Article Text

Download PDFPDF

Incidence of persistent opioid use following traumatic injury
Free
  1. Matthew C Mauck1,2,
  2. Ying Zhao2,
  3. Amy M Goetzinger2,
  4. Andrew S Tungate1,2,
  5. Alex B Spencer2,
  6. Asim Lal1,2,
  7. Chloe E Barton1,2,
  8. Francesca Beaudoin3 and
  9. Samuel A McLean1,4
  1. 1 Institute for Trauma Recovery, The University of North Carolina School of Medicine, Chapel Hill, North Carolina, USA
  2. 2 Anesthesiology, The University of North Carolina School of Medicine, Chapel Hill, North Carolina, USA
  3. 3 Department of Emergency Medicine, Brown University Warren Alpert Medical School, Providence, Rhode Island, USA
  4. 4 Emergency Medicine, The University of North Carolina School of Medicine, Chapel Hill, North Carolina, USA
  1. Correspondence to Dr Matthew C Mauck, Institute for Trauma Recovery, The University of North Carolina at Chapel Hill School of Medicine, Chapel Hill CB#7011, North Carolina, USA; Matt_Mauck{at}med.unc.edu

Abstract

Introduction Major traumatic injuries are a known risk factor for persistent opioid use, but data describing the relationship between specific traumatic injuries and opioid use is lacking.

Methods We used insurance claims data from January 1, 2001 to December 31, 2020 to estimate the incidence of new persistent opioid use in three hospitalized trauma populations: individuals hospitalized after burn injury (3809, 1504 of whom required tissue grafting), individuals hospitalized after motor vehicle collision (MVC; 9041), and individuals hospitalized after orthopedic injury (47, 637). New persistent opioid use was defined as receipt of ≥1 opioid prescriptions 90–180 days following injury in an individual with no opioid prescriptions during the year prior to injury.

Results New persistent opioid use was observed in 12% (267/2305) of individuals hospitalized after burn injury with no grafting, and 12% (176/1504) of burn injury patients requiring tissue grafting. In addition, new persistent opioid use was observed in 16% (1454/9041) of individuals hospitalized after MVC, and 20% (9455/47, 637) of individuals hospitalized after orthopedic trauma. In comparison, rates of persistent opioid use in all trauma cohorts (19%, 11, 352/60, 487) were greater than the rates of persistent opioid use in both non-traumatic major surgery (13%) and non-traumatic minor surgery (9%).

Conclusions These data demonstrate that new persistent opioid use frequently occurs in these common hospitalized trauma populations. Improved interventions to reduce persistent pain and opioid use in patients hospitalized after these and other traumas are needed.

  • CHRONIC PAIN
  • Analgesics, Opioid
  • Outcome Assessment, Health Care
  • Pain Management

Data availability statement

Data may be obtained from a third party and are not publicly available. IBM MarketScan.

Statistics from Altmetric.com

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

WHAT IS ALREADY KNOWN ON THIS TOPIC

  • Chronic pain and emotional distress are common after traumatic injury.

WHAT THIS STUDY ADDS

  • This study estimates the incidence of chronic opioid use among survivors of specific major traumatic injuries including burn injury, motor vehicle collisions and orthopedic trauma.

HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY

  • Understanding rates of persistent opioid use among common traumas is helpful to raise awareness among trauma care providers, policy-makers, and researchers in order to effect positive change.

Introduction

Traumatic injuries are common. Each year, approximately two million Americans are involved in a motor vehicle collision (MVC),1 a half million suffer burn injury,2–4 and 600,000 have an orthopedic injury.5 Traumatic injuries are frequent triggers of acute and persistent pain.6–8 Among individuals hospitalized after traumatic injury, up to 63% continue to experience chronic pain at 12 months.9 Hospitalized trauma survivors also frequently develop other post-traumatic neuropsychiatric sequelae, such as post-traumatic stress and depression.10–12

