represents whether the surgeon writing the prescription was assigned to the treatment arm or the control arm. In this model, the $\:\beta\:$s and $\:\sigma\:$s inform what the disparity looked like prior to the intervention; the $\:\beta\:$s estimate the difference between each race/ethnicity category relative to NHW in the control group in the pre-period, and the $\:\sigma\:$s estimate how the racial/ethnic differences were different between the treatment and control groups in the pre-period. The main terms of interest are the $\:{\tau\:}_{k}$ terms, which estimate the difference in the change in racial/ethnic differences between the treatment and control arms from pre- to post-intervention, which can be interpreted as the causal effect of the intervention on racial/ethnic disparities. Standard errors were clustered as described above.

For the intervention to have had an impact on mitigating race/ethnic differences in opioid prescribing two criteria must be met:

1. A persistent difference in opioid prescribing in the control arm: statistically significant differences in mean total MME prescribed to a race/ethnic minority group compared to the NHW group in the pre-intervention period ($\:\beta\:$s in Eq. 2) that remained present among surgeons randomized to the control arm in the intervention period ($\:{\beta\:}_{k}+{\omega\:}_{k}>0$). This represents a pre-condition for the intervention to have an effect.

2. A reduction in the difference in opioid prescribing due to the intervention: a statistically significant higher-order interaction term ($\:{\tau\:}_{k}s)$, representing a larger reduction in the disparity in the treatment arm compared to the control arm; this change is considered a ‘reduction’ in the difference only if it is in the opposite direction of the persistent disparity in the control group, (i.e., 1, above). This represents the effect of the intervention.

For all regression models, we conducted stratified analyses for each of the three specialties. All analysis were conducted in State 17.0 (StataCorp; College Station, TX). A p < 0.05 was considered statistically significant.

Description of cohort

Among 117,370 patients with a surgical procedure during the three-year study period, 55,637 (47.4%) met cohort eligibility criteria. Approximately 53% of all eligible patients were NHW, 24% were Hispanic, 11% were NHA, 6% were NHB, and 6% were NHO.

Compared to NHW patients, racial/ethnic minority patients were on average younger and were more frequently male, were less likely to have a pain disorder, were more likely to have evidence of diabetes, and were more likely to be Medicaid or workers’ compensation beneficiaries (Table 1). Racial/ethnic minority patients more frequently underwent Ob/Gyn procedures and less frequently underwent orthopedic procedures than NHW patients; they also less often had an elective procedure and more often had a procedure requiring an overnight stay in the hospital. Other characteristics by race/ethnicity can be found in Table A1 (see Appendix).

