An official website of the United States government
NCBI Bookshelf. A service of the National Library of Medicine, National Institutes of Health.
Dawn I. Velligan, PhD, Megan Fredrick, MA, Cynthia Sierra, MA, Kiley Hillner, BA, John Kliewer, MD, CPS, David L Roberts, PhD, and Jim Mintz, PhD.
Author Information and AffiliationsAs many as 40% of individuals with behavioral health challenges (BHCs) that require hospitalization—such as psychotic disorders, affective disorders, and severe anxiety disorders—do not attend any outpatient visits in the 30 days following discharge.1
We examined engagement-focused care (EFC) vs treatment as usual (TAU) in a university-based transitional care clinic (TCC) with a 90-day program serving individuals with BHCs who were discharged from hospitals and emergency departments (EDs). EFC included a unique group intake process, designed to get individuals into care rapidly, and a coach for shared decision-making (SDM).
We approached individuals for study either before hospital discharge or upon arrival to the TCC, following referral. Assessments of quality of life (QOL), symptomatology, and SDM preferences were conducted at an initial assessment, at 3 months (corresponding to the end of TCC treatment), and 6 months post-TCC discharge. We assessed communication among participants and providers at each TCC visit and we assessed service utilization during and post-TCC for inpatient, outpatient, emergency, and hospital services as well as in criminal justice contexts.
For the primary outcome—ie, subjective QOL from the Quality of Life Interview—465 patients were randomized, 326 completed an initial assessment, 274 were lost to follow-up (139 of these never even received an initial assessment), and 191 had an initial and at least 1 follow-up assessment and were included in the primary repeated measures analysis of covariance for mixed models. Mean subjective QOL scores were 3.88 (95% CI, 3.73-4.02) and 3.78 (95% CI, 3.56-4.00) for EFC and TAU, respectively, at initial assessment; 4.33 (95% CI, 4.14-4.52) and 4.04 (95% CI, 3.78-4.29) at 3 months; and 4.48 (95% CI, 4.26-4.70) and 4.19 (95% CI, 3.93-4.45) at 6 months. The mixed effects regression, examining impact on subjective QOL, yielded significant effects of group (ie, EFC vs TAU) at 3 months (F1,216 = 4.14; P = .04); and nonsignificant effects of time (F1,132 = 2.31; P = .13) and group by time (F1,132 = 0.32; P = .56). The effect size for the group effect was 0.28 (Cohen d), which is a small effect. The views of prescribers and consumers about communication converged as time went on, and 91% of patients wanted at least some say in decisions about their treatment.
EFC, an intervention consisting of 2 components—access to a group intake and access to an SDM coach—may improve QOL. Most people with BHCs want some say in treatment decisions.
We conducted the study at only one site, many individuals were lost to follow-up, and due to the nature of our clinic, we conducted the first assessment following randomization. Our TAU was likely superior to the typical standard care in the community, reducing the ability to find group differences between EFC and TAU.
For individuals with behavioral health challenges (BHCs), timely access to community services after discharge from a hospital or an emergency department (ED) is particularly critical.1,2 In fact, costs for treating BHCs are an estimated $317 billion annually in the United States; this translates to an estimated $1000 per year for every person in the country.3 As the health care system moves progressively toward a “pay for performance” reimbursement model, it will become increasingly important to ensure continuity of care and engagement in the long-term management of these BHCs. Coordination between inpatient and outpatient clinics increases initial outpatient attendance, and case management can improve this coordination.4-9 However, beyond these factors, little is known about psychiatric transitional care. Specifically, whether transitional services can affect an individual's long-term engagement in care, perceptions of care, or quality of life (QOL) is not known.
Enormous systemic obstacles affect individuals with BHCs. Many communities face a shortage of providers willing to treat this population. Community mental health centers are often on diversion (ie, unable to take patients due to lack of availability), and many have long waiting lists for service. In these cases, individuals with BHCs end up using the ED and the hospital for care, never engaging in outpatient services. This is exactly the situation that faced Bexar County, Texas (the location of the PCORI-funded study described here), according to a 2010 report by Capital Healthcare Planning.10 Similar shortages plague many communities. Finding ways to get individuals into mental health care rapidly after a hospitalization is a growing need, as the National Council for Behavioral Health's Same Day Access Multistate initiative attests.11 Individuals transitioning from hospitals must make decisions that greatly influence their recovery in brief and infrequent meetings with mental health providers. Assistance in making shared decisions is needed to help individuals weigh pros and cons, as well as to gain skills and comfort in sharing information with a provider in a way that maximizes treatment outcomes.