Individuals hospitalized after traumatic injury, with this high burden of chronic pain and neuropsychiatric morbidity, are at risk of transitioning to persistent opioid use.13–15 Persistent opioid use following trauma can be cause for concern, as persistent opioid use can be associated with both persistent pain and opioid misuse and addiction.16–19 Further, patients on chronic opioid prescriptions are at higher risk of initiation of injection drug use than opioid-naïve patients.20 New persistent opioid use following traumatic injury previously has been reported in relatively small samples and among single centers.21–23 In addition, more recently, larger studies have examined persistent opioid use by examining retrospective national insurance claims data.24 25 Whereas these studies are important, they are narrow in scope, do not compare different mechanisms of trauma, and do not examine the temporal patterns of persistent opioid use over time. Our investigation fills these gaps. Understanding rates of persistent opioid use among common traumatic injuries will stimulate new lines of inquiry and raise awareness of persistent opioid use among trauma care providers, policy-makers, and researchers. This study seeks to describe the incidence of new persistent opioid use following three common traumatic injuries: burn, MVC-related, and orthopedic trauma (defined as requiring open reduction and internal fixation of a long bone fracture) using aggregate national insurance claims data. As a reference, we have compared opioid use in these trauma populations to individuals who have had common major (ventral hernia repair) and minor (laparoscopic cholecystectomy) abdominal surgery.

Hospitalization following acute trauma has been linked to greater incidence of chronic pain,6–8 sociodemographic disadvantage,26 27 and substance abuse.28 29 Given these findings, we hypothesize that the incidence of new persistent opioid prescribing would be higher among hospitalized trauma patients as compared with those hospitalized following non-trauma surgery, such as individuals undergoing typical minor (laparoscopic cholecystectomy) and major surgeries (ventral hernia repair). Burn injuries were stratified by injury severity using tissue grafting as a marker for more severe burns requiring surgical intervention.

Methods

Retrospective insurance claims cohort

The patient panel used for this study was drawn retrospectively from a commercial insurance claims database (IBM MarketScan) spanning January 1, 2001 to December 31, 2020. MarketScan is a large, nationwide research database with one of the longest-running collections of deidentified insurance claims data. Individuals were included in this analysis if they were continuously insured for 1 year prior to through 180 days after injury/surgery, received no opioid prescriptions for 1 year prior to injury/surgery, and were between the ages of 18–65 years (figure 1). In order to be included in the analysis, individuals had to fill an opioid prescription within 2 weeks of their admission for injury/surgery. The sample represented aggregate data from the United States and there was no lost to follow-up. Patients needed to meet all inclusion criteria as detailed above in order to be included. Patients admitted to the hospital for burn injury were identified based on International Classification of Diseases (ICD)-9 and ICD-10 codes (for codes see online supplemental table 1). Patients who experienced a burn injury who also required a tissue graft were identified on the basis of a burn injury diagnostic code and a Current Procedural Terminology (CPT) code associated with a graft procedure (online supplemental table 1). Orthopedic injuries were defined as patients undergoing an open reduction and internal fixation of a long bone fracture (online supplemental table 1). MVC patients were identified on the basis of a diagnostic code for MVC (online supplemental table 1). These multiple methods of categorization were used for the benefit of highlighting outcomes for the different types of providers who care for these patients (eg, care teams for hospitalized MVC patients, orthopedic surgeons who perform open reduction and internal fixation). Of note, patients could be categorized in more than one group, depending on the nature of their injury. This overlap was rare in our dataset, but most common in those patients experiencing an MVC and an orthopedic trauma (n=368), followed by individuals who experienced a burn and an orthopedic trauma (n=6), followed by individuals who experienced a burn and MVC (n=2). There were no individuals who experienced a burn, MVC and an orthopedic injury. The Strengthening the reporting of observational studies in epidemiology (STROBE) checklist has been completed for this retrospective insurance claims study.

Supplemental material

Figure 1

Schematic of study design. Patients were included in the dataset if they were continuously insured for 1 year prior to injury/surgery through 180 days after injury/surgery. There was no lost to follow-up. New persistent opioid use was defined as receipt of ≥1 new opioid prescription filled 90–180 days after injury/surgery.

Comparison cohorts

To compare opioid use of trauma survivors to individuals experiencing common, non-traumatic surgical procedures, we included a comparison cohort of individuals who underwent major surgery (ventral hernia repair) and a cohort who experienced minor surgery (laparoscopic cholecystectomy). We selected individuals with CPT codes for a ventral hernia repair (eg, major surgery) and laparoscopic cholecystectomy (eg, minor surgery). All individuals in both comparison cohorts were hospitalized, although some stays were brief. Persistent opioid use after these controlled surgical traumas has been studied extensively, making these cohorts suitable reference groups to help demonstrate the validity of our methods.

Primary outcome

Incidence of new persistent opioid use was the primary outcome. Consistent with previous studies, new persistent opioid use was defined as receipt of one or more opioid prescriptions 90–180 days after surgery/injury, in an individual who did not receive an opioid prescription in the year prior to injury.30 A sensitivity analysis was performed using different definitions and is reported for reference in online supplemental table 2.

