In the United States many hospital system stakeholders have expressed concern about increasing backlogs in emergency departments of patients who are in acute need of mental health and substance abuse treatment services (1,2,3,4,5). The total volume of emergency department visits by patients with mental and substance use disorders has been increasing more rapidly than the number of visits by other patients (2,3,6,7), a trend that may intensify administrative burdens and add to the costs of emergency department care. Increased patient volume coupled with a reduced number of emergency departments may exacerbate chronic shortages of specialty mental health care providers and services, which exist in many U.S. communities (8,9), and may increase pressure on inpatient bed capacity in some acute care hospitals. As a result, the typical amount of time that patients with mental and substance use disorders spend in the emergency department may have increased during the past decade.
Longer emergency department visits by patients with mental and substance use disorders would imply that emergency departments are incurring higher costs for acute psychiatric stabilization services. Reports since 2003 (4,5,10) suggest that shortages of psychiatric inpatient beds have become severe in some locations and that managed behavioral health care organizations and hospitals have placed new, tighter restrictions on inpatient admissions (4,11). Because of limited access to beds and other factors, more acutely ill patients could require stabilizing care and may be spending additional time in emergency departments. Delivery of additional services for these visits imposes additional costs on emergency departments in the form of a greater amount of time required from clinical and administrative staff.
Another factor that could contribute to longer emergency department visits by patients with mental and substance use disorders is increased frequency of visits by persons with serious mental illnesses or substance use disorders or persons with both types of disorder (3). Patients with these conditions may require extra time and attention from staff and extensive diagnostic testing related to physical and mental health symptoms. In addition, many of these patients are transferred to psychiatric hospitals or crisis programs, a process that has become administratively more burdensome as a result of regulatory requirements that were introduced to enforce the Emergency Medical Treatment and Active Labor Act (EMTALA) (12), which establishes emergency departments' responsibilities in relation to providing access to care. No formal assessment has been done of whether the duration of visits by patients with serious mental illnesses or substance use disorders differs from the duration of visits by other patients.
This article provides estimates of national trends in the duration of emergency department visits from 2001 to 2006. Trends for visits by patients who presented with complaints related to mental health problems and by patients who presented with other, non-mental health-related complaints are compared. It was hypothesized that visit durations for patients who had mental health complaints would increase relative to the trend for patients with other, non-mental health complaints. The study also assessed the relationship of diagnosis and other factors to the duration of mental health and non-mental health visits.
Data came from the 2001 to 2006 National Hospital Ambulatory Medical Care Survey Emergency Department (NHAMCS-ED) databases, which are developed and made available by the National Center for Health Statistics (NCHS) at the Centers for Disease Control and Prevention. The NHAMCS-ED is a national probability sample of visits to the emergency departments of general and short-stay hospitals, exclusive of federal, military, and Department of Veterans Affairs hospitals, located in the 50 states and the District of Columbia. Data from hospital units of other acute care and long-term care institutions are not included. This study was determined to be exempt from institutional review board review, because the NHAMCS-ED data are deidentified and publicly available.
Data from 2001 and subsequent years were used, because 2001 was the first year that the NHAMCS-ED database included the duration of emergency department visits. The 2001 to 2006 NHAMCS-ED databases contain 218,179 visit records. Records for 467 individuals who died in the emergency department or were dead on arrival were excluded because the duration of these visits was not comparable to others. Of the remaining 217,712 records, 20,762 had no information on visit duration, which left a sample of 196,950. The raw data included some implausibly short visits (less than .02 hours) and some unusually long visits (more than 47 hours). Uncommonly short and long durations threaten robust inference and external validity because they can disproportionately influence estimates of trends in mean values. To reduce the influence of outliers, the lowest and highest 1% of visit durations (3,873 records), which corresponded to visits shorter than 20 minutes and longer than 21 hours, were removed or "trimmed." Consequently, the final sample contained 193,077 visit records, and our results pertain to the 98% trimmed mean for visit durations.
The NHAMCS sample is selected from approximately 1,900 geographically defined primary sampling units (PSUs) that cover the 50 states and the District of Columbia. The 1,900 PSUs are stratified by socioeconomic and demographic variables and then selected with a probability proportional to their size. Stratification is done within four geographical regions by metropolitan statistical area status.