Hospitals and outpatient clinics have attempted to improve the likelihood of successful transitions by ensuring that the outpatient team contacts the individual while he or she is in the hospital; this promotes shared accountability across providers and organizations, transition planning, and patient and family engagement and education Kasprow et al.68 Models such as the Care Transitions Intervention (Viggiano and Pincus70) and the Transition Care Model (Naylor et al69) have used care transition managers who meet with individuals in the hospital, engage them in managing their own care, and follow up with home visits after individuals have been discharged from the hospital.
In April 2012, we launched a novel transitional care clinic (TCC) for individuals transitioning from inpatient units and EDs in Bexar County (San Antonio), Texas, with the goal of improving linkage to outpatient care and of improving engagement in maintenance mental health care through person-centered services. The TCC is designed to offer more rapid intake and access to services than that of the local community mental health service system and provides treatment for up to 90 days, until the individual can be connected to existing services. This program uses many of the strategies discussed above as standard care (eg, meeting individuals in the hospital before discharge, having contact after discharge to address barriers to getting to their intake). Services provided at the TCC include medication management, care coordination, and a variety of evidence-based psychotherapies.12 Treatment ends with ensuring that the person makes a successful transition to long-term facilities in the community.
In this clinical setting, we compared a traditional standard care approach with engagement-focused care (EFC) during transition for individuals with BHCs. EFC had 2 components not typically found in transitional care models: (1) an access group intake process and (2) brief coaching on shared decision-making (SDM).13 Access group is a group intake in which members of the treatment team meet with several patients at once, speaking to them in turn about the reasons for hospitalization or ED use and their current needs. The impetus for access group as opposed to individual intake was the need to get individuals into intake appointments as soon as possible. When individual intakes are scheduled with show rates hovering below 50%, slots fill quickly, and people end up being scheduled out for intake months following discharge. In the meantime, patients have no access to a psychiatrist or other prescribers to refill medications, and many individuals re-present at emergency services and inpatient facilities. Moreover, a group intake process allows the TCC to prioritize individual needs. For example, some individuals have not filled a prescription they received from the hospital because they cannot pay for medication. They urgently need social workers to assist them with getting their medication as soon as possible. Other patients come from EDs, where they were not prescribed medications; many of these individuals need an appointment with a psychiatrist or other prescriber as soon as possible. Still other individuals are heading into crisis and need rapid access to appointments for psychotherapy. A group intake process allows needs to be prioritized, and each person can get his or her most pressing needs met as quickly as possible. Recent research at the TCC suggested that, although individual intakes are rated slightly more favorably by participants, both services are rated good or better, and individuals are very willing to have group intakes if it decreases the wait time for services. 14
Although rarely used in transitional programs, SDM was included in our study to improve engagement. Individuals who feel they are playing an active role in their treatment typically have better engagement and outcomes than those who experience themselves as passive recipients of care.15-17 Unfortunately, across different BHCs, individuals typically engage with providers in a passive manner, and lack of engagement is related to poor follow-through with treatment regimens.18,19 Insufficient treatment follow-through has severe consequences for individuals immediately following hospitalization or crisis care, often leading to recidivism and decreased QOL.20
The use of SDM in psychiatry has received support through the New Freedom Commission on Mental Health and other recent government policy reports.21-24 SDM may be of particular benefit to patients with BHCs because of the high incidence of poor treatment follow-through; it represents a nonthreatening approach to clarifying motivations and options for improving QOL. Adults with BHCs frequently make competent and prudent treatment decisions.25,26 Also, despite some behavioral passivity, those with BHCs express a strong desire to be informed about their behavioral health issues and treatment options and to be active participants in their treatment decisions—a desire slightly higher than that expressed by patients treated for medical conditions in primary care.27-31
The aims of this study were to examine whether a care approach during transition—one that added a group intake for faster access to services and an SDM coach to help with engagement—would improve engagement and outcomes in individuals recently transitioned from hospitals. We hypothesized that individuals in EFC, which included 2 components—access to group intake and access to an SDM coach—would be more likely than those in treatment as usual (TAU) to attend their TCC intake appointment. We hypothesized that those in EFC would report significantly greater SDM in their TCC psychiatrist/prescriber visits, greater attendance at postintake appointments, greater attendance at post-TCC mental health follow-up appointments, and greater QOL than those in TAU. Primary outcomes were subjective QOL and attendance at mental health services after discharge.
