Persistent post-discharge opioid prescribing after traumatic brain injury requiring intensive care unit admission: A cross-sectional study with longitudinal outcome.

Traumatic brain injury (TBI) is associated with increased risk for psychological and substance use disorders. The study aim is to determine incidence and risk factors for persistent opioid prescription after hospitalization for TBI. Electronic medical records of patients age ≥ 18 admitted to a neuroscience intensive care unit between January 2013 and February 2017 for an intracranial injury were retrospectively reviewed. Primary outcome was opioid use through 12 months post-hospital discharge. A total of 298 patients with complete data were included in the analysis. The prevalence of opioid use among preadmission opioid users was 48 (87%), 36 (69%) and 22 (56%) at 1, 6 and 12-months post-discharge, respectively. In the opioid naïve group, 69 (41%), 24 (23%) and 17 (19%) were prescribed opioids at 1, 6 and 12 months, respectively. Preadmission opioid use (OR 324.8, 95% CI 23.1-16907.5, p = 0.0004) and higher opioid requirements during hospitalization (OR 4.5, 95% CI 1.8-16.3, p = 0.006) were independently associated with an increased risk of being prescribed opioids 12 months post-discharge. These factors may be used to identify and target at-risk patients for intervention.

Introduction

Traumatic brain injury is a significant global and national public health concern. According to the Global Burden of Diseases, Injuries, and Risk Factors Study, an estimated 27 million new cases of TBI occurred in 2016.[1] In the United States, the Centers for Disease Control reported an estimated 2.8 million TBI-related emergency department visits, hospitalizations, and deaths in 2013 with an estimated direct and indirect economic cost of these injuries of approximately $76.5 billion (based on 2010 data).[2] Although death rates from TBI have declined by 5% between 2007 to 2013, TBI-related emergency department visits have increased by 47% over the same period [2], suggesting that more patients are living with TBI-related disability.[3, 4]

Patients with traumatic brain injury (TBI) are at increased risk for psychological disorders including suicide, post-traumatic stress and mood disorders, and for substance abuse; however, data regarding opioid use following TBI in the general population is lacking.[5] Opioids [6] together with benzodiazepines [7] are frequently used in the acute treatment of critically brain-injured patients for the management of intracranial hypertension, to provide analgesia and anxiolysis and to facilitate mechanical ventilation. Opioids and benzodiazepines reduce cerebral metabolic oxygen consumption with minimal effects on cerebral perfusion, which makes them ideal agents for the management of TBI.[8-10] However, exposure to opioids for the acute management of TBI may increase risk for dependence and opioid use disorder in these patients. We and others have shown that in-hospital exposure to opioids increases the risk for chronic opioid use in surgical patients [11-13]; with persistent use reported as much as one year after the index admission. Therefore, it is conceivable that in-hospital exposure to opioids might predispose patients with TBI to continue to use opioids after hospital discharge.

The aim of this study is to determine the incidence and risk factors for persistent opioid prescription (up to one year after admission) in patients with a primary TBI. We hypothesized that opioid use prior to hospitalization and in-hospital exposure to opioids for management of TBI would be associated with increased risk for persistent opioid prescription 1 year after hospital discharge.

Material and methods

This manuscript adheres to the STROBE guidelines. The study was approved by the Institutional Review Board (IRB) for Health Science Research (HSR-19922) and the requirement for written informed consent was waived by the IRB. All human and animal studies have been approved by the institution ethics committee and have been performed in accordance with the ethical standards laid down in the 1964 Declaration of Helsinki and its later amendments.

This was a single-center retrospective study. ICD-10 diagnosis code, demographic data and the electronic medical record were used to identify patients age ≥ 18 who were admitted between January 2013 and February 2017 to our neuroscience intensive care unit (ICU) with a primary admission diagnosis of TBI. The neuroscience ICU cares for patients with isolated traumatic brain or spine injury. Patients with polytrauma are cared for in our Surgical Trauma ICU and were not included in our patient cohort. Surgical CPT codes were used to identify any surgical procedures that were performed during the index admission of the study cohort. Patients with major chest, abdomen, pelvis, vascular or long-bone fractures with an associated TBI were excluded from this study, to remove the confounding effect of non-neurological injuries on persistent post discharge opioid prescription. The primary outcome was persistent opioid prescription through 12 months after discharge, as defined by having a prescription for one or more opioid medications at each of 1, 6, and 12-month time points. Secondary outcomes were persistent opioid prescription at 1 and 6 months post-discharge, defined as prescription at 1 month and prescription at 1 and 6 months, respectively.

Covariates

Patient demographic data (age, sex, ethnicity), weight, body mass index (BMI), psychiatric and substance abuse history, preadmission medication use including non-steroidal anti-inflammatory medications (NSAIDs), acetaminophen, benzodiazepines, muscle relaxants, antidepressants, antipsychotics, opioids and daily oral morphine-equivalent (OME) dose (for dose conversion see http://www.uptodate.com/contents/cancerpain-management-with-opioids-optimizing-analgesia) [14] (complete medication list in S1 Table), in-hospital Verbal Response Scale (VRS) and Critical Care Pain Observation Tool (CPOT) pain scores, in hospital medication administrations, opioid administration 48 hours prior to discharge (yes/no), in-hospital intubation, surgical CPT codes, intensive care unit (ICU) and hospital length of stay (LOS), discharge disposition and prescription medication data from 1, 6 and 12 months after discharge were recorded. These covariates were chosen a priori based on previous studies that demonstrated an association with persistent opioid use after surgery (e.g. antidepressant use, substance abuse history), biological plausibility (e.g. in-hospital opioid exposure) and previously unexplored variables (e.g. opioid exposure 48 hrs prior to discharge and discharge disposition).

OME was calculated by converting and summating all enteral and parenteral opioid administered within a 24-hour period, which was then standardized to BMI (mg/kg/m2). To control for duration of opioid exposure during the hospital stay the total daily OME was divided by length of hospital stay (days) and BMI.

A sensitivity analysis was performed in those patients who were taking opioids pro re nata (PRN) prior to their admission. Subjects were assumed to be taking either some (25%) or all (100%) of the PRN dose. For example, a subject with a prescription for “10 mg oxycodone every 4 hours PRN pain” was estimated to be taking between 25% of the prescribed amount (15 mg oxycodone) or 100% of the prescribed amount (60 mg oxycodone) daily.

Data validation

To ensure data accuracy for all reported variables, we used the Microsoft Excel function “RANDBETWEEN” to generate 150 random numbers between the highest and lowest case numbers. Thereafter, we manually validated all data for the associated random generated case numbers using the hospital electronic medical record.

Statistical analysis

Descriptive statistics are presented as number (n) and proportion (%) for dichotomous variables, and mean (M) and standard deviation (SD) for continuous variables.

Differences between patients using opioids prior to admission and opioid naïve patients were compared using Fisher/chi-square tests (for dichotomous or categorical variables), linear regression models (for continuous variables), and generalized linear regression models with a logit function (for count variables). Subjects with missing preadmission opioid use data were excluded from the analysis.

Generalized linear models were used to examine whether the odds of persistent opioid prescription 1, 6 or 12 months after discharge. Prior to all predictive modeling, collinearity among covariates was assessed using the variance inflation factor: if greater than 5, covariates were selectively excluded from the model. All analyses were performed in R version 3.3.2.[15]

Results

Patient demographics and preadmission characteristics

Of 528 records reviewed, preadmission opioid use data was missing for 134 subjects, 65 patients had no TBI on radiological imaging and 32 were readmissions following their TBI. A total of 297 subjects were included in the final analysis. Seventy-six were preadmission opioid users and 221 were opioid naïve prior to admission. A consort diagram is shown in Fig 1.

Fig 1

Consolidated standards of Reporting trials (CONSORT) diagram.

Patient demographics, preadmission characteristics and in-hospital exposures are shown in Table 1. A greater percentage of preadmission opioid users were women (p = 0.05). Preadmission history of depression (p < 0.001), use of acetaminophen (p < 0.001), benzodiazepines (p < 0.001), muscle relaxants (p = 0.009), and antidepressants (p = 0.013) were significantly higher among preadmission opioid users compared to opioid naïve patients. Among patients using opioids prior to admission, the mean daily OME dose was 22.8 ± 55.2 mg (minimum 0.00, maximum 407.5 mg) assuming 25% of the prescribed PRN dose and mean 28.1 ± 56.7 mg (minimum 0.0 mg, maximum 430.0 mg) assuming 100% of the prescribed PRN dose.

Table 1

Comparison of patient demographics, preadmission characteristics and in-hospital exposures between preadmission opioid users and opioid naïve patients.

