Abstract
Background:
Efforts to reduce opioid use in trauma patients are currently hampered by an incomplete understanding of the baseline opioid exposure and variation in United States. The purpose of this project was to obtain a global estimate of opioid exposure following injury and to quantify the variability of opioid exposure between and within United States trauma centers.
Study Design:
Prospective observational study calculating opioid exposure by converting all sources of opioids to oral morphine milligram equivalents (MME). To estimate variation, an intra-class correlation was calculated from a multilevel generalized linear model adjusting for the a priori selected variables Injury Severity Score (ISS) and prior opioid use.
Results:
The centers enrolled 1,731 patients. The median opioid exposure amongst all sites was 45 MME/day, equivalent to 30 mg of oxycodone or 45 mg of hydrocodone per day. Variation in opioid exposure was identified both between and within trauma centers with the vast majority of variation (93%) occurring within trauma centers. Opioid exposure increased with injury severity, in male patients, and patients suffering penetrating trauma.
Conclusion:
The overall median opioid exposure was 45 MME/day. Despite significant differences in opioid exposure between trauma centers, the majority of variation was actually within centers. This suggests that efforts to minimize opioid exposure after injury should focus within trauma centers and not on high level efforts to affect all trauma centers.
Type of Study:
Epidemiological
Level of Evidence:
Level III
Keywords: trauma, injury, opioid exposure, opioid, acute pain
Introduction
Safe and effective acute pain management after injury is critical. Inadequate acute pain relief can increase physiologic stress, result in harmful psychologic sequelae, delay mobilization, and, potentially, lead to the development of chronic pain.(1–7) For decades, opioids have been a mainstay in the treatment of acute pain after injury.(8) However, this reliance on opioid therapy for acute pain has directly contributed to the current opioid crisis, as 75% of patients seeking treatment for heroin addiction were first introduced to opioids via prescription.(9–11)
Trauma patients are particularly vulnerable to progress to persistent opioid use.(12, 13) A dearth of baseline information on opioid exposure in injured patients currently hampers the trauma community’s nascent efforts at addressing the complex problem of post-injury opioid exposure and the negative long-term consequences of said exposure.(14, 15) The limited information also hinders clinical trial planning and design for optimizing treatment of acute pain and opioid use in trauma.
To address these limitations, we performed a prospective, observational study with the overall goals of quantifying opioid exposure and evaluate the variability in opioid exposure in adult trauma patients. We hypothesized that the majority of the variation would be between trauma centers. Additionally, we aimed to evaluate exploratory, hypothesis-generating associations between opioid exposure and lengths of stay.
Methods
This manuscript was written in accordance with the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guidelines. Institutional Review Board approval was obtained at each participating trauma center.
Setting
Seven United States trauma centers participated in this prospective observational study: 1) the Red Duke Trauma Center at Memorial Hermann Hospital – Texas Medical Center and McGovern Medical School at UT Health, Houston, Texas (urban Level 1), 2) UT Health Tyler, Tyler, Texas (rural Level 1); St. Joseph’s Hospital and Medical Center, Phoenix, Arizona (urban Level 1); Indiana University Health Methodist Hospital, Indianapolis, Indiana (urban Level 1); Eskenazi Health, Indianapolis, Indiana (urban Level 1); University of Arkansas for Medical Sciences, Little Rock, Arkansas (rural Level 1); University of Colorado Health Memorial Hospital Central, Colorado Springs, Colorado (rural Level 1). Trauma center type was explicitly considered (4 urban, 3 rural; varying volumes of admission) in the determination of site inclusion to increase the heterogeneity of enrolled patients.
Design, Participants, and Study Size
Trauma centers enrolled injured patients ≥16 years old admitted to the adult trauma service over two months from May 1, 2018 through June 30, 2018. Enrolled participants were followed from admission to discharge or 30 days, whichever was sooner. Patients admitted after burn injury, admitted without a traumatic mechanism, or placed in observation were not included. A two month observational study was planned to increase the feasibility of this data-intensive project. Given the observational nature of the study, pre-specified power analyses were not performed and instead a Bayesian analytic plan utilized.
Variables
Baseline demographic and injury severity information were recorded, including age, sex, race/ethnicity, a history of prior opioid use, Abbreviated Injury Scores (AIS), Injury Severity Score (ISS), and mechanism of injury. Race/ethnicity was determined according to the medical record and how the variable was recorded in the institutional trauma registry. Additionally, granular injury characteristics, such as the number of rib fractures, flail chest, long bone fracture, vertebral body fracture, traumatic brain injury, laparotomy, thoracotomy, and pelvis/acetabulum fracture, were recorded.
The primary outcome was in-patient opioid exposure, measured as morphine milligram equivalents (MME) per day.(16) The sources of opioid exposure included intravenous and oral opioids given in all phases of care (including the emergency department, operating room, intensive care unit, intermediate units, or wards), including intravenous drips and patient controlled analgesia. The explicit goal of the study was to obtain a global estimate of opioid exposure (MME/day) from admission to discharge. Total MME was recorded for hospital day 0 (time of admission to 0700), hospital day 1 (first full day of admission, 0700–0659), and continued until hospital day 30 or discharge, whichever was sooner. MME/day was calculated by dividing total MME by the number of hospital days.
Secondary outcomes included the median of the mean Numeric Rating Scale (NRS) pain scores each day in communicative patients, discharge from the hospital with an opioid prescription, use of regional anesthesia, ileus, cardiac arrest, unplanned intubation, unplanned admission to the intensive care unit (ICU), mortality, and lengths of stay. The average daily NRS pain values were recorded for hospital day 0 (time of admission to 0700), hospital day 1 (first full day of admission, 0700–0659), and continued until hospital day 30 or discharge, whichever was sooner.
