Chapter 3. Substance Use and Mental Disorder Discharges from U.S. Community Hospitals in the Early 1990s, Revisited

Sarah Q. Duffy, Ph.D.

Introduction

Managed care and behavioral health care carve-outs proliferated during the early 1990s, and research suggests these arrangements reduce inpatient mental health services and substance abuse treatment (Callahan, Shepard, Beinecke, Larson, & Cavanaugh, 1995; Etheridge, Hubbard, Anderson, Craddock, & Flynn, 1997; Goldman, McCulloch, & Sturm, 1998; Iglehart, 1996; Ma & McGuire, 1998; Mechanic, 1997b). Based on these findings, one might expect to have seen a coinciding decline in admissions to community hospitals of patients with substance use and/or mental disorders (SU/MD). Such short-term, general, non-Federal hospitals have long been involved in SU/MD treatment and have accounted for a large share of inpatient stays for those with SU/MD, including approximately 54 percent of all such stays in 1985 and 69 percent of those of Medicare beneficiaries in 1995 (Cano, Hennessy, Warren, & Lubitz, 1997; Kiesler & Simpkins, 1993; Mechanic, 1997a).

However, much of the research on managed care has relied on methods, such as simple pre- and post-comparisons of aggregate claims from privately insured populations, that may fail to capture the experience of many with SU/MD (Callahan et al., 1995; Goldman et al., 1998; Ma & McGuire, 1998). Other reports suggest that these patients may receive inadequate substitutes for inpatient mental health services, that a treatment gap exists, and that a growing percentage of the U.S. population lacks insurance (Bae, 1997; Dana, Conner, & Allen, 1996; Hirschfeld et al., 1997; Mechanic, Schlesinger, & McAlpine, 1995; Robertson, 1997; Rouse, 1998; U.S. Bureau of the Census, 1997). If the result is inadequate or fragmented community-based specialty treatment for those with SU/MD, they may be more likely to be admitted to local community hospitals for stabilization and detoxification (Olfson, 1993; Olfson & Walkup, 1997; Walkup, 1997; Wolfe & Sorensen, 1989). We examine these concerns by analyzing trends in discharges of those with SU/MD from community hospitals nationwide during the first half of the 1990s.

Two studies, one by Maynard and Cox (1998) and the other by Mechanic, McAlpine, and Olfson (1998), examined trends in community hospitalizations of those with SU/MD in the early 1990s using the National Center for Health Statistics (NCHS) National Hospital Discharge Survey (NHDS). However, their reports provided vastly different trend estimates. According to Maynard and Cox (1998), SU/MD discharges increased only 0.5 percent between 1990 and 1994. Mechanic et al. (1998), on the other hand, reported that SU/MD discharges increased by 35 percent between 1988 and 1994. Furthermore, Maynard and Cox (1998) reported that there was no change in the number of discharges with a co-occurring disorder—one substance use and one mental disorder—during the time period (a trend that Mechanic et al. did not examine).

In this study, we reexamine trends during this time period both by explaining how these different estimates could have been generated by the NHDS data and by providing new estimates using a dataset more appropriate for examining community hospitalizations of those with SU/MD. The findings presented here will contribute to our understanding of the impact of the changes in the health care system in the early 1990s on those with SU/MD diagnoses.

Data

In this study, we use data covering 1990–1995 from the Agency for Healthcare Research and Quality's Nationwide Inpatient Sample (NIS), which contains discharge abstract records that hospitals report to State data organizations (Agency for Health Care Policy and Research, 1995). With more than 6 million records per year, it approximates a 20 percent sample of U.S. community hospitals and includes information necessary to compute national estimates and standard errors using methods for complex database designs, such as those available in SUDAAN software (Shah, Barnwell, & Bieler, 1996).

