Abstract
Headache is a common complaint in the emergency department (ED). In order to examine headache work-ups and diagnoses across the USA, we queried a representative sample of adult ED visits (the National Hospital Ambulatory Medical Care Survey) for the years 1992–2001. Headache accounted for 2.1 million ED visits per year (2.2% of visits). Of the 14% of patients who underwent neuroimaging, 5.5% received a pathological diagnosis. Of the 2% of patients who underwent lumbar puncture, 11% received a pathological diagnosis. On multivariable analysis, a decreased rate of imaging was noted for patients without private insurance [odds ratio (OR) 0.61, confidence interval (CI) 0.44, 0.86] and for those presenting off-hours (OR 0.55, CI 0.39, 0.77). Patients over 50 were more likely to receive a pathological diagnosis (OR 3.3, CI 1.2, 9.3). In conclusion, clinicians should ensure that appropriate work-ups are performed regardless of presentation time or insurance status, and be vigilant in the evaluation of older patients.
Introduction
Headache is a common condition (1, 2), accounting for 1–4% of all emergency department (ED) visits (3–6). While most are benign, some headaches are caused by life-threatening illness (5). The physician’s task is to differentiate appropriately those patients requiring only symptomatic relief from those who require further diagnostic work-up for serious but treatable causes of headache (‘pathological diagnoses’).
The prevalence of benign headache in the general population is high. Migraine headache has a prevalence of 14–17% among females and 5–6% among males (1, 7). According to one survey, 3–5% of the population has visited an ED for headache (7). Total national visits to EDs have been estimated at 800 000 per year for migraine and 600 000 for non-migrainous benign headache (8).
For patients presenting to the ED, it is important to select appropriately those who require further diagnostic evaluation (4, 9–19). A number of studies have evaluated the frequency with which secondary (pathological) causes of headache are diagnosed. These studies show some variability, with rates as low as 4% for all headache patients (3, 5, 17) to >14% for those with sudden-onset, severe headaches (20–22). Most of these studies evaluate the care provided at tertiary academic medical centres. However, the patient population and diagnostic work-ups performed at community hospitals, the source of most emergency care in the USA, are unclear.
In order to estimate frequency of headache presentation, work-ups performed and the rate of serious pathology found in the ED, we investigated the epidemiology of all headache presentations in US EDs over a 10-year period. In order to address this question, we queried the National Hospital Ambulatory Medical Care Survey, a nationally representative sample of hospital visits. While this database does not offer details of the physical examination or patient follow-up, it does offer an unparalleled strength in performing large demographic studies. We hypothesized that neuroimaging would be performed infrequently, that frequency of neuroimaging would be influenced by geography and insurance status, and that older patients would display a higher rate of pathological diagnoses.
Methods
Data from the National Hospital Ambulatory Medical Care Survey (NHAMCS), years 1992–2001, were combined for analysis. NHAMCS is a four-stage probability sample of visits to randomly selected non-institutional general and short stay hospitals, excluding federal, military, and Veterans Affairs hospitals, in the USA (23–32). NHAMCS is conducted annually and covers geographical primary sampling units, hospitals within primary sampling units, EDs within hospitals, and patients within EDs. Trained hospital staff collected data during a randomly assigned 4-week data period for each of the sampled hospitals. Completed forms were reviewed at the National Center for Health Statistics (NCHS), where data abstraction and medical coding were performed. Data were verified with a two-way independent 10% subsample, and keying and coding errors rates averaged <0.1% and <2%, respectively (24). National estimates were obtained through use of assigned patient visit weights, which accounted for probability of visit selection, non-response and ratio of sampled hospitals to hospital universe.
Cases were identified using NCHS-assigned Patient Reason-for-Visit Classification codes of headache (1210.0), migraine (2365.0), sinus headache (1410.1) or facial pain (1055.4) in any of the three reason for visit fields (33). ‘Traumatic’ visits (those with injury or poisoning ICD-9 codes specified) were excluded. Rates per 1000 US population were calculated using data from the US Census Bureau and rates per 1000 ED visits were calculated using projected NHAMCS estimates (34).
