Community Mental Health Journal

, Volume 48, Issue 5, pp 598–603

Effectiveness of a High End Users Program for Persons with Psychiatric Disorders


    • Center for Mental Health Policy and Service Research
  • Sumedha Chhatre
    • Center for Mental Health Policy and Service Research
  • Cynthia Zubritsky
    • Center for Mental Health Policy and Service Research
  • Karen Fortuna
    • Center for Mental Health Policy and Service Research
  • Steven Dettwyler
    • Mental Health ProgramsDSAMH/DHSS
  • Renata J. Henry
    • Department of Health and Mental Hygiene
    • Department of Behavioral Health and Disabilities
  • Melissa Smith
    • Division of Substance Abuse and Mental HealthDSAMH/DHSS
Brief Report

DOI: 10.1007/s10597-012-9479-0

Cite this article as:
Rothbard, A.B., Chhatre, S., Zubritsky, C. et al. Community Ment Health J (2012) 48: 598. doi:10.1007/s10597-012-9479-0


To evaluate the effectiveness of an intensive system of case management for high end users of inpatient care in reducing psychiatric inpatient utilization. A pre-post study design with a contemporaneous comparison group was employed to determine the effects of a State designed intervention to reduce inpatient care for adults with a mental health disorder who had high utilization of inpatient psychiatric care between 2004 and 2007. Logit and negative binomial regression models were used to determine the likelihood, frequency and total days of inpatient utilization in the post period as a function of the intervention. Data from administrative reporting forms and Medicaid claims were used to construct inpatient utilization histories and characteristics of 176 patients. Patients in both groups had a significant reduction in mean inpatient days. However, being in the intervention program did not result in lower odds of being re-hospitalized or in fewer episodes during the study period.


Psychiatric hospitalizationHigh end usersInpatient psychiatric recidivism


High end users generally comprise a small percentage of patients in treatment that use a disproportionate amount of psychiatric inpatient services in a defined period (Casper and Pastva 1990; Hansson 2000; Harrison-Read et al. 2002; Roick et al. 2004). Junghan and Brenner (Junghan and Brenner 2006) report that 20% of individuals with severe mental illnesses are high utilizers at any one time and use 60–80% of total inpatient resources. Other studies found that 33% of inpatient users account for 75% of total mental health costs annually (Hadley et al. 1990; Holohean et al. 1991; Hadley et al. 1992). Research findings from many studies have found no clear distinctions between high and average users of inpatient care with respect to clinical and demographic characteristics (Harrison-Read et al. 2002; Lucas et al. 2001; Carperter et al. 1985; Green 1988; Doering et al. 1998; Mortensen and Eaton 1994; Vogel and Huguelet 1997; Frederick et al. 2002). The role of prior psychiatric hospitalization and length of stay has also been uneven as a predictor of high inpatient utilization over time (Roick et al. 2004; Green 1988; Vogel and Huguelet 1997; Havassy and Hopkin 1989; Monnelly 1997; Moran et al. 2000; Hendryx et al. 2003; Appleby et al. 1993; Bezold et al. 1996; Hudson 2001; Herren et al. 2002). Several studies have shown that only a small proportion of high inpatient users in 1 year remain “high users” in the next year (Roick et al. 2004; Hadley et al. 1992; Casper and Donaldson 1991).

Despite the lack of clear evidence that a history of high utilization results in continued high use, intervention strategies have been put into practice as a means of reducing hospital care for what is considered a high risk population. Two pre-post studies of high users showed lower inpatient use following an intervention. The first, an assertive outreach intervention program for high users in Sydney, Australia, found a 29–67% reduction in inpatient length of stay depending on geographic location of patients (Teenson and Hambridge 1992). The second study reported decreased frequency and duration of hospitalizations and crisis services in a 15 month study of 39 “high users” who received ongoing treatment and aggressive outreach by a continuous care team at a community mental health center (Arana et al. 1991). However, neither study had a comparison group. The outcomes of two randomized control trials (RCTs) of high end users found no statistically significant differences between inpatient use in the intervention versus the control arm (Harrison-Read et al. 2002; Quinlivan et al. 1995).