For a number of reasons, NIS data are more appropriate for studying community hospitalizations of individuals with SU/MD during the early 1990s than are the NHDS data. First, unlike the NIS, the NHDS is a sample of all short-term hospitals, including short-term psychiatric hospitals. According to one estimate, 13 percent of the discharges with mental disorder diagnoses in the NHDS were from psychiatric hospitals. Failure to account for them in the NHDS data caused at least one team of researchers to vastly overestimate the number of individuals with mental disorders receiving care in swing beds in general hospitals (Kiesler & Simpkins, 1993). The NIS also is a much larger sample than the NHDS and allows analysis of patients by more refined diagnosis categories, which is useful because those with SU/MD diagnoses are a diverse group. Finally, the NIS has been consistently coded across years, which makes it easier to use. Inconsistencies in coding across years in the NHDS were likely responsible for the results reported by Maynard and Cox (1998). These inconsistencies will be described in the next section.

Methods

Study Sample

We study discharges coded with the Clinical Classifications for Health Policy Research (CCHPR) principal diagnosis (DCCHPR1) categories, a variable available on the NIS. CCHPRs reclassify codes from the International Classification of Diseases, 9^th Revision, Clinical Modification (ICD-9–CM) into broader reporting categories (Elixhauser, 1996). Clients with a DCCHPR1 listed in Table 3.1 were included in the analysis. This definition of SU/MD is consistent with much of the existing research on such hospitalizations (Cano et al., 1997; Kiesler & Simpkins, 1993; Maynard & Cox, 1998; Mechanic et al., 1998).¹ The sample sizes varied between 250,000 and 340,000 per year.

**Table 3.1 Distribution of Discharges from U.S. Community Hospitals, by Principal Diagnosis CCHPR, 1990 and 1995, Substance Use and Mental Disorder Sample**
Principal Diagnosis CCHPR (DCCHPR1)¹	1990		1995		% Change
Principal Diagnosis CCHPR (DCCHPR1)¹	National Estimate	% of SU/MD Sample²	National Estimate	% of SU/MD Sample²	% Change
65 Mental Retardation	887	0.06	593	0.03	-33.14
66 Alcohol-Related Mental Disorders (acute alcohol intoxication; other and unspecified alcohol dependence; nondependent alcohol abuse; other alcohol-related mental disorders)	302,821	21.80	280,651	16.17	-7.32
67 Substance-Related Mental Disorders (opioid dependence; cocaine dependence; other, combined, and unspecified drug dependence; cocaine abuse; other, mixed, or unspecified drug abuse; other substance-related mental disorders)	111,517	8.03	240,792	13.88	115.92
68 Senility and Organic Mental Disorders (senile dementia, uncomplicated; arteriosclerotic dementia; transient organic psychotic conditions; specific nonpsychotic mental disorders due to organic brain damage; presenile dementia, uncomplicated; senile dementia with delirium; other senility and organic mental disorders)	94,290	6.79	129,795	7.48	37.66
69 Affective Disorders (major depressive disorder, single episode; major depressive disorder, recurrent episode; neurotic depression; bipolar affective disorder; manic-depressive psychosis; other affective disorders)	409,126	29.45	557,445	32.13	36.25
70 Schizophrenia and Related Disorders (paranoid schizophrenia; schizo-affective type; other schizophrenia)	186,913	13.45	238,188	13.73	27.43
71 Other Psychoses	53,299	3.84	56,457	3.25	5.93
72 Anxiety, Somatoform, Dissociative, and Personality Disorders (anxiety states; personality disorders; other anxiety, somatoform, dissociative, and personality disorders)	64,294	4.63	67,627	3.90	5.18
73 Preadult Disorders	17,606	1.27	18,086	1.04	2.73
74 Other Mental Conditions (adjustment reaction; depressive disorder, not elsewhere classified)	147,965	10.65	145,145	8.36	-1.91
75 Personal History of Mental Disorder, etc.	463	0.03	407	0.02	-12.10
CCHPR = Clinical Classifications for Health Policy Research; DCCHPR1 = CCHPR principal diagnosis; SU/MD = substance use/mental disorders. ¹ See Elixhauser (1996) or http://www.ahrq.gov for more information. ² Difference in distribution of DCCHPRs over time significant at better than 1 percent level. Source: SAMHSA, Office of Applied Studies' analysis of Agency for Healthcare Research and Quality's National Inpatient Sample, 1990–1995.