Table 1

Cohort characteristics by race/ethnicity
	Total	Hispanic, Any Race	Non-Hispanic Asian	Non-Hispanic Black	Non-Hispanic Other	Non-Hispanic White
	N = 55,637	N = 13,101	N = 6,336	N = 3,448	N = 3,353	N = 29,399
Patient age, years (SD)	48.4 (17.9)	40.2 (15.1)	46.4 (16.4)	44.2 (16.5)	43.8 (16.5)	53.5 (18.0)
Sex
Male	15,289 (27.5)	2,776 (21.2)	1,274 (20.1)	696 (20.2)	793 (23.7)	9,750 (33.2)
Female	40,344 (72.5)	10,323 (78.8)	5,062 (79.9)	2,752 (79.8)	2,560 (76.3)	19,647 (66.8)
Other/Unknown	4 (0.0)	2 (0.0)	0 (0.0)	0 (0.0)	0 (0.0)	2 (0.0)
BMI, kg/m² (SD)	30.5 (27.9)	32.8 (37.9)	27.7 (5.6)	33.9 (9.1)	31.4 (7.6)	29.6 (28.4)
Pain disorder, n (%)	27,078 (48.7)	4,548 (34.7)	2,156 (34.0)	1,465 (42.5)	1,286 (38.4)	17,623 (59.9)
Diabetes, n (%)	5,020 (9.0)	1,303 (9.9)	649 (10.2)	459 (13.3)	331 (9.9)	2,278 (7.7)
Opioid in 1–12 months, n (%)	8,736 (15.7)	1,697 (13.0)	541 (8.5)	565 (16.4)	437 (13.0)	5,496 (18.7)
Opioid ≤ 30 days, n (%)	4,489 (8.1)	850 (6.5)	288 (4.5)	310 (9.0)	217 (6.5)	2,824 (9.6)
Non-opioid in ≤ 30 days, n (%)	5,351 (9.6)	1,188 (9.1)	453 (7.1)	326 (9.5)	299 (8.9)	3,085 (10.5)
Chronic opioid use, n (%)	2,382 (4.3)	436 (3.3)	152 (2.4)	179 (5.2)	108 (3.2)	1,507 (5.1)
Primary care patient, n (%)	16,898 (30.4)	3,351 (25.6)	1,655 (26.1)	810 (23.5)	890 (26.5)	10,192 (34.7)
ED visit, n (%)	3,280 (5.9)	913 (7.0)	261 (4.1)	317 (9.2)	230 (6.9)	1,559 (5.3)
Inpatient admission, n (%)	1,335 (2.4)	267 (2.0)	146 (2.3)	141 (4.1)	72 (2.1)	709 (2.4)
Primary Insurance, n (%)
Commercial	43,485 (78.2)	8,316 (63.5)	5,104 (80.6)	2,569 (74.5)	2,523 (75.2)	24,973 (84.9)
Medicare	1,913 (3.4)	343 (2.6)	229 (3.6)	156 (4.5)	108 (3.2)	1,077 (3.7)
Medicaid	4,789 (8.6)	2,617 (20.0)	356 (5.6)	301 (8.7)	349 (10.4)	1,166 (4.0)
Worker's Comp	29 (0.1)	14 (0.1)	1 (0.0)	0 (0.0)	3 (0.1)	11 (0.0)
Other	130 (0.2)	45 (0.3)	6 (0.1)	9 (0.3)	11 (0.3)	59 (0.2)
Unknown/Missing	5,291 (9.5)	1,766 (13.5)	640 (10.1)	413 (12.0)	359 (10.7)	2,113 (7.2)
Any Medicaid, n (%)	7,532 (13.5)	3,554 (27.1)	667 (10.5)	558 (16.2)	533 (15.9)	2,220 (7.6)
Any Worker’s Comp., n (%)	917 (1.6)	235 (1.8)	49 (0.8)	48 (1.4)	57 (1.7)	528 (1.8)
Procedure specialty, n (%)
General	25,044 (45.0%)	5,893 (45.0%)	2,865 (45.2%)	1,519 (44.1%)	1,401 (41.8%)	13,366 (45.5%)
Obstetrics/Gynecology	19,197 (34.5%)	5,723 (43.7%)	2,831 (44.7%)	1,507 (43.7%)	1,434 (42.8%)	7,702 (26.2%)
Orthopedics	11,396 (20.5%)	1,485 (11.3%)	640 (10.1%)	422 (12.2%)	518 (15.4%)	8,331 (28.3%)
Case type, n (%)
Elective	40,310 (72.5%)	8,224 (62.8%)	4,407 (69.6%)	2,394 (69.4%)	2,237 (66.7%)	23,048(78.4%)
Emergent	5,422 (9.7%)	1,816 (13.9%)	748 (11.8%)	382 (11.1%)	384 (11.5%)	2,092 (7.1%)
Urgent	9,842 (17.7%)	3,042 (23.2%)	1,176 (18.6%)	669 (19.4%)	730 (21.8%)	4,225 (14.4%)
Other	63 (0.1%)	19 (0.1%)	5 (0.1%)	3 (0.1%)	2 (0.1%)	34 (0.1%)
ASA classification, n (%)
Healthy	4,851 (8.7%)	1,263 (9.6%)	769 (12.1%)	140 (4.1%)	298 (8.9%)	2,381 (8.1%)
Mild Systemic Disease	32,423 (58.3%)	7,591 (57.9%)	4,071 (64.3%)	1,717 (49.8%)	1,934 (57.7%)	17,110 (58.2%)
Severe Systemic Disease	14,839 (26.7%)	3,401 (26.0%)	1,112 (17.6%)	1,312 (38.1%)	928 (27.7%)	8,086 (27.5%)
Incapacitating Disease	617 (1.1%)	156 (1.2%)	43 (0.7%)	60 (1.7%)	24 (0.7%)	334 (1.1%)
Moribund	10 (0.0%)	4 (0.0%)	0 (0.0%)	0 (0.0%)	2 (0.1%)	4 (0.0%)
Missing	2,897 (5.2%)	686 (5.2%)	341 (5.4%)	219 (6.4%)	167 (5.0%)	1,484 (5.0%)
Length of stay, n (%)
Same day discharge	22,726 (40.8%)	4,484 (34.2%)	2,229 (35.2%)	1,133 (32.9%)	1,217 (36.3%)	13,663 (46.5%)
1–2 days	18,325 (32.9%)	4,649 (35.5%)	1,757 (27.7%)	1,041 (30.2%)	1,064 (31.7%)	9,814 (33.4%)
3–4 days	10,070 (18.1%)	2,811 (21.5%)	1,609 (25.4%)	825 (23.9%)	758 (22.6%)	4,067 (13.8%)
4 or more days	4,516 (8.1%)	1,157 (8.8%)	741 (11.7%)	449 (13.0%)	314 (9.4%)	1,855 (6.3%)
Procedure site classification, n (%)
Large Central Metropolitan	23,394 (42.0%)	5,163 (39.4%)	3,785 (59.7%)	2,532 (73.4%)	1,603 (47.8%)	10,311 (35.1%)
Large Fringe Metropolitan	17,584 (31.6%)	2,992 (22.8%)	1,730 (27.3%)	434 (12.6%)	939 (28.0%)	11,489 (39.1%)
Medium Metropolitan	12,672 (22.8%)	4,635 (35.4%)	788 (12.4%)	468 (13.6%)	671 (20.0%)	6,110 (20.8%)
Micropolitan	1,339 (2.4%)	222 (1.7%)	23 (0.4%)	12 (0.3%)	113 (3.4%)	969 (3.3%)
Noncore	648 (1.2%)	89 (0.7%)	10 (0.2%)	2 (0.1%)	27 (0.8%)	520 (1.8%)
ASA, American Society of Anesthesiologists; comp., compensation; SD, standard deviation.

