Median length of stay and population rate of child and adolescent mental health discharges from community hospitals from 1990 to 2000. Results were based on Healthcare Cost and Utilization Project Nationwide Inpatient Sample weighted sampling of discharged children aged younger than 18 years with principal diagnoses 290 through 319 from the International Classification of Diseases, Ninth Revision, Clinical Modification. Population rates were calculated using annual US Census measures of civilian noninstitutionalized residents aged younger than 18 years.
Case BG, Olfson M, Marcus SC, Siegel C. Trends in the Inpatient Mental Health Treatment of Children and Adolescents in US Community Hospitals Between 1990 and 2000. Arch Gen Psychiatry. 2007;64(1):89-96. doi:10.1001/archpsyc.64.1.89
Previous work has demonstrated marked changes in inpatient mental health service use by children and adolescents in the 1980s and early 1990s, but more recent, comprehensive, nationally representative data have not been reported.
To describe trends in inpatient treatment of children and adolescents with mental disorders between 1990 and 2000.
Design and Setting
Analysis of the Healthcare Cost and Utilization Project Nationwide Inpatient Sample, a nationally representative sample of discharges from US community hospitals sponsored by the Agency for Healthcare Research and Quality.
Patients aged 17 years and younger discharged from US community hospitals with a principal diagnosis of a mental disorder.
Main Outcome Measures
Changes in the number and population-based rate of discharges, total inpatient days and average length of stay, charges, diagnoses, dispositions, and patient demographic and hospital characteristics.
Although the total number of discharges, population-based discharge rate, and daily charges did not significantly change between 1990 and 2000, the total number of inpatient days and mean charges per visit each fell by approximately one half. Median length of stay declined 63% over the decade from 12.2 days to 4.5 days. Declines in median and mean lengths of stay were observed for most diagnostic categories and remained significant after controlling for changes in background patient and hospital characteristics. Discharge rates for psychotic and mood disorders as well as intentional self-injuries increased while rates for adjustment disorders fell. Discharges to short-term, nursing, and other inpatient facilities declined.
The period between 1990 and 2000 was characterized by a transformation in the length of inpatient mental health treatment for young people. Community hospitals evaluated, treated, and discharged mentally ill children and adolescents far more quickly than 10 years earlier despite higher apparent rates of serious illness and self-harm and fewer transfers to intermediate and inpatient care.
Use of inpatient mental health treatment underwent marked change during the 1990s. In contrast to stable numbers of discharges for other medical conditions, there was an increase in population-adjusted discharges for mental disorders in the United States.1,2 Concurrently, declines in the duration of hospital stays for mental conditions far outpaced reductions for other illnesses, and total days of treatment for mental disorders fell. Because youth and adults increasingly differ in their use of inpatient medical care,3 findings from research on adult and all-ages inpatient mental health care use trends may not accurately describe the care received by young Americans.
Little information exists concerning trends in the inpatient mental health treatment of child and adolescents, and available studies document disparate findings. Work based on 1 private health insurance claims database4,5 found consistent annual decreases between 1993 and 2000 in the proportion of youth using inpatient mental health services, the length of stay (LOS), and the cost of their treatment. In contrast, analysis of a nationally representative survey of specialty public and private sector mental health service providers6 found significant increases in annual inpatient admission rates of approximately 120% for young people during the period 1986 through 1997. A nationally representative general hospital discharge survey,7,8 which excluded most psychiatric hospitals, found a more modest increase in the rate of inpatient psychiatric admission of children and adolescents between 1988 and 1995. This was accompanied by substantial declines in LOS resulting in a reduction in the total number of inpatient days.
We analyze data from the largest nationally representative community hospital discharge survey to describe changes in the rate, admission type and referral source, patient demographic and clinical characteristics, hospital characteristics, duration, cost, and disposition of child and adolescent mental health discharges from community hospitals between 1990 and 2000. Our findings update earlier reports on national trends in general hospital child and adolescent inpatient mental health care use, providing an additional 5 years of data during a period marked by dramatic expansions of both managed behavioral health care9- 12 and treatment of American youth with psychotropic medications.4,13,14 To our knowledge, our work is the first nationally representative study to present combined analyses for discharge rate, LOS, and charges. Further, the large sample sizes permit for the first time a description of secular trends in LOS by diagnostic category and other patient and hospital characteristics.