Secondary outcomes

Estimated daily morphine milligram equivalents over the reference window

Among individuals who met our criteria for new persistent opioid use, we calculated mean daily morphine milligram equivalents (MMEs). Individuals were considered to have received an opioid agonist if they either received an opioid medication over the study window delivered via a transdermal, oral, or transmucosal method. To determine this, we initially extracted all opioid medications received by insured participants in the therapeutic class 60 (opioid agonist) or 61 (opioid partial agonist) in the MarketScan REDBOOK database. Then, opioids were individually coded and reviewed to avoid misclassification. Intravenous opioids, partial agonists to treat opioid use disorder such as buprenorphine/naloxone were filtered out of our final dataset. Buprenorphine preparations to treat pain such as transmucosal buprenorphine and transdermal buprenorphine were included. Codeine-containing cough syrups were filtered and removed from the dataset and were not included as persistent opioid use. This approach ensures capture of all opioid analgesics dispensed to insured patients used to treat pain, and excluding misclassified opioids used to treat other conditions such as opioid use disorder and cough.

To calculate MMEs, we used national drug codes (NDC) linked to opioid medications filled to the 2018 Centers for Disease Control and Prevention (CDC) dataset (https://www.cdc.gov/drugoverdose/resources/data.html). This dataset provides a conversion of each opioid analgesic to MMEs. We then used package size (amount of drug dispensed) multiplied by the milligram of each dose and the opioid conversion factor in the CDC dataset and divided this value by the entire reference window, providing a daily MME estimate over the reference period. We used the entire reference window as a consistent denominator across individuals to provide a relative estimate of opioid use (eg, if a patient filled an opioid prescription at 5 months, we divided the total MMEs for that prescription across their entire reference window). To be included in this analysis, MMEs prescribed were required to be >0, and missing data, negative values or 0 for days of medication supplied were excluded from the analysis. We also examined using ‘days supplied’ as an alternative to the presented approach, but this yielded much less reliable estimates with large SD. We report these MME values using ‘days supplied’ for reference in online supplemental table 3. MMEs consumed or prescribed in the hospital stay were not available for extraction from the national claims dataset, so this variable could not be included in analyses.

Classification of opioids

We used drug names and NDCs to identify specific prescribed opioids and generated a plot (online supplemental figure 1) to display the proportion of prescriptions for different medications. We also calculated the proportion of individuals receiving long-acting or short-acting medications (using the NDC for each available drug) for MME calculation.

Covariates of interest including relevant comorbidities

Age, sex, and hospital length of stay (LOS) were extracted from the insurance claims database given the relationship to pain and opioid utilization described above. Hospital LOS in days was calculated based on date of intake and date of hospital discharge. Race was not available in the dataset. Using previously described methods,30 31 we also calculated comorbid mental health illness, chronic pain conditions (same codes as used by Santosa et al 32), benzodiazepine use as an overall marker of comorbidity, the Charlston Comorbidity Index (CCI).33 34 Given that we used ICD diagnostic codes to enumerate these comorbidities and our dataset spanned 2001–2020, both ICD-9 and ICD-10 diagnostic codes were used.35 We included tables of diagnostic codes used in our analysis in online supplemental tables 4,5. Additionally, we used both ICD-9 and ICD-10 codes and weightings provided by by Quan et al 35 to calculate the CCI.

Statistical analysis

Available sample demographic data were summarized using descriptive statistics (proportions, mean, SD, median and range). All statistical analyses and data management were conducted using R (V.1.2.133536 and plots were generated using GraphPad Prism (V.7.04). Difference of proportion tests were used to compare proportions, and unpaired t-tests were used to compare means. P values were considered significant if <0.05. We used multivariate logistic regression methods to determine the adjusted OR of developing persistent opioid use based on each characteristic and comorbidity reported in table 4. P values were two tailed, and significance was set at p<0.05.

Results

Patient characteristics

Sample characteristics and relevant comorbidities by injury/surgery type and for the overall sample are summarized in table 1. A majority of burn survivors and individuals hospitalized following MVC were middle-aged males. Individuals undergoing open reduction internal fixation (ORIF) for long bone fracture were middle-aged and balanced with regard to sex. A majority of individuals hospitalized for the selected prototypical major surgery (ventral hernia repair) and minor surgery (laparoscopic cholecystectomy) were middle-aged females. The majority, nearly three-quarters of patients in all cohorts, required hospitalization for greater than 24 hours. The CCI reveals that comorbidities were uncommon in the sample population. Overall, the majority (nearly three-quarters) of individuals included in the analysis had LOS greater than 24 hours. Although percentages ranged across trauma and surgery type, overall, 17% experienced comorbid joint pain, 2% experienced neck pain, 3% experienced low back pain, and 14% experienced other pain. The most common mental health disorders experienced by individuals in the cohort were anxiety and mood disorders. Benzodiazepine use was less than 3% in all cohorts.

