Brief Critical Time Intervention to Reduce Psychiatric Rehospitalization

Changes in behavioral health care in recent years have resulted in a transition from inpatient psychiatric care to primarily community-based treatment for individuals with serious mental illness (1–3). Although this transition has resulted in shorter psychiatric inpatient stays, up to 50% of individuals are readmitted within one year of a discharge, and some are readmitted on multiple occasions, creating a “revolving door” situation (4). The period immediately after psychiatric hospital discharge poses many challenges to the individual, including increased risk for a number of adverse events, including rehospitalization (5,6). Successfully connecting individuals to community-based resources (7) or community-based behavioral health treatment (8–11) has been shown to reduce rehospitalization and improve quality of life for individuals with serious mental illness. Fostering relationships with community-based behavioral health treatment through engagement on the inpatient unit and consistent postdischarge contact may help to facilitate the recovery process (11,12), especially for individuals for whom usual transitional models of care have not been effective.

The critical time intervention (CTI) model and its shortened adaptation, brief critical time intervention (BCTI), are evidence-based practices shown to be highly effective in helping individuals with serious mental illness to connect with community resources, to begin or continue outpatient treatment after hospital discharge, and to find emotional and practical support during periods of transition (13–17). CTI is included in the Substance Abuse and Mental Health Services Administration’s National Registry of Evidence-Based Programs and Practices (www.nrepp.samhsa.gov). Traditional CTI is a nine-month model of care coordination that was originally developed as an individual-level intervention intended to prevent recurrent homelessness among persons with serious mental illness during periods of transition (13,17,18). In samples of homeless individuals with mental illness, CTI has been associated with reductions in homelessness (13,17,18), psychiatric symptoms (14), and psychiatric readmission within 18 months of discharge (19). Jones and colleagues (18) further concluded that the CTI model is a cost-effective alternative to usual care.

BCTI, a brief, three-month version of CTI, may be an efficient alternative to the longer nine-month model (16). In a study by Dixon and colleagues (16), adult veterans with serious mental illness received either care management with BCTI or usual aftercare without BCTI after a psychiatric hospitalization. Results indicated that individuals in the BCTI group were more likely to connect with outpatient services with significantly fewer days between inpatient discharge and first outpatient service utilization and utilized more outpatient mental health and substance abuse treatment services than individuals in the control group, providing evidence that BCTI implementation at the time of inpatient discharge led to improved continuity of care postdischarge.

The purpose of this investigation was to examine the association between implementation of a BCTI intervention model and occurrence of early and long-term psychiatric hospital readmission in a sample of adults with serious mental illness. Individuals with repeated rehospitalizations are especially susceptible to early readmission, defined as rehospitalization between 30 and 90 days of discharge (20,21), and targeted care coordination during the transition from inpatient stay to community-based care may help increase community tenure. Rates of early readmissions and those after 90 days may be used as indicators of an intervention’s ability to reduce psychiatric inpatient service utilization among the “revolving door” population.

Methods

Setting and Design

Activities were approved by the University of Pittsburgh’s Institutional Review Board as a quality improvement activity. A quasi-experimental investigation was conducted at six community-based provider organizations located in western Pennsylvania within the network of the Community Care Behavioral Health Organization, a not-for-profit, managed behavioral health care organization (MBHO) that supports care for publicly funded individuals. This investigation included a prospective intervention group and a comparison cohort. BCTI was integrated into acute service coordination (ASC), an existing service developed as an aggressive engagement-and-intervention approach to coordinate services for individuals with complex psychiatric, substance abuse or dependence, social, environmental, and housing needs. The quality improvement effort was designed to improve and standardize the delivery of ASC by training and supporting ASC staff to use the BCTI model. Acute service coordinators are expected to engage the individual during psychiatric admission, identify the unmet needs that led to the admission, and develop a plan with the individual to meet those needs postdischarge. These coordinators must have at least a bachelor’s degree and three years of experience in service coordination. As a result of the improvement effort, community-based service utilization after psychiatric hospital discharge was expected to increase and rate of psychiatric hospital readmission was expected to decrease.

