Article, Emergency Medicine

Effect of opioid analgesics on emergency department length of stay among low back pain patients in the United States

a b s t r a c t

Objectives: The objective of this study was to compare emergency department (ED) length of stay (LOS) between patients treated with opioid analgesia versus non-opioid analgesia for Low back pain in the ED.

Methods: We conducted a secondary analysis of National Hospital Ambulatory Medical Care Survey data (2014-2015). Adults (age >=18 years) who presented to the ED with a reason for visit or primary diagnosis of LBP were included in the final study sample. Patient visits were categorized into two groups based on whether they received opioid analgesia (with or without non-opioid analgesia) or non-opioid analgesia only in the ED. The primary outcome measure was ED LOS, which was log-transformed (as ED LOS was not normally distrib- uted) for analysis. A multivariable Linear regression analysis was used to evaluate the association between opioid use and ED LOS.

Results: The study sample consisted of a national estimate of approximately 8.6 million ED visits for LBP (during 2014-2015), of which 60.1% received opioids and 39.9% received non-opioids only. The geometric mean ED LOS for patient visits who received opioids was longer than patient visits who received non-opioids (142 versus 92 min, respectively; p b 0.001). After adjusting for confounders in the multivariable analysis, patient visits that re- ceived opioids had a significantly longer ED LOS (coefficient 0.25; 95% CI 0.11 to 0.38; p b 0.001).

Conclusions: In a nationally Representative sample of patient visits to ED due to LBP in the US, use of opioids in the ED was associated with an increased ED LOS.

(C) 2020


United States (US) national estimates have previously shown that low back pain accounts for approximately 2.6 million annual ED visits representing 2.3% of all ED visits [1].

A recent systematic review and meta-analysis estimated that 4.39% of all emergency settings visits are due to LBP, which makes LBP a top 10 reason for ED visit [2]. In a study that previously evaluated the Cen- ters for Disease Control and Prevention’s (CDC) National Hospital Am- bulatory Medical Care Survey (NHAMCS) (2002-2006), nearly 62% of patients who presented with LBP to the ED received an opioid analgesic as treatment [1]. However, the use of opioids for this indication may not be necessary in most circumstances [3]. In 2017, the American College of Physicians (ACP) systematically evaluated the literature and published

* Corresponding author at: Department of Pharmacy Practice and Science, The University of Arizona College of Pharmacy, 1295 North Martin Avenue, Drachman Hall B306L, Tucson, AZ 85721, United States of America.

E-mail address: [email protected] (S. Bhattacharjee).

evidence-based practice guidelines for acute and chronic LBP [4]. The ACP recommends that the first line pharmacological treatments for acute LBP should be Non-steroidal anti-inflammatory drugs (NSAIDs) or Skeletal Muscle relaxants (SMR) [4]. It was also determined that there was not enough evidence to evaluate the effectiveness of opioid analgesics for treatment of acute LBP. The ACP guidelines are not specific to the ED setting, but suggest that opioids should not be the primary agent of choice. Evidence comparing opioids versus non-opioids for acute LBP in ED setting is lacking.

ED LOS serves as an outcome that is closely related to a patients’ con- dition, especially for those who are not admitted to hospital [5]. In- creased LOS in the ED has been shown to directly contribute to the high cost of emergency care and is also associated with diminished quality of care and increased adverse events (AEs) such as: increased in- fection rates, medication related errors, pressure sores, and delirium [2,5-9]. The association between opioids for the treatment of LBP and ED LOS is unknown. It is possible that patients given opioids for this in- dication would have a longer ED LOS. This is possibly due to the need for prolonged monitoring after opioid administration, functional

0735-6757/(C) 2020

impairment or need to rule out Red flags. The objective of this study was to compare emergency department (ED) length of stay (LOS) between patients treated with opioid analgesia versus non-opioid analgesia for Low back pain in the ED.