Characteristic	Total(n = 431)	Preadmit Opioid Use(n = 76)	Opioid Naïve(n = 221)	p value
Preadmission Characteristics
Age (years)	63 ± 20	64 ± 18	65 ± 19	0.64
Sex	263 (61%) Men	37 (49%) Men	139 (63%) Men	0.05
	168 (39%) Women	39 (51%) Women	83 (37%) Women
Ethnicity	389 (90%) White	68 (89%) White	199 (90%) White	1.00
	32 (8%) Black	6 (8%) Black	19 (9%) Black
	10 (2%) Other	1 (0.01%) Other	4 (0.02%) Other
BMI (kg/m2)	26.4 ± 5.5	26.0± 5.3	26.9 ± 5.6	0.25
Admission diagnosis
Concussion	21 (5%)	1 (1%)	13 (6%)	0.105
Laceration/contusion	35 (8%)	5 (7%)	23 (105)	0.322
SAH, SDH, extradural hemorrhage	217 (50%)	46 (61%)	111 (50%)	0.121
Other unspecified ICH	39 (9%)	7 (9%)	17 (8%)	0.674
Other intracranial injury	119 (28%)	17 (22%)	57 (26%)	0.549
Tobacco use	88 (23%)	19 (26%)	42 (20%)	0.435
Alcohol use	155 (43%)	25 (36%)	84 (43%)	0.430
Recreational drug use	23 (7%)	5 (8%)	8 (5%)	0.346
Depression history	73 (19%)	28 (39%)	36 (17%)	< 0.001
Anxiety history	32 (8%)	9 (13%)	16 (7%)	0.293
NSAIDs use	132 (44%)	36 (47%)	96 (43%)	0.667
Acetaminophen use	59 (20%)	43 (57%)	16 (7%)	< 0.001
Benzodiazepine use	47 (16%)	22 (29%)	25 (11%)	< 0.001
Muscle relaxant use	24 (8%)	12 (16%)	12 (5%)	0.009
Antidepressants use	72 (24%)	27 (36%)	45 (20%)	0.013
Antipsychotics use	17 (6%)	6 (8%)	11 (5%)	0.390
In-hospital exposures
Intubation	233/431 (54%)	40/76 (53%)	107/221 (48%)	0.616
Surgical Procedure	119/297 (40%)	34/76 (45%)	81/221 (37%)	0.211
Intracranial surgery	69/119 (58%)	20/34 (59%)	49/81 (60%)	0.865
Spine surgery	32/119 (27%)	9/34 (26%)	23/81 (28%)	0.834
Other orthopedic surgery	8/119 (7%)	4/34 (12%)	4/81 (5%)	0.190
Minor procedures	10/119 (8%)	1/34 (3%)	9/81 (11%)	0.156
NSAIDS (# administrations)	6.38 ± 9.34 [0.00, 98.00]	6.01 ± 6.90 [0.00, 35.00]	5.43 ± 7.32 [0.00, 73.00]	0.666
Benzodiazepines (# administrations)	1.75 ± 3.41 [0.00, 30.00]	1.82 ± 3.83 [0.00, 29.00]	1.33 ± 3.26 [0.00, 30.00]	0.002
Muscle relaxants (# administrations)	0.79 ± 3.09 [0.00, 36.00]	1.66 ± 5.25 [0.00, 36.00]	0.44 ± 1.81 [0.00, 13.00]	< 0.001
Antidepressants (# administrations)	2.90 ± 6.99 [0.00, 74.00]	3.21 ± 6.44 [0.00, 33.00]	2.32 ± 5.44 [0.00, 36.00]	< 0.001
Opioids (# administrations)	5.98 ± 7.86 [0.00, 92.00]	5.66 ± 5.12 [0.00, 26.00]	4.76 ± 5.94 [0.00, 37.00]	0.003
Average ME (mg/ (kg/m2) per day)	0.88 ± 2.19 [0.00, 21.07]	0.83 ± 1.44 [0.00, 8.11]	0.71 ± 1.92 [0.00, 21.07]	0.612
Opioid 48 hours prior to discharge	199/431 (46%)	63/76 (83%)	136/221 (62%)	0.001
ICU length of stay (days)	3.83 ± 9.58 [0.10, 179.00]	2.80 ± 3.21 [0.30, 17.80]	2.61 ± 3.49 [0.10, 31.00]	0.676
Hospital length of stay (days)	9.28 ± 13.36 [0.30, 179.00]	7.96 ± 8.02 [1.90, 49.90]	7.68 ± 8.69 [0.30, 95.50]	0.810
Discharge
Home	134/297 (45%)	37/76 (51%)	97/221 (44%)	0.471
Skilled Nursing	72/297 (24%)	16/76 (22%)	56/221 (25%)	0.453
Rehabilitation	62/297 (21%)	14/76 (19%)	48/221 (22%)	0.542
Death	20/297 (7%)	6/76 (8%)	14/221 (6%)	0.638

Data presented as number (percentage) or mean ± standard deviation [minimum, maximum].

Body mass index (BMI), intracranial hemorrhage (ICH), intensive care unit (ICU), morphine equivalent (ME), non-steroidal anti-inflammatory drug (NSAID), subarachnoid hemorrhage (SAH), subdural hemorrhage (SDH)

There were similar rates of intubation and surgery between groups. There were no significant differences in discharge disposition between groups.

Daily and weekly OME (per BMI), average daily OME (per BMI) are shown in S2 Table. VRS and CPOT pain scores were excluded from the analysis due to large amounts of missing data. Patients using opioids prior to admission received significantly more benzodiazepine (p = 0.002), muscle relaxant (p<0.001), antidepressant (p<0.001), and opioid (p = 0.003) administrations compared to opioid naïve patients. Sixty-three patients (83%) using opioids prior to admission received opioids in the 48 hours prior to discharge compared to 136 (62%) of previously opioid naïve patients, which was a statistically significant difference (p = 0.001).

The proportion of patients prescribed post-discharge opioids and other medications is shown in Table 2. Among preadmission opioid users the point prevalence of opioid prescription at 1, 6 and 12-month post-discharge was 48 (87%), 36 (69%) and 22 (56%) while among previously opioid naïve patients the prevalence rate was 69 (41%), 24 (23%) and 17 (19%). A greater number of preadmission opioid users had a prescription for opioids at 1, 6 and 12 months compared to opioid naïve patients at each of these time points (p<0.001). Preadmission opioid users were also prescribed more benzodiazepines at 1 (p = 0.009) months and more muscle relaxants at 1 (p<0.001) and 6 months (p = 0.046).

Table 2

Descriptive statistics comparing post-discharge opioid use between preadmission opioid users and opioid naïve patients.

Medication	Totaln = 431n (%)	Preadmit opioid usen = 76n (%)	Opioid naïven = 221n (%)	p value
1 month
Opioid use	155 (52%)	48 (87%)	69 (41%)	< 0.001
NSAIDs	91 (30%)	20 (36%)	45 (27%)	0.267
Acetaminophen	148 (50%)	28 (51%)	79 (48%)	0.815
Benzodiazepine	50 (17%)	17 (31%)	23 (14%)	0.009
Muscle relaxants	41 (14%)	15 (27%)	13 (8%)	< 0.001
Antidepressants	88 (29%)	22 (40%)	51 (31%)	0.283
Antipsychotics	41 (14%)	8 (15%)	21 (13%)	0.933
6 months
Opioid use	70 (35%)	36 (69%)	24 (23%)	< 0.001
NSAIDs	80 (40%)	22 (42%)	42 (40%)	0.880
Acetaminophen	79 (39%)	19 (37%)	42 (40%)	0.841
Benzodiazepine	30 (15%)	11 (21%)	14 (13%)	0.292
Muscle relaxants	27 (13%)	11 (21%)	9 (9%)	0.046
Antidepressants	72 (36%)	22 (42%)	39 (37%)	0.620
Antipsychotics	21 (10%)	8 (16%)	8 (8%)	0.195
12 months
Opioid use	47 (30%)	22 (56%)	17 (19%)	< 0.001
NSAIDs	71 (45%)	19 (49%)	41 (47%)	1
Acetaminophen	60 (38%)	17 (44%)	33 (38%)	0.687
Benzodiazepine	27 (17%)	9 (23%)	15 (17%)	0.599
Muscle relaxants	18 (13%)	8 (21%)	8 (9%)	0.089
Antidepressants	58 (37%)	16 (41%)	35 (40%)	1
Antipsychotics	20 (13%)	6 (15%)	9 (10%)	0.552

Data presented as number (percentage).

Non-steroidal anti-inflammatory drug (NSAID)

Persistent opioid prescription after hospital discharge

The number of preadmission opioid users who were persistently prescribed opioids through 1, 6 and 12 months was 48 (87%), 30 (61%) and 18 (44%), respectively. Sixty-nine (42%), 15 (11%) and 3 (2%) previously opioid naïve patients were persistently prescribed opioids through each of these timepoints. Generalized linear models with a logit function were used to examine odd of persistent opioid prescription 1, 6 and 12 months after discharge (Table 3). Testing collinearity using variance inflation factors (VIF) indicated no problematic amount of collinearity among the independent variables (all VIF < 5). Preadmission opioid use was associated with increased odds of being prescribed opioids at 1 (OR 8.7, 95% CI 3.2–27.1, p<0.0001), 6 (OR 23.4, 95% CI 6.9–97.0, p<0.0001) and 12 months (OR 324.8, 95% CI 23.1–16907.5, p = 0.0004). Administration of opioids 48 hours prior to discharge (OR 3.3, 95% CI 1.4–8.0, p = 0.007) was associated with higher odds of being prescribed opioids 1-month post-discharge. Older age was associated with higher odds of persistent opioid prescription at 6 (OR 1.1, 95% CI 1.0–1.1, p = 0.005) and 12 months after discharge, as was increased average ME (day/kg /m2) at 6 (OR 1.8, 95% CI 1.2–2.0, p = 0.008) and 12 months (OR 4.5, 95% CI 1.8–16.3, p = 0.006). History of tobacco use was associated with decreased risk of persistent opioid prescription at 1 (OR 0.2, 95% CI 0.1–0.5, p = 0.002), 6 (OR 0.2, 95% CI 0.04–0.9, p = 0.037) and 12 months (OR 0.03, 95% CI 0.00–0.36, p = 0.014). Discharge to a skilled nursing facility (OR 0.03, 90% CI 0.00–0.61, p = 0.048) was also associated with decreased risk.

