Abstract
Background
Published data from 1998 revealed that most patients treated for migraine in an emergency department received opioids. Over the intervening years, a large body of evidence has emerged demonstrating the efficacy and safety of non-opioid alternatives. Expert opinion during these years has cautioned against use of opioids for migraine. Our objectives were to compare current frequency of use of various medications for acute migraine in US emergency departments with use of these same medications in 1998 and to identify factors independently associated with opioid use.
Methods
We analyzed National Hospital Ambulatory Medical Care Survey data from 2010, the most current dataset available. The National Hospital Ambulatory Medical Care Survey is a public dataset collected and distributed by the Centers for Disease Control and Prevention. It is a multi-stage probability sample from randomly selected emergency departments across the country, designed to be representative of all US emergency department visits. We included in our analysis all patients with the ICD9 emergency department discharge diagnosis of migraine. We tabulated frequency of use of specific medications in 2010 and compared these results with the 1998 data. Using a logistic regression model, into which all of the following variables were entered, we explored the independent association between any opioid use in 2010 and sex, age, race/ethnicity, geographic region, type of hospital, triage pain score and history of emergency department use within the previous 12 months.
Results
In 2010, there were 1.2 (95% confidence interval 0.9, 1.4) million migraine visits to US emergency departments. Including opioid-containing oral analgesic combinations, opioids were administered in 59% of visits (95% confidence interval 51, 67). The most commonly used parenteral agent, hydromorphone, was used in 25% (95% confidence interval 19, 33) of visits in 2010 versus less than 1% (95% confidence interval 0, 3) in 1998. Conversely, use of meperidine had decreased markedly over the same timeframe. In 2010, it was used in just 7% (95% confidence interval 4, 12) of visits compared to 37% (95% confidence interval 29, 45) in 1998. Metoclopramide, the most commonly used anti-dopaminergic, was administered in 17% (95% confidence interval 12, 23) of visits in 2010 and 3% (95% confidence interval 1, 6) of visits in 1998. Use of any triptan was relatively uncommon in 2010 (7% (95% confidence interval 4, 11) of visits) and in 1998 (10% (95% confidence interval 6, 15) of visits). Of the predictor variables listed above, only emergency department use within the previous 12 months was associated with opioid administration (adjusted odds ratio: 2.87 (95% confidence interval 1.03, 7.97)).
Conclusions
In spite of recommendations to the contrary, opioids are still used in more than half of all emergency department visits for migraine. Though use of meperidine has decreased markedly between 1998 and 2010, it has largely been replaced by hydromorphone. Opioid use in migraine visits is independently associated with prior visits to the same emergency department in the previous 12 months.
Introduction
Migraine, an episodic headache disorder characterized by acutely disabling headaches, causes the majority of the five million headache visits to US emergency departments (EDs) annually (1,2). A large variety of different medications, and classes of medications, are used to treat migraine, including triptans, ergotamines, opioids, non-steroidal anti-inflammatory drugs, dopamine antagonists, anti-epileptics, barbituates and magnesium (1). Perhaps because of the broad range of therapeutic options available, substantial heterogeneity in practice patterns exists across EDs (3,4). Published data from 1998 revealed that 51% of patients treated for migraine in a US ED received parenteral opioids (1). Over the intervening years, a large body of evidence has emerged demonstrating the efficacy and safety of many non-opioid alternatives (5). Expert opinion during this same time period has cautioned against use of opioids for migraine, with reports linking opioid use to ED recidivism (3), ‘chronification’ of migraine (6) and refractoriness to triptan medication (7).
The objective of this analysis was to use the most current National Hospital Ambulatory Medical Care Survey (NHAMCS), a descriptive, ED-oriented US dataset, to compare current frequency of use of various medications for acute migraine with use of these same medications in 1998 and to determine factors independently associated with opioid use.