Table 1

Characteristics of the patient sample stratified by trauma type

New persistent opioid use

The proportions of individuals transitioning to new persistent opioid use after injury, according to injury type and stratified by LOS, are shown in table 2. Approximately 1 in 11 patients undergoing laparoscopic cholecystectomy transitioned to new persistent opioid use, compared with approximately 1 in 8 patients experiencing burn injury not requiring tissue graft, and 1 in 8 undergoing ventral hernia repair. Approximately one in six individuals hospitalized after MVC transitioned to new persistent opioid use, as compared with approximately one in five patients who required ORIF for long bone fracture or burn injury with tissue allograph. The percentages of individuals with new persistent opioid use over time from years 2001 to 2020 according to trauma/surgery type are shown in figure 2. Error bars were not included in this plot to reduce complexity but are included in online supplemental figure 1. Rates of persistent opioid use in all trauma cohorts (11,352/60,487, 19%) were greater (p<0.001) than the rates of persistent opioid use in both non-traumatic major surgery (4002/30,825, 13%) and non-traumatic minor surgery (7782/86,473, 9%). To examine whether LOS influenced the incidence of persistent opioid use, table 2 includes the rates of new persistent opioid use in individuals who had LOS greater than 24 and individuals with LOS less than 24 hours. Across all trauma and surgery categories, except for burn injury requiring tissue grafting, individuals who were admitted for >24 hours have significantly higher rates of new persistent opioid use compared with individuals who were admitted for ≤24 hours for all groups.

Table 2

New prescription opioid use according to trauma population and patient characteristics

Figure 2

Proportion of trauma patients with new persistent opioid use following major thermal burn injury (with and without tissue grafting), motor vehicle collision, orthopedic trauma (open reduction internal fixation (ORIF) of long bone fracture), major surgery (ventral hernia repair), and minor surgery (laparoscopic cholecystectomy). Error bars were not included to improve interpretation, but they are included for reference in online supplemental figure 1.

Average daily MME over the reference window by injury/surgery type

Cumulative average daily MME over the reference window among individuals with new persistent opioid use, according to trauma/surgery type, is summarized in table 3 and figure 3. In general, average daily MME over the reference window was highest for orthopedic injury, followed by burn injury and then MVC. Average daily MME over the reference window was higher for orthopedic trauma compared with ventral hernia repair and laparoscopic cholecystectomy. Average daily MME following orthopedic trauma and MVC was higher compared with laparoscopic cholecystectomy (p<0.05). The highest average daily opioid use was observed in individuals who underwent ORIF for long bone fracture, followed by burn injury (no tissue grafting). Individuals who experienced an orthopedic injury and surgery (both ventral hernia repair and laparoscopic cholecystectomy) with LOS greater than 24 hours had higher daily MMEs over the reference window compared with individuals with shorter LOS.

Figure 3

Average daily milligram morphine equivalents prescribed among individuals requiring persistent opioid following major thermal burn injury, burn injury severe enough to warrant tissue grafting, motor vehicle collision, orthopedic trauma (open reduction internal fixation (ORIF) of long bone fracture), major surgery (ventral hernia repair), and minor surgery (laparoscopic cholecystectomy). Error bars represent the 95% CIs.

Table 3

Average daily milligram morphine equivalent (MME) over the reference window dose among individuals requiring persistent opioids after injury/surgery between years 2001 and 2020

Type of opioid analgesic prescribed to individuals requiring persistent opioids

Nearly all patients with new persistent opioid use were prescribed only short-acting preparations (95.9%, 11,517/12,013). Only 3.4% (405/12,051) were prescribed both short-acting and long-acting preparations. The most commonly prescribed opioid analgesics were hydrocodone (49%; 7119/14,483), oxycodone (20%; 2917/14,483, and tramadol (12%; 1821/14,483). The proportion of each opioid prescribed is shown in online supplemental figure 2.