Table 3

Generalized linear model comparing odds of remaining on opioids after discharge between preadmission opioid users and naive opioid patients.

Covariate	1 monthn = 117		6 monthsn = 45		12 monthsn = 21
	OR(95% CI)	p value	OR(95% CI)	p value	OR(95% CI)	p value
Preadmission opioid use	8.68 (3.24–27.05)	<0.0001	23.44 (6.86–96.98)	< 0.0001	324.77 (23.10–16907.45)	0.0004
Age	0.99 (0.97–1.01)	0.444	1.06 (1.02–1.11)	0.0046	1.14 (1.05–1.30)	0.012
Sex (male)	0.64 (0.30–1.37)	0.251	0.92 (0.29–3.01)	0.892	0.09 (0.01–0.87)	0.053
Race (non-white)	1.18 (0.34–4.08)	0.796	4.00 (0.76–22.41)	0.105	1.22 (0.06–19.89)	0.888
Prior tobacco use	0.18 (0.06–0.51)	0.002	0.21 (0.04–0.87)	0.037	0.03 (0.00–0.36)	0.014
Prior alcohol use	1.40 (0.6–3.05)	0.391	1.39 (0.44–4.54)	0.577	2.52 (0.21–36.47)	0.462
Prior recreational drug use	1.34 (0.10–13.16)	0.810	7.90 (0.64–191.27)	0.144	0.52 (0.00–1504.57)	0.831
Prior benzodiazepine use	1.00 (0.31–3.22)	0.996	0.86 (0.17–4.71)	0.853	3.82 (0.08–1399.21)	0.552
Prior antidepressant use	1.30 (0.49–3.52)	0.603	0.27 (0.05–1.29)	0.100	4.01 (0.12–190.16)	0.444
Prior antipsychotic use	2.23 (0.40–12.26)	0.348	19.36 (0.66–2780.46)	0.178	0.34 (0.00–449.00)	0.685
History of depression	0.69 (0.24–1.93)	0.477	5.10 (0.95–33.76)	0.071	0.25 (0.01–8.21)	0.419
History of anxiety	2.00 (0.44–9.15)	0.365	0.55 (0.06–4.71)	0.580	0.49 (0.02–13.12)	0.659
Inpatient intubation	1.42 (0.62–3.33)	0.417	2.05 (0.67–6.58)	0.213	1.89 (0.24–16.07)	0.535
Hospital stay (days)	0.98 (0.92–1.04)	0.495	1.02 (0.96–1.11)	0.687	0.99 (0.78–1.17)	0.916
Average ME (mg/(kg/m2) per day)	1.25 (0.94–1.84)	0.186	1.80 (1.19–2.87)	0.008	4.48 (1.78–16.34)	0.006
Opioids in last 48 hours prior to discharge	3.27 (1.41–7.99)	0.007	3.50 (1.04–13.36)	0.052	0.25 (0.01–3.05)	0.292
Discharge to skilled nursing facility	1.79 (0.64–5.09)	0.269	0.56 (0.11, 2.71)	0.480	0.03 (0.00–0.61)	0.048
Discharge to rehabilitation	1.57 (0.60–4.24)	0.361	2.25 (0.58, 9.38)	0.248	1.48 (0.12–19.23)	0.749

Data reported as odds ratio (95% confidence interval).

Body mass index (BMI), intensive care unit (ICU), non-steroidal anti-inflammatory drug (NSAID), morphine equivalent (ME).

Comparison of surgical versus non-surgical patients

Generalized linear models were used to determine the influence of surgery on risk for persistent opioid prescription at 1, 6 and 12 months among patients with traumatic brain injury. Results in Table 4 show that there was a statistically significant main effect of pre-admission opioid use on the likelihood of persistent opioid prescription at 1, 6, and 12 months; patients with pre-admission opioid use were more likely to be prescribed opioids at 1, 6, and 12 months. The main effect of surgery on the likelihood of chronic opioid use was not statistically significant; there were no statistically significant differences in the likelihood of chronic opioid use at 1, 6, and 12 months between surgical and non-surgical cases. The interaction effect between pre-admission opioid use and surgical cases was not statistically significant.

Table 4

Generalized linear model comparing the effect of preadmission opioid use and surgery on persistent opioid prescription 1, 6 and 12 months post-discharge.

Covariate	1 month		6 months		12 months
	OR(95% CI)	p value	OR(95% CI)	p value	OR(95% CI)	p value
Preadmission opioid use	5.54 (2.18–16.12)	< 0.001	11.50 (4.25–33.74)	< 0.001	24.46 (6.60–199.91)	< 0.001
Surgery	0.97 (0.52–1.80)	0.913	0.81 (0.26–2.39)	0.709	0.64 (0.03–6.82)	0.716
Preadmission opioid use X surgery	6.21 (0.86–127.91)	0.116	1.30 (0.27–6.58)	0.751	1.63 (0.11–42.85)	0.725

Patients with missing opioid data

Among the 134 patients with no preadmission opioid data, post-discharge opioid usage data was missing for 55 patients at 1 month, 67 patients at 6 months and 69 patients respectively. Of those with valid post-discharge data, 38 (48%) were prescribed opioids 1 month post discharge; 7 (10%) were prescribed opioids at 6 months post-discharge and 3 (5%) were prescribed opioids at 12 months post-discharge. To evaluate for potential sources of bias, we compared preadmission characteristics of patients with and without preadmission opioid data. There were no statistically significant differences in sex (p = 0.22), race (p = 0.30), and BMI (p = 0.16). Patients with preadmission opioid data were statistically significantly older (mean age 65 years versus 57 years, p = 0.001), had a higher incidence of recreational drug use (10 (13%) versus 13 (5%), p = 0.04), a lower incidence of depression history (9 (9%) versus 64 (22%), p<0.001) and were more likely to be intubated (86 (64%) versus 147 (49%), p = 0.01).

Discussion

Our study is the first to examine risk factors for persistent opioid prescription following ICU admission for TBI. A major finding of this study, which is of significant clinical importance, is that prior opioid use before an unanticipated ICU admission is associated with persistent opioid use up to a year after surgery. This is congruent with other perioperative studies investigating persistent opioid use after elective surgery. However, our findings are significantly different than Wang et al., who showed that amongst chronic opioid users, hospitalization with critical illness was not associated with substantial increases in opioids prescribed in the 6 months following hospitalization.[16] The implications of this on a practical level is improved vigilance, especially after discharge, of the high risk preoperative opioid user who sustains a TBI.

The amount of opioids patients receive during hospitalization (average OME per BMI per day) and opioid use prior to discharge (opioids in the last 48 hours) are significantly associated with persistent opioid prescription. This appears to be independent of injury severity as intubation, length of hospital stay, and discharge disposition were not associated with increased risk. These findings are similar to the work by Harbaugh et al, who showed that in patients undergoing third molar (wisdom tooth) extraction, persistent opioid use occurs at an adjusted rate of 13 (95% CI, 9–19) per 1000 patients who filled an opioid prescription compared with 5 (95% CI, 3–7) per 1000 patients who did not fill a perioperative prescription.[17] Our study is the first to demonstrate this important risk factor in a critical care cohort.

Together these findings suggest that ICU management of patients influences risk for persistent opioid prescription and may help identify patients at risk for chronic opioid use and misuse after hospital discharge. Reducing opioid exposure through use of non-opioid sedatives [18], multimodal analgesia [19] and limiting opioid prescription at discharge may play a key role in reducing the risk for chronic opioid dependence after TBI. This is consistent with the Society for Critical Care Medicine’s 2018 Clinical Practice Guidelines for Management of Pain, Agitation/Sedation, Delirium, Immobility and Sleep Disruption which recommend systemic assessment with validated pain and sedation scales to reduce opioid use, use of multimodal analgesia including non-opioid adjuncts and consideration of sedation agents, such as propofol or dexmedetomidine.[20]

TBI is a significant global health concern; however, few studies have investigated opioid use in this population. To date the majority of studies examining opioid use after TBI have focused on military service members.[21] This study extends these findings in a broader population group and identifies risk factors for persistent opioid prescription in this population.

Our study has several limitations: the study design was retrospective in nature with significant amounts of missing data. We characterized and compared the missing data with the complete cohort data, which demonstrated lower depression history and higher recreational drug use among patients with missing data. However, rates of chronic opioid prescription at 1, 6 and 12 months were similar and we do not believe that eliminating them from the analysis would have altered the results significantly.

Our cohort was intended to include only patients with a primary TBI; however, 40% of patients required a surgical intervention during the hospitalization, primarily intracranial or spine procedures. We do not believe that this significantly affected our results, as surgical cases were not associated with increased odds for persistent opioid use at 1, 6 or 12 months compared to non-surgical cases. The impact of poly-trauma requiring critical care needs to be investigated separately.