Methods
Overview
This is an analysis of migraine management in EDs across the USA, drawn from the NHAMCS. NHAMCS is a retrospective probability sample from randomly selected US EDs, designed through a four-tiered sampling strategy, to be representative of all US ED visits. We compared the frequency of use of specific medications for migraine in 2010, the most current dataset available at the time of our analysis, with the frequency of use of the same medications in 1998, the only other time a similar analysis was published (1). We were particularly interested in changes in frequency and type of opioid use for migraine over this time interval. We also examined these data for independent associations between opioid use and the various socio-demographic and clinical variables of interest discussed below. The Albert Einstein College of Medicine Institutional Review Board reviewed our research protocol administratively and determined it to be exempt from further review.
Subject selection
We included in our analysis all patients in the NHAMCS 2010 dataset with the ED discharge diagnosis of ‘migraine’ (ICD9 346). NHAMCS is a dataset available in the public domain, collected and distributed by the National Center for Health Statistics of the Centers for Disease Control and Prevention (http://www.cdc.gov/nchs/ahcd.htm). NHAMCS personnel identify random ED patient visits from randomly selected emergency service areas within randomly selected hospitals, drawn from randomly selected geographic regions of the USA. Trained analysts extract from the medical record the patient’s socio-demographic characteristics, characteristics of the patient’s chief complaint and treatment (8).
Variables of interest
The NHAMCS patient record form is available at http://www.cdc.gov/nchs/data/ahcd/nhamcs100ed_2010.pdf.
Opioids
In 2010, NHAMCS used the Multum classification of therapeutics, a proprietary system of classifying individual medications. Each medication administered or prescribed can be assigned up to four separate therapeutic classes. We counted a medication as an opioid if it was categorized as ‘060: narcotic analgesic’ and as an oral opioid analgesic combination if it was categorized as ‘191: narcotic analgesic combination’. Unfortunately, there is no mechanism in NHAMCS for differentiating route of medication administration. We therefore assumed that most of the opioids were administered parenterally and all of the opioid analgesic combinations were administered orally, since they are not generally available for parenteral use. The 1998 dataset does not contain a similar classification scheme. Therefore, summary data for medication classes in that dataset could be obtained only by summing frequency of use of all the specific opioids.
Medications
We created a list of medications drawn from the published 1998 NHAMCS data and those medications with known efficacy or frequent usage in migraine. Opioids of interest included meperidine, hydromorphone, morphine, nalbuphine and butorphanol. Oral opioid analgesic combinations included those containing oxycodone, hydrocodone and codeine. Anti-emetics of interest included metoclopramide, prochlorperazine, droperidol and promethazine. We counted a medication as an anti-emetic if it was categorized as one of the following: ‘195: 5HT3 receptor antagonist’, ‘196: phenothiazine antiemetics’, ‘197: anticholinergic anti-emetics’ or ‘198: miscellaneous anti-emetics’. Triptan medications were aggregated into a single variable called ‘any triptan’. Also included were the non-steroidal anti-inflammatory drug ketorolac and the corticosteroid dexamethasone, both medications with established efficacy for acute migraine (9,10). Other than promethazine, an anti-histamine with some anti-dopaminergic anti-emetic activity, two other anti-histamines, diphenhydramine and hydroxyzine, are commonly used as adjunctive therapy in migraine and are included in this analysis as well. In NHAMCS, all prescribed medications are assigned at least two five digit numeric codes based on generic and brand name. For example, acetaminophen has a code distinct from Tylenol. To identify all possible names of a particular medication, we searched for different brand names of generic medications using the Wikipedia (http://www.wikipedia.org) and Epocrates (http://www.epocrates.com) websites. We then created variables in which all the different generic and brand names for each medication were coded as a single medication variable.
Socio-demographic variables
NHAMCS reports age, sex and race/ethnicity for each patient visit. In our analysis, we use the race/ethnicity variable reported as non-Hispanic white, non-Hispanic black, Hispanic and non-Hispanic other.
Geographic region
In NHAMCS, the USA is divided into four regions: the Northeast; South; Midwest; West.
Type of hospital
We used two NHAMCS variables to describe type of hospital. The first variable is hospital ownership, which is categorized as: (1) voluntary non-profit; (2) Government, non-Federal; (3) proprietary. Federal hospitals, which represent only 4% of US hospitals, are not included in NHAMCS. The second variable is ‘Seen by resident/ intern?’, which is categorized as yes or no.