When patients were unable to provide a NRS pain score, no pain data were recorded. We excluded non-NRS pain scores for three reasons. First, centers used different pain scales for non-communicative patients. Second, there is no currently validated manner to combine the 11 point NRS pain scale and any of the pain scales meant for non-communicative patients (e.g. the 3 domain, 12 point Behavioral Pain Scale; the 4 domain, 8 point Critical Care Pain Observation Tool) into a longitudinal quantification of the patient’s overall pain experience. Lastly, the majority of pain assessments occur in communicative patients, making the NRS pain scale values the largest collection of pain quantification available.
Statistical Methods
Descriptive statistics were performed presenting continuous data as median with interquartile ranges (IQR) and categorical data as the number and percent.
Opioid exposure:
Given that trauma patients are a heterogeneous population, both stratification and regression were performed. The stratified analytic plan was chosen as the primary method of analysis as the authors felt it best illustrated and emphasized the variability observed between centers. To illustrate variability, MME/day were estimated for the following strata: sex, prior opioid use, race/ethnicity, mechanism of injury, ISS, and the presence of individual injuries, including flail segment, long bone fracture, vertebral body fracture, traumatic brain injury, pelvis or acetabulum fracture, laparotomy, and thoracotomy. Differences between centers and strata were then assessed using Kruskal-Wallis, Chi Square, and Fisher’s Exact test for continuous, binary, and sparse binary variables, respectively.
To understand the independent effects of demographic characteristics on opioid exposure, a frequentist and Bayesian multi-level generalized linear model accounting for clustering due to trauma center was created to predict MME/Day including the following variables: Injury Severity Score, age, sex, race/ethnicity, and prior opioid use. Finally, the longitudinal trajectory of opioid exposure was investigated by modeling daily MME using a zero-inflated negative binomial model.
Variability in opioid exposure:
To evaluate variability in opioid exposure amongst the participating trauma centers, a multilevel generalized linear model was created accounting for clustering due to trauma center and adjusting for the a priori selected variables ISS and prior opioid use. This model accounted for heterogeneity both within and across centers.(17) From this model, an intra-class correlation coefficient (ICC) was calculated. As a sensitivity analysis, similar multilevel generalized linear models were created accounting for clustering due to trauma center and using different combinations of the following included variables: ISS, age, sex, prior opioid use, and race/ethnicity. An ICC was then calculated from each of these models and the range of ICC values evaluate the precision of our estimate.
Hypothesis-generating associations between opioid exposure and lengths of stay:
To assess the association between opioid exposure and lengths of stay, frequentist and Bayesian multilevel generalized linear models were created adjusting for the a priori selected variables ISS and prior opioid use accounting for clustering due to trauma center. For all Bayesian models, priors for regression coefficients were specified as ~Normal (μ=0, σ2=1000) on the identity or log-scale depending upon the model, level one error variances were specified as ~Half-T (df = 3, mean = 0, standard deviation = 100). Prior distribution for level two variances used ~Half-T (df = 3, mean = 0, standard deviation = 100). Priors for the comparison of proportions were specified as ~Beta (α=0.5, β=0.5). Results are presented as Incidence Rate Ratio with 95% confidence intervals for frequentist models or 95% credible intervals with posterior probabilities for Bayesian models. Frequentist and Bayesian models were performed in parallel to optimize the information available in the data and because Bayesian posterior probabilities more readily inform clinical decision making.(18)
Pain scores:
Due to the absence of information regarding temporal relationship between NRS pain scores and opioid exposure, we deemed the associations non-interpretable and chose to only provide measures of central tendency and dispersion along with opioid exposure.
Analyses were performed on STATA (version15.1; College Station, Texas) and R (version 3.6.1; Vienna, Austria).
Data Sources and Integrity
Depending on institutional capability, recorded data came from one of three sources – research personnel manual entry, the electronic medical record, and the institutional trauma registry. Data were entered into a secure, central REDCap® database, which also allowed for assessment of data quality and evaluation of missingness.
Results
The sites enrolled 1,772 patients over the two month study period. Of these, 41 were excluded (1 snake bite, 3 hangings, 28 burns, 1 unknown trauma, and 8 other/unspecified mechanisms) leaving at total of 1,731 patients included in the analysis.
Trauma Centers
Trauma center volume ranged from 152 patients admitted over the study period to 533 (Supplemental Table 1). Four centers had a written pain management protocol and five had an electronic order set to facilitate the ordering of pain medication. Five of the seven centers had an acute pain service available for consultation.
Demographic and Injury Characteristics
Demographic and injury characteristics of patients treated at the seven trauma centers are provided in Table 1. Clinically and statistically significant differences between centers were seen in age, sex, race/ethnicity, and prior opioid use. Regional AIS values and ISS also had clinically and statistically significant differences between centers, including AIS values for Chest region, ISS, and percentage of trauma admissions with an ISS > 15.
Table 1.
