Intravenous acetaminophen does not have an opioid sparing effect in Emergency Department patients with painful conditions

a b s t r a c t

Background: Pain in the Emergency Department is common and is frequently treated with opioids. Due to the opi- oid epidemic, it is clinically helpful to decrease opioid usage. The purpose of this study was to evaluate opioid re- quirement in Emergency Department patients with painful conditions who receive intravenous acetaminophen. Methods: In this prospective cohort study, patients aged 18 years and older and treated with opioids in the acute phase were included. Patients receiving additional intravenous acetaminophen were compared to patients who did not. Primary outcome was opioid requirement, measured in Morphine Equivalent Units (MEU) during Emer- gency Department stay. Secondary outcomes were opioid requirement after discharge; decrease in pain scores; occurrence of adverse events and patient satisfaction.

Results: A total of 116 patients were included of whom 76 received intravenous acetaminophen. Opioid con- sumption in the acute phase was not significantly different (p=0.53) between patients receiving (10.0 MEU (IQR 7.5; 15.0)) and those not receiving acetaminophen: 10.0 MEU (IQR 7.1; 15.0). After discharge these numbers were 15.0 MEU (IQR 7.5; 30.0) versus 30.0 MEU (IQR 15.0; 43.8), respectively (p=0.059). In both groups median NRS pain scores decreased from 9.0 to 4.0 and N80% of patients were satisfied regarding pain treatment. Nine minor adverse events were recorded, equally divided among the groups.

Conclusions: The additional use of intravenous acetaminophen did not decrease opioid requirement in adult pa- tients with acute pain during Emergency Department stay. There was a trend towards decreased opioid require- ment during 24 h after discharge.

(C) 2020 The Authors. This is an open access article under the CC BY license (http://


Currently, the United States is facing a tragic opioid epidemic, which has been referred to as being “the most consequential preventable pub- lic health problem” [1]. Many factors have contributed to the current opioid crisis, that have been outlined extensively in a recent review by Manchikanti, et al. [1] N33.000 deaths annually are due to both prescrip- tion and illegal opioids [2]. Although opioid use shifted from prescribed towards illicit opioids, such as fentanyl and heroin, a thoughtful opioid use in the clinical setting is warranted [3]. However, 61 to 71% of pa- tients present with a painful condition to the Emergency Department and opioids are frequently used to treat acute pain [4,5]. Finding a bal- ance between unrelieved pain and preventing opioid misuse and conse- quently, potential overdose related deaths is challenging [6]. More frequent use of multimodal analgesia, meaning the combination of anal- gesics from different classes that act by different mechanisms in order to

* Corresponding author.

E-mail address: (M.L. Ridderikhof).

obtain additive effects, is one of the proposed solutions to tackle the opi- oid epidemic [7]. Moreover, multimodal pain treatment, including acet- aminophen, might decrease opioid requirements and thus reduce opioid side effects, as has been observed in the perioperative setting [8,9]. However, the evidence for the use of multimodal pain treatment in Emergency Department patients is scarce [10,11].

We hypothesized that administration of additional intravenous acet- aminophen, in adult patients with a painful condition, reduces opioid requirement both in the Emergency Department and during the first 24 h after discharge.


This prospective cohort study was performed and reported in con- formance with the Strengthening the Reporting of Observational Stud- ies in Epidemiology (STROBE) statement [12]. As the Medical Research Involving Humans Subjects Act did not apply, an official approval of the Institutional Review Board was not necessary (waiver number

0735-6757/(C) 2020 The Authors. This is an open access article under the CC BY license (

W18_015#18.026). The study was registered prospectively in the Dutch Trial Register (registration number NTR6958).

Study population

All adult patients, aged 18 years and older, presenting to the Emer- gency Department of a level one trauma center were considered eligible for recruitment between January 30th to March 18th 2018. All patients receiving opioids in the ambulance or during their stay in the Emer- gency Department were included in the study. The exclusion criterion was chronic use of any analgesics. Moreover, as only intravenous acet- aminophen was evaluated, patients who received acetaminophen orally were excluded as well.

Data collection and study outcomes

A trained research assistant collected all data prospectively in stan- dard case report forms during office hours. Treating Emergency Depart- ment staff collected data outside office hours in order to recruit all consecutive patients. After discharge, data were collected by telephone or in person, in case patients were admitted to the hospital. When nec- essary, additional data were acquired from (electronic) ambulance or Emergency Department patient charts.