Trends in opioid prescribing overall and by race/ethnicity

Over time, opioid prescribing quantities decreased, with an overall mean change of

-15.2 MMEs in Year 3 versus Year 1, representing an approximate 10% decline (Table 2). Largest declines in opioid prescribing were among racial/ethnic minority groups in both absolute and relative terms compared to NHW patients. Guideline-concordant prescribing increased overall from Year 1 (26% of prescriptions) to Year 3 (40.3% of prescriptions) and increased within each racial/ethnic group. Similar results by race/ethnicity were observed after adjusting for potential confounding variables (Appendix Table A2).

Table 2

Post-operative opioid prescribed by race/ethnicity and year
	Total	Hispanic, Any Race	Non-Hispanic Asian	Non-Hispanic Black	Non-Hispanic Other	Non-Hispanic White
	N = 55,637	N = 13,101	N = 6,336	N = 3,448	N = 3,353	N = 29,399
Mean Total MME (SD)
1. October 2019 – September 2020	149.9 (96.8)	135.4 (78.6)	130.2 (81.6)	152.7 (96.7)	147.9 (92.2)	160.1 (105.7)
2. October 2020 – September 2021	147.8 (97.4)	128.8 (76.7)	131.9 (84.5)	149.2 (94.0)	143.6 (91.9)	159.9 (107.0)
3. October 2021 – October 2022	134.7 (98.2)	115.8 (75.4)	114.8 (81.8)	127.4 (87.5)	127.2 (85.8)	149.1 (110.0)
Mean Change (% Change), Year 3 to 1	-15.2 (-10.1)	-19.5 (-14.4)	-15.4 (-11.8)	-25.3 (-16.6)	-20.6 (-14.0)	-11.0 (-6.9)
Guideline Concordant Rx, %
October 2019 – September 2020	26.0	21.6	23.6	19.4	21.5	29.7
October 2020 – September 2021	29.4	23.9	26.8	22.9	26.2	33.1
October 2021 – October 2022	40.3	36.3	40.0	36.0	34.5	43.4
Mean Change, Year 3 to 1	+ 14.2	+ 14.7	+ 16.4	+ 16.6	+ 13.0	+ 13.7