Demographics and Injury Characteristics
| Institution | 1 (n=200) |
2 (n=252) |
3 (n=533) |
4 (n=178) |
5 (n=168) |
6 (n=248) |
7 (n=152) |
p-value |
|---|---|---|---|---|---|---|---|---|
| Demographics | ||||||||
| Age, years | 51 (33, 67) | 38 (26, 55) | 55 (35, 69) | 44 (29, 58) | 51 (33, 65) | 59 (37, 74) | 43 (30, 60) | <0.001 |
| Male | 120 (60%) | 183 (73%) | 320 (60%) | 127 (71%) | 112 (67%) | 143 (58%) | 103 (68%) | |
| White | 150 (75%) | 119 (47%) | 448 (84%) | 86 (48%) | 123 (73%) | 191 (77%) | 59 (39%) | |
| Unknown | 38 (19%) | 109 (43%) | 71 (13%) | 35 (20%) | 1 (1%) | 75 (30%) | 6 (4%) | |
| Injury Characteristics | ||||||||
| Penetrating | 22 (11%) | 68 (27%) | 28 (5%) | 30 (17%) | 10 (6%) | 20 (8%) | 25 (16%) | |
| AIS Head | 1 (0, 3) | 1 (0, 1) | 0 (0, 2) | 0 (0, 3) | 0 (0, 2) | 0 (0, 3) | 0 (0, 2) | <0.001 |
| AIS Face | 0 (0, 0) | 0 (0, 1) | 0 (0, 0) | 0 (0, 0) | 0 (0, 0) | 0 (0, 0) | 0 (0, 0) | 0.460 |
| AIS Chest | 0 (0, 2) | 0 (0, 2) | 0 (0, 2) | 0 (0, 2) | 0 (0, 2) | 0 (0, 2) | 2 (0, 3) | <0.001 |
| AIS Abdomen | 0 (0, 0) | 0 (0, 2) | 0 (0, 0) | 0 (0, 0) | 0 (0, 0) | 0 (0, 0) | 0 (0, 2) | <0.001 |
| AIS Extremity | 2 (0, 2) | 1 (2, 2) | 2 (0, 2) | 0 (0, 2) | 0 (0, 2) | 0 (0, 2) | 2 (0, 3) | <0.001 |
| AIS External | 1 (0, 1) | 0 (0, 0) | 1 (0, 1) | 1 (1, 1) | 0 (0, 0) | 1 (0, 1) | 1 (0, 1) | <0.001 |
| ISS | 10 (5, 21) | 9 (5, 17) | 10 (5, 17) | 10 (5, 17) | 9 (4, 13) | 10 (5, 17) | 17 (9, 23) | <0.001 |
| ISS>15 | 75 (37%) | 66 (26%) | 174 (33%) | 57 (32%) | 34 (20%) | 68 (27%) | 77 (51%) | <0.001 |
Continuous date presented as: median (interquartile range)
ISS – Injury Severity Score; AIS – Abbreviate Injury Scale
The rates of penetrating mechanism also differed. Specific injury patterns and operations differed, including flail chest (range 1% to 12%, p<0.001), long bone fractures (range 11% to 34%, p<0.001), vertebral body fractures (range 13% to 25%, p<0.001), traumatic brain injury (range 9% to 43%, p<0.001), pelvis or acetabulum fracture (range 9% to 23%, p<0.001), laparotomy (range 2% to 15%, p<0.001), thoracotomy (range 0% to 12%, p<0.001), and extremity amputation (range 0% to 4%, p<0.001) (Supplemental Table 2).
Opioid Exposure
The overall median MME per day was 45 MME/day (IQR 16, 82), with individual trauma center values ranging from 18 MME/day (IQR 6, 35) to 64 MME/day (IQR 31, 111) (p<0.001) (Table 2). When stratified, the ISS >15 subgroup had a higher opioid exposure than the ISS≤15 group. This difference was seen in each center except for one, Site 7.
Table 2.
Opioid Exposure and Demographic Strata
| 1 (n=200) |
2 (n=252) |
3 (n=533) |
4 (n=178) |
5 (n=168) |
6 (n=248) |
7 (n=152) |
All patients (n=1,731) |
|
|---|---|---|---|---|---|---|---|---|
| Total MME | 176 (34, 486) | 221 (61, 580) | 148 (48, 376) | 182 (54, 532) | 119 (40, 285) | 86 (27, 215) | 241 (118, 504) | 153 (45, 422) |
| MME/day | 48 (16, 82) | 64 (31, 111) | 51 (18, 89) | 61 (26, 119) | 45 (14, 80) | 18 (6, 35) | 45 (27, 72) | 45 (16, 82) |
| Median of average daily pain scores | 5 (3, 6) | 5 (4, 6) | 4 (3, 6) | 4 (1, 5) | 5 (3, 6) | 6 (3, 7) | 3 (2, 5) | 5 (3, 6) |
| Regional anesthesia | 6 (3%) | 59 (23%) | 13 (2%) | 0 (0%) | 50 (30%) | 10 (4%) | 20 (13%) | 158 (9%) |
| MME/Day by Stratum: Injury Severity Score | ||||||||
| ISS ≤15 | 34 (6, 70) n=126 |
60 (31, 96) n=186 |
48 (18, 84) n=360 |
53 (24, 83) n=121 |
43 (10, 74) n=134 |
15 (6, 31) n=180 |
47 (23, 80) n=75 |
41 (14, 77) n=1,182 |
| ISS >15 | 68 (31, 103) n=74 |
84 (35, 191) n=66 |
56 (20, 104) n=173 |
102 (31, 155) n=57 |
69 (26, 149) n=34 |
21 (8, 52) n=68 |
44 (28, 64) n=77 |
54 (22, 104) n=549 |
| p-value | <0.001 | 0.023 | 0.043 | 0.004 | 0.015 | 0.029 | 0.447 | <0.001 |
| MME/Day by Stratum: Mechanism of Injury | ||||||||
| Blunt | 45 (10, 80) n=178 |
52 (28, 100) n=184 |
50 (18, 89) n=505 |
56 (23, 112) n=148 |
44 (13, 79) n=158 |
17 (6, 34) n=228 |
41 (22, 64) n=127 |
42 (15, 79) n=1,528 |
| Penetrating | 75 (30, 104) n=22 |
78 (56, 135) n=68 |
61 (34, 82) n=28 |
72 (42, 130) n=30 |
80 (55, 113) n=10 |
23 (3, 47) n=20 |
72 (51, 90) n=25 |
69 (37, 112) n=203 |