Sample

Individuals were included in the BCTI cohort if they were over 18 years of age, had a mental disorder or co-occurring mental and substance use disorder diagnosis, and were referred to ASC at one of the six community-based providers. Individuals were referred to ASC by the MBHO if they had two or more psychiatric admissions within 30 days and were not connected to any other intensive level of care, such as assertive community treatment or intensive case management. Individuals referred to other levels of service or with abrupt discharges against medical advice or who were discharged to a substance abuse treatment facility outside of the service area were not referred to ASC. From February 1, 2011, to April 30, 2012, a total of 260 inpatient discharges were referred to ASC; 163 discharges resulted in the initiation of ASC, representing 160 unique individuals. Multiple readmissions within 30 days were verified through claims data of 149 individuals. A comparison cohort was identified with the use of claims data. Inclusion in the comparison cohort was based on the following criteria: over 18 years of age, having a mental health diagnosis or a diagnosis of co-occurring disorders, and utilization of ASC because of multiple readmissions within 30 days at the same six provider organizations as used by the intervention cohort during the year prior to implementation of BCTI (January 1, 2010, through December 31, 2010). The intervention and comparison cohorts were mutually exclusive.

Training and Implementation of BCTI

The MBHO coordinated implementation of the BCTI model with community-based ASC providers under the direction and consultation of two members of the research team (RWG and DH). The authorization for ASC, which was previously 45 days, was extended to 90 days to coincide with the recommended time frame of BCTI. The phased model of BCTI was used as a guide for service coordinators to focus on assessment of immediate needs and resources (phase 1), ongoing connection to other community-based resources (phase 2), and transition from ASC to community-based mental health services (phase 3). BCTI enhanced the delivery of ASC by adding or changing several elements, including focus on person-centered care; assessment of the individual’s strengths; more intensive recovery planning; emphasis on connections to community-based resources, such as housing options and food banks; and building individual autonomy over the course of the three phases. Also, the implementation process for BCTI provided additional clinical support and training for acute service coordinators in motivational approaches, recovery and crisis planning, and identifying community resources. Ongoing teleconsultation sessions with consultants and acute service coordinators were held throughout the intervention.

Study Variables

To track implementation, a face sheet, completed together by the provider and the individual, was added to the health record to monitor the occurrence of activities important in the delivery of BCTI, such as predischarge meetings, recovery planning, needs assessments, contact with aftercare providers, and treatment planning. The face sheet was used to monitor priority areas identified in the treatment plan: treatment engagement, housing and income assistance, involvement of family and social network, legal assistance, motivation concerning substance abuse treatment, medication or integrated medical care, practical needs assistance, or other priority goal area. Demographic information for individuals receiving ASC was obtained from state administrative data. To further describe the population, we looked at behavioral health service utilization in the 30 days prior to the inpatient readmission and diagnoses during treatment, which were obtained from administrative claims data.

Statistical Analysis

To evaluate the impact of BCTI on behavioral health services, including psychiatric hospital readmission within 30 days (early) and within 31–180 days (long term), we used chi square tests to compare utilization rates for individuals receiving BCTI with early and long-term utilization rates in the comparison cohort. For psychiatric readmission, considering that Tomita and Herman (19) examined readmission at twice the length of the CTI intervention (18 months), we also examined readmission at 180 days, which was twice the length of the BCTI intervention.

Next, we developed a logistic regression model using as predictors variables that had a significant association with readmission in our sample and using the dichotomous variable of readmission (yes=1, no=0) as the outcome and cohort (BCTI versus comparison) as the independent variable, with controls for any variables that had significant differences. Regression models for early and long-term readmissions were performed separately. The Hosmer-Lemeshow (H-L) test statistic was used as a goodness-of-fit test, and the likelihood ratio, Lagrange multiplier, and Wald tests were examined to evaluate the model (22). Two additional measures of fit were produced: the Cox and Snell R² (23) and the Nagelkerke R² (24). The c statistic measure of association was assessed to determine the degree to which predicted probabilities agreed with actual outcomes; a c statistic of 1 represents perfect model discrimination (25). Odds ratios with 95% Wald confidence intervals were examined and interpreted. The level of significance was set to .05 for the analysis. All statistical procedures were completed with SAS version 9.3 (26).