We analyzed data from the Nationwide Inpatient Sample (NIS) of the Healthcare Cost and Utilization Project for the years 1990 to 2000. The NIS, conducted by the Agency for Healthcare Research and Quality, is a nationally representative sample survey of US community hospitals, defined annually as all hospitals that were open during any part of the calendar year and were designated as community hospitals in the American Hospital Association's Annual Survey of Hospitals.15 The American Hospital Association identifies as community hospitals any nonfederal, short-term, general, or specialty hospitals, including public hospitals and academic medical centers. Excluded are hospital units of institutions; the few freestanding psychiatric hospitals and alcoholism/chemical dependency treatment facilities; long-term hospitals; and, beginning in 1998, short-term rehabilitation hospitals.
Annual NIS databases contain data for 5 to 8 million hospital stays from approximately 1000 hospitals. Inpatient stay records in the NIS include clinical and resource use information typically available from discharge abstracts. Charge information is provided for all discharges, regardless of payer, including persons covered by Medicare, Medicaid, private insurance, and the uninsured. The NIS uses a stratified probability sample design to approximate a 20% stratified sample of all US community hospitals with sampling probabilities proportional to the number of US community hospitals in each stratum. To obtain national estimates, the Agency for Healthcare Research and Quality provides sampling weights and strata that we used in all analyses.16
We limited analyses to children and adolescents aged younger than 18 years with a principal diagnosis of a mental disorder (codes 290 to 319 from the International Classification of Diseases, Ninth Revision, Clinical Modification [ICD-9-CM]) in each year of the NIS between 1990 and 2000. Our trend analyses used the 2 yearly end points in which there were 21 450 child and adolescent mental disorder discharges in the 1990 sample and 29 590 in the 2000 sample.
Principal ICD-9-CM mental health diagnoses were partitioned into 8 diagnostic categories using a crosswalk developed in prior research.7 Comorbid substance-use disorders, intentional self injuries, and general medical conditions were identified among the 14 available secondary diagnoses. Demographic characteristics included age, sex, and race/ethnicity. Hospital characteristics included hospital region, state, urban/rural location, teaching status, and bed size. Hospital stay characteristics examined were admission type, admission source, primary payer, day and month of admission, and discharge disposition. Hospital region was coded by US Census convention into Northeast, Midwest, South, or West.17 Hospitals located in metropolitan statistical areas were considered urban, and hospitals with a residency program approved by the American Medical Association, with a membership in the Council of Teaching Hospitals, or with a ratio of full-time–equivalent interns and residents to beds of 0.25 or higher were considered teaching facilities.18 Urban/rural location and teaching status were combined into a single descriptor coded as rural, urban nonteaching, or urban teaching. Bed size was coded consistent with NIS convention as small, medium, or large with cut points chosen so that approximately one third of the hospitals in a given region and location/teaching combination would be included in each bed-size category. Day of admission was defined as a weekday (Monday through Friday) or weekend (Saturday and Sunday). Month of admission was coded as winter (December through February), spring (March through May), summer (June through August), or fall (September through November) as well as school vacation (December, July, and August) or school year (all other months). Variables addressing median household income in patients' home ZIP codes and hospital ownership were inconsistent across study years, precluding meaningful comparison.
We calculated national population-based rates of child and adolescent mental health discharges and associated charges per 1000 children using US Census measurements of the number, age, sex, and race/ethnicity of civilian noninstitutionalized residents aged 17 years and younger for the years 1990 through 2000.19,20 Population-based discharge rates by insurance status were calculated using adjusted estimates for noninstitutionalized youth from the Current Population Survey published by the US Census Bureau.21,22 Charges were calculated in constant year-2000 dollars after adjustment according to the US Bureau of Labor Statistics annual all-urban consumer price index for medical care.23 Length of stay was calculated by subtracting the admission date from the discharge date except in cases of same-day discharges, which were assigned an LOS of 0.5 days. Mean or median changes or differences in discharge rate, number, charges, or LOS were considered statistically significant if the corresponding 95% confidence interval (CI) excluded zero. Changes in the distribution of discharge characteristics were tested for significance at the 5% level using a χ2 statistic. If significant, the change of each category of the variable vs all other categories was tested for significance at a 5% Bonferroni adjusted level (based on the number of categories) using a χ2 statistic. Differences in population-based rates between years or variable categories were considered statistically significant if the 95% CIs of the estimated rates did not overlap. These analyses were conducted with SAS-callable SUDAAN version 9.0 (RTI International, Research Triangle Park, NC) using programming code provided by the Agency for Healthcare Research and Quality to accommodate complex sample design and weighting.