Presenting level of pain
NHAMCS records the triage pain score, as reported on a standard verbal numerical rating scale from zero to 10.
Previous ED visits
NHAMCS personnel determine how many visits the patient had made to the ED under examination in the previous 12 months. For the purpose of our analysis, this variable was dichotomized as ‘previous visits’ or ‘no previous visits’.
Length of visit
NHAMCS abstractors calculated the time elapsed between ED arrival and ED discharge in minutes.
Number of medications administered in ED
NHAMCS reports the number of different medications administered.
Outcomes
The outcomes of interest for this study were the frequency of use of opioid and non-opioid medications for acute migraine in US EDs.
Analysis
We used SPSS v.21’s complex samples module to incorporate the relative weight of each patient visit. NHAMCS determines the relative weight by adjusting for sampling strategy, missing data and the relative contribution of various regions of the country. More information on NHAMCS methodology is available at the website (http://www.cdc.gov/nchs/ahcd.htm). We reported frequency of use of each medication in 2010 bounded by 95% confidence interval (CI). We compared the frequency of use of these medications with the same medications in 1998. We re-analyzed the 1998 data to include an updated weighting procedure and to report 95% CI. We also examined the independent association of each of the following eight candidate predictor variables with opioid use: sex; age; race/ethnicity; geographic region of the USA; hospital type; seen by resident/intern; presenting level of pain; previous use of ED in the past 12 months. For each variable, we report the frequency of opioid or opioid analgesic combination use for each response item for the variable. We report raw frequency count, estimated frequency count, and frequency percentage with 95% CI. We computed adjusted odds ratios with 95% CI using a logistic regression model in which all of the variables listed above were included as independent candidate variables. Length of stay is reported as median with inter-quartile range. Number of medications administered in the ED is reported for all migraine patients and for migraine patients who received an opioid. For the continuous variables, age and presenting level of pain, we divided the data at clinically relevant cut-points. For age, this was at 18 years and 50 years. We had hoped to use 65 years as a cut-point, but there was an insufficient number of elderly patients. For pain, we used the categories ‘0’, ‘1–7’ and ‘8–10’. These categories represent no pain, mild to moderate pain and severe pain. Mild and moderate pain were lumped together due to a paucity of patients with mild or moderate pain.
Results
Frequency of use of medications in 2010 and 1998.
In 2010, anti-emetics were administered in 77% (95% CI 70, 83) of migraine visits. Promethazine was the anti-emetic used most commonly in both 1998 and 2010, while use of metoclopramide increased more than five-fold over the intervening period (Table 1). Use of the non-steroidal anti-inflammatory drug ketorolac and the corticosteroid dexamethasone also increased markedly over the intervening years. The frequency of use of these and other medications for acute migraine is reported in Table 1.
Frequency of opioid + opioid analgesic combination use in 2010 by candidate predictor variable.
Estimates based on 30 or fewer individual cases in NHAMCS are considered unreliable.
Federal hospitals are excluded from NHAMCS.
Number of medications administered in 2010.
Median length of stay in the ED in 2010.
Limitations
Limitations of working with NHAMCS have been well described previously (11,12). Primary among these is the use of data that were collected primarily for clinical care. Our definition of opioids and opioid analgesic combinations relies upon correct classification of these medications by NHAMCS personnel, a process that is subject to their own quality control measures. Questions have been raised about the ability of NHAMCS personnel to abstract data from medical records accurately (11,12). We believe the variables of primary interest in this analysis including medications administered, socio-demographic characteristics, type of hospital and previous visits are readily ascertainable and thus not subject to inaccurate chart abstraction.
Factors leading to the choice of administered medication could not be explored fully using available data – we were forced to rely on variables available in NHAMCS. Also, whether or not the listed medication was actually administered cannot be verified. We do not, however, believe these limitations bias our results in a meaningful manner.