A significant limitation of our study is that we were unable to stratify patients by injury severity as measures of injury severity (i.e. Glascow Coma Scale, Revised Trauma Score, APACHE score) are not routinely documented in our electronic medical record. We attempted to control for opioid exposure (mild injury/less exposure versus severe injury/greater exposure) by normalizing the mg OME per BMI per day in our generalized in linear models. Whether or not a patient was intubated on admission, the length of hospital stay and discharge disposition were used as surrogates for injury severity, before admission, during hospitalization and after discharge, respectively. However, this represents a significant limitation of our study.

Finally, we observed that 44% of preadmission opioid users were not prescribed opioids 12 months after admission. Many of these patients were discharged to a rehabilitation or skilled nursing facility (57% (8/14) compared to 39% (7/18) preadmission opioid users who continued to be prescribed opioids). We were unable to reliably track opioid prescription as patients transitioned from in-patient hospitalization to SNF or rehabilitation care, thus the veracity of this data is in question. Interestingly, discharge to a skilled nursing facility was associated with decreased odds of persistent opioid prescription at 12 months. It is possible that patients discharged to SNF/rehabilitation facilities received opioids that were not documented in our study. The Virginia Prescription Monitoring Program could be used to validate opioid prescription data from multiple care facilities; however, this data is not available for research purposes.

Our results raise several important unanswered questions: Is decreasing opioid use in previous opioid requiring patients associated with worse modified Rankin Scale or Glasgow Coma Outcome Scales where opioids are not administered per patient request? Does re-establishing healthcare amongst previous opioid users following a TBI reduce opioid use? Our findings suggest areas for further exploration to determine factors associated with opioid cessation in this group of patients.

Conclusions

Among primary TBI patients requiring critical care, history of opioid use, depression, higher average daily opioid dose during hospitalization, and receiving opioids 48h prior to discharge are associated with increased risk of being prescribed opioids up to one year after discharge. These factors may be used to identify TBI patients who are at highest risk and target them for intervention and counseling.

Medication list.

(DOCX)

Click here for additional data file.

Average total morphine equivalent dose administered per day and week between preadmission opioid users and opioid naïve patients.

(DOCX)

Click here for additional data file.

Dataset.

(CSV)

Click here for additional data file.

Transfer Alert

This paper was transferred from another journal. As a result, its full editorial history (including decision letters, peer reviews and author responses) may not be present.

5 Jul 2019

PONE-D-19-15116

Persistent Post-Discharge Opioid Prescribing after Traumatic Brain Injury Requiring ICU Admission: A Cross-Sectional Study with Longitudinal Outcome

PLOS ONE

Dear Dr. Dunn,

Thank you for submitting your manuscript to PLOS ONE. After careful consideration, we feel that it has merit but does not fully meet PLOS ONE’s publication criteria as it currently stands. Therefore, we invite you to submit a revised version of the manuscript that addresses the points raised during the review process.

The reviewers suggest several modifications to the analysis that would improve its novelty, interpretation and alignment with your stated objectives. For example:

1. Please consider Reviewer 1's suggestions for additional analyses to explore how TBI is specifically a risk factor for prolonged opioid use.

2. Please address the concerns related to model overfitting

3. A major issue relates to the stated objective compared to the modeling approach taken (as described by the second reviewer). In particular, it appears that the models should be set-up with in-hospital opioid exposure as the main independent variable, with preadmission opioid use considered as an effect modifier (ie in a stratified analysis).

We would appreciate receiving your revised manuscript by Aug 19 2019 11:59PM. When you are ready to submit your revision, log on to https://www.editorialmanager.com/pone/ and select the 'Submissions Needing Revision' folder to locate your manuscript file.

If you would like to make changes to your financial disclosure, please include your updated statement in your cover letter.

To enhance the reproducibility of your results, we recommend that if applicable you deposit your laboratory protocols in protocols.io, where a protocol can be assigned its own identifier (DOI) such that it can be cited independently in the future. For instructions see: http://journals.plos.org/plosone/s/submission-guidelines#loc-laboratory-protocols

Please include the following items when submitting your revised manuscript:

A rebuttal letter that responds to each point raised by the academic editor and reviewer(s). This letter should be uploaded as separate file and labeled 'Response to Reviewers'.

A marked-up copy of your manuscript that highlights changes made to the original version. This file should be uploaded as separate file and labeled 'Revised Manuscript with Track Changes'.

An unmarked version of your revised paper without tracked changes. This file should be uploaded as separate file and labeled 'Manuscript'.

Please note while forming your response, if your article is accepted, you may have the opportunity to make the peer review history publicly available. The record will include editor decision letters (with reviews) and your responses to reviewer comments. If eligible, we will contact you to opt in or out.

We look forward to receiving your revised manuscript.

Kind regards,

Tara Gomes

Academic Editor

PLOS ONE

Journal Requirements:

1. When submitting your revision, we need you to address these additional requirements.

Please ensure that your manuscript meets PLOS ONE's style requirements, including those for file naming. The PLOS ONE style templates can be found at

http://www.journals.plos.org/plosone/s/file?id=wjVg/PLOSOne_formatting_sample_main_body.pdf and http://www.journals.plos.org/plosone/s/file?id=ba62/PLOSOne_formatting_sample_title_authors_affiliations.pdf

2.

We note that you have indicated that data from this study are available upon request. PLOS only allows data to be available upon request if there are legal or ethical restrictions on sharing data publicly. For information on unacceptable data access restrictions, please see http://journals.plos.org/plosone/s/data-availability#loc-unacceptable-data-access-restrictions.

In your revised cover letter, please address the following prompts:

a) If there are ethical or legal restrictions on sharing a de-identified data set, please explain them in detail (e.g., data contain potentially identifying or sensitive patient information) and who has imposed them (e.g., an ethics committee). Please also provide contact information for a data access committee, ethics committee, or other institutional body to which data requests may be sent.

b) If there are no restrictions, please upload the minimal anonymized data set necessary to replicate your study findings as either Supporting Information files or to a stable, public repository and provide us with the relevant URLs, DOIs, or accession numbers. Please see http://www.bmj.com/content/340/bmj.c181.long for guidelines on how to de-identify and prepare clinical data for publication. For a list of acceptable repositories, please see http://journals.plos.org/plosone/s/data-availability#loc-recommended-repositories.

We will update your Data Availability statement on your behalf to reflect the information you provide.

3. We note that you have included the phrase “data not shown” in your manuscript. Unfortunately, this does not meet our data sharing requirements. PLOS does not permit references to inaccessible data. We require that authors provide all relevant data within the paper, Supporting Information files, or in an acceptable, public repository. Please add a citation to support this phrase or upload the data that corresponds with these findings to a stable repository (such as Figshare or Dryad) and provide and URLs, DOIs, or accession numbers that may be used to access these data. Or, if the data are not a core part of the research being presented in your study, we ask that you remove the phrase that refers to these data.

[Note: HTML markup is below. Please do not edit.]

Reviewers' comments:

Reviewer's Responses to Questions

Comments to the Author

1. Is the manuscript technically sound, and do the data support the conclusions?

The manuscript must describe a technically sound piece of scientific research with data that supports the conclusions. Experiments must have been conducted rigorously, with appropriate controls, replication, and sample sizes. The conclusions must be drawn appropriately based on the data presented.

Reviewer #1: No

Reviewer #2: Partly

**********

2. Has the statistical analysis been performed appropriately and rigorously?

Reviewer #1: Yes

Reviewer #2: No

**********

3. Have the authors made all data underlying the findings in their manuscript fully available?

The PLOS Data policy requires authors to make all data underlying the findings described in their manuscript fully available without restriction, with rare exception (please refer to the Data Availability Statement in the manuscript PDF file). The data should be provided as part of the manuscript or its supporting information, or deposited to a public repository. For example, in addition to summary statistics, the data points behind means, medians and variance measures should be available. If there are restrictions on publicly sharing data—e.g. participant privacy or use of data from a third party—those must be specified.

Reviewer #1: Yes

Reviewer #2: Yes

**********

4. Is the manuscript presented in an intelligible fashion and written in standard English?

PLOS ONE does not copyedit accepted manuscripts, so the language in submitted articles must be clear, correct, and unambiguous. Any typographical or grammatical errors should be corrected at revision, so please note any specific errors here.

Reviewer #1: Yes

Reviewer #2: Yes

**********

5. Review Comments to the Author

Please use the space provided to explain your answers to the questions above. You may also include additional comments for the author, including concerns about dual publication, research ethics, or publication ethics. (Please upload your review as an attachment if it exceeds 20,000 characters)

Reviewer #1: Please see uploaded review for :

Manuscript Number: PONE-D-19-15116

Article Type: Research Article

Full Title: Persistent Post-Discharge Opioid Prescribing after Traumatic Brain Injury Requiring

ICU Admission: A Cross-Sectional Study with Longitudinal Outcome

Short Title: Persistent Opioid Use After Traumatic Brain Injury

Corresponding Author: Lauren Kingsley Dunn

University of Virginia

Charlottesville, VA UNITED STATES

Keywords: traumatic brain injury; opioid dependence; opioid use disorder; Subarachnoid

hemorrhage; subdural hemorrhage

Reviewer #2: Thank you for the opportunity to review this interesting article. The authors performed a retrospective analysis of patients admitted to the neurological intensive care unit to evaluate risk factors for prolonged opioid use after discharge. They hypothesized that increased ICU opioid exposure would result in increased risk of prolonged opioid prescription. They separated their analysis into two groups: patients on pre-admission opioids and patients who were opioid naïve. They found that pre-admission opioid use and higher opioid requirements during hospitalization were associated with increased risk of being prescribed opioids 12 months post discharge.