Study design and data source

This study is a secondary analysis of NHAMCS data of ED visits col- lected during 2014 and 2015. The NHAMCS is a national survey col- lected annually by the CDC and was designed to collect data on healthcare utilization in both hospital emergency departments (EDs), outpatient departments (OPDs). For the purpose of our study, we used only the ED component of NHAMCS data. The survey has a complex de- sign where each patient visit is assigned a weight to provide national es- timates based on probabilistic sampling. Additional information about the scope and design of NHAMCS is available from the CDC [10]. The University’s Institutional Review Board (IRB) considered the study ex- empt from IRB oversight. The study was conducted after the exemption was obtained.

Study sample

The LBP population was identified by using both International Clas- sification of Diseases, Ninth Revision, Clinical Modification codes (ICD- 9-CM) as well as documented reason for visit. Specifically, patient visits were included in the sample if they were 18 years or older, had any pri- mary ICD-9-CM code, or any reason for visit corresponding to LBP. Only the primary ICD-9-CM code fields were used in this analysis to capture the most relevant diagnosis for each patient visit. The reason for visit fields was used to further identify patients with LBP. The included ICD-9-CM codes are in Supplementary Table 1 and reason for visit codes are in Supplementary Table 2 [1]. Visits were excluded from the analysis if they had a reason for visit related to motor vehicle accidents (MVAs), did not receive any analgesic during the ED visits, transferred to psychiatric hospital, returned or transferred to a nursing home, trans- ferred to another hospital, left against medical advice, or died while in the ED. MVAs were excluded from study because these patients are likely candidates for opioids and did not represent our population of in- terest [1]. Patients who met the above inclusion criteria were catego- rized by one of two treatment groups, opioid treatment or non-opioid analgesic group. The treatment group was identified by using the NHAMCS variable of drug category, identifying drug by their corre- sponding therapeutic class. The NHAMCS database utilizes Multum’s Lexicon Drug Database for drug categorization, and includes informa- tion on up to 30 drugs per visits [11]. Therapeutic classification reflects Multum’s 3-level nested category system and was used to identify the following therapeutic classes: opioids, skeletal muscle relaxants, skele- tal muscle relaxant combinations, nonsteroidal Anti-inflammatory agents, analgesics combinations, COX-2 inhibitors, benzodiazepine anti- convulsants, and miscellaneous analgesics Supplementary Table 3 [1,11]. The opioid group consisted of patient visits where opioid analge- multivariable regression analysis“>sics for LBP were prescribed (with or without a non-opioid analgesic). The comparison group consisted of patients who only received non- opioid analgesics. These national estimates reflect a sample of all US ED visits during 2014 and 2015 and consisted of a total of 44,905 un- weighted visits (23,844 from 2014 and 21,061 from 2015) representing 278,363,460 total ED visits over the two-year period. The LBP subpopu- lation represented n = 1363 unweighted visits corresponding to 8,564,059 weighted ED visits.

Outcome measures and variables

The primary outcome measure was ED LOS (minutes). This was cal- culated from the NHAMCS data variable length of visit after subtracting

the triage wait time. Thus, ED LOS represents the time from being in an ED room to discharge. Other variables of interest were considered to be potential confounders and included: patient age, sex, race/ethnicity (white, black, other), immediacy with which patient should be seen, if diagnostic services were ordered or provided (yes/no), if procedures were performed (yes/no), pain scale (0-10), metropolitan statistical area, year (2014 or 2015), and the number of chronic comorbidities (0

-12) [12,13].

Data analysis

Normally distributed continuous data were analyzed using the student’s t-test and categorical data were analyzed using the chi squared test with an a-priori alpha of 0.05 for all analyses. A multivari- able linear regression analysis was used to determine the association between opioid use and LOS in the ED. Due to the non-normally distrib- uted characteristic of the LOS, this variable was log transformed for analysis. The following assumptions of the multivariable linear regres- sion were tested: linearity, normality of the residuals, and homoscedas- ticity. Homoscedasticity was analyzed by using White test. Multi- collinearity of independent variables was analyzed using the variance inflation factor (VIF) with a value of >=5 indicating collinearity. Missing data were imputed using the median substitution method [14].

Analysis estimates were obtained by accounting for the complex survey design of the NHAMCS by utilizing the weighting and strata within the database.