| p-value | 0.057 | 0.005 | 0.373 | 0.076 | 0.120 | 0.677 | <0.001 | <0.001 |
| MME/Day by Stratum: Prior Opioid Use | ||||||||
| No | 40 (11, 80) N=159 |
58 (25, 110) N=102 |
49 (18, 84) N=377 |
46 (15, 80) N=127 |
45 (15, 89) N=99 |
17 (6, 31) N=104 |
46 (28, 72) N=138 |
44 (16, 79) N=1,106 |
| Unknown | 57 (17, 90) N=38 |
60 (32, 96) N=109 |
49 (15, 90) N=71 |
122 (72, 285) N=35 |
85 (85, 85) N=1 |
21 (7, 46) N=75 |
35 (18, 51) N=6 |
50 (18, 91) N=335 |
| Yes | 115 (91, 126) N=3 |
90 (53, 142) N=41 |
65 (20, 113) N=85 |
105 (35, 158) N=16 |
46 (13, 75) N=68 |
13 (6, 30) N=69 |
31 (20, 112) N=8 |
46 (13, 94) N=290 |
| p-value | 0.060 | 0.021 | 0.061 | <0.001 | 0.527 | 0.240 | 0.548 | 0.064 |
Continuous date presented as: median (interquartile range)
MME – morphine milligram equivalents
Patients with prior opioid use did not have a higher in-hospital opioid exposure, though this finding is potentially biased by a lack of information about prior opioid use in 335 (19%) patients. Overall, men were found to have had a higher opioid exposure than women, however, this difference was not observed in 4 of the trauma centers, Sites 1, 2, 3, and 4 (Supplemental Table 3). Opioid exposure by race/ethnicity was variable, with 2 centers demonstrating differences and the remainder not. Patients who suffered a penetrating mechanism had significantly higher opioid exposure than those injured by a blunt mechanism (Table 2).
Opioid exposure by injury patterns and operations also varied (Supplemental Tables 4 and 5). While no difference in opioid exposure was seen in patients with flail segment compared to those without, patients with a long bone fracture, vertebral body fracture, or pelvis/acetabulum fracture had a higher opioid exposure than those without those injuries. Patients with a traumatic brain injury had a lower opioid exposure than those without. Patients who underwent laparotomy, thoracotomy, or extremity amputation also had a higher opioid exposure than those who did not undergo those operations.
Differences in rate of opioid prescribing at discharge and the type of opioid prescribed also were seen (Table 3). The overall rate of opioid prescribing at discharge was 73% (range 48% to 83% across centers p<0.001). The most commonly prescribed opioids at discharge were oxycodone, hydrocodone, and tramadol. Three centers prescribed oxycodone the most at discharge (range 54% to 92% of discharges with an opioid), two centers prescribed hydrocodone the most at discharge (range 62% to 66% of discharges with an opioid), and one center prescribed tramadol the most (89% of discharges with an opioid). Codeine, methadone, fentanyl patch, hydromorphone, and morphine were infrequently prescribed at discharge.
Table 3.
Opioid Prescribing at Discharge
| 1 (n=200) |
2 (n=252) |
3 (n=533) |
4 (n=178) |
5 (n=168) |
6 (n=248) |
7 (n=152) |
All patients (n=1,731) |
|
|---|---|---|---|---|---|---|---|---|
| Opioid Given at Discharge | 96 (48%) | 209 (83%) | 425 (80%) | 128 (72%) | 113 (67%) | 185 (75%) | 111 (73%) | <0.001 |
| Opioid Prescribed at Discharge | ||||||||
| Tramadol | 0 (0%) | 5 (2%) | 54 (13%) | 12 (9%) | 15 (13%) | 44 (24%) | 99 (89%) | <0.001 |
| Hydrocodone | 35 (36%) | 82 (39%) | 263 (62%) | 1 (1%) | 35 (31%) | 123 (66%) | 13 (12%) | <0.001 |
| Oxycodone | 56 (58%) | 122 (58%) | 138 (32%) | 118 (92%) | 61 (54%) | 2 (1%) | 8 (7%) | <0.001 |
| Codeine | 5 (5%) | 0 (0%) | 2 (0%) | 0 (0%) | 0 (0%) | 16 (9%) | 10 (9%) | <0.001 |
| Methadone | 0 (0%) | 0 (0%) | 4 (1%) | 0 (0%) | 0 (0%) | 0 (0%) | 1 (1%) | 0.484 |
| Fentanyl patch | 0 (0%) | 0 (0%) | 4 (1%) | 0 (0%) | 1 (1%) | 1 (1%) | 0 (0%) | 0.712 |
| Hydromorphone | 0 (0%) | 0 (0%) | 1 (0%) | 0 (0%) | 1 (1%) | 0 (0%) | 2 (2%) | 0.135 |
| Morphine | 0 (0%) | 0 (0%) | 8 (2%) | 1 (1%) | 0 (0%) | 1 (1%) | 2 (2%) | 0.196 |
In the multilevel generalized linear modelling of MME/Day, Injury Severity Score and prior opioid use were independently associated with increased opioid exposure. Increasing age and “Other” race/ethnicity were independently associated with lower opioid exposure (Table 4).
Table 4.