We present trends for all discharges with SU/MD and for five, mutually exclusive subgroups based on principal and all secondary diagnoses. To create these subgroups, we grouped all secondary diagnoses with the CCHPR software program, which is freely available for downloading at http://www.ahrq.gov/data/hcup/ccs.htm (Elixhauser, 1996). These subgroups, which appear in Table 3.2, reflect differences in complexity and in the ability and willingness of community-based providers to treat patients (Etheridge et al., 1997; Mechanic, 1997b). For example, those with both a substance use and another mental disorder diagnosis are a distinct subgroup here because, during the early 1990s, the substance abuse treatment and mental health services systems often were separate. Changes in the health care system during that time may have made it especially difficult for patients with both types of disorders to navigate two separate systems.

**Table 3.2 Diagnosis Subgroup Definitions**
Diagnosis Subgroup	Includes records with...
Substance Use Only	Only substance use diagnoses (DCCHPR codes 66 or 67).
Substance Use and Mental Disorder	At least two diagnoses—at least one mental disorder (DCCHPR 65, 68–75) and at least one substance use, either one of which may be principal.
Substance Use and Medical	At least two diagnoses—a substance use principal and one non-SU/MD (DCCHPR not in 65–75).
Mental Disorder Only	Only mental disorder diagnoses.
Mental Disorder and Medical	At least two diagnoses—a mental disorder principal and at least one non-SU/MD.
DCCHPR = Clinical Classifications for Health Policy Research diagnosis code; SU/MD = substance use/mental disorders.

These methods are similar to those used by Maynard and Cox (1998), except that those authors appear to have missed a change in the way diagnoses were coded in the 1994 NHDS that required a modification to the CCHPRS formatting program.² Failure to account for this change likely resulted in their undercounting both the number of discharges with SU/MD and the number of those with a co-occurring disorder in the Nation's short-term hospitals reported by the NHDS in that year. This can be verified by referencing several NCHS publications (e.g., Gillum, Graves, & Kozak, 1996; Graves & Gillum, 1997) that report counts of discharges by disease category. According to these publications, the number of discharges with SU/MD diagnoses increased from approximately 1,538,000 to 2,112,000, or 37 percent, between 1990 and 1994.

Statistical Methods

We present weighted means, percentages, and age- and gender-adjusted discharge rates per 10,000 population. We discuss in the text differences that are significant at or better than the 5 percent level. For comparisons among groups of diagnoses in the same year, we computed t tests for continuous variables and chi-square (²) tests for categorical variables using SUDAAN (Shah et al., 1996), along the lines of the example provided with the NIS documentation (Duffy & Sommers, 1999). To examine trends in discharge rates, we computed the Estimated Annual Percentage Change (EAPC) (Ries et al., 1997). The EAPC is 100(e^m-1), where m is the coefficient on a regression of the natural logarithm of the standardized discharge rates on calendar year. A negative EAPC indicates that the standardized rate has declined, while a positive EAPC indicates that it has increased. We used the standard error (SE) from the regression to compute t statistics. To determine whether differences over time were significant for other variables, we computed test statistics based on the differences in value between 1990 and 1995 using a method that accounted for hospitals that appear in the sample both years. Although we focus our discussion on differences between 1990 and 1995, statistics computed using data from all 6 years confirm the trends we report.