Two Sensitivity analyses were conducted to examine the robustness of the base case analysis. These sensitivity analyses were: (i) opioid an- algesics alone vs non-opioid analgesics alone (i.e. patients in the opioid group who received concurrent non-opioids were excluded). This was conducted to understand the difference between opioid analgesics alone compared to non-opioid analgesics alone. The second sensitivity analysis was conducted by removing missing observations from the analysis to determine if it changed the results due to data that may not be missing at random.


In 2014 and 2015, there were 8,564,059 ED patient visits for LBP where analgesics were administered. Of these, 5,147,820 (60.12%) pa- tient visits involved administration of an opioid analgesic during the ED encounter. Characteristics of the ED visit patient included: average age of 45 (SE = 1) years, 59.1% female, 72.1% white, 98.92% with five or less chronic conditions, 71.48% with a pain level of six or more, 82.75% residing in a metropolitan statistical area (additional details are in Table 1). Approximately 51% LBP ED visits in our study were re- corded in 2015. Most (90.52%) of the ED visits were classified as urgent/semi-urgent, 38.05% had a procedure performed, and 56.60% re- ceived diagnostic imaging. ED visits where opioid analgesics were used, had more likelihood of receiving diagnostic imaging (p = 0.008) or have a procedure performed (p = 0.002). The geometric mean ED LOS for patient visits who received opioids was longer than patient visits who received non-opioids (142 versus 92 min, respectively; p b 0.001).

Multivariable regression analysis

Table 2 shows the results of the ordinary least squares (OLS) multi- variable regression analysis. The percentage of variation explained by independent variables used in this model (R2) was 26.91%. After con- trolling for other covariates, the primary variable of interest, opioid an- algesic use was associated with increased logLOS (? =0.25, exp. (?)= 1.28, p b 0.001, CI: 0.11, 0.38). This result shows a 28% increase in the LOS for LBP individuals who received any opioid analgesic as compared to patients who only receive only non-opioid analgesics. Results of the white’s test for homoscedasticity were not significant indicating that there is constant variance in the residuals (p b 0.804). Multi-

Table 1

Comparison between opioid and non-opioid groups.

directly from the ED. Thus, our results are less likely to be affected by throughput issues and Patient boarding that is common in the ED. Pa-

Variable Total

(n =

Opioid (n =

Non-opioid (n =


tients who had imaging and those who received a procedure in the ED, had longer ED LOS. These are expected and are known confounders.







However, after adjusting for this, our multivariable model still showed

that opioid use was independently associated with increased ED LOS.

Agea (years), mean (SE)

45 (1)

48 (1)

41 (1)


The current literature does not support the use of opioid analgesics

Female, N (%)

Pain scale, N (%)

Pain level of five or less

5.06 (59)

2.44 (28)

3.00 (58)

1.43 (28)

2.06 (60)




as first line treatment for Acute back pain due to insufficient evidence as compared to more compelling evidence for NSAIDs and skeletal mus-

Pain level of six or more

6.12 (72)

3.72 (72)

2.40 (70)

cle relaxants [15-21]. However, patients who present to the ED may

Length of staya (min), (SE)

164 (8)

184 (7)

135 (8)


have pain that is more severe than the Primary care settings. For exam-

Immediacy with patient was

seen, N (%) Non-urgent

0.31 (5)

0.14 (4)

0.17 (7)


ple, a large proportion of the cohort in our study had severe pain, which may prompt clinicians to prescribe opioids. As recommended in the


5.95 (90)

3.61 (91)

2.34 (90)

American College of Physicians guidelines, from an ED perspective, opi-


0.31 (5)

0.23 (6)

0.08 (3)

oids should not be prescribed upon discharge from the ED [4]. For exam-

ple, even short-term prescribing can result in persistent Opioid utilization [22]. However, it is less clear if a single dose of an opioid in the ED setting is harmful. We acknowledge that there are circumstances in the ED, when opioids will be warranted for acute back pain. This in-

Image received, N (%)

4.85 (57)

3.13 (60)

1.72 (50)


Race, N (%)