The authors are to be commended on a well-written analysis of opioid prescribing patterns after admission for TBI. This is a timely topic, and of interest to readers. However, there are several areas that could be improved upon. Please see below for comments:

Page 6 – The authors lay out their argument for their interest in opioid use after TBI – 1) TBI is a significant global health concern with increasing incidence and significant cost associated 2) patients with TBI have been shown to have increased substance abuse disorders, including increased chronic use of opioids 3) however, opioids are ideal drugs to treat TBI in the acute phase 4) but they are also concerned that opioid use in the ICU will lead to prolonged prescription, as they themselves have already shown this to be a risk factor. Therefore 5) they want to see if exposure to opioids is a risk factor for prolonged prescription after TBI. It is unclear what new information about the topic they are offering, if they have already shown that TBI increases risk of chronic opioids, and opioid exposure in the hospital increases chronic opioids. Can the authors please more explicitly state what new information they are adding to the literature? (I.e., highlight the differences in you cohort – civilian vs military?, looking at only opioid-naïve patients?, etc.)

P9, L149 – Can the authors please comment on the reason for choosing the covariates?

P10, L183 – The model is confusing to me – why was a linear regression instead of logistic regression used for a dichotomous outcome? Opioid rx at 1,6,12 months (yes/no)? Additionally, considering only 48 patients still had an opioid rx at 12 months, the model is overfit with 19 covariates.

P14, L 222 – It might be reasonable to combine all the covariates of interest in Table 1. This would allow the authors to simplify the first paragraph on p14 into something more streamlined such as, “There were similar rates of intubation and surgery between the two groups. There was no difference in the types of surgeries between the two groups (see Table 1).” This will allow the text to be easier read and prevent redundant numbers.

P16, Table 2 – It is unclear why the model, which was supposed to determine whether the independent variable of ICU opioid-exposure is associated with the outcome, is presented in terms of pre-admission opioid users vs opioid-naïve patients? The question posited originally was “does ICU-opioid exposure increase prolonged opioid prescription?” Pre-admission opioid use should have been considered as a covariate in this model. In order to answer the original question, it seems the analysis should be redone, with the same outcome, the independent variable of ICU-opioid exposure, and only those covariates that influence the relationship between the independent variable and dependent variable, eg, pre-admission opioid exposure, without overfitting the model.

P 19 – Table 3-5 can be combined with columns for 1, 6, and 12 months for more expedient reading. This will also highlight easily across all time points that pre-admission opioid use was consistently significant.

P28, L350 – Is it not clear that in-hospital opioid exposure is independently associated with prolonged prescription. Presumably this effect is due to pre-admission opioid use, which led to higher inpatient use, which then led to continued post-ICU use. This should be addressed by either using an interaction term, or analyzing naïve and non-naïve patients separately to examine the effect of ICU-opioid exposure on outcome.

P28 Discussion – much of the introduction is repeated in the discussion. The authors have already made the case that opioid exposure after surgical procedures has been associated with increased risk of prolonged opioid use. A more effective use of this space would be to discuss how the author’s new data will influence practice. Is it feasible to avoid opioids in critically ill opioid-dependent patients? Are there other drugs that can be used in TBI care if you identify a patient to be at risk for prolonged prescription? How can one interpret the SCCM analagosedation guidelines given this data?

Minor:

Page 8, L 137 – Consistency with naming of the Neuro ICU – neurological (as in the abstract) or neuroscience?

P10, L186 – “Models were used to examine whether the odds of receiving opioid prescriptions at 1, 6, or 12 months after discharge…” This should be changed to AND to match the abstract and covariate section, which list the outcome as 1, 6, and 12 months. Otherwise the outcome is confusing, since the authors seem to be looking at a continuous exposure from 1-12months.

P14, line 222 – Intubation is not listed as one if the covariates on page 9.

**********

6. PLOS authors have the option to publish the peer review history of their article (what does this mean?). If published, this will include your full peer review and any attached files.

If you choose “no”, your identity will remain anonymous but your review may still be made public.

Do you want your identity to be public for this peer review? For information about this choice, including consent withdrawal, please see our Privacy Policy.

Reviewer #1: No

Reviewer #2: No

[NOTE: If reviewer comments were submitted as an attachment file, they will be attached to this email and accessible via the submission site. Please log into your account, locate the manuscript record, and check for the action link "View Attachments". If this link does not appear, there are no attachment files to be viewed.]

While revising your submission, please upload your figure files to the Preflight Analysis and Conversion Engine (PACE) digital diagnostic tool, https://pacev2.apexcovantage.com/. PACE helps ensure that figures meet PLOS requirements. To use PACE, you must first register as a user. Registration is free. Then, login and navigate to the UPLOAD tab, where you will find detailed instructions on how to use the tool. If you encounter any issues or have any questions when using PACE, please email us at figures@plos.org. Please note that Supporting Information files do not need this step.

Submitted filename: Manuscript Number-PONE-D-19-15116.doc

Click here for additional data file.

9 Sep 2019

Response to reviewers

The reviewers suggest several modifications to the analysis that would improve its novelty, interpretation and alignment with your stated objectives.

We thank the Editor and Reviewers for the opportunity to submit a revised manuscript. We have attempted to address each of the reviewers concerns below.

For example:

1.Please consider Reviewer 1's suggestions for additional analyses to explore how TBI is specifically a risk factor for prolonged opioid use.

Please see our response to Reviewer 1’s suggestions. Results of our generalized linear models show other factors associated with prolonged opioid use in this patient population, specifically in hospital opioid exposure, opioid administration < 48 prior to discharge.

2. Please address the concerns related to model overfitting

To limit model overfitting, we reduced the number of covariates in each model by removing BMI, NSAIDs, acetaminophen, and muscle relaxants. In addition, we performed 3 separate analyses. 1) all covariates, 2) patient demographics and preadmission characteristics, 3) patient demographics and in hospital exposures. Results of models 2 and 3 were similar to model 1, thus we have included only model 1, with all the covariates in the current manuscript. Please see our response to Reviewer 2 comment

3. A major issue relates to the stated objective compared to the modeling approach taken (as described by the second reviewer). In particular, it appears that the models should be set-up with in-hospital opioid exposure as the main independent variable, with preadmission opioid use considered as an effect modifier (ie in a stratified analysis).

We respectfully note that Reviewer 2 was referring to Table 2 in the prior manuscript which displayed descriptive statistics. In-hospital opioid exposure (mg ME/kg/day) and preadmission opioid use are included as separate covariates in our generalized linear models shown in Tables 3 (1 month), 4 (6 months) and 5 (12 months). We have clarified this in the text.

We would appreciate receiving your revised manuscript by Aug 19 2019 11:59PM. When you are ready to submit your revision, log on to https://www.editorialmanager.com/pone/ and select the 'Submissions Needing Revision' folder to locate your manuscript file.

If you would like to make changes to your financial disclosure, please include your updated statement in your cover letter.

To enhance the reproducibility of your results, we recommend that if applicable you deposit your laboratory protocols in protocols.io, where a protocol can be assigned its own identifier (DOI) such that it can be cited independently in the future. For instructions see: http://journals.plos.org/plosone/s/submission-guidelines#loc-laboratory-protocols

Please include the following items when submitting your revised manuscript:

• A rebuttal letter that responds to each point raised by the academic editor and reviewer(s). This letter should be uploaded as separate file and labeled 'Response to Reviewers'.

• A marked-up copy of your manuscript that highlights changes made to the original version. This file should be uploaded as separate file and labeled 'Revised Manuscript with Track Changes'.

• An unmarked version of your revised paper without tracked changes. This file should be uploaded as separate file and labeled 'Manuscript'.

Please note while forming your response, if your article is accepted, you may have the opportunity to make the peer review history publicly available. The record will include editor decision letters (with reviews) and your responses to reviewer comments. If eligible, we will contact you to opt in or out.

We look forward to receiving your revised manuscript.

Kind regards,

Tara Gomes
Academic Editor
PLOS ONE

Journal Requirements:

1. When submitting your revision, we need you to address these additional requirements.

Please ensure that your manuscript meets PLOS ONE's style requirements, including those for file naming. The PLOS ONE style templates can be found at

http://www.journals.plos.org/plosone/s/file?id=wjVg/PLOSOne_formatting_sample_main_body.pdf and http://www.journals.plos.org/plosone/s/file?id=ba62/PLOSOne_formatting_sample_title_authors_affiliations.pdf

2.

We note that you have indicated that data from this study are available upon request. PLOS only allows data to be available upon request if there are legal or ethical restrictions on sharing data publicly. For information on unacceptable data access restrictions, please see http://journals.plos.org/plosone/s/data-availability#loc-unacceptable-data-access-restrictions.