Consumers and other stakeholders routinely participated in the design and conduct of the study. Twenty-five consumer and 10 provider stakeholders had various roles during the project; of that group, 9 consumer stakeholders and 3 providers are part of our ongoing community advisory board. In addition, 1 consumer participant was a certified peer specialist who assisted in developing and providing the SDM intervention; another was a certified peer specialist with an MD who assisted in writing this document. We obtained stakeholder views via interviews and focus groups throughout the project. Views from stakeholders were discussed with researchers and incorporated into study procedures. Suggestions that would cause practical barriers were discussed, and feasible compromises were reached that took into account the perspectives of those with lived experience. We identified potential consumer stakeholders from individuals previously treated at the center; from those who had participated in past research studies who were not patients of University of Texas Health Science Center San Antonio (UTHSCSA); from VIA Hope, the peer certification network in Texas; and from patients known to staff members, such as the case with our peer specialist.
We invited provider stakeholders from local hospital and community agencies. Stakeholders contributed to all aspects of study design, including the content of treatment, methods of recruitment, and selection of outcome variables. They participated in producing the videos used in the shared SDM coaching; provided guidance on important components of SDM coaching; and contributed to the development of the TAC-Review manual (TAC stands for Tell-Ask-Choose). Stakeholders assisted with language in the consent form, asking permission for outreach to increase participation in follow-up assessments, as well as in the assessment description. Stakeholders chose the primary outcome as QOL. They reviewed outcomes and gave feedback on manuscript drafts, and they will continue to be involved in dissemination as findings are presented. They were highly instrumental in getting university policy changed so that we now have a job description for—and can continue to hire—peer specialists. They were also active in making the advisory board an ongoing board that generates research ideas, evaluates proposed projects, and participates in all studies within the Division of Community Recovery, Research, and Training.
It is important to note that stakeholders also contributed to the way TAU was delivered and structured at the TCC before this study. Based on feedback from focus groups and advisory boards, we knew that consumers wanted more rapid access to treatment following hospitalization and ED use, as well as wraparound services, evidence-based psychotherapies, longer prescriber visits, etc. Stakeholders were a force in making the TCC deliver the best care possible. Thus, EFC was added to a treatment program that had already been substantially groomed by consumer input.
We conducted a randomized parallel group design investigating 2 treatments, TAU and EFC, with assessments immediately following intake and at three months postintake for each treatment option. We chose this design because randomization is the best way to equate groups on key demographic and symptom factors at initial assessment; however, because of the practical issues of clinic flow, individuals who agreed to randomization either by signing a consent form while hospitalized or by giving verbal assent on the day of intake were randomized to group intake (part of EFC) or individual intake (part of TAU) based on their verbal assent. This made it necessary for the first assessment (of symptomatology, SDM preferences, and QOL) to be conducted after randomization (as this was the first time the person was physically in the clinic, and intake was the first thing to happen.) Though unconventional, this approach was a compromise for a busy real-world postacute setting.
Before the start of randomization, all physicians and residents at the TCC were trained in SDM by Laurie Curtis, senior program manager of Advocates for Human Potential, a peer advocacy organization. Ms Curtis served as project director for the Substance Abuse and Mental Health Services Administration Shared Decision Making in Mental Health Project, a 3-year initiative to develop resources and tools to facilitate and support collaborative decision-making among mental health consumers and practitioners. She trained all providers in a 1.5-day seminar. This session was taped, and 1-hour refreshers on SDM were held at least once every 6 months during the trial.