In 2004, the Delaware Division of Substance Abuse and Mental Health’s (DSAMH) identified approximately ~7% of their public sector mental health population as “at risk”. This group had either multiple hospitalizations in a 12 month period or long stays of 30 days or more during a single episode. A high end user (HEU) program was implemented by the Delaware Psychiatric Center (DPC), the state’s single long term state hospital care facility, to reduce re-hospitalization employing an intensive system of care management for individuals who met the criteria described above.

The purpose of this current study was to evaluate whether the intensive case management program was an effective means of reducing psychiatric inpatient utilization, when compared to usual care. A study was designed at the request of DSAMH using information routinely collected by the State on all inpatient stays for public sector patients paid for by the State. Those individuals enrolled in the HEU program and those patients whose psychiatric inpatient histories matched the HEU criteria were selected for comparison. A pre- post study design was employed to evaluate the changes in inpatient days following participation in the program. Because this was not a randomized trial, statistical tools were employed to control for any differences between the groups. The study was approved by the University of Pennsylvania’s Institutional Review Board.


Description of Treatment System

Delaware has a single 200 bed State Psychiatric Hospital (DPC) that serves the entire state. Most of these individuals are initially treated for “short term” crisis stabilization of 7 days or less in one of three community psychiatric hospitals known as Institutes for Mental Diseases (IMD). The state contracts with the IMDs to provide acute care services for those individuals involuntarily committed for care. When more care is required, they are sent to the state hospital facility, DPC which has average lengths of stay of over a year. The State Department of Substance Abuse and Mental Health (DSAMH) is financially responsible for both those who are involuntarily committed to the IMDs and to the State Hospital. In 2005, 793 individuals had an inpatient psychiatric stay financed by DSAMH.

The HEU program was established at DPC in 2004 for patients with very high or frequent inpatient utilization at the state hospital (DPC) or multiple episodes at the IMDs. The HEU program is administered by the eligibility and enrollment unit (EEU) at DPC which routinely reviews involuntary psychiatric admissions for all state funded clients who use DPC or the IMDs.

Individuals who meet one of the following three criteria are considered to be high users and eligible for a targeted intervention: (1) four or more inpatient episodes in 1 year; (2) 30 days or more of inpatient care in 1 year; or (3) three separate inpatient admissions in 90 days. The HEU program offers enhanced support in transitioning individuals from inpatient to community outpatient programs and monitoring their care once they were in the community.

Based on input from the clinical director and staff at DPC and community providers, individuals who meet preliminary eligibility criteria and whose clinical picture indicates the need for more careful monitoring than can be provided by the community treatment system are referred to the HEU program. Clinicians interviewed by study investigators stated that they sometimes suggest individuals for HEU program admission that they believe are at “high risk,” or who are in need of enhanced support following hospital discharge. However, the staff was unable to identify specific characteristics or factors that were used, despite extensive interviewing.

A HEU Case Coordinator from DPC works with staff from the three IMDs and four continuous community care programs (CCCPs), which provide comprehensive intensive outpatient services, to transition individuals into the community. They also assist community providers in enrolling patients in residential programs and ensure follow-up for outpatient services, post discharge.


The intervention group comprises adult clients with a primary mental health diagnosis admitted to the Delaware HEU program between April, 2004 and June 2006. A contemporaneous group of individuals who had an inpatient stay and met the HEU admission criteria, but were not in the program during the same period, were selected for comparison. These individuals were receiving treatment as usual in the community following discharge from a hospital stay(s). The inpatient selection criteria were identical for both groups.

Data Sources

All patients hospitalized in DPC or an IMD and paid for by DSAMH have a claims or event record found in the Mental Health Consumer Reporting File (CRF) which contains administrative records from all state owned or funded providers. Both the HEU intervention and the comparison group were selected from these files based on service histories constructed by investigators. Each service record included socio-demographic information, clinical history at baseline and dates of hospitalization. Medical Assistance claims for inpatient stays during the pre-post period were obtained from the State to ensure comprehensive utilization information for those on Medical Assistance during the pre-period. Medicare and private insurance data were not available.