Relative differences in total MME prescribed by race/ethnicity over time

We next examined differences in opioid prescribing for each racial/ethnic minority group relative to the NHW group by year, after statistical adjustment for potential confounding variables. Figure 1A illustrates that across all specialties, independent of randomized assignment, NHA and Hispanic patients received on average lower opioid quantities than NHW patients within each year (NHA: -7.3, -8.0, -10.4 MMEs in Year 1, 2 and 3; Hispanic: -5.1, -9.0, and − 8.9 MMEs in Year 1, 2 and 3), whereas NHB patients received higher opioid quantities in Year 1 (+ 11.0 MMEs) and Year 2 (+ 4.9 MME) and lower opioid quantities in Year 3 (-6.0 MMEs) relative to NHW patients. Stratified by specialty, differences were most prevalent and persistent over time for NHA and Hispanic patients undergoing general surgical procedures (Fig. 1B) and Hispanic patients undergoing Ob/Gyn procedures (Fig. 1C). For orthopedic procedures, differences in opioid prescribing for racial/ethnic minority patients relative to NHW patients were not prevalent in Years 1 and 2 but emerged in Year 3 (NHA: -22.3 MMEs; NHB: -19.0 MMEs; Hispanic: -14.1 MMEs) (Fig. 1D). No differences in opioid prescribing were observed for NHO versus NHW patients overall or by surgical specialty.

Impact of the intervention on opioid prescribing differences

To assess if the intervention had an impact on opioid prescribing differences by race/ethnicity, we calculated mean differences in total MME prescribed to each racial/ethnic minority group relative to the NHW group by period (pre-intervention vs. intervention) and randomized assignment (control vs. treatment arms). For all surgeries combined, NHAs received lower opioid quantities than NHW patients among surgeons randomized to the control arm prior to the intervention (-8.5 MME, p < 0.05; Column A, Table 3) and this difference persisted in the control arm during the intervention (-12.8 MME, p < 0.05, Column C, Table 3), thus fulfilling Criteria 1 (pre-condition for an effect of the intervention); however, the interaction term representing the differences in opioid quantities between the treatment and the control arms, before and during the intervention was not statistically significant (Column E, Table 3). For all other race/ethnic minority groups, Criteria 1 and 2, were not met.