Results

Discharges of those with SU/MD grew substantially between 1990 and 1995, and Table 3.3 shows that this growth, from 1.39 million to 1.74 million (t test, p <0.0001, df = 1,302), contrasts with the stability of total discharges.³ Figure 3.1, which displays age- and gender-adjusted discharge rates, reveals that those with both a substance use and mental disorder diagnosis accounted for most of the increase. Discharges of individuals with both diagnoses increased from 9.4 to 17.22 per 10,000 population (EAPC t value = 14.774, p = 0.0001, df = 5). Discharges with both a mental disorder and medical diagnosis increased as well, but at a lower rate, from 19.3 to 22.5 per 10,000 population (EAPC t value = 4.222, p = 0.0135, df = 5). A small decline in the rate of discharges with mental disorders alone, from 14.0 to 11.8 per 10,000 population (EAPC t value = -6.288, p = 0.0033, df = 5), only partially offsets these increases.

As the number and discharge rate of those with SU/MD grew, their average length of stay (ALOS) declined by 25 percent (t = 7.17, p < 0.001, df = 1,302) compared with a 13 percent decline for all discharges (t = 9.984, p < 0.0001, df =1,337). Although ALOS declined for all subgroups, and the decline was most pronounced for the three substance-related subgroups, the ALOS ranking remained the same over time. Discharges with mental disorder and medical diagnoses had the longest ALOS throughout the period, which declined 20 percent (t = 6.28, p < 0.001, df = 1,277) while those with substance use diagnoses alone had the shortest, which declined 36 percent (t = 4.47, p < 0.0001, df = 927).

The age distribution of SU/MD discharges also changed substantially during this time (² = 242.23, p < 0.0001, df = 6). Table 3.3 displays information on those aged 35 to 45 years, who increased the most among the groups analyzed (< 12, 12 to 17, 18 to 25, 26 to 34, 35 to 45, 46 to 64, 65 or older). They comprised the largest share of discharges with SU/MD diagnoses in 1995 at 36 percent, replacing the 26 to 34 year olds, who had the largest share of discharges in 1990. The growth of those aged 35 to 45 years occurred among all SU/MD subgroups, but was most noticeable within the substance-related subgroups.

Although the age distribution changed between 1990 and 1995, Table 3.3 also shows that the gender distribution did not. However, there were differences in these distributions between those with SU/MD and all discharges (² = 183.79, p < 0.0001, df = 1), as well as across diagnosis groups (² = 866.18, p < 0.0001, df = 4). Slightly over 51 percent of those with SU/MD diagnoses were male compared with fewer than 42 percent of all discharges. Among the subgroups, most likely to be male were those with substance use diagnoses, varying from 59.1 percent for those with co-occurring mental and substance use disorders to 73.7 percent for those with only substance use diagnoses.