Data has been made available.

In your revised cover letter, please address the following prompts:

a) If there are ethical or legal restrictions on sharing a de-identified data set, please explain them in detail (e.g., data contain potentially identifying or sensitive patient information) and who has imposed them (e.g., an ethics committee). Please also provide contact information for a data access committee, ethics committee, or other institutional body to which data requests may be sent.

b) If there are no restrictions, please upload the minimal anonymized data set necessary to replicate your study findings as either Supporting Information files or to a stable, public repository and provide us with the relevant URLs, DOIs, or accession numbers. Please see http://www.bmj.com/content/340/bmj.c181.long for guidelines on how to de-identify and prepare clinical data for publication. For a list of acceptable repositories, please see http://journals.plos.org/plosone/s/data-availability#loc-recommended-repositories.

We will update your Data Availability statement on your behalf to reflect the information you provide.

3. We note that you have included the phrase “data not shown” in your manuscript. Unfortunately, this does not meet our data sharing requirements. PLOS does not permit references to inaccessible data. We require that authors provide all relevant data within the paper, Supporting Information files, or in an acceptable, public repository. Please add a citation to support this phrase or upload the data that corresponds with these findings to a stable repository (such as Figshare or Dryad) and provide and URLs, DOIs, or accession numbers that may be used to access these data. Or, if the data are not a core part of the research being presented in your study, we ask that you remove the phrase that refers to these data.

This phrase has been removed and appropriate data provided.

[Note: HTML markup is below. Please do not edit.]

Reviewers' comments:

Reviewer's Responses to Questions

Comments to the Author

1. Is the manuscript technically sound, and do the data support the conclusions?

The manuscript must describe a technically sound piece of scientific research with data that supports the conclusions. Experiments must have been conducted rigorously, with appropriate controls, replication, and sample sizes. The conclusions must be drawn appropriately based on the data presented. 

Reviewer #1: No

Reviewer #2: Partly

2. Has the statistical analysis been performed appropriately and rigorously? 

Reviewer #1: Yes

Reviewer #2: No

3. Have the authors made all data underlying the findings in their manuscript fully available?

The PLOS Data policy requires authors to make all data underlying the findings described in their manuscript fully available without restriction, with rare exception (please refer to the Data Availability Statement in the manuscript PDF file). The data should be provided as part of the manuscript or its supporting information, or deposited to a public repository. For example, in addition to summary statistics, the data points behind means, medians and variance measures should be available. If there are restrictions on publicly sharing data—e.g. participant privacy or use of data from a third party—those must be specified.

Reviewer #1: Yes

Reviewer #2: Yes

4. Is the manuscript presented in an intelligible fashion and written in standard English?

PLOS ONE does not copyedit accepted manuscripts, so the language in submitted articles must be clear, correct, and unambiguous. Any typographical or grammatical errors should be corrected at revision, so please note any specific errors here.

Reviewer #1: Yes

Reviewer #2: Yes

5. Review Comments to the Author

Please use the space provided to explain your answers to the questions above. You may also include additional comments for the author, including concerns about dual publication, research ethics, or publication ethics. (Please upload your review as an attachment if it exceeds 20,000 characters)

Reviewer #1: Please see uploaded review for :

Manuscript Number: PONE-D-19-15116
Article Type: Research Article
Full Title: Persistent Post-Discharge Opioid Prescribing after Traumatic Brain Injury Requiring
ICU Admission: A Cross-Sectional Study with Longitudinal Outcome
Short Title: Persistent Opioid Use After Traumatic Brain Injury
Corresponding Author: Lauren Kingsley Dunn
University of Virginia
Charlottesville, VA UNITED STATES
Keywords: traumatic brain injury; opioid dependence; opioid use disorder; Subarachnoid
hemorrhage; subdural hemorrhage

The authors have chosen to review a timely and relevant issue, opioid exposure and the sequelae. Their aim,

“to determine incidence and risk factors for persistent opioid prescription after hospitalization for TBI.”

The methodology employed was a retrospective review. The statistical analysis is noted to be appropriate.

There are however some limitations posed by the protocol design, the methodology and the conclusions, that must be noted/addressed:

#1. The premise of the study is an extrapolation of risk from the military TBI cohort. This is however difficult, as the TBI military cohort, also known as the Polytrauma cohort does not have a high specificity in identifying TBI. For many reasons, patients are included as TBI, allowing a high sensitivity in the diagnosis at the cost of specificity. These patients also have significant co-morbid injuries that impact comparisons of risk for opioid use. Thus any extrapolation is significantly limited.

We thank Reviewer 1 for this comment. It was not our intention to compare risk for opioid use in our TBI cohort to previous studies in military veterans. Military veteran data were discussed only to highlight that prior studies have focused only on this cohort and data regarding risk for opioid use in the general population after TBI is lacking, and as the Reviewer notes represents a very different group. We have limited discussion of the military veteran population to the discussion.

#2. The exclusion of major injuries is noted,

“to remove the confounding effect of non- neurological injuries on persistent post discharge opioid prescription…”

However, there is no confirmation that these patients are isolated TBI. There are many other conflating injuries that might significantly impact pain and subsequent opioid use. These are not noted to be excluded. Further supporting this concern, is that spine surgical patients are included in the cohort.

No injuries outside of intracranial and spinal injuries in this ICU. Sensitivity analysis removing patients having surgical procedure.

We thank Reviewer #1 for making this important point. Our patient cohort was selected based on admission to our Neuroscience Intensive Care Unit which cares for patients with isolated traumatic brain injury or spine injury. Patients presenting with polytrauma are admitted to our Surgical Trauma Intensive Care Unit. This has been added to the methods page 5 line 104-106.

Number of surgical procedures by type has been added to Table 1. Of 115 surgical procedures, a majority were intracranial or spine procedures. Orthopedic fractures and other minor surgical procedures (skin debridement, tracheostomy) accounted for <10% each.

We have also included comparison of the effect of surgery on risk for persistent opioid use in the Results page 19, lines 511-520. “There was a statistically significant effect of pre-admission opioid use on the likelihood of persistent opioid prescription at 1, 6, and 12 months; patients with pre-admission opioid use were more likely to be prescribed opioids at 1, 6, and 12 months. The main effect of surgery on the likelihood of chronic opioid use was not statistically significant; there were no statistically significant differences in the likelihood of chronic opioid use at 1, 6, and 12 months between surgical and non-surgical cases. The interaction effect between pre-admission opioid use and surgical cases was not statistically significant.”

#3. The TBI patients are not stratified by severity.

“Occasionally high doses of opioids are employed for sustained periods of time in patients with a severe TBI requiring mechanical ventilation…”

Severely injured patients are alluded to as needing sedation as a risk factor, but without stratification, it is unclear how the identified risk factors impacts the specific cohorts. As an example, mild patients would have less exposure and presumably less risk????

Again, Reviewer #1 makes an excellent point. Unfortunately, measures of injury severity (i.e. Glasgow Coma Scale, Revised Trauma Score, APACHE score) are not routinely documented in our electronic medical record. We attempt to control for opioid exposure (mild injury/less exposure vs severe injury/greater exposure) by normalizing mg oral morphine equivalent per BMI per day in our generalized in linear models. In addition, we included 3 covariates in our generalized linear models 1) whether or not a patient was intubated on admission, 2) length of hospital stay and 3) discharge disposition as surrogates for injury severity before admission, during hospitalization and after discharge, respectively. We have emphasized this a limitation in the discussion section Page 23 lines 699-706.

#4. The stratification of patients with pre versus post injury opioid use is of benefit. The findings however do not differ from the established understanding of risk associated with previous use. Unclear how TBI becomes an additional risk factor beyond previous use.

Preadmission opioid use is one factor associated with risk for chronic opioid prescription. Key findings from our study are that, after controlling for measures of injury severity and hospital length of stay, the amount of opioids that patients are exposed to during their hospitalization and whether or not they are weaned off opioid prior to discharge is significantly associated with their risk for being prescribed opioids one year after discharge. This suggests that ICU management of patients may influence their risk for chronic opioid use. Strategies aimed at reducing opioid exposure, such as use of non-opioid sedative infusions and multimodal analgesia and limiting opioid prescription at discharge may play a key role in reducing the risk for chronic opioid dependence after TBI and warrant future study. We clarify the importance of these findings in our discussion. Please also see our response to Reviewer 2 comment 1.

#5. The discussion section focuses on the perils of opioid use established, but fails to either clearly establish the risks from the study nor the link to the devastating sequelae highlighted.

Please see our response to Comment #4 above. We have revised the discussion section in order to more clearly highlight the important findings from our study which are that in-hospital management of these patient may contribute to long-term risk for chronic opioid prescription. Potential risk reduction strategies such as use of non-opioid sedative agents (i.e. dexmedetomidine) or multimodal analgesia and limiting opioid prescribing at discharge are important areas for future study.

In summary, it is agreed that the risk associated with opioid use in TBI patients is of concern. The manuscript however does not clearly identify risk factors specific to TBI beyond the known pre- injury use and further does not account for concomitant injuries.

Would recommend a cohort of isolated TBI versus those with additional injuries to better understand opioid use in TBI.