Table 3

Differences in opioid prescribing by race/ethnicity, intervention period and randomized assignment.
All surgeries combined	Pre-Intervention Period				Intervention Period		Intervention vs. Pre-Intervention Period
	A	B			C	D	E
	Difference in total MME vs. NHW in control group (main term)	Difference in difference total MME vs. NHW in treatment vs. control group (main term)			Difference in total MME vs. NHW in control group	Difference in difference in total MME vs. NHW in treatment vs. control group	Difference in effect for treatment vs. control group (interaction term; column D - B)
Hispanic	-6.5 (-13.6, 0.6)	-0.8 (-9.5, 8.0)			-6.7 (-16.3, 3.0)	-3.5 (-16.5, 9.4)	-2.7 (-13.4, 8.0)
NHA	-8.5 (-16.7, -0.3)*	1.3 (-10.5, 13.1)			-12.8 (-25.4, -0.3)*	3.4 (-12.9, 19.8)	2.1 (-8.5, 12.8)
NHB	1.4 (-6.4, 9.3)	11.8 (-1.7, 25.3)			-6.3 (-25.7, 13.1)	0.7 (-23.3, 24.8)	-11.0 (-35.1, 13.0)
NHO	-3.4 (-13.6, 0.6)	5.6 (-4.7, 15.9)			-2.6 (-10.4, 5.1)	-3.2 (-14.1, 7.8)	-8.8 (-21.3, 3.8)
General surgeries	Pre-Intervention Period				Intervention Period		Pre- vs. Intervention Period
	A	B			C	D	E
	Difference in total MME vs. NHW in control group (main term)	Difference in difference total MME vs. NHW in treatment vs. control group (main term)			Difference in total MME vs. NHW in control group	Difference in difference in total MME vs. NHW in treatment vs. control group	Difference in effect for treatment vs. control group (interaction term; column D - B)
Hispanic	-5.0 (-12.0, 2.0)	-5.5 (-15.5, 4.6)			-8.3 (-15.6, -1.0)*	4.3 (-5.4, 14.1)	9.8 (-1.9, 21.5)
NHA	-10.2 (-18.2, -2.2)*	2.1 (-8.1, 12.4)			-12.1 (-19.8, -4.4)**	8.0 (-2.0, 18.1)	5.9 (-8.0, 19.8)
NHB	1.5 (-6.3, 9.4)	4.8 (-6.6, 16.3)			-12.5 (-21.2, -3.8)**	17.8 (1.8, 33.7)*	12.9 (-2.5, 28.3)
NHO	0.1 (-6.0, 6.1)	-4.0 (-14.8, 6.7)			-3.2 (-13.6, 7.2)	1.2 (-9.9, 12.4)	5.3 (-11.0, 21.5)
Oby/Gyn surgeries	Pre-Intervention Period				Intervention Period		Pre- vs. Intervention Period
	A		B		C	D	E
	Difference in total MME vs. NHW in control group (main term)		Difference in difference total MME vs. NHW in treatment vs. control group (main term)		Difference in total MME vs. NHW in control group	Difference in difference in total MME vs. NHW in treatment vs. control group	Difference in effect for treatment vs. control group (interaction term; column D - B)
Hispanic	-2.7 (-12.8, 7.4)		-4.1 (-17.8, 9.5)		-2.0 (-8.4, 4.3)	-5.3 (-18.1, 7.5)	-1.2 (-9.7, 7.4)
NHA	0.4 (-5.3, 6.0)		-2.0 (-8.7, 4.7)		-7.9 (-20.3, 4.6)	10.7 (-2.8, 24.2)	12.7 (4.8, 20.7)**
NHB	5.9 (0.8, 11.0)*		8.6 (-1.8, 18.9)		8.7 (2.0, 15.4)*	-12.3 (-28.5, 3.0)	-20.8 (-40.1, -1.5)*
NHO	-0.6 (-10.9, 9.7)		2.8 (-10.8, 16.5)		-4.8 (-11.1, 1.5)	0.2 (-12.5, 12.8)	-2.7 (-20.9, 15.5)
Orthopedic surgeries	Pre-Intervention Period				Intervention Period		Pre- vs. Intervention Period
	A			B	C	D	E
	Difference in total MME vs. NHW in control group (main term)			Difference in difference total MME vs. NHW in treatment vs. control group (main term)	Difference in total MME vs. NHW in control group	Difference in difference in total MME vs. NHW in treatment vs. control group	Difference in effect for treatment vs. control group (interaction term; column D - B)
Hispanic	-9.7 (-25.7, 6.3)			4.6 (-10.8, 20.0)	-22.6 (-55.6, 10.4)	11.5 (-26.9, 49.9)	6.8 (-27.5, 41.2)
NHA	-9.0 (-28.0, 10.1)			-3.2 (-22.1, 15.8)	-38.4 (-64.5. -12.2)**	18.2 (-16.1, 52.4)	21.3 (-24.7, 67.4)
NHB	4.8 (-9.3, 18.9)			0.2 (-26.8, 27.0)	-10.3 (-70.3, 49.6)	-7.0 (-71.7, 57.7)	-7.2 (-88.0, 73.5)
NHO	-4.6 (-18.0, 8.7)			16.0 (1.08, 30.8)*	-21.5 (-36.7, -6.3)**	14.0 (-7.2, 35.2)	-2.0 (-28.2, 24.2)
MME, morphine milligram equivalent; NHA, non-Hispanic Asian; NHB, non-Hispanic Black; NHO, non-Hispanic Other; NHW; non-Hispanic White. p < 0.05; *p < 0.01.

In stratified analysis by surgical specialty, effects of the intervention were observed only for NHB receiving Ob/Gyn procedures, whereby Criteria 1 (Column A and C, Table 3) and Criteria 2 (Column E, Table 3) were fulfilled, indicating that the effect of the intervention on prescribing differences between NHB vs NHW patients was by an average of -20.8 total MME. This appeared to eliminate prescribing differences for NHB vs. NHW patients in the treatment arm during the intervention period (see Appendix, Panel C, Figure A1).