As the objective of the study was to evaluate opioid requirement in patients in whom intravenous acetaminophen was administered, we aimed to compare patients with or without additional intravenous acet- aminophen administration next to opioids.

The primary study outcome was opioid requirement during the Emergency Department stay. Secondary outcomes included opioid re- quirement after discharge until 24 h later and decrease of the corre- sponding Numerical Rating Scale (NRS) pain scores, both in the acute phase as well as during the first 24 h after discharge from the Emer- gency Department. Additionally, patient satisfaction regarding pain treatment in the Emergency Department and occurrence of adverse events were recorded.

In order to calculate opioid requirement and to compare different opioid analgesics, Morphine Equivalent Units (MEU) were utilized, as described previously and recommended by the Centers for Disease Con- trol and Prevention (CDC) [13,14].

Pain was measured and recorded as NRS pain scores, in which zero means no pain and ten is the worst pain imaginable. The NRS pain score has been validated previously for use in the Emergency Depart- ment [15]. Pain scores were recorded at four different time points. The baseline pain score was recorded just before administration of the (first dose of) opioid (NRS T1). In case the first dose of analgesics was administered prehospitally, the NRS pain score recorded in the ambu- lance was retrieved from the ambulance chart and documented as NRS T1. The second pain score was measured between 15 and 30 min after the initial opioid administration (NRS T2). The third time point was at discharge from the Emergency Department (NRS T3) and the final pain score was measured at 24 h after discharge (NRS T4) and was asked for by telephone.

Patient satisfaction was measured using a 5-point Likert scale with the following options for reply: very dissatisfied; dissatisfied; neutral; satisfied and very satisfied with the pain treatment.

The following parameters were collected as well: age; sex; painful condition and need for additional analgesics (besides acetaminophen or opioids). The triage category was recorded as well and documented utilizing the Dutch Triage System (NTS), which consists of five catego- ries: U1 means immediate care required; U2 within 15 min; U3 within 30 min; U4 within 1 h and U5 within 4 h. In order to compare painful conditions among groups, patients were categorized utilizing the cate- gories traumatic blunt or penetrating injuries; non-traumatic abdomi- nal, thoracic, skin or muscle/joint disorders; headache and vascular extremity disorders.

Statistical analysis and sample size calculation

A sample size calculation based on a standard deviation of 2.5 (un- published pilot data) showed that a group size of 40 patients in each group would have 80% power to detect a difference of 2 MEU between the two groups, utilizing a p-value of 0.05. It was estimated beforehand that both groups would be of unequal size and that less patient would not receive acetaminophen intravenously, as it is common clinical prac- tice in the Netherlands to do so. Therefore, the length of the recruitment period would depend on the total number of included patients who did not receive additional acetaminophen, due to the observational nature of this study.

For descriptive purposes, categorical data were characterized in terms of frequencies with percentages and continuous data as mean values with 95%-Confidence Intervals (95%-CI) when distributed nor- mally. Otherwise, median values with corresponding interquartile ranges (IQR) were used. The two-tailed Student’s t-test or the non- parametric Mann-Whitney U test were used in continuous variables, depending on whether data met the assumptions of normality. For paired groups, the paired t-test or the Wilcoxon signed rank test were utilized, depending on normality. For categorical variables the Chi- Square test was applied. All data were analyzed using IBM SPSS Statis- tics (version 25; 2017; Chicago, Ill).


During recruitment, 316 adult patients (18 years and older) received an opioid prehospitally or in the Emergency Department and were po- tentially eligible for inclusion in the study. A total of 200 patients were excluded. Reasons for exclusion are shown in the Consort diagram (Fig. 1). Morphine was administered as part of acute heart failure man- agement in two patients and as palliative sedative in two patients with massive intracerebral hemorrhages.

Of the remaining 116 patients, 76 received additional acetamino- phen intravenously and 40 did not. Acetaminophen was administered in 31 patients in the prehospital phase and in 47 patients in the Emer- gency Department. Two patients received acetaminophen both in the ambulance as well as in the Emergency Department and were included in the acetaminophen group in the study.