Associations between Demographics and Opioid Exposure
| Frequentist | Bayesian | |||||
|---|---|---|---|---|---|---|
| Coefficient | 95% Confidence Interval | p value | Coefficient | 95% Credible Interval | Posterior Probability Change>1 | |
| Morphine Milligram Equivalents per Day | ||||||
| Constant | 84.413 | 70.348, 98.478 | <0.001 | 84.403 | 55.171, 113.879 | >99% |
| Injury Severity Score | 1.655 | 0.829, 2.482 | <0.001 | 1.659 | 1.220, 2.098 | >99% |
| Age | −0.884 | −1.080, −0.688 | <0.001 | −0.882 | −1.100, −0.664 | <1% |
| Woman | 1.082 | −7.027, 9.192 | 0.794 | 1.161 | −7.938, 10.191 | 60% |
| Other | −22.868 | −36.477, −9.260 | 0.001 | −23.153 | −48.091, 1.836 | 4% |
| Yes | 29.617 | 15.469, 43.758 | <0.001 | 29.603 | 17.528, 41.820 | >99% |
| Unknown | 10.335 | −2.276, 22.947 | 0.108 | 10.201 | −1.220, 21.846 | 96% |
Frequentist coefficients mean; Bayesian coefficients are median
The longitudinal trajectory of opioid exposure from hospital day 0 through 30 was graphically represented in Figure 1. Opioid exposure decreases by a factor of 0.94 per day (95% confidence interval 0.93–0.95). Additionally, there was an increase in the odds of observing zero MME/day for every additional day (OR 1.11, 95% confidence interval 1.09–1.13). Simply, as time passes, the count of MME per day decreases and the odds of have zero MME in a day increases. Interestingly, the curve modelling the daily opioid exposure for the entire cohort (solid black line in Figure 1) fell below the overall median opioid exposure (45 MME/Day) on approximately hospital day 3. This suggests that the majority of opioid exposure occurs the first three hospital days after admission.
Figure 1. Zero-Inflated Negative Binomial Modelling of the Longitudinal Trajectory of Opioid Exposure.
In this zero-inflated negative binomial model to investigate the longitudinal trajectory of morphine milligram equivalent (MME) per day, opioid exposure decreases by a factor of 0.94 per day (95% confidence interval 0.93–0.95). Additionally, there was an increase in the odds of observing zero MME/day for every additional day (OR 1.11, 95% confidence interval 1.09–1.13).
Between and Within Trauma Center Variation
The ICC of the multilevel generalized linear model adjusting for the a priori selected variables ISS and prior opioid use was 0.071 (95% confidence interval 0.052 – 0.097). This means that 7.1% of variation in opioid exposure was attributable to center membership while the majority of variation in opioid exposure (92.9%) occurred within trauma centers (Supplemental Table 6). A graphical representation how the majority of variation can occur within trauma centers when significant differences were also observed between centers was provided in the Supplemental Figure 1. The ICC of the models created as a sensitivity analysis ranged from 0.057 to 0.072. This means that our estimates of the percent of variation in opioid exposure attributable to trauma center membership ranged from 5.7% to 7.2% and our estimates of the percent of variation in opioid exposure occurring within trauma centers ranged from 92.8% to 94.3%.
Morbidity, Mortality, and Lengths of Stay
Differences in the rate of ileus (range 0% to 8%, p<0.001), cardiac arrest (range 0% to 3%, p=0.007), unplanned intubation (range 0% to 4%, p=0.048), and mortality (range 0% to 4%, p=0.035) were observed between centers (Supplement Table 7). No difference in unplanned return to the intensive care unit (range 1% to 3%, p=0.845) was seen. There were also differences in hospital days, intensive care unit days, and ventilator days.
Increased opioid exposure was associated with small increases in hospital, intensive care unit, and ventilator days (Table 5). For every 1 MME/day, hospital, intensive care, and ventilator days increased by 0.1%, 0.2%, and 0.8%, respectively. For example, the difference between the institution with the highest and lowest median MME/day was 46 MME/day. If true, this additional 46 MME/day would be associated with an increase in hospital, intensive care unit, and ventilator days of 0.05, 0.09, and 0.37 days, respectively.
Table 5.
Lengths of Stay
| Frequentist | Bayesian | |||||
|---|---|---|---|---|---|---|
| IRR | 95% Confidence Interval | p value | IRR | 95% Credible Interval | Posterior Probability IRR>1 | |
| Hospital Days | ||||||
| Constant | 2.97 | 2.572, 3.424 | <0.001 | |||
| MME/Day | 1.001 | 1.001, 1.001 | <0.001 | 1.001 | 1.001, 1.002 | >99% |
| ISS | 1.046 | 1.042, 1.050 | <0.001 | 1.046 | 1.042, 1.050 | >99% |
| Prior opioid use | 1.001 | 0.897, 1.116 | 0.993 | 1.001 | 0.899, 1.117 | 51% |
| ICU Days | ||||||
| Constant | 0.337 | 0.234, 0.486 | <0.001 | |||
| MME/Day | 1.001 | 1.000, 1.003 | 0.008 | 1.002 | 1.001, 1.003 | >99% |
| ISS | 1.092 | 1.092, 1.115 | <0.001 | 1.096 | 1.084, 1.108 | >99% |
| Prior opioid use | 0.854 | 0.655, 1.113 | 0.244 | 0.779 | 0.595, 1.026 | 4% |
| Ventilator Days | ||||||
| Constant | 0.053 | 0.034, 0.083 | <0.001 | |||
| MME/Day | 1.008 | 1.005, 1.011 | <0.001 | 1.008 | 1.005, 1.011 | >99% |
| ISS | 1.125 | 1.107, 1.144 | <0.001 | 1.126 | 1.110, 1.146 | >99% |
| Prior opioid use | 0.598 | 0.382, 0.938 | 0.025 | 0.601 | 0.381, 0.958 | 2% |
All models adjusted for: ISS (Injury Severity Score) and prior opioid use.
MME – morphine milligram equivalents
The median of the average daily NRS pain scores also varied between trauma centers, ranging from median 3 (IQR 2, 5) to 6 (IQR 3, 7) (p=0.001). The rates of utilization of regional anesthesia ranged from 0% to 30% (p<0.001).