In this secondary analysis of a clustered RCT, in the two years prior to the start of the intervention, NHB patient received on average higher opioid quantities than NHW patients for all surgical procedures combined, whereas NHA and Hispanic patients received lower opioid quantities; these differences persisted for NHA and Hispanic patients in the year that the intervention was active; however, for NHB patients the differences was inversed, such that they received lower opioid quantities than NHW patients. Nevertheless, opioid safety reports, based on social norms, designed to improve postoperative guideline-concordant prescribing among surgeons, did not have an overall, independent effect on mitigating opioid prescribing differences by race/ethnicity.

While the RCT showed that these safety reports, overall, had a positive impact on the percentage of opioid prescriptions that were guideline concordant and that on average patients received 12 total MMEs less as a result of the intervention [12], this did not, across surgical procedures combined, translate into closing the prescribing gaps for racial/ethnic minority groups compared to the NHW groups. In fact, Criteria 1 as a pre-condition for the intervention to have an effect was only met for NHA versus NHW patients, for all surgeries combined; despite this, Criteria 2 was not met.

Other studies from the literature have shown mixed results for interventions that are designed to reduce opioid prescribing levels overall but have secondary effects on prescribing by race/ethnicity. For example, a study conducted in the same health system as the current study showed that an RCT designed to reduce the number of opioid pills prescribed at discharge from the ED through individual audit, peer comparison feedback, or the combination of the two, was effective [17]; however, the intervention resulted in similar reductions in the number of pills prescribed to NHB and NHW patients and, thus, baseline differences, in which NHW patients received on average more opioid pills per prescription than NHB patients, persisted [18]. In a retrospective study, Herbs and colleagues examined the impact of a standardized opioid prescribing schedule that was introduced for general surgical procedures within a large, academic medical center [11]. Using a quasi-experimental, interrupted time-series design, the authors found that NHB patients were prescribed higher MME quantitative at hospital discharge compared with NHW patients prior to the intervention and that these differences were mitigated after the intervention was introduced and, ostensibly, as a result of the intervention.

The results found by Herb et al. are consistent with findings from our sub-analysis by surgical specialties, in which the intervention appeared to have an impact on differences in opioid prescribing for NHB versus NHW patients undergoing Ob/Gyn procedures. Specifically, our analysis indicates that the intervention reduced opioid prescribing for NHB patients, who initially received higher total MME quantities, by an average of -20.8 MME compared to NHW patients. This difference translates to an approximate reduction of 2.8 oxycodone 5 mg pills per prescription for NHB patients, a seemingly small quantity at the patient level. Nevertheless, extrapolated to the population level, for every 100 NHB undergoing an Ob/Gyn procedures, 280 fewer opioid pills would be available in the community for potential misuse, abuse, or diversion. Given the recent rise in opioid-related deaths in among Black Americans, this reduction could help to curb these trends, but future studies would be needed to quantify this.

While we cannot know why the intervention may have had a select effect among NHB patients undergoing Ob/Gyn procedures, there are hypotheses that warrant consideration. First, Ob/Gyn surgeons and prescribers may have been more amendable to prescribing lower opioid quantities in response to the intervention. This was a trend in the original RCT, but one that did not reach statistical significance [12]. Second, racial/ethnic minority patients, especially NHB patients, experienced larger decreases in prescribed opioid quantities than NHW patients over time, independent of the intervention, and NHB patients, unlike NHA or Hispanic patients, were on average initially more likely to receive higher opioid quantities than NHW patients (which was observed only in the Ob/Gyn specialty). Thus, it follows that larger decreases in opioid quantities prescribed to NHA and Hispanic patients would potentially perpetuate differences compared to NHW patients, whereas larger decreases in opioid quantities prescribed to NHB patients would potentially mitigate these differences. These two hypotheses are not mutually exclusive and, in fact, may have congruently contributed to the selective effect of the intervention. Future research is needed to validate these hypotheses.