Of the 116 patients that were included in this study, the mean age was 48.2 years and 53.4% of patients were male. Most patients were triaged in category U2 (39.7%) and U3 (31.9%). The majority of patients presented with fractures, strains or sprains or with non-traumatic ab- dominal pain. Table 1 shows that patients receiving acetaminophen were slightly younger and most were male, without significant

Fig. 1. Patient flow chart. Flow chart depicting patient flow. Abbreviations: ED = Emergency Department; APAP = acetaminophen.

Table 1

Baseline characteristics.

Parameter Without APAP With APAP p-value

15.0)) and those not receiving acetaminophen (10.0 MEU (IQR 7.1; 15.0)). When assessing only those patients with opioid administration in the Emergency Department, there were no significant differences in

Mean age – years (95%-CI) 50.6 (44.4 to



47.0 (43.2 to



opioid requirements as well: 7.0 MEU (IQR 2.1; 10.0) compared to 7.5 MEU (IQR 3.1; 13.4), respectively (p=0.32).

3.2. Secondary outcomes

Male sex – n (%)

20 (50.0)

42 (55.3)


Triage category – n (%)?



8 (21.1)

8 (12.3)


17 (44.7)

29 (44.6)


12 (31.6)

25 (38.5)


1 (2.6)

2 (3.1)


0 (0)

1 (1.5)

Pain categories – n (%)



– Blunt

14 (35.0)

35 (46.1)

– Penetrating/wounds

0 (0)

5 (6.6)

After discharge from the Emergency Department, nearly half of all patients (44.0%) were admitted to the hospital, without a significant dif- ference (p=0.13) between patients who did not receive acetaminophen (61.1%) and patients who did receive acetaminophen (45.3%). Forty- five patients (38.8%) required opioid analgesics after disposition and among these patients the median opioid requirement was 15.0 MEU (IQR 11.0; 34.8). Specifically, patients receiving acetaminophen re-

– Abdominal pain 12 (30.0) 29 (38.2)

– Headache 4 (10.0) 4 (5.3)

– Chest pain 5 (12.5) 1 (1.3)

  • Muscle and joints 3 (7.5) 1 (1.3)
  • Skin and soft tissue 2 (5.0) 1 (1.3)

quired 15.0 MEU (IQR 7.5; 30.0) and patients who didn’t receive acet- aminophen required a median of 30.0 MEU (IQR 15.0; 43.8) after disposition (p=0.059).

In total, NSAIDs were administered in 11 patients in the acute phase,

Initial NRS pain score – median (IQR)

9.0 (8.0; 10.0) 9.0 (8.0; 10.0) 0.18

both in the prehospital phase and in the Emergency Department, mainly in patients who received acetaminophen (10 patients).

* Data were missing in 13 patients regarding the triage category. Abbreviations: APAP = acetaminophen; 95%-CI = 95% confidence interval; NRS = Numerical Rating Scale; IQR = interquartile range.

differences. There were no differences in acuity between both groups. Significantly more patients who received acetaminophen presented with wounds. Overall, the initial median NRS pain score was 9.0 (IQR 8.0; 10.0) and both baseline pain scores in the two groups were identical.

Primary outcome

The median dose of opioids administered in the acute phase (i.e. in the ambulance and in the Emergency Department combined) was

10.0 MEU (IQR 7.5; 15.0) in the complete study population. Table 2 shows that opioid consumption in the acute phase was not significantly different (p=0.53) between patients receiving (10.0 MEU (IQR 7.5;

Table 2


Opioid use in ED only – MEU (IQR)


7.5 (3.1; 13.4)


7.0 (2.1;



Patients receiving opioids after

16 (40.0)

29 (38.2)


discharge – n (%)

Opioid use after discharge – MEU (IQR)

30.0 (15.0;

15.0 (7.5;





Abbreviations: APAP = acetaminophen; MEU quartile range; ED = Emergency Departme drugs.

a As conditions of Chi Square test were no

1 (2.5)

= Morphine Equ nt; NSAID = non-s

t met, the Fisher E

10 (13.2) 0.094

ivalent Units; IQR = inte teroid anti-inflammatory

xact Test was utilized.

Study outcomes.