**Table 3.3 Substance Use and Mental Disorder Discharges Compared with All Discharges from U.S. Community Hospitals, 1990 and 1995**
Sample	Year	Count in 1,000s	Rate¹ per 10,000 Population	Length of Stay (days)^a	% ^a,b Male	% ^a,b,c Age 35–45	Discharge Status^b,c		Expected Primary Pay Source^a,b
Sample	Year	Count in 1,000s	Rate¹ per 10,000 Population	Length of Stay (days)^a	% ^a,b Male	% ^a,b,c Age 35–45	% Died	% AMA	% Medicare	% Medicaid	% Private	% UCC	% Other
All Discharges	1990	35,215	1,420^d	6.1^e	42.2	9.6	2.8	0.8
All Discharges	1995	34,802	1,328^d	5.3^e	41.5	10.7	2.6	0.9	36.5	18.0	37.0	5.3	3.4
SU/MD	1990	1,389^e	56.1^d	12.7^e	50.0	22.1	0.3	6.3
SU/MD	1995	1,735^e	66.3^d	9.5^e	51.3	27.5	0.2	6.3	29.9	30.0	25.8	10.3	4.5
Substance Use Only	1990	144.3	5.8	8.6^e	73.4	26.7	0.01	15.2
Substance Use Only	1995	163.4	6.7	5.5^e	73.7	36.2	0.01	19.3	6.7	38.5	27.6	22.0	5.2
Substance Use and Mental Disorder	1990	233.6^e	9.4^d	11.5^e	58.1	25.8	0.03	8.6
Substance Use and Mental Disorder	1995	446.2^e	17.2^d	8.3^e	59.1	34.1	0.03	6.6	21.5	34.4	26.5	12.6	4.9
Substance Use and Medical	1990	185.2	7.5	10.0^e	71.6	28.2	0.32	10.2
Substance Use and Medical	1995	223.8	8.5	6.1^e	70.6	36.1	0.21	12.1	18.0	36.7	24.0	17.5	3.8
Mental Disorder Only	1990	346.8	14.0^d	13.1^e	43.1	22.4	0.05	4.52
Mental Disorder Only	1995	306.5	11.8^d	10.6^e	43.7	24.6	0.02	3.62	23.0	31.4	32.2	7.7	5.7
Mental Disorder and Medical	1990	479.2^e	19.3^d	15.2^e	35.7	16.3	0.63	2.37
Mental Disorder and Medical	1995	595.3^e	22.5^d	12.1^e	35.9	18.4	0.41	1.78	49.2	21.0	22.3	4.1	3.5
AMA = against medical advice; SU/MD = all substance use and mental disorder discharges; UCC = uncompensated care. Note: The following symbols represent significant differences at or better than the 5 percent level: ^a Distributions across subgroups in 1995 (²). ^b Distribution between all discharges and SU/MD discharges in 1995. ^c Distribution over time (except for discharge status for substance use disorder only). ^d Rates over time. ^e Within groups across years. ¹ Age- and gender-adjusted discharge rate. Source: SAMHSA, Office of Applied Studies' analysis of Agency for Healthcare Research and Quality's National Inpatient Sample, 1990–1995.

Figure 3.1 Age- and Gender-Adjusted Discharge Rates, by Subgroup, 1990 to 1995

Figure 3.1 D

Source: SAMHSA, Office of Applied Studies' analysis of Agency for Healthcare Research and Quality's National Inpatient Sample, 1990–1995.

We examined two rough indicators of outcomes based on patient disposition at discharge: the in-hospital mortality rate and the percentage who leave against medical advice (AMA). The distribution of SU/MD patients at discharge was significantly different from that of all discharges (1995: ² = 827.46, p < 0.0001, df = 3), and across SU/MD diagnosis groups (² = 818.52, p < 0.0001, df = 12). Table 3.3 reveals that the in-hospital mortality rate for SU/MD discharges was substantially lower than that for all discharges. It varied among subgroups, from a low of 0.01 percent for those with substance use diagnoses alone in 1995 to a high of 0.41 percent for those with both mental disorder and medical diagnoses. However, discharges with SU/MD diagnoses, especially those with substance-related disorders, were much more likely to leave AMA than were other discharges, varying from 1.78 percent of the mental disorder and medical subgroup to 19.3 percent of the substance use only subgroup.

"Expected primary payer" is defined as the payer who is expected, at the time of the admission, to pay the hospital bill. It would be most informative to analyze this variable over time and examine separately those insured under managed care arrangements. Unfortunately, neither of these is possible due to data limitations, so the following categories are examined for 1995 only: Medicare, Medicaid, private (Blue Cross, PPO, commercial, HMO, prepaid health plan), uncompensated care (UCC: self-pay, no charge), and other coverage (Title V, workers' compensation, CHAMPUS/CHAMPVA, other government).⁴ Table 3.3 reveals that those with SU/MD diagnoses were more likely than all discharges to receive uncompensated care or have Medicaid coverage and were less likely to be covered by Medicare or private insurance (² = 202.87, p < 0.0001, df = 4).

Substantial differences existed among SU/MD subgroups (² = 1,010.01, p < 0.0001, df = 16). Almost 50 percent of discharges with both a mental disorder and medical diagnosis had Medicare coverage, while those diagnosed with mental disorders alone most frequently had private coverage, and those with substance-related disorders alone most frequently had Medicaid coverage. Only 4.1 percent of those with both mental disorder and medical diagnoses had no coverage compared with 22 percent of those in the substance use only subgroup.