Independent of the intervention, the average total MME prescribed to patients decreased over time, which is likely due to shifting norms and uptake of prescribing guideline recommendations. However, as we note earlier, racial/ethnic minority patients had larger decreases in total MME over time relative NHW patients, suggesting that prescribers may have been – implicitly or explicitly -- more likely to prescriber lower opioid quantities to racial/ethnic minority patients during this period. Strategies are needed to mitigate such unwarranted variation in opioid prescribing practices through both direct and indirect means. For example, implicit bias training may be helpful to make clinicians more aware of their unconscious behaviors, but education may also be needed to address historic misconceptions of differences in pain tolerance by race and ethnicity [5–7].

There are several limitations to this analysis that should be considered when interpreting the findings. First, this was an exploratory analysis of a clustered RCT, and the study was not specifically powered to test our stated hypothesis. Samples sizes by race/ethnicity and surgical specialty vary and non-statistically significantly results may be due to a lack of statistically power, rather than lack of an effect. Second, unmeasured confounding by race/ethnicity is still possible, as randomization occurred at the level of the hospital-specialty. Nevertheless, this study has important strengths. It was conducted on a large, diverse patient population from a community-based healthcare system. Further, analyses took into considerate a large number of potential confounding variables available from the EHR.

In conclusion, opioid safety reports based on social norms, designed to decrease guideline-discordance opioid prescribing, did not appear to have an effect on differences in prescribing between racial/ethnic groups across surgical specialties. While a select effect was observed among NHB versus NHW patients undergoing Ob/Gyn procedures, this was concomitant with an overall trend in larger reductions in opioid prescribing for racial/ethnic minority groups compared to NHW patients. Future research is needed to identify strategies to mitigate racial/ethnic postoperative opioid prescribing differences at hospital discharge.

Acknowledgements

The authors acknowledge grant support for this work from the National Institute on Drug Abuse (Grant number: 1R01DA046226; PI: Watkins). The authors also wish to acknowledge the helpful contributions of Carolina Ibarra in the preparation and submission of this manuscript and the Sutter Health executive leadership for their support.

Author contributions

Robert J. Romanelli was involved in the design of the work, analysis, interpretation of the data, and wrote the manuscript. Zachary Wagner was involved in the design of the work, analysis, interpretation of the data, and provided substantial revisions to the manuscript. Louis T. Mariano the design of the work, analysis, interpretation of the data, and provided substantial revisions to the manuscript. Rivfka Shenoy was involved in the interpretation of the data. Allison Kirkegaard was involved in the interpretation of the data. Katherine E. Watkins was involved in the design of the work, interpretation of the data, and provided substantial revisions to the manuscript.

Funding

This work was supported by the National Institute on Drug Abuse of the National Institutes of Health under Award Number R01DA046226.

Data Availability

The datasets analyzed during the current study may be available from the corresponding author on reasonable request.

Ethics approval and consent to participate

All methods were carried out in accordance with relevant guidelines and regulations. This study was approved by the Sutter Health Institutional Review Board.

Consent for publication

All authors approved of the final version of the manuscript and consent to publication.