Without APAP



Opioid use in complete acute phase –

10.0 (7.1;

10.0 (7.5;


NSAID use – n (%)

NRS pain scores – median (IQR)

– After 30 min

4.0 (2.5; 7.5)

6.0 (4.0; 8.0)


– At discharge from the ED

4.0 (3.0; 7.0)

5.0 (3.8; 7.0)


– 24 h later

4.0 (1.0; 7.5)

4.0 (2.0; 6.0)


Patient satisfaction with pain treatment – n (%)

– Very unsatisfied

2 (7.4)

1 (2.0)


– Somewhat unsatisfied

2 (7.4)

2 (3.9)

– Neutral

0 (0.0)

1 (2.0)

– Somewhat satisfied

6 (22.2)

11 (21.6)

– Very satisfied

17 (63.0)

36 (70.6)


At disposition, median pain scores in the acetaminophen group de- creased from 9.0 to 5.0 (IQR 3.8; 7.0) and to 4.0 (IQR 3.0; 7.0) in the non-acetaminophen group. After disposition these scores were 4.0 (IQR 2.0; 6.0) and 4.0 (IQR 1.0; 7.5), respectively.

There were no significant differences between the two groups re- garding NRS pain scores at any time point. The complete course of pain scores is shown in Fig. 2.

There were no significant differences regarding pain treatment satis- faction between the two study groups. In the acetaminophen group, 21.6% were somewhat satisfied and nearly 71% were very satisfied and in the non-acetaminophen group, 22.2% were somewhat satisfied and 63% were very satisfied (Table 2).

The median length of stay in the Emergency Department was 209 min (IQR 141; 294) in the total study population. There was no sig- nificant difference (p=0.42) between the acetaminophen and the non- acetaminophen group: 218 min (IQR 143; 307) and 189 min (IQR 140; 283), respectively.

A total of nine patients experienced adverse events in the acute phase, five in the acetaminophen group and four in the non- acetaminophen group. None of the adverse events were serious and they consisted of a dry mouth, nausea with or without vomiting, dizzi- ness and sleepiness.


The results of this study showed that the administration of addi- tional intravenous acetaminophen did not reduce acute opioid require- ment in adult patients with pain presenting to the Emergency Department. However, after disposition there was a noticeable trend to- wards lower opioid use in patients who received additional acetamino- phen in the acute phase versus patients who did not receive this. However, due to the relatively small sample size of patients who re- ceived opioid analgesics after discharge, it cannot be concluded that this effect can be contributed to the administration of acetaminophen alone. Moreover, patients in the acetaminophen group patients received NSAIDs more frequently in the acute phase than patients who did not receive acetaminophen: 13% versus 2.5%. As these patient numbers were small, this difference could partially be explained by the unequal sizes of the patient groups and therefore by the fact that the study was not powered to detect differences in NSAID co-administration. The- oretically, it could have been that in case health care providers adminis- tered acetaminophen besides opioids, a mindset existed regarding the use of multimodal analgesia, which might have increased the chances that NSAIDs were co-administered as well. It should be noted that these differences did not lead to differences in decreases in NRS pain

Fig. 2. NRS pain scores during study. Box plots of NRS pain scores representing interquartile ranges. The bold horizontal line represents the median NRS pain score and the circle the mean score. The vertical lines indicate the upper and lower adjacent values. Abbreviations: NRS = Numerical Rating Scale.

scores between the two groups, although the study was not powered to detect these potentially minor differences.

In the current study, moments of Emergency Department crowding were not evaluated. Crowding could have played a role in not detecting differences in opioid requirements between patients who received acet- aminophen and patients who did not. It has previously been described that crowding is associated with suboptimal pain treatment in the Emergency Department [16].

The opioid crisis in the United States has led to increased interest and use of multimodal pain treatment strategies in the Emergency De- partment, however evidence regarding multimodal pain management in the acute setting is scarce [10]. For this purpose several non-opioid analgesics, such as Non-Steroidal Anti-Inflammatory Drugs (NSAIDs), acetaminophen, low-dose ketamine, intravenous lidocaine as well as re- gional anesthesia could be used [17]. Based on a retrospective chart re- view study of 102 patients, those receiving adjunct analgesia (mainly NSAIDs) required less opioids (7 vs 8 MEU) [11].