Referring to Table 3.1, one can see that the vast majority of the SU/MD discharges had principal mental disorder diagnoses, with smaller but almost equal percentages with alcohol-related and substance-related diagnoses in 1995. However, there were changes in the distribution over time (² = 86.65, p < 0.0001, df = 10). Although in 1995 the top two DCCHPRs remained Affective Disorders and Alcohol-Related Mental Disorders, the third most prevalent in 1995, Substance-Related Mental Disorders, had been fifth in 1990.

Discussion

Community hospitals remained important in caring for individuals with SU/MD diagnoses in 1995, and such patients were a growing part of community hospitals' inpatient business. While total discharges remained stable during the first half of the 1990s, we found, as did Mechanic et al. (1998), that discharges of those with SU/MD diagnoses increased substantially. Affective disorders, among mental disorders, and alcohol-related disorders, among substance use disorders, remained among the most frequent diagnoses, although abuse of other substances increased. This increase appears to have become permanent, as, according to more recent NIS data, the number of discharges from community hospitals of those with SU/MD has continued to be above 1.7 million through the year 2001 (the most recent year for which data are available), when they topped 1.9 million.

Contrary to previously published reports (Maynard & Cox, 1998), the percentage of discharges with at least one substance use and one mental disorder increased substantially during the 1990s. Although this growth may reflect more accurate diagnosis and coding, the negligible offsetting reduction in the single diagnosis categories argues against that explanation. At the same time, although a smaller percentage of patients with SU/MD diagnoses died in the hospital compared with all patients, a much larger percentage left the hospital against medical advice (AMA).

As with all discharges in 1995, Medicare and Medicaid paid for more than 50 percent of discharges for those with SU/MD diagnoses. Unfortunately, due to data limitations, we cannot compare this figure with earlier years of the NIS data. As a point of comparison, we can turn to estimates based on 1985 NHDS data, which, although they suffer from the shortcomings described earlier, were the only nationwide hospital discharge abstract data publicly available before 1988. According to these 1985 data, commercial insurance (then consisting mostly of fee-for-service plans) paid for 44 percent of inpatients with SU/MD, followed by Medicare at 20 percent and Medicaid at 16 percent (Kiesler & Simpkins, 1993). This comparison suggests that the Federal Government's role in paying for these patients may have increased substantially since the mid-1980s.

Sharp declines in length of stay suggest that hospitals may provide short-term lifesaving services, such as detoxification and stabilization, but not treatment for their chronic underlying disorders (Jayaram, Tien, Sullivan, & Gwon, 1996; National Institute of Mental Health, 1998, 2003). Discharged patients subsequently may receive outpatient treatment, which may be entirely appropriate (Kiesler & Simpkins, 1993). However, the extent to which this is occurring is unclear given some evidence of hospitals' infrequent follow-up of patients referred to outpatient aftercare and the reluctance of many outpatient mental health providers to treat those with co-occurring mental and substance use disorders or those with medical complications, who showed the greatest increases in hospitalizations reported here (Etheridge et al., 1997; Mechanic, 1997b; Olfson, 1993; Olfson & Walkup, 1997; Walkup, 1997). Shorter stays may mean patients are being discharged or leaving AMA in sicker condition and may need to be rehospitalized (Olfson & Walkup, 1997). One limitation of the NIS (as well as the NHDS) is that it does not allow linkages across individuals, so we cannot determine whether patients are being rehospitalized.

Although these results cannot prove causality because they are based on a series of cross-sectional observations rather than following specific individuals through time, they do not diminish concern that changes during the early 1990s adversely affected those with SU/MD and may continue to affect them today. They suggest that further study into the causes of the increases in community hospital discharges of those with SU/MD and a more thorough study of effects on outcomes are warranted. Clearly, in 1995, U.S. community hospitals remained important in caring for those with SU/MD.