Within each demographic and clinical stratum of interest, we used backward, stepwise, multivariate linear regression analysis in SAS version 9.1 (SAS Institute, Cary, NC) to identify the impact of year of discharge on LOS, controlling for all other patient, hospital, and discharge characteristics that were significant at the P<.25 level. We refit the resulting models in SUDAAN to produce regression coefficients (βs) and appropriate 95% CIs for year of discharge, accounting for the sample weights and complex sample design.
The number and population-based rate of child and adolescent mental disorder discharges from community hospitals did not significantly change between 1990 and 2000 (Table 1) or between any years over the period (Figure). However, the total number of associated inpatient days fell by approximately one half over this period with a somewhat greater decline in the mean mental disorder inpatient days per 1000 children and adolescents. These declines were accounted for entirely by a marked reduction in LOS. Mean charge per discharge fell significantly, reflecting reductions in LOS.
Table 2 presents the distribution of patient demographic, hospital, and other associated characteristics of child and adolescent mental health discharges in the years 1990 and 2000. Significant changes were observed for age, admission source, and discharge disposition. The proportion of child discharges aged 6 to 13 years rose significantly over the period while the proportion of discharges in other age categories fell. Population-based discharge rates by age, sex, race/ethnicity, and insurance status did not differ between years. In 2000, population-based discharge rates per 1000 youth aged under 5 years, 5 to 13 years, and 14 to 17 years were 0.1 (95% CI, 0.08 to 0.13), 1.3 (95% CI, 1.1 to 1.6), and 5.7 (95% CI, 4.8 to 6.7), respectively. Rates for male individuals were 1.9 (95% CI, 1.6 to 2.3) and for female individuals, 2.1 (95% CI, 1.7 to 2.4). Population-based discharge rates for white, black, Hispanic, and other children and adolescents were 2.4 (95% CI, 1.9 to 2.9), 1.7 (95% CI, 1.2 to 2.3), 0.9 (95% CI, 0.6 to 1.3), and 1.6 (95% CI, 1.2 to 2.0), respectively. Rates were higher for publicly insured than privately insured or uninsured youth in both years, and in 2000, they were 3.4 (95% CI, 3.1 to 3.8), 1.5 (95% CI, 1.4 to 1.6), and 1.5 (95% CI, 1.1 to 1.9), respectively.
Characteristics of the hospitals from which youth with principal mental health diagnoses were discharged showed no significant changes over the period. Among other discharge characteristics, the proportion of discharges initially referred from other health facilities rose while the proportion discharged to short-term, nursing, and other inpatient facilities fell.
In both years, a great majority of the children and adolescents with principal mental disorder discharge diagnoses were discharged to home, and this proportion increased over the period. The proportion who left against medical advice declined over time.
Clinical characteristics of child and adolescent mental disorder discharges in 1990 and 2000 are presented in Table 3. Principal diagnoses, principal and comorbid substance-use disorders, and comorbid intentional self-injuries differed between the years. Depressive disorders were the most common principal mental disorders in both years and increased over the decade. The proportion of discharges diagnosed with principal bipolar disorders rose dramatically from 2.9% to 15.1%. Increases were also observed in the proportion diagnosed with psychotic disorders. These changes were accompanied by decreases in principal adjustment disorders. While principal substance-use diagnoses accounted for a markedly reduced proportion of discharges over the period, comorbid substance use diagnoses increased. The proportion of discharges in which intentional self-injury was diagnosed rose but remained relatively uncommon. Rates of comorbid general medical conditions were unchanged.
Table 4 presents changes in median and mean LOS for child and adolescent mental disorder discharges between 1990 and 2000 by selected characteristics as well as adjusted estimates of the change in mean LOS associated with year 2000 vs year 1990 discharges. Median LOS for all discharges fell 63% over the decade, from 12.2 days (95% CI, 10.5 to 15.1) to 4.5 days (95% CI, 4.1 to 5.0). The 25th percentile of the median LOS declined significantly from 4.0 to 2.3 days, a reduction of 43%, while the 75th percentile fell significantly from 27.2 to 7.7 days, a decline of 72%. The Figure presents intermediate year changes. Declines of 5.3 days (44%) over the years 1990 to 1995 and of 2.3 days (34%) from 1995 to 2000 were each significant. Median LOS declined for all variable categories except age less than 5 years, principal diagnosis of anxiety or “other psychiatric disorders,” and intentionally self-injured. Declines were most dramatic for depressive, substance, and bipolar disorders, which fell 76%, 76%, and 72%, respectively.
