Trends in U.S. Emergency Department Visits for Mental Health Conditions, 1992 to 2001

Begun in 1992 as part of the ambulatory component of the National Health Care Survey, the NHAMCS' emergency department survey measures use of emergency health care services by employing a four-stage probability sample of visits to noninstitutional general and short-stay hospitals, excluding federal, military, and Veterans Affairs facilities, in the United States (16,17,18,19,20,21,22,23). Conducted annually, the NHAMCS covers geographic primary sampling units, hospitals within primary sampling units, emergency departments within hospitals, and patients within emergency departments. Data are collected by hospital staff during annual, randomly assigned, four-week data periods and are coded by using ICD-9-CM (24). We used the most recent emergency department component of the NHAMCS available (1992 to 2001).

To maximize case finding, mental health-related emergency department visits were included if their records met any one of three criteria: DSM-IV-TR-based, major mental health problems (ICD-9-CM diagnoses 290.0 to 305, 307 to 310, or 311 to 319.0 or V codes 61.1 to 71.02); National Center for Health Statistics (NCHS)-assigned reason-for-visit classification codes related to mental health (25) (1100.0 to 1199.9); or injury E codes related to suicide (E950.0 to E959.9). Otherwise, visits that did not meet at least one of the above criteria were deemed non-mental health visits. ICD-9-CM codes in the range of 290 to 319 were excluded if they lacked corresponding DSM-based diagnosis—for example, psychosexual disorders (ICD code 302), sleeping disturbances (code 307.4), physiological malfunction (code 306), postconcussive syndrome (code 310.2), nondependent tobacco use disorder (code 305.1), and enuresis and encopresis (codes 307.6 and 307.7).

Mental health-related visits were assigned specific DSM-IV-compatible categories as follows: mood disorders (ICD or DSM-IV codes 296 to 296.9, 300.4, or 311)—for example, major depressive disorder, depression not otherwise specified, dysthymia, bipolar disorder, other mood disorders, and NCHS reason-for-visit code 1110.0 (depression); anxiety disorders (ICD or DSM-IV codes 300.00 to 300.61 except 300.4 [dysthymia])—for example, generalized anxiety disorder, panic disorder, obsessive-compulsive disorder, posttraumatic stress disorder, acute stress disorder, agoraphobia, anxiety not otherwise specified, and NCHS reason-for-visit code 1100.0 (anxiety and nervousness) and 1105.0 (fears and phobias); psychotic disorders (ICD or DSM-IV codes 295 to 295.9, 297.3, 298.8, or 298.9)—for example, schizophrenia, psychosis, and NCHS reason-for-visit code 1155.0 (delusions or hallucinations); substance-related conditions (ICD or DSM-IV codes 290.44 to 292.94 or 302.89 to 305.98)—for example, alcohol and other substance abuse—and NCHS reason-for-visit code 1145.0 (alcohol-related problems) or 1150.0 (abnormal drug use); suicide attempts or ideation (ICD-9 E codes 950 to 959) and NCHS reason-for-visit code 5820.0 (suicide attempt) or 5820.1 (intentional overdose); miscellaneous disorders (ICD-9 V codes 61.1 to 71.09)—for example, adjustment disorders and problems in living—and NCHS reason-for-visit classification 1130.0 (behavioral disturbances) and 1165 (other symptoms or problems relating to psychological and mental disorders not elsewhere classified).

Absolute numbers of emergency department visits were estimated by using census-based, NCHS-assigned patient weights rounded to the nearest thousand. Emergency department visit rates per population were calculated by using denominator estimates of the civilian, noninstitutionalized U.S. population from the U.S. Census Bureau, which were adjusted for underenumeration (26).

We analyzed visit rates by age, sex, race or ethnicity (white non-Hispanic, black non-Hispanic, other, and Hispanic). Although NHAMCS data contain imputed ethnicity for 1992 to 1996 in the public use files, we did not include any imputed or missing data in the ethnicity subanalysis. We also analyzed mental health-related visits by insurance status, location in a metropolitan statistical area (MSA), and region of the country (the Northeast, the Midwest, the South, and the West). MSA and U.S. region categories represent standardized geographic divisions defined by the U.S. Census Bureau (26).

Visits were further analyzed by urgency at triage. For 1992 to 1996, visits were classified as either urgent or nonurgent, but this coding system changed in 1997. For consistency, we coded visits that occurred from 1997 to 2001 as nonurgent or nonemergent if the expected triage time was recorded as more than one or two hours.

Confidence intervals (95 percent CIs) for visit rates were calculated by using the relative standard error of the estimate, controlling for weighting, four-stage sampling, and cluster effects with use of generalized estimating equations from SUDAAN-8.0. In accordance with NCHS recommendations (27,28), only estimates for which the relative standard error was less than 30 percent and for which there were more than 29 raw data records in the cell are reported here. The least-squares method of linear regression was used for analysis of trends with use of STATA 7.0, with p<.05 considered statistically significant. Differences in continuous variables were assessed by using two-tailed independent-samples t tests or repeated-measures analysis of variance. Significance testing for multiple comparisons between groups were corrected by using the method of Bonferroni.