We relied on a primary ICD9 diagnostic code to identify our patient population. We only included patient visits if the ICD9 code was ‘migraine’. This no doubt excluded many patients with migraine who received non-specific headache codes or patients who may have incorrectly been coded as sinus headache or tension-type headache, which is a relatively common occurrence (13). Although our strategy lacked sensitivity for identifying cases of migraine, we sacrificed this intentionally for increased specificity. We believe it much less likely for an emergency clinician to code migraine when in fact the patient had some other cause of headache. Also, the reliability of ED migraine diagnoses and the validity of these diagnoses versus criterion standards has been questioned (14). Ultimately, our study population consists of patients whom the emergency clinician believed had migraine, a population that serves well our primary goal of describing prescribing patterns.
Yet another limitation, which we did not anticipate, was the discovery that we lacked sufficient power to perform robust comparisons of opioid use among the various subpopulations of interest. For example, only 19 pediatric patients were included in the sample. We believe this represents under-diagnosis of migraine in the pediatric population. Similarly, only 23 patients were seen by house staff. NHAMCS advises investigators that variables with fewer than 30 cases are considered unreliable when used to create population estimates. Therefore, while our estimates for these variables are the best available, they should be interpreted with caution until validated in future NHAMCS samples. Because data on previous ED visits were available for only two-thirds of the sample, the precision of the regression model is less than it might have been.
Discussion
Our analysis of the NHAMCS demonstrates that despite recommendations to the contrary (3), opioids are still used in the majority of ED visits for migraine. Indeed, the use of opioids for migraine in the ED remains largely unchanged since a similar analysis of national data from 1998 showed that opioids were used in more than half of all migraine patient visits (1). During the 12-year period between 1998 and 2010, use of meperidine, an opioid with a well-known toxic metabolite, decreased from more than one-third of visits to 7% of visits. Another notable change has occurred in the use of hydromorphone, which seems to have taken the place of meperidine in ED-based migraine management. In 1998, hydromorphone was used in fewer than 1% of migraine visits, in contrast to 2010, when hydromorphone has become the most commonly used opioid in migraine, utilized in about 25% of ED visits.
Opioid administration for migraine appears to be independently associated with patients who made at least one other visit to the same ED in the previous 12 months. We were not surprised to find an association between repeated ED use and administration of opioids. This has been reported elsewhere (3,15). We were surprised to find no association between house staff involvement and opioid use. We assumed that treatment of migraine with non-opioid medications was a more current concept and therefore more likely to be seen in academic medical settings. This assumption was not borne out in the house staff data, nor do our data demonstrate a decrease in overall prescribing of opioids between 1998 and 2010, which we would expect if practice patterns were changing gradually.
It is not clear why ED use of opioids for migraine has not decreased since 1998. ED patients accustomed to receiving opioids for migraine may insist on therapy with opioids or may have a history of unsuccessful treatment with non-opioid alternatives. Thus, the emergency clinician may be limited by a lack of alternative management strategies. It is also possible that emergency physicians are less willing to contest requests for opioids to ensure good ratings on satisfaction surveys, a practice that is not evidence-based (16). For patients presenting
Experts advise against the use of opioids for acute migraine (17). Opioids offer less short-term relief than certain anti-dopaminergics and combinations of dihydroergotamine with anti-dopaminergics (18) and are considered less likely to produce sustained headache relief (19). Administration of opioids in the ED is associated with increased ED recidivism (3) and less subsequent responsiveness to sumatriptan (7). Initiation of opioids in the ED for migraine may contribute to the ongoing epidemic of over-prescription of opioid analgesics (20). Outpatient administration of opioids appears to be associated with ‘chronification’ of migraine, i.e. the multi-factorial process by which a patient who suffers from intermittent, episodic migraines begins to experience migraine more frequently (6). Opioid use in the ED may also generate the expectation that the patient will continue to be treated with opioids, which may complicate ongoing care.
On the other hand, opioids are an evidence-based therapy, more effective than placebo with regard to short-term headache relief (5) and supported in this role, albeit weakly, by the US Headache Consortium acute migraine treatment guidelines (21). The US Headache Consortium treatment guidelines, now more than 15 years old, are the only migraine treatment guidelines endorsed by a US emergency medicine organization (21). Clearly, emergency clinicians feel comfortable using this class of medication for acute migraine, just as they would for any cause of acute, severe pain. Serious complications of parenteral opioids are rare in a monitored setting such as the ED and worrisome side effects can be rapidly reversed with a commonly available antidote. High quality evidence is needed to demonstrate the long-term harm associated with episodic use of parenteral opioids in the ED and to determine whether the risk of these medications is greater than with non-opioid alternatives, each of which comes with its own set of adverse effects.