Changes in mean LOS, a measure more influenced by outlier values than median LOS, generally followed patterns similar to those observed in median LOS. Divergent findings included relatively long median LOSs in 2000 for anxiety disorders and other psychiatric disorders, reflecting small overall numbers of discharges influenced by a few large outlying LOSs. Controlling for all other significant covariates, year of discharge was associated with a decline in mean LOS of 12.3 days, an approximate reduction of 68% from the observed mean in 1990 and similar to the observed reduction of 60%.
Between 1990 and 2000, child and adolescent inpatient mental health care in the United States was transformed. Mental health treatment in community hospital settings sustained a reduction in population total patient days of over one half, driven almost exclusively by declines in patient LOS. Inpatient clinicians who on average evaluated, treated, and discharged mentally ill children over the course of 12 days in 1990 routinely accomplished these tasks in 4½ days by 2000. They did so with patients who, by principal diagnosis and rates of intentional self-harm, appeared more severely ill than those 10 years earlier. And despite treating apparently more seriously ill patients in far less time, hospital staff in 2000 did not rely as heavily on transfers to other inpatient facilities or intermediate levels of institutional care. Correspondingly, per-discharge charges declined.
Overall, our findings suggest that during a decade marked by the emergence of managed mental health care9- 12 and increased use of psychotropic medications in the management of young people's illnesses,5,13,24 inpatient mental health professionals appear to be doing more with less. Nonetheless, numerous questions remain, and interpreting our findings in clinical and health policy contexts presents serious challenges. First, the impact of changes in the community hospital inpatient mental health system on readmission to community hospitals and hospitals in other sectors is unclear. The absence of patient identifiers in our data prevents us from distinguishing index admissions from readmissions so that our finding of stable discharges over the study period may reflect rising readmission rates among a declining population young people admitted to community hospitals. Further, findings elsewhere of dramatically increased admissions of youth to private mental hospitals, from approximately 43 000 in 1986 to 149 000 in 1997,25,26 may represent an expansion of overall inpatient use and a shift of care to these settings, potentially through readmission of children and adolescents treated in community hospitals. Private psychiatric hospitals are not included in our sample, complicating our interpretation of this apparent trend and underlining the need for comprehensive data that span specialty and community inpatient treatments.
Rising adult discharge rates from community and general hospitals reported previously1,2 differ from our findings of stable population-adjusted discharge rates for youth. The increase in adult community hospital discharges may reflect a both shift away from state and county mental hospital care, which sustained a 48% decline in admissions between 1986 and 1997 compared with a 17% increase for youth, and a more measured shift toward treatment in private psychiatric hospitals, where admissions rose 90% between 1986 and 1997 for adults compared with 250% for youth.25,26 As for youth, inpatient LOS for adults in community hospitals appears to have declined consistently over the decade. However, reductions for adults were smaller in absolute and relative magnitude than those we observed, and adults began the decade with a shorter average LOS.1
Existing evidence on the impact of changing LOS on readmission is mixed. Although some research in adult patient populations has found that declines in LOS were not associated with altered rates of psychiatric inpatient readmission,27- 32 one study of children and adolescents has demonstrated increased rates of readmission after introduction of managed care practices.33 Work with a mixed population of privately insured children and adults in 1997 and 1998 found that readmission was inversely correlated with index admission LOS. In that study, slight decreases in LOS were associated with significant increases in risk of readmission.34
The observed reduced variation in the range of mean and median LOS by diagnostic group over the period raises questions about quality of inpatient care. Declines in LOS over the period were generally greatest for diagnoses and other characteristics associated with the longest LOS in 1990. Rather than a targeted reduction of LOS for treatment of patients with less severe illnesses and, presumably, less complex clinical needs, this trend in LOS suggests the emergence of a more uniform standard of inpatient treatment duration irrespective of patient need. A similar phenomenon has been observed in earlier work with child7 and adult2 patients. Alternately, the dramatic decrease in the 75th percentile of LOS may suggest a decline in the use of short-term hospital units for long-term or residential mental health treatment of youth, a practice that has received more attention in the popular than academic press.35- 38 The absence of clinical information concerning patient outcomes following hospital discharge precludes establishing conclusions about quality of care.