Discussion
In this prospective, multicenter observational study in trauma patients, the median opioid exposure amongst all patients was 45 MME/day, equivalent to 30 mg of oxycodone or 45 mg of hydrocodone per day. Increasing injury severity was associated with increased opioid exposure, while increasing age was associated with decreased opioid exposure. The majority of opioid exposure occurred in the first three days following admission. Variation in opioid exposure was due to differences between and within trauma centers, with the vast majority of variation being within trauma centers. Increasing opioid exposure was independently associated with increasing hospital, intensive care unit, and ventilator days.
To our knowledge, this study is the first to provide a global opioid exposure estimate for patients admitted to trauma centers in the United States. Other peer-reviewed studies estimating opioid exposure after injury have used different units of measurement, times of measurement, and/or types of injured patients. Nevertheless, when using similar criteria, our estimate of opioid exposure appears consistent. For example, Carver et al reported a cumulative mean opioid exposure of 208 ± 227 MME (mean ± SD) in the first 48 hours after admission in adult patients with rib fractures.(14) In our study, the mean and standard deviation of opioid exposure in all patients during the first 48 hours was 211 ± 385 MME, respectively.
Two important findings were identified in this study. First, opioid-minimizing treatment of acute pain after injury is possible. Second, there are multiple potential areas to target in order to minimize opioid exposure, including the first 72 hours after injury and within center variation.
It is becoming increasingly clear that non-opioid medications can effectively treat acute pain after injury. Acetaminophen and non-steroidal anti-inflammatory drugs appear to provide a similar degree of acute pain relief after extremity injury compared to opioids.(19) As the unintended consequences of untreated acute pain are untenable, trauma surgeons have the duty to relieve acute pain after injury and it appears that doing so in an opioid-minimizing fashion is entirely possible.(1, 4, 20, 21) For example, at the Red Duke Trauma Center at Memorial Hermann Hospital–Texas Medical Center, the implementation of a multi-modal pain regimen decreased opioid exposure by 31% and concomitantly reduced self-reported NRS pain scores.(22)
However, the additional, potential non-opioid interventions to be implemented, particularly within the first 72 hours after admission, are currently unknown. As demonstrated in this study, patients suffer varying injuries of different severities and undergo differing types of procedures. Patients may receive multiple procedures on different areas of the body, precluding the effective utilization of regional anesthesia. Indeed, only 9% of all patients included in this study received any type of regional anesthesia. Ketamine and lidocaine infusions for pain have inconsistent evidence to support their use. Clearly, a lack of high quality clinical trials in acute pain after injury have left us wanting for more interventions to effectively and responsibly treat pain without a reliance on opioids. High quality trials in acute pain management might also minimize between center variation if best practices are identified.
In the interim, efforts to address within center variation appear likely to be effective in minimizing opioid exposure after injury. While we cannot specifically determine the source of within center variation, the data from this study suggest that it was not largely due to patient differences for two reasons. Quantitatively, the ICC estimations in the sensitivity analysis had a quite narrow range (Supplemental Table 6). Indeed, removing ISS from the model changed the estimated percent variation within and between centers by only 1%. Qualitatively, there was no discernable pattern in significant opioid exposure by stratum. Some strata had significant differences in opioid exposure while the individual site may or may not have had significant differences. Some strata had no significant difference in opioid exposure while the individual sites may or may not have had significant differences. No rational explanation emerged to explain these various patterns. The stratified analysis allowed us to cluster patients who are more alike, leaving provider differences in acute pain management to account for the remainder of the variation.
Despite trauma patients being a heterogeneous population in terms of demographics, injury patterns, and injury severity, opioid-minimizing acute pain management strategies appear feasible. When evaluating the entire cohort, patients who were more severely injured and who had a prior opioid use had a higher opioid exposure than less severely injured patients and those without prior opioid use. However, in the stratified analyses, individual centers appear to be effectively treating these patients without increased opioid exposure. Site 7 had the most severely injured patients in the cohort, yet the opioid exposure when stratified by injury severity was not different. Sites 5, 6, and 7 had no difference in opioid exposure between opioid naïve patients and those with prior opioid use. Thus, reducing opioid exposure in these high-risk populations appears possible.
Current strategies to combat the opioid epidemic include high-level state and federal regulations and standardizing opioid prescribing at discharge after common procedures.(23–25) Since the majority of variation was within a center, high-level state or federal regulations are less likely to be effective in reducing opioid exposure after injury than interventions focused on the unique barriers to opioid-minimization that exist at each trauma center.(26) Indeed, such high-level changes have not been associated with consistent results.(27) Sometimes, there are even unintended consequences, such as paradoxical increases in the filling of opioid prescriptions.(28)
While standardized prescribing has been highly effective in reducing opioid diversion following elective surgery, it is unclear if these efforts could address the heterogeneity in the acute pain needs of injured patients who have suffered varying injuries of different severities and undergo differing types of procedures. In this study, 73% of patients received a prescription for opioids at discharge, with a range of 48%−83% among centers. Not only did the number of patient receiving opioids at discharge vary, but so did the actual types of opioids being prescribed. The use of tramadol at discharge varied from 0% to 89%, the use of hydrocodone at discharge varied from 1% to 66%, and the use of oxycodone at discharge varied from 1% to 92%.
Despite these many challenges in providing responsible and effective acute pain relief, the wide variation observed in both in-hospital opioid exposure and discharge opioid prescribing suggests that a process similar to the enhanced recovery after surgery protocols may actually be possible after injury.(29, 30) If all trauma centers minimize in-patient opioid exposure and variation similar to the best performing centers, then the number of opioid prescriptions at discharge should be less. In those who still require an opioid at discharge, a standardized approach to discharge prescribing focused on the in-patient opioid requirements as opposed to specific procedures may indeed be an effective manner to reduce diversion in this vulnerable patient population. The processes, interventions, and, potentially, surgeon buy-in to realize an enhanced recovery after trauma pathway, however, are currently lacking.