There were 2 phases of recruitment in this comparative effectiveness trial. In the first phase, participants were recruited at the hospitals referring to TCC, where they were randomized to either individual or group intake at the time of scheduling their appointment. Because of significant logistical problems—including opening TCC referrals to many more facilities and lacking sufficient staff to be physically present at all of them, as well as many people in the hospital signing consent who subsequently did not attend the TCC—we stopped recruiting in the hospital and began recruiting just before the intake at the TCC. Because of this change in recruitment, our first hypothesis—regarding show rates at intake after randomization at discharge from the hospital—could be tested only with the first recruiting phase in the subset of individuals recruited in the hospital. For recruitment in the hospital, individuals were approached by a UTHSCSA research staff member credentialed at participating sites in accordance with HIPAA requirements, told about the program, and asked to sign informed consent. They were then randomized using a computer program that scheduled them for TAU or EFC. For recruitment at the TCC, individuals were approached by a research staff member after a group orientation to the clinic and asked if they would verbally assent to randomization into either individual or group intake. Following the intake, individuals who had agreed verbally to randomization were then asked if they would be willing to consent to the study as a whole, including all assessments and the SDM component if offered. All participants signed a written consent form approved by an Institutional Review Board, and procedures were consistent with internationally recognized standards for ethical conduct of human research. Randomization was 2:1 for EFC to TAU because of limited staff availability for individual intake.
All participants received medication follow-up, psychotherapy, and care coordination based on a preliminary treatment plan completed at intake. After intake, individuals were given a first assessment and followed for approximately 9 months. (This was not called baseline because participants had been previously randomized.) Formal assessments were repeated at 3 months (the modal treatment termination time at the TCC), or TCC discharge and 6 months following discharge from the TCC. In addition, attempts were made at each visit with a psychiatrist or other prescriber to have both participants and providers rate the process of the visit. Finally, calls to track service utilization after discharge from the TCC were made monthly.
In addition to individual intake, participants in the TAU group received standard care. Standard care provided by the TCC was more robust than standard care at other agencies serving this population; it included medication management, in-home visits using cognitive adaptation training,32 case management to connect individuals to resources, and evidence-based psychotherapies including cognitive behavior therapy,33 dialectical behavior therapy,34 solution-focused therapy,35 and group psychotherapies addressing specific issues (substance use, depression, etc).
This treatment included group intake as well as access to all the treatments described above as our standard care. EFC also included SDM, and this component involved meeting with an SDM coach before and/or following prescriber visits. SDM focused on how recovery goals could be met and provided training using a simple acronym to help participants learn their role in advocating for their care. This acronym, TAC-Review, helps people learn how to Tell-Ask-Choose (or not) and Review so they can be active in selecting and evaluating their treatment. In addition, SDM involved role playing to ensure that people with BHCs could clearly express their values and preferences to the provider; information on options and SDM tools were also provided. Participants had access to videos showing people with different BHCs and how they coped, as well as to videos of individuals engaged in SDM. Finally, SDM tools were available to weigh pros and cons, plan a visit, or explore different treatment options. All of these resources were available in an SDM resource center, which was an office on the same floor as the clinic, equipped with computer and internet, handouts explaining various parts of SDM, and worksheets. The resource center was staffed by a coach who was either a counselor or a certified peer specialist.
Of the 326 individuals participating in initial assessments, 147 were male and 179 were female; 160 were Hispanic, 135 were White, 24 were African American, 3 were Asian, and the remainder were of mixed ethnic background. The mean (SD) age of participants was 38.0 (12.1) years. At initial assessment, mean (SD) level of symptomatology as rated by the expanded version of the Brief Psychiatric Rating Scale (BPRS) was 53.9 (12.3). The BPRS scores ranged from 27 to 97, with higher scores indicating more severe pathology.