Results

Sample

We identified 149 individuals who had an inpatient stay and received BCTI. As shown in Table 1, 52% of the BCTI sample consisted of women. The age of this sample ranged from 20 to 65, with a mean±SD age of 38.8±11.5 years. For the entire sample, 56% were white, 42% were black, and 3% were Asian or of other race-ethnicity. Common diagnoses in the BCTI group were anxiety, 13%; bipolar disorder, 17%; depression, 52%; and schizophrenia, 18%. Fifteen percent had other diagnoses, which included dementia, eating, neurologic, paranoia, personality, sleep, and sexual disorders. Thirty-three percent of the sample had co-occurring mental and substance use diagnoses. Diagnostic categories were not mutually exclusive. The most commonly used behavioral health services before inpatient readmission of the BCTI sample were outpatient mental health services, 58%; crisis services, 28%; and outpatient substance use disorder treatment, 8%.

The comparison cohort consisted of 224 individuals who received ASC in the year prior to BCTI implementation at one of the six providers of the study. Women comprised 38% of this cohort, and age ranged from 18 to 65 years, or 40.2±12.0 years. Fifty-six percent of individuals were white, 43% black, and 1% were of other race-ethnicity. Common diagnoses for the comparison cohort were anxiety, 6%; bipolar, 13%; depression, 31%; and schizophrenia, 23%, with 10% of individuals having co-occurring disorders. The most commonly used behavioral health services before inpatient readmission were outpatient mental health, 55%; crisis, 20%; and substance use disorder services, 4%. As shown in Table 1, the proportion of individuals with co-occurring disorders was greater in the BCTI cohort than in the comparison cohort (p<.001), as were the proportions with anxiety (p=.02) and depression (p<.001). There was a significantly higher proportion of women in the BCTI cohort than in the comparison cohort (p=.01).

Implementation of BCTI in ASC

Information on the face sheets showed that most BCTI activities were completed at a high rate during the three phases. Service coordinators met with referred individuals before psychiatric inpatient discharge in 89% of cases. Rates of conducting treatment planning, needs assessment, and recovery planning were high in all phases (78%−100%) and on average were completed within seven days of discharge during phase 1, within 50 days during phase 2, and during transition from ASC to community after 90 days in phase 3. The three priority areas most commonly identified as goals were housing and income, treatment engagement, and substance abuse treatment motivation.

Service Utilization and Readmission

As shown in Table 2, utilization rates for commonly used behavioral health services after hospital discharge were similar between cohorts for crisis, outpatient mental health, and substance use disorder services. The proportion of individuals with an early psychiatric hospital readmission was significantly lower for the BCTI cohort versus the comparison cohort (p<.001). Twenty-eight percent of individuals in the BCTI cohort had a psychiatric hospital readmission within 30 days, compared with 47% of individuals in the comparison cohort. Although not statistically significant, the observed longer-term (31–180 days) readmission rate was lower for the BCTI cohort (44%) than for the comparison cohort (52%).

As shown in Table 3, the characteristics of the individuals with early (N=147) or longer-term (N=181) readmission regardless of the intervention cohort were compared with those of nonreadmitted individuals, with no between-group differences found. The logistic regression models for early and longer-term readmission included the intervention cohort (BCTI or comparison) adjusted for four variables, with significant differences between cohorts on gender and in diagnoses of anxiety, depression, and co-occurring disorders (Table 4). For the early readmission model, the H-L goodness-of-fit test was not significant (p=.05), which indicated that the model fit was appropriate for the data (22). In addition, the likelihood ratio, Lagrange multiplier, and Wald tests were significant (p<.001), which revealed that the model containing the covariates was superior to the null model. The Cox and Snell and Negelkerke R² statistics were .06 and .09, respectively. The c statistic was 65%, representing the proportion of individuals who were correctly classified as having a higher probability of readmission (25).

For early readmission, after controlling for between-group differences, we found that those in the comparison cohort were 2.83 times more likely than those receiving BCTI to be readmitted (p<.001). After we controlled for the other variables in the model, no other factor was associated with early readmission.