For each emergency department, visits are selected over a randomly assigned four-week reporting period. NCHS staff exclude from the data visits solely for administrative purposes, such as payment of a bill, and visits in which no medical care is provided, such as visits to deliver a specimen. NHAMCS-ED information on visits is extracted by hospital staff from patients' charts and from hospitals' administrative files. A brief, one-page patient record form is completed for each sample visit. Data entries are independently checked by the NCHS for quality control purposes.
Visit duration (hours) was the dependent variable in the analysis. The 2001 to 2006 NHAMCS-ED databases include the number of minutes the patient spent in the emergency department, from admission to discharge. Minutes were converted to hours by dividing by 60.
Mental health-related visits
Trends in the duration of visits made primarily for mental health reasons were compared with trends in the duration of visits made for other, non-mental health reasons. Reasons for the visit are entered as text by hospital staff members and are later assigned numeric codes by the NCHS. Up to three reasons for the visit can be listed (52% of records listed only one reason); respondents are instructed to list the primary reason first. Reasons were classified as mental health related if they included common symptoms of mental disorders (for example, distress, hopelessness, anxiety, or hostility), problems and symptoms resulting from use of alcohol or drugs, mental health and substance use disorders, need for a mental health or psychological evaluation, suicidal behaviors and intentional self-injuries, and need for counseling for various social problems. All other reasons were classified as non-mental health related. Visits were classified as having been associated with a mental health complaint if the primary reason for the visit corresponded to an issue related to mental health or substance use. Visits related to substance use were included in the mental health category because patients' mental health and substance use complaints frequently co-occur and because mental health assessments are usually required when substance use is the primary reason for the visit.
Diagnosis categories. On the basis of ICD-9 codes listed in the visit record, visits were assigned to one of four diagnosis categories by use of a hierarchy (highest to lowest order): serious mental illness, substance use disorders, other mental disorders, and no mental disorder diagnosis. Up to five diagnoses are listed on a visit record. The visit was assigned to the highest-order category among the diagnoses listed. The serious mental illness category included schizophrenia spectrum disorders (ICD-9 295.x), bipolar disorders (296.0, 296.1, and 296.4—296.9), major depression with psychotic features (296.24 and 296.34), and other psychotic disorders (297.x and 298.x). The substance use disorder category included abuse and dependence disorders for alcohol and illicit substances (303.x—305.x, except 305.1). All other mental disorder diagnoses (290.x—319.x, excluding the serious mental illness and substance use disorder diagnoses) were grouped together. All other diagnoses were grouped under no mental disorder diagnosis.
Discharge status. NHAMCS-ED discharge status codes (for example, inpatient admission, transfer, and leaving the emergency department without being seen) are revised annually. Before 2005, visits in which a patient was seen by a provider and then was referred or returned to the referring provider were subclassified into five categories depending on the type of referral. Beginning in 2005, these categories were consolidated, which resulted in time inconsistency. To create a time-consistent classification, discharge status was recoded into the following categories: admitted to the hospital as an inpatient, transferred to another facility, admitted for an observation stay (that is, less than 24 hours), seen by a provider and released (includes all outpatient referrals, return if needed, return by appointment, and no follow-up planned), left before being seen or against medical advice, and other (includes discharges that were coded "no answer" and "other" by NCHS).
Source of payment. The primary source of payment for the visit was categorized as private, Medicare, Medicaid, other coverage, self-pay or charity, or unknown.
Demographic indicators. Patient age, gender, and race-ethnicity (white, black, Hispanic, or other) were used in analyses.
Medical diagnostic services and interventions. Persons seeking mental health care in emergency departments may receive medical diagnostic services (for example, blood glucose tests or CAT scans) or medical interventions (for example, intravenous fluids) before receiving any specialty mental health care (13). Receipt of medical care plus mental health care might account for differences in the duration of mental health and non-mental health visits. To adjust for receipt of medical care, we included three covariates: the total number of common medical diagnostic services provided (range 0 to 21), the total number of common medical procedures (range 0 to 11), and whether the visit was related to an injury or poisoning.
Immediacy of need. Patients who are judged to have a more immediate need for medical attention at the time of admission may be seen sooner, and consequently they may have shorter visits. Four indicators derived from the admission record were included: the admitting provider judged that the patient should be seen within one hour from the time of admission; the patient was not oriented to time, place, and person; the patient's mode of arrival was by ambulance or by public service (police, social service, or beach patrol or escorted by another public service official) versus a "walk-in" (the patient arrived by car, taxi, bus, or foot); and the admitting provider rated the patient as being in moderate or severe pain.