This study has several limitations. Most importantly, the longitudinal, global assessment of pain from admission to discharge on a large scale is currently infeasible. In this study, we utilized the ubiquitous, subjective NRS pain scores collected in the usual care of the patient. These scores, however, require the patient to be verbal. For non-verbal patients, other pain scores such as the Behavioral Pain Scale or Critical Care Pain Observation Tool have a different scale and are objective in nature, precluding the interchangeable use of them as the patient progresses from admission to discharge. In addition, the unclear temporal nature of the assessment of pain and the opioid exposure made modeling association between the two uninterpretable. Secondly, the observational design of the study limits us to stating associations as opposed to definitively determining the effects of increasing opioid exposure on outcomes, such as pain and lengths of stay. Lastly, to ensure feasibility of data collection, many variables that would have been informative were not collected, such as all procedures (e.g. spine surgery, orthopedic surgery).
Conclusion
The median opioid exposure for patients admitted to one of seven United States trauma centers was 45 MME/day, equivalent to 30 mg of oxycodone or 45 mg of hydrocodone per day. Increasing injury severity and age was associated with increased and decreased opioid exposure, respectively. The majority of opioid exposure occurred in the first 72 hours following admission.
Significant variation in opioid exposure was identified both between and within trauma centers with the vast majority of variation being within trauma centers. The implications of the results of this study are that interventions and cultural changes are needed within individual centers to reduce the opioid exposure of this vulnerable population.
Supplementary Material
Acknowledgements:
Stephanie A Savage MD MS
Jennifer L Hartwell MD
Brian L Brewer MD
Peter C Jenkins MD MS
Stephen Gale MD
Funding:
Dr. Harvin is supported by a KL2 grant from the Center for Clinical and Translational Sciences, which is funded by National Institutes of Health Clinical and Translational Award UL1 TR000371 and KL2 TR000370 from the National Center for Advancing Translational Sciences. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Center for Advancing Translational Sciences or the National Institutes of Health.
References
- 1.Joshi GP, Ogunnaike BO. Consequences of Inadequate Postoperative Pain Relief and Chronic Persistent Postoperative Pain. Anesthesiol Clin North Am. Nar 2005;23(1):21–36. [DOI] [PubMed] [Google Scholar]
- 2.Apfelbaum JL, Chen C, Mehta SS, Gan TJ. Postoperative Pain Experience: Results from a National Survey Suggest Postoperative Pain Continues to be Undermanaged. Anesth Analg. August 2003;97(2):534–40. [DOI] [PubMed] [Google Scholar]
- 3.Baratta JL, Schwenk ES, Viscusi ER. Clinical Consequences of Inadequate Pain Relief: Barriers to Optimal Pain Management. Plast Reconstr Surg. October 2014;134(4 Suppl 2):15s–21s. [DOI] [PubMed] [Google Scholar]
- 4.Jones J Jr, Southerland WB. The Importance of Optimizing Acute Pain in the Orthopedic Trauma Patient. Orthop Clin North Am. October 2017;48(4):445–65. [DOI] [PubMed] [Google Scholar]
- 5.Grace PM, Strand KA, Galer EL, Urban DJ, Wang X, Baratta MV, Fabisiak TJ, Anderson ND, Cheng K, Greene LI, et al. Morphine Paradoxically Prolongs Neuropathic Pain in Rats by Amplifying Spinal NLRP3 Inflammasomes Activation. Proc Natl Acad Sci U S A. June 2016;113(24):E3441–50. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Perkins FM, Kehlet H. Chronic Pain as an Outcome of Surgery. A review of predictive factors. Anesthesiology. October 2000;93(4):1123–33. [DOI] [PubMed] [Google Scholar]
- 7.Clay FJ, Watson WL, Newstead SV, McClure RJ. A Systematic Review of Early Prognostic fFactors for Persisting Pain following Acute Orthopedic Trauma. Pain Res Manag. Jan-Feb 2012;17(1):35–44. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Devlin JW, Skrobik Y, Gelinas C, Needham DM, Slooter AJC, Pandharipande PP, Watson PL, Weinhouse GL, Nunnally ME, Rochwerg B, et al. Clinical Practice Guidelines for the Prevention and Management of Pain, Agitation/Sedation, Delirium, Immobility, and Sleep Disruption in Adult Patients in the ICU. Crit Care Med. September 2018;46(9):e825–e73. [DOI] [PubMed] [Google Scholar]
- 9.Cicero TJ, Ellis MS, Surratt HL, Kurtz SP. The Changing Face of Heroin Use in the United States: a Retrospective Analysis of the Past 50 Years. JAMA Psychiatry. July 2014;71(7):821–6. [DOI] [PubMed] [Google Scholar]
- 10.Brummett CM, Waljee JF, Goesling J, Moser S, Lin P, Englesbe MJ, Bohnert ASB, Kheterpal S, Nallamothu BK. New Persistent Opioid Use After Minor and Major Surgical Procedures in US Adults. JAMA Surg. June 2017;152(6):e170504. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Bicket MC, Murimi IB, Mansour O, Wu CL, Alexander GC. Association of New Opioid Continuation with Surgical Specialty and Type in the United States. Am J Surg. November 2019;218(5):818–27. [DOI] [PubMed] [Google Scholar]