Multivariate logistic regression of new persistent opioid use stratified by trauma/surgery type

Multivariate logistic regression was used to examine the risk factors that may underlie persistent opioid use following traumatic injury and laparoscopic cholecystectomy and ventral hernia repair. Adjusted ORs were calculated to determine factors that confer risk to persistent opioid use across trauma types and are reported in table 4. As a reference, risk factors for persistent opioid use following laparoscopic cholecystectomy, ventral hernia repair and overall are included. In every cohort, except burn injury requiring tissue grafting, the CCI was associated with persistent opioid use. In surgical cohorts and orthopedic trauma, female sex was associated with persistent opioid use; however, there was no association in individuals experiencing burn or MVC. In the majority of cohorts, advancing age was associated with persistent opioid use. There were some associations between mental health and pain disorders throughout the trauma cohorts. LOS over 24 hours was associated with persistent opioid use following orthopedic trauma, MVCs, and ventral hernia repair.

Table 4

Multivariate logistic regression demonstrating the adjusted odds of persistent opioid use by comorbidity

Discussion

Results demonstrate that a substantial proportion of hospitalized trauma patients transition to new persistent opioid use after injury. Approximately one in six individuals hospitalized after MVC trauma develop new persistent opioid use. Approximately one in five patients who required ORIF for long bone fracture, and one in eight burn injury survivors requiring tissue allograft for burn injury develop new persistent opioid use. These rates were higher than the reference non-trauma minor surgery, laparoscopic cholecystectomy (1 in 11 patients). Following orthopedic trauma, individuals were prescribed more MMEs over the reference window than laparoscopic cholecystectomy or ventral hernia repair, although the mean differences were small. Nearly all new persistent opioids prescribed in the cohort were short acting.

Results from this study are consistent with previous studies. For example, our estimate that 20% of individuals undergoing ORIF for long bone fracture receive new persistent opioids is consistent with studies identifying persistent opioid use in 9%–43% of individuals with traumatic musculoskeletal injury,37–40 and our estimate that 16% of patients following MVC receive new persistent opioids is consistent with previous MVC studies.41 For burn injury, our estimates of new persistent opioid use in 12% of hospitalized burn patients not receiving a tissue graft, and 12% of those receiving a tissue graft, are comparable to a previous small study identifying chronic opioid use in 14% at 90 days and 10% at 1 year following injury,21 and another report indicating that over 12% of burn injury patients filled >3 opioid prescriptions.22

Of the trauma types included, patients having orthopedic traumas were most likely to receive new persistent opioid prescriptions (20%). This trend could be indicative of severity of pain caused by bone injuries, and/or the particular lack of alternative treatments for these injuries, given the concern for non-steroidal anti-inflammatory (NSAID) effects on bone formation.42 This finding is consistent with previous work which demonstrated that burn survivors were less likely to receive opioids compared with survivors of non-burn trauma (eg, fractures, dislocations).43 For the purposes of comparison, we examined two non-orthopedic surgical procedures as a reference, ventral hernia repair and laparoscopic cholecystectomy. We found rates of new persistent opioid use of 13% and 9% after these major and minor surgeries, respectively. These rates were somewhat higher compared with a 2017 population-based study that used nationwide insurance claims from 36 177 privately-insured surgical patients between the years 2013 and 2014. The 2017 study found the incidence of new opioid use after surgical procedures to be 7% and 6% after major and minor surgeries, respectively.30 The previous study included a wider variety of surgical procedures, which may have contributed to the lower prescribing rates. Our study results are also different in that the receipt of persistent opioids was higher in the major surgery group (ventral hernia repair, 13%) versus minor surgery (laparoscopic cholecystectomy, 9%), and also varied with respect to trauma type.

Recent studies have similarly employed national insurance claims data to evaluate the incidence of persistent opioid use after trauma. One of these studies24 also made use of the IBM MarketScan database. However, it was much narrower in focus and only examined opioid use among adolescents aged 11–17 who underwent facial fracture repair. Another study, which used the Optum Clinformatics Database,25 examined postdischarge opioid use among trauma survivors from 2010 to 2017. Whereas we compared mechanism of trauma, these authors focused on anatomical location of injury and injury severity. Given our study’s larger size, longer observation period (2001–2020) and use of non-trauma comparison groups, our results are more generalizable, highlight trends over time, and compare several specific traumatic injuries in which chronic pain development is common.