Competing interests

The authors declare no competing interests

Howard R, Fry B, Gunaseelan V, Lee J, Waljee J, Brummett C, et al. Association of opioid prescribing with opioid consumption after surgery in Michigan. JAMA Surg. 2019;154(1):e184234.
Ward R, Taber D, Gonzales H, Gebregziabher M, Basco W, McCauley J. Risk factors and trajectories of opioid use following total knee replacement. Knee Surg Relat Res. 2022;34(1):18.
Aalberg JJ, Kimball MD, McIntire TR, McCullen GM. Long-term outcomes of persistent post-operative opioid use: A retrospective cohort study. Ann Surg. 2022.
Wylie JA, Kong L, Barth RJ. Jr. Opioid dependence and overdose after surgery: Rate, risk factors, and reasons. Ann Surg. 2022;276(3):e192–8.
Santoro TN. D. SJ. Racial bias in the us opioid epidemic: A review of the history of systemic bias and implications for care. Cureus. 2018;10.
Anderson KO, Green CR, Payne R. Racial and ethnic disparities in pain: causes and consequences of unequal care. J Pain. 2009;10(12):1187–204.
Burgess DJ, Crowley-Matoka M, Phelan S, Dovidio JF, Kerns R, Roth C, et al. Patient race and physicians' decisions to prescribe opioids for chronic low back pain. Soc Sci Med. 2008;67(11):1852–60.
Kaiser Family Foundation. Opioid overdose deaths by race/ethnicity, 2019 [updated 2021. https://www.kff.org/other/state-indicator/opioid-overdose-deaths-by-raceethnicity/.
D'Orsogna MR, Böttcher L, Chou T. Fentanyl-driven acceleration of racial, gender and geographical disparities in drug overdose deaths in the United States. PLOS Glob Public Health. 2023;3(3).
Romanelli RJ, Shenoy R, Martinez MC, Mudiganti S, Mariano LT, Zanocco KA, Wagner Z, Kirkegaard A, Watkins KE. Disparities in postoperative opioid prescribing by race and ethnicity: an electronic health records-based observational study from Northern California, 2015–2020. Archives Public Health. 2023;81(1):83.
Herb JN, Williams BM, Chen KA, Young JC, Chidgey BA, McNaull PP. The impact of standard postoperative opioid prescribing guidelines on racial differences in opioid prescribing: A retrospective review. Surgery. 2021;170(1):180–5.
Wagner Z, Kirkegaard A, Mariano LT, Doctor JN, Yan X, Persell SD. Peer comparison or guideline-based feedback and postsurgery opioid prescriptions: A randomized clinical trial. JAMA Health Forum. 2024;5(3):e240077.
Kirkegaard A, Wagner Z, Mariano LT, Martinez MC, Yan XS, Romanelli RJ. Evaluating the effectiveness of email-based nudges to reduce postoperative opioid prescribing: study protocol of a randomised controlled trial. BMJ Open. 2022;12.
U.S. Department of Health and Human Services. Calculating Total Daily Dose of Opioids for Safer Dosage. Atlanta, GA:Centers for Disease Control and Prevention.
Committee on Economics. ASA Physical Status Classification System Schaumburg, IL: American Society of Anesthesiologists. 2014 [updated December 13, 2020. https://www.asahq.org/standards-and-guidelines/asa-physical-status-classification-system.
Coleman C, Lennon RP, Robinson JM, Tuan WJ, Sehgal N, Zgierska AE. Provider and patient sex disparities in opioid prescribing and patient outcomes. J Opioid Manag. 2022;18(5):435–45.
Navathe AS, Liao JM, Yan XS, Delgado MK, Isenberg WM, Landa HM. The effect of clinician feedback interventions on opioid prescribing: study examines the effect of clinician feedback interventions on opioid prescribing. Health Aff. 2022;41(3):424–33.
Crowley AP, Sun C, Yan XS, Navathe A, Liao JM, Patel MS. Disparities in emergency department and urgent care opioid prescribing before and after randomized clinician feedback interventions. Acad Emerg Med. 2023;30(8):809–18.

No competing interests reported.

Appendix.docx

Download PDF

Reviewers agreed at journal
18 Sep, 2024
Reviewers agreed at journal
18 Sep, 2024
Reviewers agreed at journal
03 Sep, 2024
Reviewers agreed at journal
29 Aug, 2024
Reviewers invited by journal
30 Jul, 2024
Editor assigned by journal
22 Jul, 2024
Submission checks completed at journal
22 Jul, 2024
First submitted to journal
20 Jun, 2024

You are reading this latest preprint version

The impact of opioid safety reports on racial and ethnic differences in post- operative opioid prescribing practices at hospital discharge: exploratory analysis of a clustered randomized controlled trial

Status:

Version 1

Abstract

Background:

Methods:

Results:

Conclusions:

Figures

Contributions to the Literature

Introduction

Methods

Study design

Data collection

Intervention

Cohort eligibility

Outcome variables

Explanatory variables

Statistical analysis

Results

Description of cohort

Trends in opioid prescribing overall and by race/ethnicity

Relative differences in total MME prescribed by race/ethnicity over time

Impact of the intervention on opioid prescribing differences

Discussion

Declarations

References

Additional Declarations

Supplementary Files

Status:

Version 1