A descriptive analysis was done to compare sociodemographic, clinical and inpatient service histories between the groups using Chi Square and t-statistics. Each study participant had his/her own pre-post time line based on an ‘index-date’. For the HEU intervention group, the index-date was the date of admission to the HEU program, which typically followed discharge from a hospital stay. Since individuals in the comparison group did not have a HEU program admission date, they were selected based on: (1) the discharge date associated with their first hospitalization in 2005; and (2) inpatient use meeting HEU criteria in the 12 month period prior to the index date. The timeline for each individual consisted of a 12 month period prior to the index date period which we called the ‘pre-period’ and a 12 month ‘post’ period beyond the index date. Group means for length of stay was constructed using the total number of inpatient days for each of the periods divided by the number of individuals in the group.

Three regression models were used to examine the effect of the intervention (yes = 1 for HEU group membership). A simple logit model was employed to examine the likelihood of having an inpatient admission (0,1) in the post period and two negative binomial models were used to explain the extent to which the HEU program affected the frequency of episodes and inpatient days in the post period. Negative binomial regression models were chosen because the data had a large number of zeros and a large variance in days (Mullahy 1986).

Since the individuals were not randomized, we controlled for any known differences between groups using a set of covariates obtained from the administrative records. These control variables included: (1) socio-demographic characteristics (age, gender, race); (2) clinical characteristics (functional level based on Global Assessment Functioning-GAF-scores) and diagnosis during the pre-period or baseline; (3) service history in pre period (state hospital stay, total number of psychiatric inpatient days, long episode versus multiple shorter ones, participation in a designated intensive outpatient program (CCCP), insurance status (Medicaid).


Sample characteristics of intervention and comparison groups: Sixty one (n-61) individuals met the HEU criteria for enrollment in the intervention group. The comparison group consisted of 115 individuals who met the diagnostic and utilization criteria for HEU status, but were not enrolled in the program. These 115 clients comprised about 15% of the total inpatient psychiatric population served by DSAMH in 2005 (n = 793). The HEU group was not part of the 793 patients.

Sociodemographic Characteristics Statistically significant differences between groups was found for the following characteristics: the HEU intervention group was younger than the comparison group (mean age 39.6 vs. 44.6; p = 0.0268) and had fewer individuals with Schizophrenic psychoses and a higher proportion of those with Affective disorder, compared to the comparison group (p = 0.0056).

Prior Service History A smaller percentage of patients in the HEU intervention group had a state hospital admission during the pre period (68.3 vs. 89.5%; p = 0.0005) as well as fewer inpatient days (68.7 vs. 94.8 days; p = 0.0342). The proportion of the HEU intervention group who had ever participated in the Community Continuum of Care Program (CCCP) was higher than the comparison group (75.5 vs. 60.0%; p = 0.0409). The greatest difference between groups was the number of episodes in the pre period; 11% of the HEU intervention group had a single long episode versus 72% of the comparison group (p < 0.0001). Thus, the intervention group had many more individuals with multiple episodes of a shorter nature.

Post Service History During the 12 month post period following the index study date, the mean number of inpatient days declined for both groups, however the decline was greater for the comparison group, than that for the HEU intervention group (78.7 vs. 27.7 days; p = 0.0002). Although both groups had a reduction in inpatient use, the likelihood of having an inpatient re-admission was significantly higher in the HEU intervention group (68.9 vs. 24.3%, p < 0.0001). The mean number of inpatient days also was significantly higher in the HEU group (41.0 vs. 16.1 days, p = 0.0021). The mean number of episodes in the post period was 1.7 for the HEU group and 0.3 for the comparison group (p < 0.0001). Also the proportion of those having two or more inpatient episodes in the post period was higher for the HEU group, than that for the comparison group (45.9 vs. 5.2%; p < 0.0001).

Regression Results The results show that individuals in the HEU intervention group, after controlling for group differences had (1) a higher probability of being re-hospitalized (OR = 4.3; 95% CI = 1.72, 10.74); and (2) a higher probability of having more episodes (OR = 2.99; CI = 1.83, 4.88).