Conclusion

In sum, the above data demonstrate a high rate of new persistent opioid use among patients commonly hospitalized after trauma. Whereas these findings could be viewed as concerning from the standpoint of opioid stewardship, we believe that it is far more accurate to consider that severe pain in the setting of trauma is challenging to treat. In an era in which the dangers of opioid use/misuse are fully apparent, it is challenging to manage severe pain after traumatic injury, and alternative non-opioid treatments for severe pain remain sparse. These data provide useful benchmarks of current limitations in contemporary trauma care. As trauma practitioners, policy-makers, and researchers, we must continue to make progress in the development of more effective treatments to prevent and treat chronic pain.

Limitations

There are a number of limitations of this study which should be considered when interpreting our results. First, our analyses are retrospective, not prospective, using a nationally representative insurance claims database of patients in the USA. It is also possible that opioid prescriptions filled and billed to insurance were not actually consumed. This dataset only includes individuals who were privately insured, and thus findings may not be generalizable to both non-insured and/or Medicare/Medicaid populations. In addition, new opioid prescribing may not include prescriptions filled without insurance billing.44 We assessed comorbidities in our dataset, which depends on providers coding these comorbidities in the record. Therefore, it is possible that our analyses underestimated comorbid health conditions. We performed multivariate logistic regression in secondary analyses to illustrate the relationship between medical comorbidities and persistent opioid use. Our study is likely underpowered for some of the cohorts (eg, burn injury requiring graft (n=1504)) to demonstrate these associations. Future prospective studies examining relationships among these key variables are needed. These basic questions, about new persistent opioid use following trauma, highlight limitations in current trauma pain care and the importance of benchmarking. Our study also is limited in that pain severity and provider perception of injury were not available in the insurance claims database, and therefore, not included in hypotheses or analyses. Also, our calculation of MMEs is based on opioid prescription receipts with the prescription being averaged over time leading to a potential overestimation of consumed MMEs. Because our data are from an insurance claims dataset, we cannot suggest whether there is an association between: (1) pain and receipt of opioids versus (2) receipt of opioids and substance use disorder.

Another major limitation of our study is that there is a chance that another event (outside of the referenced trauma or surgery) could have accounted for opioid prescribing. It is possible that the persistent opioid use could be related to persistent pain or inappropriate opioid prescribing. This possibility could be addressed in future prospective studies in which patients are monitored for recurrent events that may lead to opioid prescribing. Finally, our dataset lacked information about race and the extent of injury. In addition, our dataset did not include non-hospitalized trauma patients as a comparison. To approximate this, we did include results from patients who had a LOS less than 24 hours versus greater than 24 hours. Future prospective studies should address these limitations.

Data availability statement

Data may be obtained from a third party and are not publicly available. IBM MarketScan.

Ethics statements

Patient consent for publication

Ethics approval

The study objectives and methods were reviewed and deemed to be exempt by the University of North Carolina Institutional Review Board (18-2349).

References

Footnotes

  • Contributors MCM contribution to the manuscript was conception of idea, writing, analysis, critical revision, and decision to publish. MCM serves as the guarantor for the work presented in the manuscript. AST and YZ contribution was data management and interpretation, and statistical analysis. AL contributed with literature search, writing, and critical analysis. ABS contributed to writing and critical revision of the manuscript. CEB contribution was literature search, writing, and analysis. FB contributed to writing. AMG contributed to writing and critical revisions. SAM contribution to the manuscript was conception of idea, writing, and critical revision.

  • Funding Research reported in this publication was supported by the University of North Carolina Department of Anesthesiology. Research reported in this publication was supported by the Eunice Kennedy Shriver National Institute Of Child Health & Human Development of the National Institutes of Health under Award Number K12HD001441. The database infrastructure used for this project was funded by the Department of Epidemiology, UNC Gillings School of Global Public Health; the Cecil G. Sheps Center for Health Services Research, UNC; National Center for Advancing Translational Sciences (NCATS), National Institutes of Health, through Grant Award Number UL1TR002489.

  • Competing interests None declared.

  • Provenance and peer review Not commissioned; externally peer reviewed.

  • Supplemental material This content has been supplied by the author(s). It has not been vetted by BMJ Publishing Group Limited (BMJ) and may not have been peer-reviewed. Any opinions or recommendations discussed are solely those of the author(s) and are not endorsed by BMJ. BMJ disclaims all liability and responsibility arising from any reliance placed on the content. Where the content includes any translated material, BMJ does not warrant the accuracy and reliability of the translations (including but not limited to local regulations, clinical guidelines, terminology, drug names and drug dosages), and is not responsible for any error and/or omissions arising from translation and adaptation or otherwise.

Linked Articles