- 12.von Oelreich E, Eriksson M, Brattstrom O, Sjolund KF, Discacciati A, Larsson E, Oldner A. Risk Factors and Outcomes of Chronic Opioid Use following Trauma. Br Jo Surg. March 2020;107(4):413–21. [DOI] [PubMed] [Google Scholar]
- 13.Beyer CA, Poltavskiy E, Walker LE, Pettey W, Suo Y, Redd A, Rittel AG, Kazanis WH, Dunham JR, Zarzabal LA, et al. Persistent Opioid Use After Combat Injury and Subsequent Long-term Risk of Abuse: A Retrospective Cohort Study. Ann Surg. November 2019. [Epub ahead of print]. [DOI] [PubMed] [Google Scholar]
- 14.Carver TW, Kugler NW, Juul J, Peppard WJ, Drescher KM, Somberg LB, Szabo A, Yin Z, Paul JS. Ketamine Infusion for Pain Control in Adult Patients with Multiple Rib Fractures: Results of a Randomized Control Trial. J Trauma Acute Care Surg. February 2019;86(2):181–8. [DOI] [PubMed] [Google Scholar]
- 15.Moskowitz EE, Garabedian L, Hardin K, Perkins-Pride E, Asfaw M, Preslaski C, Leasia KN, Lawless R, Burlew CC, Pieracci F. A Double-Blind, Randomized Controlled Trial of Gabapentin vs. Placebo for Acute Pain Management in Critically Ill Patients with Rib Fractures. Injury. September 2018;49(9):1693–8. [DOI] [PubMed] [Google Scholar]
- 16.Harvin JA, Green CE, Vincent LE, Motley KL, Podbielski J, Miller CC, Tyson JE, Holcomb JB, Wade CE, Kao LS. Multi-modal Analgesic Strategies for Trauma (MAST): Protocol for a Pragmatic Randomized Trial. Trauma Surg Acute Care Open. 2018;3(1):e000192. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Tyson JE, Parikh NA, Langer J, Green C, Higgins RD, Network NIoCHaHDNR. Intensive Care for Extreme Prematurity--Moving beyond Gestational Age. N Engl J Med. April 2008;358(16):1672–81. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Wijeysundera DN, Austin PC, Hux JE, Beattie S & Laupacis A. Bayesian Statistical Inference Enhances the Interpretation of Contemporary Randomized Controlled Trials. J Clin Epidemiol. January 2009:62(1), 13–21. [DOI] [PubMed] [Google Scholar]
- 19.Chang AK, Bijur PE, Esses D, Barnaby DP, Baer J. Effect of a Single Dose of Oral Opioid and Nonopioid Analgesics on Acute Extremity Pain in the Emergency Department: A Randomized Clinical Trial. JAMA. November 2017;318(17):1661–7. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Schreiber S, Galai-Gat T. Uncontrolled Pain following Physical Injury as the Core-Trauma in Post-Traumatic Stress Disorder. Pain. July 1993;54(1):107–10. [DOI] [PubMed] [Google Scholar]
- 21.Hamrick KL, Beyer CA, Lee JA, Cocanour CS, Duby JJ. Multimodal Analgesia and Opioid Use in Critically Ill Trauma Patients. J Am Coll Surg. May 2019;228(5):769–75.e1. [DOI] [PubMed] [Google Scholar]
- 22.Wei S, Green C, Thanh Truong VT, Howell J, Ugarte SM, Albarado R, Taub EA, Meyer DE, Adams SD, McNutt MK, et al. Implementation of a Multi-modal Pain Regimen to Decrease Inpatient Opioid Exposure after Injury. Am J Surg. December 2019;218(6):1122–27. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.Howard R, Fry B, Gunaseelan V, Lee J, Waljee J, Brummett C, Campbell D Jr., Seese E, Englesbe M, Vu J. Association of Opioid Prescribing With Opioid Consumption After Surgery in Michigan. JAMA Surg. January 2019;154(1):e184234. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24.Howard R, Waljee J, Brummett C, Englesbe M, Lee J. Reduction in Opioid Prescribing Through Evidence-Based Prescribing Guidelines. JAMA Surg. March 2018;153(3):285–87. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25.Lillemoe HA, Marcus RK, Day RW, Kim BJ, Narula N, Davis CH, Gottumukkala V, Aloia TA. Enhanced Recovery in Liver Surgery Decreases Postoperative Outpatient Use of Opioids. Surgery. July 2019;166(1):22–7. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26.Stulberg JJ, Schafer WLA, Shallcross ML, Lambert BL, Huang R, Holl JL, Bilimoria KY, Johnson JK. Evaluating the Implementation and Effectiveness of a Multi-Component Intervention to Reduce Post-Surgical Opioid Prescribing: Study Protocol of a Mixed-Methods Design. BMJ Open. June 2019;9(6):e030404. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27.Reid DBC, Shapiro B, Shah KN, Ruddell JH, Cohen EM, Akelman E, Daniels AH. Has a Prescription-limiting Law in Rhode Island Helped to Reduce Opioid Use After Total Joint Arthroplasty? Clin Orthop Relat Res. February 2020;478(2):205–15. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28.Habbouche J, Lee J, Steiger R, Dupree JM, Khalsa C, Englesbe M, Brummett C, Waljee J. Association of Hydrocodone Schedule Change With Opioid Prescriptions Following Surgery. JAMA Surg. December 2018;153(12):1111–9. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29.Singh M, Askari R, Stopfkuchen-Evans M. Enhanced Recovery After Surgery: Are the Principles Applicable to Adult and Geriatric Acute Care and Trauma Surgery? Anesthesiol Clin. March 2019;37(1):67–77. [DOI] [PubMed] [Google Scholar]
- 30.Moydien MR, Oodit R, Chowdhury S, Edu S, Nicol AJ, Navsaria PH. Enhanced Recovery after Surgery (ERAS) in Penetrating Abdominal Trauma: a Prospective Single-Center Pilot Study. S Afr J Surg. November 2016;54(4):7–10. [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.