We thank the Reviewer for this recommendation. Please see our response to Comment #1. We hope that the inclusion of an analysis of the effect of surgery on risk for persistent opioid use appropriately addresses this concern.

Would also recommend stratification of injury type, for better understanding of the risk assumed to exist in the severe cohort.

We sincerely thank the Reviewer for this recommendation. Unfortunately, we are limited in our ability to stratify patients by injury severity as the necessary measures (GCS, Revised trauma score, APACHE score) are not documented in our electronic medical record. We recognize that this is a significant limitation of our study, which we have attempted to clearly articulate in the discussion section. We control for opioid exposure by normalizing for mg ME per BMI per day and account for injury severity to the best of our ability by using surrogate measures (intubation, length of stay, and discharge disposition) in our models.

Reviewer #2: Thank you for the opportunity to review this interesting article. The authors performed a retrospective analysis of patients admitted to the neurological intensive care unit to evaluate risk factors for prolonged opioid use after discharge. They hypothesized that increased ICU opioid exposure would result in increased risk of prolonged opioid prescription. They separated their analysis into two groups: patients on pre-admission opioids and patients who were opioid naïve. They found that pre-admission opioid use and higher opioid requirements during hospitalization were associated with increased risk of being prescribed opioids 12 months post discharge.

The authors are to be commended on a well-written analysis of opioid prescribing patterns after admission for TBI. This is a timely topic, and of interest to readers. However, there are several areas that could be improved upon. Please see below for comments:

Page 6 – The authors lay out their argument for their interest in opioid use after TBI – 1) TBI is a significant global health concern with increasing incidence and significant cost associated 2) patients with TBI have been shown to have increased substance abuse disorders, including increased chronic use of opioids 3) however, opioids are ideal drugs to treat TBI in the acute phase 4) but they are also concerned that opioid use in the ICU will lead to prolonged prescription, as they themselves have already shown this to be a risk factor. Therefore 5) they want to see if exposure to opioids is a risk factor for prolonged prescription after TBI. It is unclear what new information about the topic they are offering, if they have already shown that TBI increases risk of chronic opioids, and opioid exposure in the hospital increases chronic opioids. Can the authors please more explicitly state what new information they are adding to the literature? (I.e., highlight the differences in you cohort – civilian vs military?, looking at only opioid-naïve patients?, etc.)

Thank you for the succinct overview of the study.

1. One of the major findings of this study, which is of significant clinical importance, is that prior opioid use before an unanticipated ICU admission is associated with persistent opioid use up to a year after surgery. This finding is congruent with other perioperative studies investigating persistent opioid use after elective surgery. However, our findings are significantly different than Wang et al. (Crit Care Med. 2018 Dec;46(12):1934-1942), who showed that amongst chronic opioid users, hospitalization with critical illness was not associated with substantial increases in opioids prescribed in the 6 months following hospitalization. The implications of this finding on a practical level is improved vigilance, especially after discharge, of the high risk preoperative opioid user who sustains a TBI.

2. A second important finding of our study is the significance of opioid use prior to discharge (Opioids last 48 hrs). This finding is similar to the work by Harbaugh et al (JAMA. 2018 Aug 7;320(5):504-506) who showed that in patients undergoing third molar (wisdom tooth) extraction, persistent opioid use occurs at an adjusted rate of 13 (95% CI, 9-19) per 1000 patients who filled an opioid prescription compared with 5 (95% CI, 3-7) per 1000 patients who did not fill a perioperative prescription. Our study is the first to demonstrate this important risk factor in a critical care cohort.

We have limited our discussion of the military cohort in the introduction and revised the discussion to highlight what new information our study is providing. Please also see our response to Reviewer 1 comment 4.

P9, L149 – Can the authors please comment on the reason for choosing the covariates?

Covariates were chosen a priori based on previous studies that demonstrated an association with persistent opioid use after surgery (e.g. antidepressant use, substance abuse history), biological plausibility (e.g. in-hospital opioid exposure) and previously unexplored variables (e.g. opioid exposure 48 hrs prior to discharge and discharge disposition). This has been added to the Methods Page 6, lines 128-132. 

P10, L183 – The model is confusing to me – why was a linear regression instead of logistic regression used for a dichotomous outcome? Opioid rx at 1,6,12 months (yes/no)? Additionally, considering only 48 patients still had an opioid rx at 12 months, the model is overfit with 19 covariates.

In response to your first question on linear vs. logistic regression:

1. P10 L183 refers to the models used to examine differences between opioid naïve and chronic opioid patents. As described in P10 L185, we use generalized linear models to examine whether the odds of receiving opioid prescriptions (yes/no) differ between opioid naïve and chronic opioid patients. When the outcome variable is dichotomous, generalized linear models with a logit function can also be referred as logistic regression models. In the revised manuscript, we clarified that we used a logit function in our generalized linear models.

With regard to the question of over-fitting, the reviewer makes an important observation which we have addressed in the following way:

1. We have reduced the number of co-variates in our full model for each time period by removing BMI, muscle relaxants, NSAIDS, and Acetaminophen

2. To further reduce the risk of overfitting we created 2 additional models for the three time points (1, 6 and 12 months). They include:

a. Demographic and Preadmission Exposures

b. Demographics + In-hospital Exposures

Results of the two additional models were similar so we elected to include only the results of the full model for each time point in Table 3. Please see results of these additional models below:

Persistent Opioid Prescription (1 month)

Full model (1 month)

Demographics and preadmit exposures (1 month):

Demographics and in hospital exposures (1 month)

\f

Persistent Opioid Prescription (6 month)

Full model (6 month)

Demographics and preadmit exposures (6 month):

\f

Demographics and in hospital exposures (6 month)

\f

Persistent Opioid Prescription (12 month)

Full model (12 month)

Demographics and preadmit exposures (12 month):

Demographics and in hospital exposures (12 month)

\f

P14, L 222 – It might be reasonable to combine all the covariates of interest in Table 1. This would allow the authors to simplify the first paragraph on p14 into something more streamlined such as, “There were similar rates of intubation and surgery between the two groups. There was no difference in the types of surgeries between the two groups (see Table 1).” This will allow the text to be easier read and prevent redundant numbers.

Thank you for this suggestion. We have made the appropriate changes in the manuscript and added In-hospital Characteristics and Exposures to Table 1

P16, Table 2 – It is unclear why the model, which was supposed to determine whether the independent variable of ICU opioid-exposure is associated with the outcome, is presented in terms of pre-admission opioid users vs opioid-naïve patients? The question posited originally was “does ICU-opioid exposure increase prolonged opioid prescription?” Pre-admission opioid use should have been considered as a covariate in this model. In order to answer the original question, it seems the analysis should be redone, with the same outcome, the independent variable of ICU-opioid exposure, and only those covariates that influence the relationship between the independent variable and dependent variable, eg, pre-admission opioid exposure, without overfitting the model.

Table 3 on Page 16 (Table 2 in revised manuscript) is a univariate analysis on post-discharge medications between patients with Preadmit Opioid Use vs. Opioid Naïve. There is no multivariable modeling reported in this table. Preadmission opioid use is used as a covariate in all our multivariable modeling (current Table 3). As discussed earlier, to reduce overfitting we have provided additional restricted models.

P 19 – Table 3-5 can be combined with columns for 1, 6, and 12 months for more expedient reading. This will also highlight easily across all time points that pre-admission opioid use was consistently significant.

Done

P28, L350 – Is it not clear that in-hospital opioid exposure is independently associated with prolonged prescription. Presumably this effect is due to pre-admission opioid use, which led to higher inpatient use, which then led to continued post-ICU use. This should be addressed by either using an interaction term, or analyzing naïve and non-naïve patients separately to examine the effect of ICU-opioid exposure on outcome.

We respectfully disagree with the reviewer regarding his concern about in-hospital opioid exposure correlating to preadmission opioid use. For each full model we tested for collinearity using variance inflation factors (VIF) which indicated no problematic amount of collinearity among the independent variables (all VIF < 5) for Average ME (day/kg/m2) and Chronic opioid user. This has been noted on page 15, lines 326-328. 

P28 Discussion – much of the introduction is repeated in the discussion. The authors have already made the case that opioid exposure after surgical procedures has been associated with increased risk of prolonged opioid use. A more effective use of this space would be to discuss how the author’s new data will influence practice. Is it feasible to avoid opioids in critically ill opioid-dependent patients? Are there other drugs that can be used in TBI care if you identify a patient to be at risk for prolonged prescription? How can one interpret the SCCM analagosedation guidelines given this data?

Thank you for this suggestion. We have modified our discussion based on your excellent suggestion (Page 21-22, lines 577-588). 



Minor:

Page 8, L 137 – Consistency with naming of the Neuro ICU – neurological (as in the abstract) or neuroscience?

This has been corrected. 

P10, L186 – “Models were used to examine whether the odds of receiving opioid prescriptions at 1, 6, or 12 months after discharge…” This should be changed to AND to match the abstract and covariate section, which list the outcome as 1, 6, and 12 months. Otherwise the outcome is confusing, since the authors seem to be looking at a continuous exposure from 1-12months.

This has been corrected.

P14, line 222 – Intubation is not listed as one if the covariates on page 9.

This has been added.

Submitted filename: Response to Reviewers .docx

Click here for additional data file.