In 1980 the National Medical Care Utilization and Expenditure Survey showed that approximately 11 percent of the U.S. population visited an emergency department annually (30). Twenty years later, the National Health Interview Survey reported that this proportion had doubled (31). Although overall use of U.S. emergency department services increased by 8 percent from 1992 to 2001, the number of documented mental health-related visits increased at an even faster rate—by 38 percent, from 4,371,112 to 6,721,540 visits across the decade.

Current trends in emergency department oversubscription and overcrowding stem from a complex array of micro- and macroeconomic forces. At the patient level, a lack of insurance, social support, and alternatives to care, coupled with the convenience of 24-hour accessibility to the emergency department, are major incentives for both the worried well and the acutely psychotic patient to rely on emergency-based care (32,33).

At the macroeconomic level, there is a parallel reduction in the intensity of treatment that is commonly available for mental illness and in the proportion of the U.S. health care budget spent on mental health care (34). In addition, under the Emergency Medical Treatment and Active Labor Act (EMTALA), any hospital that accepts Medicare funding must screen and stabilize all patients who present for care, regardless of the patients' ability to pay (35,36). Because of this unfunded mandate, many hospitals have closed their emergency departments, resulting in fewer resources and more overcrowding for all concerned (37,38). Because patients with mental illness may be particularly challenged to negotiate a health care system that is increasingly crowded (39), outpatient (40), and plagued with resource scarcity, the rates of both mental health-related visits and patients who leave the emergency department before being seen have increased significantly over time.

There are many possible explanations for a disproportionate increase in mental health-related emergency visits, including diminishing access to mental health care, increasing rates of mental illness in the overall population, enhanced diagnosis of mental illness through increasing use of board-certified emergency physicians and emergency department-based mental health professionals, shrinking alternative treatment resources, and increasing fragmentation and out-of-pocket payments for mental health care under managed care.

In an unweighted, non-DSM-based analysis of raw numbers of emergency mental health-related visits, McAlpine and Mechanic (41) suggested that the uninsured were largely responsible for the increasing burden of such visits. However, our data show that when population rates are examined, the greatest proportional increases in visits over the past decade were actually among individuals who had insurance. Increasing visit rates for insured and uninsured patients are more likely explained by systemwide shifts to managed mental health care and increasing reliance on emergency departments for the only 24-hour universally open-access health care in the United States (34,40,42,43,44).

In regional terms, the Northeast experienced an inverse relationship between supply of mental health providers (psychiatrists and psychologists) and demand for emergency department mental health care. The Midwest, having the highest number of psychiatrists per capita, experienced no increase in mental health-related visits across the decade, suggesting possible threshold effects at 11 providers per 1,000 population. Further explication of this phenomenon would require extending Wennberg's analyses of local norms to regional populations (45,46), but diminished access to psychiatric services may account for some of the observed regional variation in mental health-related use of the emergency department.

This analysis had several limitations, including the reporting biases resulting from the fact that 20 to 45 percent of U.S. emergency departments are still staffed by non-board-certified emergency physicians or midlevel providers; others have shown that even highly trained physicians are notoriously poor at making accurate mental health diagnoses (11, 47,48,49,50,51). Second, the NHAMCS data on mental health diagnoses are obtained by retrospective review of records (52). In the absence of validated screening instruments, inaccuracies are likely, and only the most obvious diagnoses may ever be recorded. It is also a limitation that we could not determine the frequency of multiple visits. It is well known that a small proportion of patients may account for a large number of visits, a result of both dual diagnoses and recidivism (53). To control for undercounting, we also used reason-for-visit and injury E codes to attempt to include as many mental health patients as the data would allow. Previous work in assessing occult mental health problems suggests that undercounting is more likely to occur than overcounting in U.S. emergency departments (49).

Despite its shortcomings, this is one of the first studies to show prevalence and trends in mental illness in a representative national sample of U.S. emergency department visits. Consistent increases in use of emergency departments by mental health patients show no sign of abatement. As the nation's emergency safety net frays in the wake of severe overcrowding, vulnerable persons with mental illness are likely to suffer the most. In addition, an increasing reliance on emergency care has important policy implications for the entire health care system. Future work must build on these national data to find cross-cutting solutions that address important barriers to care for persons with mental illness and the gaps between the fields of emergency medicine and mental health.

Dr. Larkin is affiliated with the department of surgery and Dr. Claassen with the department of psychiatry of the University of Texas Southwestern Medical Center, 5323 Harry Hines Boulevard, Charles Sprague 2.122, Dallas, Texas 75390-8579 (e-mail, [email protected]). Ms. Emond, Ms. Pelletier, and Dr. Camargo are with the department of emergency medicine of Massachusetts General Hospital in Boston.