We measured QOL using the Quality of Life Interview (QOLI).36 This 45-minute structured interview assesses subjective satisfaction and objective factors (ie, functioning, access to resources) in the domains of family, social relations, leisure activities, finances, legal/safety issues, work/school, and health. It is one of the most psychometrically sound instruments for assessing QOL in individuals with mental illness.37 We used the mean of all subjective items as our primary outcome to emphasize the person's perspective as one of our primary outcome variables. With our planned recruitment of 300 (note that we recruited additional participants due to high levels of loss to follow-up), the study was powered to find a medium effect (Cohen d = 0.50) for the primary outcome of QOL.
There is little consensus on the definition and measurement of treatment engagement.38,39 Our primary metric was behavioral attendance at treatment appointments, because this can be reliably measured and is correlated with clinical change.40,41 Our primary measure of treatment engagement was attendance at mental health visits following discharge from the TCC. We also examined other metrics for engagement. For individuals recruited at the hospital (first phase of recruiting) we examined the proportion in each group who came to their initial intake appointment. For all study participants we measured engagement by looking at the proportion of postintake appointments kept vs those scheduled for all services at the TCC (other than SDM coaching visits), based on data from the electronic medical record.
To characterize the sample and examine changes in symptoms over time, we used the expanded BPRS,42 a 24-item interview-based measure that captures a range of psychiatric symptoms, which are rated on 7-point Likert scales. These scales are summed to yield a total score.
We made monthly calls to participants using a brief version of an interview adapted from the National Institute of Mental Health Epidemiologic Catchment Area Program43 and the HIV Cost and Services Utilization Study.44 This abridged version of the interview has been used in multiple published studies.32,45 We examined rates of negative service utilization outcomes, including rehospitalization for mental health, presentation at emergency or crisis services for mental health, and incarceration, by treatment group based on information obtained in the monthly calls. We made at least 3 attempts to the telephone number on file. We attempted to contact alternate telephone numbers given by the participant (for family, etc) as well. When calls were missed, we attempted to capture data for the outcomes listed above in the next monthly call so that all data were captured up to the last completed call. We totaled all occurrences of these events for each participant across the follow-up period. In addition, we totaled the number of days covered by the information obtained to equate data for those with missing information. This led to a measure of incidents per covered days (days in which we know outcomes and the outcomes that occurred over that number of days).
We were interested in measuring SDM behavior, to capture the process and preferences for the type of SDM that individuals recently discharged from a hospital experience as they begin a relationship with a provider. Measurement of SDM is complex because it involves 2 parties, their interaction, and their perceptions of each other. Additionally, there are individual differences in the degree of information and participation that consumers prefer, and—in psychiatry in particular—symptom experiences and cognitive impairments may affect an individual's ability to participate in SDM, independent of his or her desire or the skill of the practitioner. For these reasons, we assessed aspects of SDM using several measures.
We collected participant preference for SDM at initial assessment and at follow-up using the Control Preferences Scale (CPS).46 This scale presents participants with 5 cards that each present a statement describing a different role in decision-making. Statements include the following: (1) I prefer to make decisions about which treatment I receive; (2) I prefer to make the final decision about my treatment after seriously considering my doctor's opinion; (3) I prefer that my doctor and I share responsibility about which treatment is best for me; (4) I prefer that my doctor make the final decision about which treatment will be used but seriously consider my opinion; and (5) I prefer to leave all decisions regarding my treatment to my doctor. Cards are presented to the participant in a prespecified order (2, 4, 3, 1, and 5), and each statement is compared with the previous one until the 5 statements are rank ordered. The 2 top choices selected are categorized based on their order into 6 possible categories, reflecting how active vs collaborative vs passive the participants wanted to be in the interaction with the provider. The reliability and validity of this method have been established.46 For purposes of determining whether interventions differentially impacted SDM preferences, we divided the 6 categories of preference at each time point (first assessment and posttreatment) into 2 larger groups; 1 that includes the bottom and top categories, reflective of a preference for 1-sided decision-making (either solely patient or solely the doctor), and 1 category that contains the middle 4 categories, which reflect a preference for collaboration in making decisions.