Multivariable regressions for adjusted average durations were estimated with the generalized linear model (GLM) with a gamma-distributed error distribution and a natural logarithm link function. GLM regression is appropriate when the distribution of the dependent variable is skewed (14). Data from all years were pooled, and a year time trend was included as a covariate. All estimates were adjusted for survey design effects and population weights by using Stata, version 10. Test statistics were evaluated at the 5% level of significance. The hypothesis that the trend in duration for mental health visits differed from the trend in duration for non-mental health visits was tested with a t test of the coefficient of the interaction between year and a binary indicator for mental health visits.
Trends in visits and duration
The descriptive data on trends in Table 1 suggest that the total number of mental health-related and non-mental health-related emergency department visits were increasing from 2001 to 2006. The descriptive data also suggest an increase in the average duration of visits for both groups. Among patients who were admitted to an inpatient bed, the average durations of mental health and non-mental health visits were similar. However, among patients who were not admitted to an inpatient bed, the durations of mental health visits appear to have exceeded the durations of non-mental health visits.
In Figure 1 trends in visit duration are disaggregated by diagnosis category and visit type. The trends in the two panels suggest that in each year, the duration of visits by persons with serious mental illness, substance use disorders, or other mental disorders exceeded the duration of visits by persons with no mental disorders.
Characteristics of visits
Table 2 summarizes the characteristics of mental health and non-mental health visits. The frequencies are for all years (2001—2006). For nearly all variables, significant group differences were found. Mental health visits differed from non-mental health visits, especially by higher rates of inpatient admission (24% compared with 12%, p<.001) and transfer (16% compared with 1%, p<.001), a higher percentage of self-pay or charity care (22% compared with 16%, p<.001), a lower prevalence of moderate or severe pain at admission (7% compared with 28%, p<.001), a higher rate of arrival by public service vehicle (31% compared with 10%, p<.001), and a higher percentage of patients not oriented to time, person, and location (10% compared with 1%, p<.001). Mental health visits were also associated with a greater number of diagnostic tests (4.03 compared with 3.01, p<.001), and a higher percentage were injury-related visits (48% compared with 36%, p<.001).
Multivariable regression estimates
Table 3 shows GLM estimates of the average durations of visits. Model 1 included only year, an indicator for visit type, and a visit type × year interaction term. The coefficient estimates can be interpreted as percentages. The estimated year coefficient indicates that the mean duration of emergency department visits increased 2.3% annually from 2001 to 2006 (p=.001). Mental health visits were 42.1% longer than were non-mental health visits (p<.001). This 42% difference implies a difference of approximately 1.25 hours (4.25 hours for mental health visits compared with 3.0 hours for non-mental health-related visits, on the basis of averages reported in Table 1). Tests of the interaction between year and type of visit (coefficient=.009) indicated no difference in the time trends for mental health and non-mental health visits.
Model 2 included the full set of covariates. Adjustment for covariates reduced the magnitude of the year coefficient, which was no longer statistically significant.
Instead of a single dichotomous indicator for mental health visit type, several indicators were included in model 2 for the various combinations of visit type and discharge status. This approach generated estimates of differences in visit duration depending on visit type and discharge status. Confirming the pattern in Table 1, the analysis found no significant difference in the average duration of mental health and non-mental health visits among patients admitted to inpatient beds (coefficient=.035). However, for visits that ended in transfer and those that ended in the patient's release after being seen, mental health visits had longer durations than non-mental health visits. The coefficients for visits that ended in transfer were .351 (p<.001) for mental health visits and -.054 (p=.052) for non-mental health visits. The test of the difference in coefficients (p<.001) implies that mental health transfers on average took longer to complete than did non-mental health transfers. The coefficients for visits that ended with the patient being released after being seen were .030 (p=.295) for mental health visits and —.203 (p<.001) for non-mental health visits. The test of the difference (p<.001) implies that mental health visits with this disposition took longer than non-mental health visits with the same disposition.
Diagnosis category was also strongly related to visit duration. Visits by persons with diagnoses of serious mental illness were 25.0% longer on average than visits by persons with no diagnosis of a mental disorder (p<.001). Similarly, visits by persons with substance use disorder diagnoses were 33.1% longer, and visits by persons with other mental disorders were 9.7% longer than visits by persons with no mental disorder diagnosis (p<.001 for both).