Trends in the diagnostic case mix toward more serious mental illnesses raise questions about the interaction of managed care practices and inpatient diagnostic evaluation. Under scrutiny of managed care review, inpatient providers may be indicating more serious diagnoses to justify admission or secure greater reimbursement, a process termed diagnostic upcoding. We observed a marked shift of substance use disorders from primary to secondary diagnoses over the period, as well as a dramatic reduction in the proportion of patients admitted with adjustment disorders mirrored by a rise in mood disorder diagnoses. Increases in rates of intentional self-harm among discharged children, while ostensibly less influenced by subjective factors than criteria-driven diagnoses, may reflect changing practice in characterizing and documenting child behaviors to conform with managed care criteria for appropriateness of inpatient care. Nonetheless, increased clinician awareness of serious mental illness in children has likely played a role in changing diagnostic patterns, especially in the remarkable rise in diagnosis of bipolar disorders. The emergence in the 1990s of work describing the presentation, prevalence, and comorbidity of bipolar disorder in child and adolescent clinical samples39- 43 may have contributed to the increase in bipolar diagnoses we observed. The difficulty of diagnosis, especially in young children44; ongoing development of diagnostic tools45; and findings of overdiagnosis in some clinical settings46 suggest diagnosis of pediatric bipolar disorder in community hospitals will continue to evolve. Implementation of more stringent admission criteria associated with managed care may have also influenced inpatient diagnostic case mix independent of upcoding or changing diagnostic practices. Use of these criteria may explain stable rates of behavioral disorders as principal diagnoses among discharged youth, despite the increasing prevalence of diagnoses of attention-deficit/hyperactivity disorder in outpatient settings47 and as comorbidities in young inpatients.48
Our finding of a markedly lower population-based discharge rate for Hispanic youth relative to white, non-Hispanic youth is consistent with results of some previous population studies and clinical samples of young people.49,50 Although our data cannot identify unmet need for mental health services, our results contribute to longstanding concerns that undertreatment of mentally ill youth may be concentrated in minority populations.51 The population-based community hospital discharge rate for publicly insured youth was more than twice rates for privately insured or uninsured children and adolescents, supporting previous findings in outpatient and mixed service settings of increased use by the publicly insured, but not privately insured, when compared with uninsured youth.8,49,52
The current study has several limitations. Information regarding treatment was restricted to LOS and diagnosis, preventing an exhaustive portrayal of care received or assessment of illness severity. We could not identify the type of inpatient unit that provided care, precluding comparisons by treatment setting. Many community hospitals treat children and adolescents with mental disorders outside of psychiatric units, and even where dedicated child and adolescent psychiatric units exist, a significant proportion of children and adolescents may be received as boarders on pediatric medicine or other nonpsychiatric services.53 Further, use of pharmacotherapies, which rose among children and adolescents during this period,5,13,24 cannot be identified in our sample. Outcomes data in our study were limited to a measure of disposition in which referral to outpatient providers was not consistently specified. The potential role of managed behavioral care in facilitating intermediate intensity and outpatient follow-up after inpatient hospitalization, which has been demonstrated in some publicly and privately insured pediatric populations54,55 but not others,56,57 also could not be evaluated using our data.
Inpatient mental health professionals now routinely evaluate, treat, and discharge depressed children and adolescents in 4 days, well before the onset of response to selective serotonin reuptake inhibitor pharmacotherapy58 or the emergence of adverse effects.59 To do so with assurance requires access to extensive clinical resources, including specialized inpatient mental health services and intensive outpatient follow-up. It remains unclear whether the changes we identified represent a more efficient use of inpatient mental health resources, a withdrawal of necessary services, or some combination of both. Further research can clarify the extent to which community hospitals are meeting the ever-mounting clinical demands inherent in constricted periods of diagnosis and treatment by identifying the content and outcomes of treatment for mental illness in children and adolescents.
Correspondence: Brady G. Case, MD, Child Psychiatry Training Office, Rhode Island Hospital–Coro West 2, 593 Eddy St, Providence, RI 02903 (firstname.lastname@example.org).
Submitted for Publication: June 27, 2005; final revision received February 27, 2006; accepted March 30, 2006.
Financial Disclosure: None reported.
Funding/Support: This study was supported in part by grant 1 R2 MH60496 (Dr Case, trainee) from the National Institute of Mental Health and the American Psychiatric Institute for Research and Education Janssen Psychiatric Research Scholar Program (Dr Case).
Acknowledgment: We thank Kathleen J. Pottick, PhD, and Ross B. Andelman, MD, for use of their diagnostic cross-walk and Eugene M. Laska, PhD, for his comments.