With regards to the control variables, younger individuals (OR = 0.97, CI = 0.96, 0.99) and those with lower functioning scores (OR = 0.98, CI = 0.96, 0.99) in the pre period were more likely to have frequent re-hospitalizations in the post period. In contrast, the odds of re-hospitalization was lower if patients had a state hospital admission in the pre period (OR = 0.27; CI = 0.09, 0.79) or one long versus multiple short episodes (OR = 0.43; CI = 0.23, 0.80). A high number of inpatient days in the pre period was associated with a higher number of inpatient episodes in the post period (OR = 1.01; CI = 1.01, 1.0) but not with higher number of days in the post period.


The findings reported in this article provide evidence in support of other research on high end users employing a pre post design that show a decline in inpatient days following a period of high inpatient utilization. In addition, the results are consistent with several randomized trials, described previously, that find that the HEU intervention program was not more effective in reducing inpatient use than treatment as usual. In fact, the members of the high end user group in this current study were more likely to be re-hospitalized than the comparison group. This suggests that the population chosen for the HEU program may have been at higher risk than the comparison group, despite our use of statistical controls. The fact that the vast majority of individuals in the intervention group had multiple versus one long hospital stay during the pre-period suggests that these individuals may not have been stabilized at the time of their enrollment in the study. There may have been other subjective factors as well, not in the administrative data or specifically identified by staff involved in the selection process, that led them to select these individuals into the program.

Although a potential limitation to our study was the use of a pre post design, evaluating the effectiveness of programs in public sector settings does not generally allow for randomization trials. We therefore strengthened our design by employing a comparison group that had inpatient histories that met the HEU intervention criteria and used the same period of time in order to eliminate any service system differences. Also, by using multivariate analysis techniques we were able to control for differences in demographics, diagnosis, etc., between those in the intervention program and those out of the program. Our preliminary analysis results, using propensity scoring techniques, had the same results.

These findings are also consistent with a prior analysis of administrative data by the authors that found that during the time period before the implementation of the HEU program in DE in 2004, 70% of inpatient users identified with a high user profile were not re-hospitalized during the year following their HEU episode (Rothbard, Zubritsky, et al. 2009). Thus being a high user in one period did not necessarily predict whether you would be a high user in the next period. In fact, the majority of high users in any 1 year are not users in the previous year. Another study by authors examining psychiatric re-hospitalization patterns in Pennsylvania using hospital discharge data from general and private psychiatric hospitals (1990 and 2006), showed that 75% of discharged patients had only one episode of psychiatric hospitalization annually. The 25% that were re-hospitalized were not distinguishable from the recidivists using demographic or clinical factors found in administrative data files (Rothbard, Lee, et al. 2009).

Predicting who will be a high user, prior to, or following an episode of heavy inpatient use, has to date, proven elusive. If the majority of annual high users cannot be identified before their high use episode, an intervention that is effective will need to identify within the high user group those at risk for continued inpatient admissions. Although this study confirmed some previous findings that being in a high end user program did not make a significant difference in reducing future re-hospitalization rates (Harrison-Read et al. 2002; Quinlivan et al. 1995), we do not know whether inpatient use would have been even higher for those selected into the program. More research is needed to identify what these factors may be.

Implications for Behavioral Health

Our current study confirms key findings in the literature showing that predicting inpatient utilization, even among known high users, is challenging. The results of this study suggest that individuals who have a long inpatient stay and are connected to the outpatient system following discharge have a lower likelihood and lower frequency of readmission. Earlier and better discharge planning for individuals who have two or more inpatient episodes in a year, with a focus on community support and adequate housing, may likely be a more effective strategy for reducing future hospital stays since many individuals return to the hospital due to inadequate placement opportunities. Alternatively, longer stays in the hospital for those with severe mental illness and other co-morbid conditions may be necessary as evidenced by the lack of re-hospitalization for those individuals discharged after a single long episode. From a policy perspective, these findings may also provide guidance in selecting program participants who will best respond to the intervention.

Copyright information

© Springer Science+Business Media, LLC 2012