2.11 (25)

1.02 (20)

1.09 (32)


6.17 (72)

3.89 (76)

2.28 (67)


0.28 (3)

0.23 (4)

0.05 (2)

Year 2015, N (%)

4.39 (51)

2.58 (50)

1.81 (52)


cludes patients who may have failed a trial of non-opioid analgesics

Procedure performed, N (%)

3.26 (38)

2.22 (43)

1.04 (30)


and remain in severe pain. The results of this study should not be used

Metropolitan area, N (%)

Region, N (%)

7.09 (83)

4.31 (84)

2.77 (81)



as an ammunition to avoid opioids entirely even in circumstances


0.13 (15)

0.69 (13)

0.60 (18)

when their use may be appropriate and warranted. However, our re-


2.78 (33)

1.81 (36)

0.96 (28)

sults highlight another potential consequence of opioid administration


2.77 (32)

1.58 (31)

1.19 (35)

even when used within the ED.


1.73 (20)

1.06 (21)

0.66 (19)

Number of chronic


conditions, N (%)

Five or less

8.47 (99)

5.06 (98)

3.41 (99)

Six or more

0.09 (1)

0.09 (2)

0.01 (b1)

Note: Based on unweighted N = 1363 (nationally representative weighted N = 8.6 mil- lion) ambulatory visits using NAMCS 2014-2015 data. Data for the individual categories have been presented in millions of visits.

a Continuous variables were analyzed using t-test, all other variables were analyzed using Chi-squared test.

collinearity was not observed in the regression model as all co-variates had a VIF b 2.

In the sensitivity analyses the association between opioid use and logLOS remained significant. In the multivariable linear regression anal- ysis of opioid only versus non-opioid only group (?=0.27, exp. (?)= 1.31, p b 0.001, CI: 0.12, 0.43; see Supplementary Table 4) as well as re- moving missing observations (?=0.20, exp. (?)= 1.23, p b 0.0001, CI: 0.16, 0.25; see Supplementary Table 5), the findings were consistent with the base case analysis.


Our key study finding suggests that there was approximately 28% in- crease in ED LOS for LBP visits where opioid analgesics were given as compared to visits where only non-opioid analgesics were given. On av- erage, this equates to a 38 min (average opioid group LOS 184 +- 7 min) longer stay for individuals who receive opioid analgesics. A Clinically meaningful difference in ED LOS is unknown, however on an aggregate level it amounts to a ED stay utilization increase of 1.6 million hours more per year when opioids are given. This finding indicates that pa- tients administered opioid analgesics in the ED may utilize greater re- sources such as treatment rooms and clinician time. In addition to increased use of limited ED resources, studies that have examined the effect of extended ED LOS on outcomes, have found associations with di- minished quality of care and increased adverse events such as: in- creased infection rates, medication related errors, pressure sores, and delirium [5-9].

ED LOS can be affected by many factors. In some cases, it may be due to issues that are unrelated to the patients’ condition. We tried to min- imize this by only including patients who were discharged home

There were other variables that were also associated with ED LOS. This included triage urgency, region, procedures being performed, imaging, pain scale, area type, and number of chronic conditions. While these confounders were not the focus of this report, we did ex- pect some of these to be associated with ED LOS. For example, we would expect that non-urgent patients would have a shorter ED LOS, and those who require imaging or a procedure, or who have more comorbidities would have a longer ED LOS. Similarly, metropolitan region could possibly have a longer ED LOS due to greater population density and crowding in ED. However, we cannot explain why higher pain score was associated with a shorter ED LOS. This is opposite of what we would have expected. Also, we do not have any theories to explain the differences seen between geographical regions.