The results of this study indicate that between 2001 and 2006 the average duration of emergency department visits among persons who sought mental health care increased approximately 2.3% per year. Among persons who sought non-mental health care, durations of emergency department visits increased at a similar rate. Throughout the period, persons who sought mental health care had approximately 42% longer visits, on average, than persons who sought other types of care when the analyses controlled for insurance coverage, mode of arrival, demographic characteristics, recorded number of medical tests and procedures, and other factors. This 42% difference implies a difference of approximately 1.25 hours. Durations were especially long for the 16% of mental health-related visits that ended in transfer to a different facility and in the 31% of mental health-related visits by persons with a diagnosis of serious mental illness or a substance use disorder.
At least since the early 1990s, the number of visits to emergency departments by persons with serious mental illness or substance use disorders has been increasing, and there have been numerous reports of overcrowding (2,3,6,7). The finding of longer visits for patients with serious mental illness and substance use disorders suggests that administrative costs associated with visits by these persons have increased as the volume of their visits has increased.
Several factors could account for the comparatively longer visits, especially for patients with serious mental illness and substance use disorders. Managed care restrictions on psychiatric admissions to hospital beds, hospitals' reluctance to admit patients, and bed shortages could lengthen the process of mental health evaluation and treatment (4,5,10,11,14,15,16,17), especially for patients with severe and complex conditions. Restrictions on inpatient admission tend to force greater responsibility for stabilization of acutely ill patients onto emergency departments and other outpatient programs. Perhaps as a result, many emergency departments now provide an array of specialized acute care interventions (for example, acute stabilization, crisis management, and emergency family therapy), and these processes may lengthen mental health visits. Requirements for medical evaluation or "clearance" of persons with mental health complaints and the comparatively greater amount of staff time that patients with disabilities often need (15) could further lengthen visits for patients with severe and complex conditions.
At the same time, capacity constraints that exist in many emergency departments and in outpatient programs that coordinate care with emergency departments could worsen backlogs of emergency department patients and increase hospitals' administrative costs. In many hospitals, shortages of openings in outpatient mental health and substance abuse treatment programs, shortages of specialty mental health and substance abuse treatment providers, and shortages of psychiatric inpatient and residential detoxification programs (8,9) may delay the discharge of emergency department patients with mental health and substance use conditions. As a result, emergency departments may be holding patients longer while linkages to needed services are arranged. This suggests that hospital administrators should periodically assess whether constraints on access to appropriate outpatient psychiatric stabilization services or inpatient care are causing unnecessary delays in patients' emergency department care. Remedies for delays might include expansion of capacity to provide services, increased reimbursement of services, or improved communication with clinical care programs.
Transfers of patients to psychiatric hospitals and crisis programs may also have become more time consuming and costly to complete as a result of regulatory enforcement of EMTALA, which prohibits "dumping" of emergency department patients (12). Under EMTALA, emergency departments are now required to complete an examination and to provide stabilizing treatment for acute conditions before a patient is transferred. They also must document services provided and the justification for the transfer. The need to complete additional assessments and paperwork could explain this study's finding that between 2001 and 2006 mental health transfers took approximately 35% longer to complete than did inpatient admissions. However, some observers have also expressed concern that transfers of patients with mental health conditions from emergency departments to other hospitals are often delayed because providers cannot find an institution or program that will accept the transfer (10). The frequency of delays attributable to this cause is not well documented, and consequently the significance of delays remains an issue being actively debated among observers.
A number of limitations of this study deserve mention. Diagnoses in the NHAMCS-ED database are not based on structured diagnostic assessments; a single patient may have made multiple visits; and very little information is available on hospital characteristics, on whether specialty psychiatric services were provided, and on many other aspects of the care provided during the visit, including the costs of care. In addition, local and hospital-specific influences are masked because the data are aggregated to large regions and visits to particular hospitals cannot be tracked over time. Finally, the database does not contain information about staffing levels or availability of community services, which are critical for assessing the consequences of capacity constraints.
From 2001 to 2006 the duration of emergency department visits made by persons with mental health complaints increased over time and generally exceeded the duration of visits made by patients with non-mental health complaints. Throughout the period, visits by persons with diagnoses of serious mental illnesses, substance use disorders, or both conditions and mental health-related visits had comparatively long durations, as did visits that that ended in transfers to other facilities. Longer visits in these categories suggest that hospital emergency departments incur higher average costs in connection with the delivery of services to persons in need of acute stabilization.