Meperidine use is becoming much less common nationally as pharmacy committees restrict the use of this medication because of concern about toxic side effects (22–24). For patients with migraine, hydromorphone has largely filled the void previously occupied by the more toxic opioid. However, the effect of this change on migraine outcomes has never been measured.
Monotherapeutic management of migraine was exceedingly uncommon in 2010. Most patients received two or more medications. Our analysis revealed that opioids were commonly administered with the anti-histamines promethazine or diphenhydramine, presumably used as a prophylaxis against pruritis, or as agents to enhance the analgesic effect of opioids synergistically. It is as yet unclear whether or not anti-histamines effectively mitigate pruritis in patients administered opioids peripherally, particularly because mechanisms other than histamine release mediate the pruritis (25). Similarly, whether or not anti-histamines augment analgesia or euphoria has not been convincingly established (26,27). The anti-dopaminergics metoclopramide and prochlorperazine were also frequently co-administered with diphenhydramine. Diphenhydramine can decrease extra-pyramidal side effects when co-administered with prochlorperazine or larger doses of metoclopramide (28,29).
Our data revealed that patients who received triptans had the shortest median length of stay in the ED, followed by ketorolac. On the other end of the spectrum were metoclopramide and hydromorphone, both of which were associated with longer stays in the ED. These data are biologically plausible because both hydromorphone and metoclopramide are centrally acting medications known to cause drowsiness. Furthermore, as we have demonstrated, both of these medications are commonly administered with centrally acting anti-histamines, which also cause drowsiness. However, opioids have not been associated with longer ED lengths of stay consistently – one similar analysis demonstrated shorter lengths of stay for opioid versus non-opioid medications (3,30). The triptan data are supported by a large, multi-center, ED-based study of subcutaneous sumatriptan, which reported a median time to meaningful pain relief of 34 minutes (31). However, randomized comparisons of sumatriptan versus combinations of the dopamine antagonists with diphenhydramine revealed that, on average, the dopamine antagonists have sedation scores and return to activity scores no worse than sumatriptan (32,33). Given these conflicting data, we wonder whether unrecognized confounders are at play, such as the severity of an individual’s underlying migraine disorder.
Metoclopramide use has increased five-fold between 1998 and 2010, reflecting an accumulation of published data demonstrating this agent’s safety and efficacy in the ED (32, 34–36). Intermittent production shortages of prochlorperazine likely contributed to this trend. Droperidol, despite a track record of excellent safety and efficacy, is not commonly used, presumably because the US FDA has strongly cautioned against its use (37–39). Like metoclopramide, usage of ketorolac has also increased, perhaps reflecting an increased awareness of safety and efficacy (10). Data supporting the efficacy of dexamethasone for prevention of migraine recurrence after ED discharge have emerged more recently (9).
In conclusion, through an analysis of NHAMCS datasets, we have found that the majority of migraine patients presenting to US EDs receive opioids, a class of medication that has been associated with long-term harm. This practice was prevalent in 1998 and remains unchanged more than a decade later, in spite of current consensus recommendations to the contrary. Although use of meperidine – once the preferred treatment of acute migraine in the ED – has substantially diminished over time, it has largely been replaced by hydromorphone, a less toxic opioid, but one that is no less likely to cause long-term harm such as recurrent ED visits and ‘chronification’ of migraine.
Clinical implications
Use of opioids for migraine in US emergency department has increased. Use of meperidine for migraine is much less common, though it has largely been replaced by hydromorphone. Use of opioids for migraine is associated with prior visit to the same emergency department within the previous 12 months. Use of migraine-specific medication in US emergency departments is uncommon. Most patients treated for migraine in US emergency departments receive more than one medication
Footnotes
Funding
This research received no specific grant from any funding agency in the public, commercial, or not-for-profit sectors.
Conflict of interest
None declared.