Limitations of this study include: potential for errors during the data collection and coding process [23]. However, the CDC has made several attempts to improve the accuracy of the NHAMCS data. Another limita- tion to the NHAMCS is missing data. Due to the complex survey struc- ture of the NHAMCS data, some variables are associated with N10% of missing data, which may not be missing completely at random. To ac- count for this, we used imputation techniques and used the complete case analysis for our sensitivity analysis. ICD-9-CM Diagnosis codes were used for identification of the LBP sample. However, it is possible that coding of ICD-9-CM is not completely accurate. In an attempt to im- prove the accuracy of identification of individuals with LBP in the ED, we supplemented this using the reason for visit field. The NHAMCS data does not include when pain scores are measured, and we do not know the pain control at time of ED discharge. This could have provided a comparison of pain control between groups. We also did not have infor- mation about opioid related adverse events that could be a contributing factor to increased ED LOS. There are other variables that can contribute to ED LOS that are not captured in the NHAMCS. These include severity of illness based on physiological measures, functional impairment, the identification of red flags by clinicians that would warrant additional monitoring, or failure of treatments that would result in delays in achieving pain control. However, this has been partly overcome by using variables such as immediacy with which patients were seen, pain scale, and need for imaging and procedures. These latter variables also serve as useful surrogates for severity of illness and the identifica- tion of red flags by clinicians. Since this is a cross-sectional study, we can only identify associations and cannot claim causality.

Table 2

Multivariable linear regression model.


? estimatea

95% confidence Interval

Exponentiated ?b

Standard error



Non-opioid only




0.11, 0.38






-0.00, 0.01









-0.18, 0.05









-0.53, 0.02






-0.84, -0.05




Procedure performed No




0.21, 0.47









-0.39, -0.02






-0.48, -0.12






,0.17, 0.21




Pain scale


-0.05, -0.00




Imaging received





0.32, 0.57









0.02, 0.37









-0.22, 0.08






-0.22, 0.32









-0.14, 0.09




Number of chronic conditions


0.01, 0.09




Note: Based on unweighted N = 1363 (nationally representative weighted N = 8.6 million) ambulatory visits using NAMCS 2014-2015 data.

a ? estimate = coefficient for log length of stay.

b Exponentiated ? converts effect size to the original units because length of stay was log transformed.


The use of opioid analgesics for treatment of LBP in the ED is associ- ated with a significant increase in ED LOS. Future studies are needed to compare opioids versus non-opioids for acute LBP in the ED with regard to outcomes such as pain control, adverse effects, and long-term opioid utilization.

Ethics approval and consent to participate

Not applicable.

Conference presentation

A part of this study has been accepted as a Poster presentation at the 24th Annual International Meeting of the International Society for Pharmacoeconomics and Outcomes Research held in May 18-22, 2019 in New Orleans, LA, USA.


This research did not receive any grant from funding agencies in the public, commercial, or not-for-profit sectors.

CRediT authorship contribution statement

Seth W. Anderson: Conceptualization, Data curation, Formal analy- sis, Investigation, Methodology, Software, Writing – original draft. Sandipan Bhattacharjee: Conceptualization, Data curation, Formal analysis, Investigation, Methodology, Project administration, Resources,

Software, Supervision, Validation, Writing – review & editing .Asad E. Patanwala: Conceptualization, Data curation, Formal analysis, Investi- gation, Methodology, Project administration, Resources, Software, Supervision, Validation, Writing – review & editing.

Declaration of competing interest


Appendix A. Supplementary data

Supplementary data to this article can be found online at https://doi. org/10.1016/j.ajem.2020.06.002.