Visits were classified by urgency (‘urgent/emergent’ or ‘non-urgent’) of the visit at triage. Visits before 1997 were coded as urgent/emergent or non-urgent. To keep analyses between earlier and later years consistent, visits that occurred after a change in coding in 1997 were coded as ‘urgent/emergent’ if expected wait time was ‘less than 15 min’ or ‘15–60 min’, and as ‘non-urgent’ if expected wait time was ‘>1–2 h.’ Screening and diagnostic procedures were recorded, with ‘any X-ray or Imaging’ defined as any chest, extremity or other unspecified X-ray; or any computed tomography (CT), magnetic resonance imaging (MRI), ultrasound or other unspecified imaging. Up to three patient ICD-9 diagnoses were recorded, and headache cases were further analysed by specific ICD-9 codes that we selected a priori.
Final diagnoses were classified as ‘benign’ or ‘pathological’ based upon the threat to life or permanent injury. Diagnoses classified as ‘benign’ included migraine, vascular headache, tension, anxiety, psychiatric, viral syndrome, hypertension, sinusitis, post-lumbar puncture (LP) headache, otitis media, pharyngitis, upper respiratory infection (URI), periapical abscess, gastroenteritis and fever. Diagnoses classified as pathological included CNS infection (meningitis, encephalitis), stroke, transient ischaemic attack, intracerebral haemorrhage, subarachnoid haemorrhage, aneurysm, glaucoma, benign intracranial hypertension, temporal arteritis and hypertensive encephalopathy. No patients in our cohort received the following diagnoses: malignant hypertension, trigeminal neuralgia, cerebral venous sinus thrombosis, intracranial abscess, cerebral vasculitis, carotid or vertebral dissection.
In order to estimate diagnostic yield of LP and imaging, we evaluated the frequency with which a pathological diagnosis was associated with each test. Only diagnoses that can potentially be made with the study in question were scored. For example, a CT scan in a patient ultimately diagnosed with temporal arteritis was considered negative, while an LP in a patient diagnosed with meningitis was considered positive.
Analyses were performed using STATA 7.0 (StataCorp., College Station, TX, USA). Confidence intervals (CIs) were calculated using the relative standard error of the estimate, using a method approved of by the NCHS. All P-values <0.05 are considered statistically significant.
Results
Demographics
Non-traumatic headache accounted for approximately 21 million visits to US EDs during the years 1992–2001, or 2.1 million visits per year. Table 1 shows the demographic and clinical characteristics of this population. Overall, non-traumatic headache accounted for 2.2% of total ED visits. When compared with the non-headache ED population, patients with headache were disproportionately females (71% vs. 52%, P < 0.001) and aged 18–49 (71% vs. 47%, P < 0.001). Even when migraine and vascular headaches were excluded, females were still disproportionately represented (64% vs. 52%, P < 0.0001). Finally, headache patients were disproportionately privately insured (46% vs. 39% of the general ED population, P < 0.001).
US emergency department (ED) visits for headache, 1992–2001
CI, Confidence interval; N/A, not available, as population data are not available by insurance status.
Work-up
The national rate of imaging was approximately 14% of visits, with CT scan accounting for 95% of these studies. The number of patients receiving an MRI was too low to analyse separately; we therefore combined CT and MRI (Table 2). Only 2.4% of patients underwent LP. On multivariable analysis [controlling for gender, age, diagnosis, time and day of presentation, Metropolitan Statistical Area (MSA), and region of the country], lack of private insurance was associated with a decreased likelihood of receiving a CT or MRI (OR 0.61, CI 0.44, 0.86). Similarly, patients were less likely to receive a CT scan if they presented on a weekday evening (OR 0.64, CI 0.45, 0.91) or on a weekend or overnight (OR 0.55, CI 0.39, 0.77) (multivariable analysis controlling for gender, age, diagnosis, MSA, insurance status and region of the country).
Diagnostic work-up for headache in US emergency departments, 1992–2001
Total number of patients and proportion of headache visitors receiving each test are listed. Neuroimaging data did not become available until 1995. Lumbar puncture data were not available for 2001. Due to low MRI rates, MRI and CT scans were collapsed into one category. Confidence intervals were not calculated (‘nc’) for absolute number of visits less than 30.