4 Nov 2019

PONE-D-19-15116R1

Persistent post-discharge opioid prescribing after traumatic brain injury requiring intensive care unit admission: a cross-sectional study with longitudinal outcome

PLOS ONE

Dear Dr. Dunn,

Thank you for submitting your manuscript to PLOS ONE. After careful consideration, we feel that it has merit but does not fully meet PLOS ONE’s publication criteria as it currently stands. Therefore, we invite you to submit a revised version of the manuscript that addresses the points raised during the review process.

We would appreciate receiving your revised manuscript by Dec 19 2019 11:59PM. When you are ready to submit your revision, log on to https://www.editorialmanager.com/pone/ and select the 'Submissions Needing Revision' folder to locate your manuscript file.

If you would like to make changes to your financial disclosure, please include your updated statement in your cover letter.

To enhance the reproducibility of your results, we recommend that if applicable you deposit your laboratory protocols in protocols.io, where a protocol can be assigned its own identifier (DOI) such that it can be cited independently in the future. For instructions see: http://journals.plos.org/plosone/s/submission-guidelines#loc-laboratory-protocols

Please include the following items when submitting your revised manuscript:

A rebuttal letter that responds to each point raised by the academic editor and reviewer(s). This letter should be uploaded as separate file and labeled 'Response to Reviewers'.

A marked-up copy of your manuscript that highlights changes made to the original version. This file should be uploaded as separate file and labeled 'Revised Manuscript with Track Changes'.

An unmarked version of your revised paper without tracked changes. This file should be uploaded as separate file and labeled 'Manuscript'.

Please note while forming your response, if your article is accepted, you may have the opportunity to make the peer review history publicly available. The record will include editor decision letters (with reviews) and your responses to reviewer comments. If eligible, we will contact you to opt in or out.

We look forward to receiving your revised manuscript.

Kind regards,

Tara Gomes

Academic Editor

PLOS ONE

[Note: HTML markup is below. Please do not edit.]

Reviewers' comments:

Reviewer's Responses to Questions

Comments to the Author

1. If the authors have adequately addressed your comments raised in a previous round of review and you feel that this manuscript is now acceptable for publication, you may indicate that here to bypass the “Comments to the Author” section, enter your conflict of interest statement in the “Confidential to Editor” section, and submit your "Accept" recommendation.

Reviewer #2: (No Response)

**********

2. Is the manuscript technically sound, and do the data support the conclusions?

The manuscript must describe a technically sound piece of scientific research with data that supports the conclusions. Experiments must have been conducted rigorously, with appropriate controls, replication, and sample sizes. The conclusions must be drawn appropriately based on the data presented.

Reviewer #2: Yes

**********

3. Has the statistical analysis been performed appropriately and rigorously?

Reviewer #2: Yes

**********

4. Have the authors made all data underlying the findings in their manuscript fully available?

The PLOS Data policy requires authors to make all data underlying the findings described in their manuscript fully available without restriction, with rare exception (please refer to the Data Availability Statement in the manuscript PDF file). The data should be provided as part of the manuscript or its supporting information, or deposited to a public repository. For example, in addition to summary statistics, the data points behind means, medians and variance measures should be available. If there are restrictions on publicly sharing data—e.g. participant privacy or use of data from a third party—those must be specified.

Reviewer #2: Yes

**********

5. Is the manuscript presented in an intelligible fashion and written in standard English?

PLOS ONE does not copyedit accepted manuscripts, so the language in submitted articles must be clear, correct, and unambiguous. Any typographical or grammatical errors should be corrected at revision, so please note any specific errors here.

Reviewer #2: Yes

**********

6. Review Comments to the Author

Please use the space provided to explain your answers to the questions above. You may also include additional comments for the author, including concerns about dual publication, research ethics, or publication ethics. (Please upload your review as an attachment if it exceeds 20,000 characters)

Reviewer #2: The authors have done a good job responding to most of my concerns. The table updates make the article easier to read. However, there is an outstanding issues:

My confusion regarding the modeling can be clarified if the introduction is changed. While I still do not think the current model answers their stated hypothesis (ICU opioid exposure as an independent variable affecting outcome) – it does answer a question. It appears that the authors are just searching for risk factors for persistent opioid prescription in patients with TBI. However, they couch all their data and discussion in terms of pre-admission opioid users and opioid naïve patients. The hypothesis in the introduction could be changed to the following:

“The aim of this study is to determine the incidence and risk factors for persistent opioid prescription (up to one year after admission) in patients with a primary TBI. We hypothesized that opioid use prior to hospitalization and in-hospital exposure to opioids for management of TBI would be associated with increased risk for persistent opioid prescription 1 year after hospital discharge.”

This would clarify why they present all the data stratified by pre-admission exposure vs naïve.

With this, it should be ready for publication from my standpoint.

**********

7. PLOS authors have the option to publish the peer review history of their article (what does this mean?). If published, this will include your full peer review and any attached files.

If you choose “no”, your identity will remain anonymous but your review may still be made public.

Do you want your identity to be public for this peer review? For information about this choice, including consent withdrawal, please see our Privacy Policy.

Reviewer #2: No

[NOTE: If reviewer comments were submitted as an attachment file, they will be attached to this email and accessible via the submission site. Please log into your account, locate the manuscript record, and check for the action link "View Attachments". If this link does not appear, there are no attachment files to be viewed.]

While revising your submission, please upload your figure files to the Preflight Analysis and Conversion Engine (PACE) digital diagnostic tool, https://pacev2.apexcovantage.com/. PACE helps ensure that figures meet PLOS requirements. To use PACE, you must first register as a user. Registration is free. Then, login and navigate to the UPLOAD tab, where you will find detailed instructions on how to use the tool. If you encounter any issues or have any questions when using PACE, please email us at figures@plos.org. Please note that Supporting Information files do not need this step.

6 Nov 2019

Response to Reviewers

Reviewer #2: The authors have done a good job responding to most of my concerns. The table updates make the article easier to read. However, there is an outstanding issue:

My confusion regarding the modeling can be clarified if the introduction is changed. While I still do not think the current model answers their stated hypothesis (ICU opioid exposure as an independent variable affecting outcome) – it does answer a question. It appears that the authors are just searching for risk factors for persistent opioid prescription in patients with TBI. However, they couch all their data and discussion in terms of pre-admission opioid users and opioid naïve patients. The hypothesis in the introduction could be changed to the following:

“The aim of this study is to determine the incidence and risk factors for persistent opioid prescription (up to one year after admission) in patients with a primary TBI. We hypothesized that opioid use prior to hospitalization and in-hospital exposure to opioids for management of TBI would be associated with increased risk for persistent opioid prescription 1 year after hospital discharge.”
This would clarify why they present all the data stratified by pre-admission exposure vs naïve.

With this, it should be ready for publication from my standpoint.

We thank Reviewer 2 for their review of our revise manuscript. We have made the recommended change to the hypothesis stated in the Introduction Page 4, line 66-70.

Submitted filename: Response to Reviewers.docx

Click here for additional data file.

13 Nov 2019

Persistent post-discharge opioid prescribing after traumatic brain injury requiring intensive care unit admission: a cross-sectional study with longitudinal outcome

PONE-D-19-15116R2

Dear Dr. Dunn,

We are pleased to inform you that your manuscript has been judged scientifically suitable for publication and will be formally accepted for publication once it complies with all outstanding technical requirements.

Within one week, you will receive an e-mail containing information on the amendments required prior to publication. When all required modifications have been addressed, you will receive a formal acceptance letter and your manuscript will proceed to our production department and be scheduled for publication.

Shortly after the formal acceptance letter is sent, an invoice for payment will follow. To ensure an efficient production and billing process, please log into Editorial Manager at https://www.editorialmanager.com/pone/, click the "Update My Information" link at the top of the page, and update your user information. If you have any billing related questions, please contact our Author Billing department directly at authorbilling@plos.org.

If your institution or institutions have a press office, please notify them about your upcoming paper to enable them to help maximize its impact. If they will be preparing press materials for this manuscript, you must inform our press team as soon as possible and no later than 48 hours after receiving the formal acceptance. Your manuscript will remain under strict press embargo until 2 pm Eastern Time on the date of publication. For more information, please contact onepress@plos.org.

With kind regards,

Tara Gomes

Academic Editor

PLOS ONE

Additional Editor Comments (optional):

Reviewers' comments:

18 Nov 2019

PONE-D-19-15116R2

Persistent post-discharge opioid prescribing after traumatic brain injury requiring intensive care unit admission: a cross-sectional study with longitudinal outcome

Dear Dr. Dunn:

I am pleased to inform you that your manuscript has been deemed suitable for publication in PLOS ONE. Congratulations! Your manuscript is now with our production department.

If your institution or institutions have a press office, please notify them about your upcoming paper at this point, to enable them to help maximize its impact. If they will be preparing press materials for this manuscript, please inform our press team within the next 48 hours. Your manuscript will remain under strict press embargo until 2 pm Eastern Time on the date of publication. For more information please contact onepress@plos.org.

For any other questions or concerns, please email plosone@plos.org.

Thank you for submitting your work to PLOS ONE.

With kind regards,

PLOS ONE Editorial Office Staff

on behalf of

Dr. Tara Gomes

Academic Editor

PLOS ONE