For longer-term readmission, the H-L goodness-of-fit test showed appropriateness of the model (p=.87), but the likelihood ratio, Lagrange multiplier, and Wald tests were not significant. The Cox and Snell and Negelkerke R² statistics were .02 and .03, respectively, and the c statistic was 59%. After controlling for between-group differences, we found that those in the comparison cohort were 1.51 times more likely than those receiving BCTI to be readmitted in days 31–180 after hospital discharge, but the difference was not significant. Individuals with co-occurring disorders were significantly more likely to be readmitted than those with either mental illness or a substance use disorder (OR=1.84, p=.04).

Discussion

In this study, incorporation of BCTI into an existing acute level of service coordination was associated with significantly lower rates of early readmission. Integration of the BCTI model enhanced and standardized an existing service, ASC, through the addition of more recovery-focused care, improved monitoring of individual strengths and needs, and focus on connection to community-based resources and treatment. These results are consistent with research suggesting that addressing individual needs in a person-centered approach can lead to better outcomes after psychiatric hospital discharge (11), and it is likely that these enhancements to ASC contributed to the lower early readmission rate. Although not statistically significant, the decreased risk of longer-term readmission in the BCTI cohort versus the comparison cohort is promising.

To our knowledge, this study was the first to examine early (within 30 days) and long-term (within 180 days) psychiatric hospital readmission for individuals receiving service coordination with BCTI. Prior studies have shown that the traditional nine-month CTI model was successful in reducing the rates of long-term psychiatric readmission rates (19); this investigation provides limited evidence that the abbreviated three-month BCTI model may also be beneficial. We would expect the implementation of the BCTI model to have the greatest impact on psychiatric admissions in the higher-risk group, the “revolving door” population of individuals who have experienced repeated psychiatric hospitalizations. BCTI’s efficacy in other populations receiving behavioral health services would need to be studied in order to determine its usefulness in preventing or delaying psychiatric admissions among the broader population of individuals receiving behavioral health services.

Our findings lend support to the previous results of Dixon and colleagues (16), with some differences. Dixon and colleagues found that BCTI supported continuity of care through increased utilization of behavioral health services after psychiatric hospital discharge, whereas we found no differences in service utilization but did find lower early hospital readmission rates. In this study, both intervention and comparison cohorts received ASC, a more intense level of care than standard care, which may have served as the sole source of community-based treatment for some of these individuals. Despite this connection, a large proportion of individuals receiving ASC with BCTI experienced a readmission within 180 days of a prior discharge. This finding is consistent with previous research (27) showing higher rates of readmission among individuals with serious mental illness who use mental health services. A better understanding of how connection to community-based resources and treatment may affect readmission is needed to help determine how providers delivering BCTI might focus efforts on treatment engagement in this population.

The study had several limitations. Despite our best efforts to identify a comparable historical cohort, we had to use a nonrandomized, quasi-experimental design, and interpretation of results should bear this in mind. Although controlled for in the analyses, several differences observed between the cohorts, for example, in diagnoses and gender, indicate that the groups differed somewhat; thus unobserved factors may have contributed to differences in rehospitalization outcomes. Also, inherent in the design is the change in ASC authorization from 45 days in the comparison cohort to 90 days in the BCTI cohort, which, in itself, may have improved longer-term outcomes. In addition, administrative claims data were used to define both population characteristics and hospitalization outcomes. These data are limited in that their primary use is not defined for this purpose but instead to meet business needs. Furthermore, use of existing administrative data prevented us from assessing important outcomes other than rehospitalization and service utilization, such as quality of life, homelessness, and other functional outcomes. Future research should examine the use of BCTI plus ASC by using a randomized design with larger samples and should measure the engagement of each individual with service coordinators. In addition, cost and cost-effectiveness should be studied to determine whether using the BCTI model offsets the cost of psychiatric inpatient stays in groups at high risk of psychiatric readmission as well as the wider population of individuals receiving behavioral health services.

Conclusions

In this study, BCTI was associated with decreased early psychiatric hospital readmission rates for individuals at high risk of readmission. On the basis of data from individuals receiving ASC, community-based providers in this effort were able to successfully implement the BCTI model into care, which included efforts to connect individuals to multiple community-based mental health services and other resources. These results support use of BCTI for individuals with serious mental illness during transition from inpatient to community living.