  1. Friedman BW, Chilstrom M, Bijur PE, Gallagher EJ. Diagnostic testing and treatment of low back pain in United States emergency departments: a national perspective. Spine (Phila Pa 1976). 2010;35:E1406-11.
  2. Edwards J, Hayden J, Asbridge M, Gregoire B, Magee K. Prevalence of low back pain in emergency settings: a systematic review and meta-analysis. BMC Musculoskelet Disord. 2017;18:143.
  3. Innes GD, Croskerry P, Worthington J, Beveridge R, Jones D. Ketorolac versus acetaminophen-codeine in the emergency department treatment of acute low back pain. J Emerg Med. 1998;16:549-56.
  4. Qaseem A, Wilt TJ, McLean RM, Forciea MA. Clinical Guidelines Committee of the American College of P. Noninvasive treatments for acute, subacute, and chronic low back pain: a clinical practice guideline from the American College of Physicians. Ann Intern Med. 2017;166:514-30.
  5. Guttmann A, Schull MJ, Vermeulen MJ, Stukel TA. Association between waiting times and short term mortality and hospital admission after departure from emergency de- partment: population based cohort study from Ontario, Canada. BMJ. 2011;342:d2983.
  6. Schull M, Vermeulen M, Guttmann A, Stukel T. Better performance on length-of-stay benchmarks associated with reduced risk following emergency department dis- charge: an observational cohort study. CJEM. 2015;17:253-62.
  7. Bashkin O, Caspi S, Haligoa R, Mizrahi S, Stalnikowicz R. organizational factors affect- ing length of stay in the emergency department: initial observational study. Isr J Health Policy Res. 2015;4:38.
  8. Ackroyd-Stolarz S, Read Guernsey J, Mackinnon NJ, Kovacs G. The association be- tween a prolonged stay in the emergency department and adverse events in older patients admitted to hospital: a retrospective cohort study. BMJ Qual Saf. 2011;20: 564-9.
  9. Diercks DB, Roe MT, Chen AY, Peacock WF, Kirk JD, Pollack Jr CV, et al. Prolonged emergency department stays of non-ST-segment-elevation myocardial infarction patients are associated with worse adherence to the American College of Cardiol- ogy/American Heart Association guidelines for management and increased adverse events. Ann Emerg Med. 2007;50:489-96.
  10. National Center for Health Statistics. Ambulatory health care data: scope and sample design. [Accessed 20 Sep- tember 2019].
  11. National Center for Health Statistics. NHAMCS micro-data file documentation 2015. 2015

    doc15_ed.pdf; 2015. [Accessed 21 February 2019].

    Asplin BR, Magid DJ, Rhodes KV, Solberg LI, Lurie N, Camargo Jr CA. A conceptual model of emergency department crowding. Ann Emerg Med. 2003;42:173-80.

  12. Altyar A, Bakhsh HT, Mohammed A, Skrepnek G, Patanwala AE. Administration of first dose antibiotic in the ED in patients with minor skin and soft tissue infections. Am J Emerg Med. 2015;33:1225-8.
  13. Bennett DA. How can I deal with missing data in my study? Aust N Z J Public Health.


    Hoy D, March L, Brooks P, Woolf A, Blyth F, Vos T, et al. Measuring the global burden of low back pain. Best Pract Res Clin Rheumatol. 2010;24:155-65.

  14. Scott NA, Moga C, Harstall C. Managing low back pain in the primary care setting: the know-do gap. Pain Res Manag. 2010;15:392-400.
  15. Hoy D, Bain C, Williams G, March L, Brooks P, Blyth F, et al. A systematic review of the global prevalence of low back pain. Arthritis Rheum. 2012;64:2028-37.
  16. Roelofs PD, Deyo RA, Koes BW, Scholten RJ, van Tulder MW. Non-steroidal anti- inflammatory drugs for low back pain. Cochrane Database Syst Rev. 2008; CD000396.
  17. Herrmann WA, Geertsen MS. Efficacy and safety of lornoxicam compared with pla- cebo and diclofenac in acute sciatica/lumbo-sciatica: an analysis from a randomised, double-blind, multicentre, parallel-group study. Int J Clin Pract. 2009;63:1613-21.
  18. Basmajian JV. Acute back pain and spasm. A controlled multicenter trial of combined analgesic and antispasm agents. Spine (Phila Pa 1976). 1989;14:438-9.
  19. Goldie I. A clinical trial with indomethacin (indomee(R)) in low back pain and sciat- ica. Acta Orthop Scand. 1968;39:117-28.
  20. Webster BS, Verma SK, Gatchel RJ. Relationship between early opioid prescribing for acute occupational low back pain and disability duration, Medical costs, subsequent surgery and late opioid use. Spine (Phila Pa 1976). 2007;32:2127-32.
  21. Cooper RJ. NHAMCS: does it hold up to scrutiny? Ann Emerg Med. 2012;60:722-5.

Leave a Reply

Your email address will not be published. Required fields are marked *