References

Agency for Health Care Policy and Research. (1995). Technical supplement 5: Design of the Nationwide Inpatient Sample, Release 1. Documentation for Nationwide Inpatient Sample, Release 1 (NIS documentation for 1990–1995). Rockville MD: Author. [Documentation for 1997–2001 available at http://www.ahcpr.gov/data/hcup/]

Bae, J. P. (1997). Assessing the need, use, and developments in mental health/substance abuse care. Health Care Financing Review, 18(3), 1–4.

Callahan, J. J., Shepard, D. S., Beinecke, R. H., Larson, M. J., & Cavanaugh, D. (1995). Mental health/substance abuse treatment in managed care: The Massachusetts Medicaid experience. Health Affairs (Millwood), 14, 173–184.

Cano, C., Hennessy, K. D., Warren, J. L., & Lubitz, J. (1997). Medicare Part A utilization and expenditures for psychiatric services: 1995. Health Care Financing Review, 18(3), 177–193.

Dana, R. H., Conner, M. G., & Allen, J. (1996). Quality of care and cost-containment in managed mental health: Policy, education, research, advocacy. Psychological Reports, 79(3 Pt 2), 1395–1422.

Duffy, S. Q., & Sommers, J. P. (1999). Technical supplement 11: Calculating variances using data from the HCUP Nationwide Inpatient Sample (technical supplements for the HCUP Nationwide Inpatient Sample, release 6). Rockville, MD: Agency for Health Care Policy and Research.

Elixhauser, A. (1996). Clinical Classifications for Health Policy Research, Version 2: Software and user's guide (AHCPR Publication No. 96–0046, Healthcare Cost and Utilization Project [HCUP-3], Research Note 1). Rockville, MD: Agency for Health Care Policy and Research. [Updated version, Clinical Classifications Software (ICD-9–CM) available at http://www.hcup-us.ahrq.gov/toolssoftware/ccs/ccs.jsp]

Etheridge, R. M., Hubbard, R. L., Anderson, J., Craddock, S. G., & Flynn, P. M. (1997). Treatment structure and program services in the Drug Abuse Treatment Outcome Study (DATOS). Psychology of Addictive Behaviors, 11, 244–260.

Gillum, B. S., Graves, E. J., & Kozak, L. J. (1996). Trends in hospital utilization: United States, 1988–92. Vital and Health Statistics 13(124), 1–71 [Available as a PDF at http://www.cdc.gov/nchs/data/series/sr_13/sr13_124.pdf]

Goldman, W., McCulloch, J., & Sturm, R. (1998). Costs and use of mental health services before and after managed care. Health Affairs (Millwood), 17, 40–52. Comments in 17, 241–244.

Graves, E. J., & Gillum, B. S. (1997). National hospital discharge survey: Annual summary, 1994. Vital and Health Statistics 13(128), i-v; 1–50. [Available as a PDF at http://www.cdc.gov/nchs/data/series/sr_13/sr13_128.pdf]

Hirschfeld, R. M., Keller, M. B., Panico, S., Arons, B. S., Barlow, D., Davidoff, F., Endicott, J., Froom, J., Goldstein, M., Gorman, J. M., Marek, R. G., Maurer, T. A., Meyer, R., Phillips, K., Ross, J., Schwenk, T. L., Sharfstein, S. S., Thase, M. E., & Wyatt, R. J. (1997). The National Depressive and Manic-Depressive Association consensus statement on the undertreatment of depression. Journal of the American Medical Association, 277, 333–340. Comments in 277, 1433.

Iglehart, J. K. (1996). Managed care and mental health. New England Journal of Medicine, 334, 131–135. Comment in 335, 56; discussion 57–58.

Jayaram, G., Tien, A. Y., Sullivan, P., & Gwon, H. (1996). Elements of a successful short-stay inpatient psychiatric service. Psychiatric Services, 47, 407–412.