Diagnosis
Final diagnoses for patients with headache are shown in Table 3. Migraine or vascular headaches accounted for 63% of all patients. The overall frequency of pathological diagnosis was low (2%; 95% CI 0.5, 3.4). The diagnostic yield of each study (estimated as described in Methods) was approximately 5.5% for CT scanning and 16.7% for LP.
Final diagnosis documented for headache in US emergency departments
Total number of patients with each diagnosis, proportion of headache patients with each diagnosis and 95% confidence intervals are listed. Definitions of ‘pathological’ and ‘benign’ diagnoses are listed in Methods. Some confidence intervals are not calculated (‘nc’) due to low numbers.
CNS, Central nervous system; CVA, cerebrovascular accident; TIA, transient ischaemic attack; ICH, intracerebral haemorrhage; SAH, subarachnoid haemorrhage.
Patients 50 and over had a 6% rate of any pathological finding, vs. 1% for the rest of the headache population (P < 0.001). On multivariable analysis (controlling for gender, region of the country, MSA, time and day of presentation), the odds ratio of a patient 50 and over to receive a pathological diagnosis was 3.3 (95% CI 1.2, 9.3). Figure 1 shows the frequency with which pathological diagnoses are made over a range of age categories. In addition, while the admission rate was 4% for all patients, it was 11% for those 50 and over (P < 0.001).
Frequency of pathological diagnosis by age group.
Discussion
This study represents the largest-scale analysis of the extent to which headache is worked up in US emergency departments. The strength of the NHAMCS database lies in its sheer size and in its representation of all types of hospitals—rural and urban, small and large, and teaching and community. This database therefore provides the most complete picture available of the true prevalence and demographics of headache in the ED. Accordingly, we found that headache accounts for approximately 2.1 million ED visits per year, or approximately 2% of all visits. As expected (6, 17), females with migraine and vascular headaches accounted for the majority of patients.
Beyond simple demographics, however, we focused on diagnostic work-up. We hypothesized that neuroimaging may be underutilized in this country. A number of studies have addressed this question, most commonly at or near academic settings (3, 5, 6, 17, 35, 36). In fact, we found that imaging was performed on 14% of patients, which on first glance appeared high given that 98% of patients were ultimately diagnosed with a benign process. However, 5.5% of those who underwent CT were ultimately diagnosed with a ‘pathological’ process. Despite suggestions that imaging can be overused (37), our national data suggest that current rates are not excessive.
CT scan rates were influenced by time of day and by insurance status, despite the expectation that an ED functions at all times of the day and all days of the week and treats all patients equally.The most worrisome interpretation of this finding is that the decision to perform emergent imaging may be influenced by ability to pay and by time of presentation. Given the yield of CT shown here, we propose that efforts should be focused on increasing the off-hours availability of CT, and improving access to socioeconomically disadvantaged populations. As devastating diagnoses continue to be missed (38), efforts should be made to ensure that appropriate work-ups are available to all who need them.
The rate of LP in this study was surprisingly low. Many patients are likely to undergo CT scan to evaluate for subarachnoid haemorrhage, and LP is typically recommended as the next step in this setting (39, 40). While the number of subarachnoid haemorrhages diagnosed by LP was low (2.1% of LPs performed, data not shown), the overall yield of this procedure was relatively high (17% rate of any pathological diagnosis that can be made by LP). Lumbar puncture may therefore be underutilized in the ED.
Finally, patient age had a significant impact on rate of pathological diagnosis. Patients aged 50 and over demonstrated four times the rate of pathology of younger patients. This supports guideline recommendations that older patients be aggressively evaluated (41).
Limitations
While the NHAMCS database offers tremendous value as a broad-based mechanism for capturing nationwide trends, it comes with several potential limitations. Patient follow-up is not available; therefore, we could not capture patients who ultimately received a pathological diagnosis which was missed in the ED. Diagnosis was determined from the ED record, which may not ultimately have been accurate. Specific features of the history and physical examination were not available, and therefore decision-making could not be analysed directly. Therefore, while broad trends in headache demographics, diagnosis and treatment can be determined, we can draw only limited conclusions regarding appropriateness of patient workups.
We report the largest analysis of headache in US EDs. These results highlight the need for many EDs to increase access to neuroimaging for socioeconomically disadvantaged populations and during off hours. In addition, they support the practice of particular vigilance in the evaluation of older patients.