However, evidence regarding an opioid sparing effect of intravenous acetaminophen administered in an Emergency Department setting is scarce. Several randomized controlled trials comparing intravenous acetaminophen to other analgesics have been published. Some of these studies reported the need for rescue analgesia as secondary end- point, however, none of the previously published trials reported on opi- oid requirement [18]. Strikingly, in the peri-operative setting, intravenous acetaminophen decreased opioid consumption. Based on a network meta-analysis, co-administration of acetaminophen together with opioids led to a mean decrease of 10.5 mg morphine per 24 h in 699 patients after major surgery [9]. This difference was even larger (N20 mg morphine per 24 h) when acetaminophen was combined with an NSAID or with nefopam. A 2016 Cochrane review of 7200 pa- tients from 75 studies showed that patients receiving intravenous acet- aminophen (or propacetamol, the precursor drug of acetaminophen) required up to 26% less opioids than those receiving placebo over 4 h and 16% less over 6 h [19].

With respect to the analgesic efficacy in the Emergency Department, intravenous acetaminophen has been compared to other analgesics in

several randomized controlled trials with conflicting results. In 2016, a systematic review summarized 14 randomized controlled trials includ- ing 1472 patients with pain from various origin and found in only three trials a significant reduction in pain scores after co-administration of in- travenous acetaminophen [18]. Moreover, the majority of included studies suffered from serious methodological shortcomings, resulting in a low level of evidence.

The largest trial so far, focusing on analgesic efficacy of intravenous acetaminophen compared to other analgesics, assessed a total of 1644 adult patients with acute renal colic randomly treated with 1000 mg acetaminophen intravenously, 75 mg diclofenac intramuscularly or

0.1 mg/kg morphine intravenously [20]. The primary outcome of 50% reduction of the initial pain score at 30 min after administration of the analgesic was achieved in 66%, 68% and 61%. Regarding the comparison of intravenous versus oral administration of acetaminophen, there is no evidence that the intravenous route is superior [21]. Taken together, based on available literature, the analgesic effect of intravenous acet- aminophen is debatable and in line with our results an opioid sparing action of acetaminophen has not been demonstrated previously in the acute setting.


First of all the current study was a non-experimental cohort study and this design has its inherent limitations. Patients who received intra- venous acetaminophen could have differed in important ways from the patients who did not. Therefore, although measured variables did not show major differences between the groups, not all relevant differences might have been captured. Besides, we cannot rule out that our sample size was too small to show meaningful clinical differences in opioid re- quirements. However, we performed an a priori sample size analysis and given the obvious lack of effect, we are convinced that increasing the sample size would not have led to a different outcome. In this sam- ple size analysis, we chose to use a difference of 2 mg MEU between the groups. This was not based on previous literature and more patients

would have been required if a smaller difference was to be detected. Ad- ditionally, the study was too small to detect meaningful adverse events. Because all patients were interviewed within 24 to 48 h after dispo- sition from the Emergency Department, recall bias was minimized. Uti- lizing interviews by telephone could have led to socially desirable answers, which could explain the high patient satisfaction rate. Addi- tionally, using the telephone led to absence of visual or nonverbal cues, that might have influenced study results. On the other hand, these forms of bias are very hard to eliminate from interview tech- niques. In our study, we did not explicitly consider the use of anti- epileptics, such as gabapentin or pregabaline, as well as other drugs with analgesic properties, however it should be emphasized that in our study population these drugs were not administered in the Emer- gency Department. Lastly, analgesics that were taken prior to arrival of

the ambulance at home, were not accounted for as well.


No significant differences in total opioid requirement in the acute phase were detected between patients receiving additional intravenous acetaminophen compared to patients treated with opioids alone. How- ever, there was a non-significant trend towards lower opioid require- ment during 24 h after discharge from the Emergency Department. Decreases in pain scores were comparable as was patient satisfaction re- garding acute pain management. Prospective intervention studies are required to draw firm conclusions regarding potential opioid-sparing effects of intravenous acetaminophen in the Emergency Department.

Sources of support


CRediT authorship contribution statement

Z. Blok: Investigation, Formal analysis, Writing – original draft. M.L. Ridderikhof: Conceptualization, Methodology, Formal analysis, Writing

– review & editing. H. Goddijn: Conceptualization, Methodology. M. Berendsen: Validation, Writing – review & editing. M.W. Hollmann: Su- pervision, Writing – review & editing.

Declaration of competing interest



We would like to thank Susanne Van Dieren for the design of Fig. 2.