Kiesler, C. A., & Simpkins, C. G. (1993). The unnoticed majority in psychiatric inpatient care (Plenum Series on Stress and Coping). New York: Kluwer Academic/Plenum Publishers.

Ma, C. A., & McGuire, T. G. (1998). Costs and incentives in a behavioral health carve-out. Health Affairs (Millwood), 17, 53–69.

Maynard, C., & Cox, G. B. (1998). Psychiatric hospitalization of persons with dual diagnoses: Estimates from two national surveys. Psychiatric Services, 49, 1615–1617.

Mechanic, D. (1997a). The challenges of managed care. New Directions for Mental Health Services (73), 5–10.

Mechanic, D. (1997b). The future of inpatient psychiatry in general hospitals. New Directions for Mental Health Services (73), 103–108.

Mechanic, D., McAlpine, D. D., & Olfson, M. (1998). Changing patterns of psychiatric inpatient care in the United States, 1988–1994. Archives of General Psychiatry, 55, 785–791.

Mechanic, D., Schlesinger, M., & McAlpine, D. D. (1995). Management of mental health and substance abuse services: State of the art and early results. Milbank Quarterly, 73(1), 19–55.

National Institute of Mental Health. (1998, November 2). Clinical Antipsychotic Trials of Intervention Effectiveness (RFP No. NIMH-99–DS-0001). Rockville, MD: Author. [Project description available at http://www.catie.unc.edu/]

National Institute of Mental Health. (2003). Home page for Clinical Antipsychotic Trials of Intervention Effectiveness (CATIE): A research program studying treatment effectiveness and outcomes in schizophrenia and Alzheimer's disease. Retrieved October 22, 2003, from http://www.catie.unc.edu/

Olfson, M. (1993). The array of psychiatric services in general hospitals. General Hospital Psychiatry, 15, 277–283.

Olfson, M., & Walkup, J. (1997). Discharge planning in psychiatric units in general hospitals. New Directions for Mental Health Services (73), 75–85.

Ries, L. A. G., Kosary, C. L., Hankey, B. F., Miller, B. A., Harras, A., & Edwards, B. K. (Eds.). (1997). SEER cancer statistics review, 1973–1994 (NIH Publication No. 97–2789). Bethesda, MD: National Cancer Institute.

Robertson, E. B. (1997). Introduction to mental health service delivery in rural areas. In E. B. Robertson, Z. Sloboda, G. M. Boyd, & et al. (Eds.), Rural substance abuse: State of knowledge and issues (pp. 413–417, NIH Publication No. 97–4177, NIDA Research Monograph 168). Rockville MD: National Institute on Drug Abuse.

Rouse, B. A. (1998). Substance abuse and mental health statistics source book 1998 (DHHS Publication No. SMA 98–3170, Analytic Series A-4). Rockville, MD: Substance Abuse and Mental Health Services Administration, Office of Applied Studies. [Available at http://www.oas.samhsa.gov/sb99a.zip].

Shah, B. V., Barnwell, B. G., & Bieler, G. S. (1996). SUDAAN: Software for the statistical analysis of correlated data. User's manual release 7.0. Research Triangle Park, NC: Research Triangle Institute.

U.S. Bureau of the Census. (1997). Statistical abstract of the United States: 1997 (117^th ed.). Washington, DC: U.S. Government Printing Office, Superintendent of Documents. [Available as a PDF at http://www.census.gov/prod/www/statistical-abstract-us.html]

Walkup, J. (1997). Family involvement in general hospital inpatient care. New Directions for Mental Health Services (73), 51–64.

Wolfe, H. L., & Sorensen, J. L. (1989). Dual diagnosis patients in the urban psychiatric emergency room. Journal of Psychoactive Drugs, 21, 169–175.

Woodward, A., Epstein, J., Gfroerer, J., Melnick, D., Thoreson, R., & Willson, D. (1997). The drug abuse treatment gap: Recent estimates. Health Care Financing Review, 18(3), 5–17.