  1. Manchikanti L, Sanapati J, Benyamin RM, Atluri S, Kaye AD, Hirsch JA. Reframing the prevention strategies of the opioid crisis: focusing on prescription opioids, fentanyl, and heroin epidemic. Pain Physician 2018;21:309-26.
  2. Soelberg CD, Brown Jr RE, Du Vivier D, Meyer JE, Ramachandran BK. The US opioid crisis: current federal and state legal issues. Anesth Analg 2017;125:1675-81.
  3. Rose ME. Are prescription opioids driving the opioid crisis? Assumptions vs facts. Pain medicine (Malden, Mass), vol. 19. ; 2018. p. 793-807.
  4. Cordell WH, Keene KK, Giles BK, Jones JB, Jones JH, Brizendine EJ. The high preva- lence of pain in emergency medical care. Am J Emerg Med 2002;20:165-9.
  5. Johnston CC, Gagnon AJ, Fullerton L, Common C, Ladores M, Forlini S. One-week sur- vey of pain intensity on admission to and discharge from the emergency depart- ment: a pilot study. J Emerg Med 1998;16:377-82.
  6. Paice JA. Managing pain in patients and survivors: challenges within the United States opioid crisis. Journal of the National Comprehensive Cancer Network: JNCCN 2019;17:595-8.
  7. Adams J, Bledsoe GH, Armstrong JH. Are pain management questions in patient sat- isfaction surveys driving the opioid epidemic?, vol. 106; 2016; 985-6.
  8. Practice guidelines for acute pain management in the perioperative setting: an up- dated report by the American Society of Anesthesiologists Task Force on Acute Pain Management, Anesthesiology 2012;116:248-73.
  9. Martinez V, Beloeil H, Marret E, Fletcher D, Ravaud P, Trinquart L. Non-opioid anal- gesics in adults after major surgery: systematic review with network meta-analysis of randomized trials. Br J Anaesth 2017;118:22-31.
  10. Todd KH. A review of current and emerging approaches to pain management in the emergency department. Pain Ther 2017;6:193-202.
  11. Priano J, Faley B, Procopio G, Hewitt K, Feldman J. Adjunct analgesic use for acute pain in the Emergency Department. Hosp Pharm 2017;52:138-43.
  12. von Elm E, Altman DG, Egger M, Pocock SJ, Gotzsche PC, Vandenbroucke JP. The strengthening the reporting of observational studies in epidemiology (STROBE) statement: guidelines for reporting observational studies. Lancet (London, England) 2007;370:1453-7.
  13. Dowell D, Haegerich TM, Chou R. CDC guideline for prescribing opioids for chronic pain–United States, 2016. Jama 2016;315:1624-45.
  14. Opioid oral morphine milligram equivalent (MME) conversion factors. Accessed May 25, 2019, at

    PrescriptionDrugCovContra/Downloads/Oral-MME-CFs-vFeb-2018.pdf; 2019.

    Bijur PE, Latimer CT, Gallagher EJ. Validation of a verbally administered numerical rating scale of acute pain for use in the emergency department. Acad Emerg Med 2003;10:390-2.

  15. Morley C, Unwin M, Peterson GM, Stankovich J, Kinsman L. Emergency department crowding: a systematic review of causes, consequences and solutions. PloS one 2018;13:e0203316.
  16. Cisewski DH, Motov SM. Essential pharmacologic options for acute pain manage- ment in the emergency setting. Turkish journal of emergency medicine 2019;19: 1-11.
  17. Sin B, Wai M, Tatunchak T, Motov SM. The use of intravenous acetaminophen for acute pain in the Emergency Department. Acad Emerg Med 2016;23:543-53.
  18. McNicol ED, Ferguson MC, Haroutounian S, Carr DB, Schumann R. Single dose intra- venous paracetamol or intravenous propacetamol for postoperative pain. Cochrane Database Syst Rev 2016:Cd007126.
  19. Pathan SA, Mitra B, Straney LD, et al. Delivering safe and effective analgesia for man- agement of renal colic in the emergency department: a double-blind, multigroup, randomised controlled trial. Lancet (London, England) 2016;387:1999-2007.
  20. Jibril F, Sharaby S, Mohamed A, Wilby KJ. Intravenous versus oral acetaminophen for pain: systematic review of current evidence to support clinical decision-making. Can J Hosp Pharm 2015;68:238-47.