We obtained consumer and prescriber ratings of in-session communication following visits with the prescriber, using the Matched-Pairs Instrument (MPI).47 The MPI is a dyadic instrument made up of 19 statements that assess the content and process of a prescriber's communication skills from the perspective of each person in the encounter. Each skill on the MPI is rated on a 5-point scale, with higher scores indicating stronger agreement with the statement. A version is completed by the prescriber and a version is completed by the patient participant immediately following the encounter. We calculated a difference score between provider and patient ratings. The absolute value of this score reflected the average distance between participants and providers in their perceptions of communication during the session.
All participants completed a face-valid feedback questionnaire that addressed various domains of their treatment experience on a series of 7-point, Likert-type scales; responses ranged from “completely dissatisfied” to “completely satisfied.” We adapted items from a measure used in a recent study of an SDM decision aid by Woltmann and colleagues.48 In addition, participants rated their intent to participate in post-TCC mental health care on a 5-point scale, with responses ranging from “definitely want future mental health care” to “definitely do not want future mental health care.”
Based on information ascertained in monthly telephone calls, we examined the percentage of individuals who participated in mental health treatment in the month following discharge from the TCC vs those who did not.
To maintain blinding of raters to treatment assignment, all participants were asked at the beginning of each outcome assessment not to divulge information about any visits at the TCC. If blinds were broken, alternative raters blinded to group assignment completed the remaining assessments.
Assessments occurred at an initial time point just following clinic intake and then again at approximately 3 months, or end of treatment at the TCC, and then again 6 months following TCC discharge. We entered all information into a HIPAA-compliant database using double entry and range restrictions. We examined distributions for normality and homogeneity of variance, and used transformations where necessary to meet the assumptions of the statistical models. In cases in which distributions could not be normalized we used nonparametric statistics. Participants were randomized and then assessed because the first contact at our facility was an intake in a very busy postacute clinic. Therefore, the first assessment is called first assessment rather than baseline.
Because participants had one component of treatment (intake in a group) before the first assessment, we examined group differences on the first assessment. We included all participants with a first assessment and at least one assessment after the first assessment in modified intent-to-treat analyses. So, regardless of whether individuals participated in the SDM coaching component of EFC, we included them in analyses. (We conducted modified intent-to-treat analyses on available cases, not based on nonadherence to the intervention.)
We examined group differences in subjective QOL, BPRS symptomatology, satisfaction with treatment, and difference scores on the MPI over time by treatment group (EFC, TAU), using mixed-effects regression with repeated measures (SAS PROC MIXED).49 We used first assessment scores as covariates where relevant (ie, satisfaction with care could not be rated in the first assessment). In addition, because of considerable missing data at follow-up, we conducted additional analyses using all individuals with an initial assessment (not just those with at least one follow-up in addition to baseline) to examine the reliability of the results. We did this only when the primary analysis of covariance was significant. We examined the assumption of missing completely at random using the missing completely at random (MCAR) test, which examines whether the pattern of missing data depends on the data values. When the test of statistical significance is <0.05, one can conclude the data are not missing completely at random.50
Furthermore, we conducted multiple imputation (n = 50) to estimate missing data to determine the reliability of significant results using different methods. For engagement as assessed by appointments kept after leaving the TCC, we examined data from the first service utilization call following TCC discharge, which asked about whether the participant had a visit with a mental health practitioner following discharge from TCC. To include all patients in this analysis we conducted a maximum likelihood analysis (3 × 2, χ2) measuring frequency of engaged patients (those who attended this appointment), nonengaged patients (those who did not attend the first post-TCC appointment), and patients for whom we had no information by treatment group. For unwanted service utilization outcomes including hospitalization, ED use, and incarcerations, we counted number of unwanted events and examined group differences as a function of days covered, using the log of exposure days using χ2 analysis. For SDM preferences on the CPS, we used a generalized linear model with repeated measures specifying an unstructured covariance matrix with fixed effects of group and visits and a binomial outcome (unilateral decision-making vs collaborative). Unless otherwise specified, all P values are 2 sided. CIs are reported.
To avoid missing data, we contacted by phone those individuals who missed assessments. In addition, patients consented to participate in active outreach for participants if they could not reached by phone. In these cases, research assistants went to homes, leaving generic notes to contact the study coordinator. This was not possible with individuals who were homeless. Every phone call was attempted a minimum of 3 times, and notes were dropped off at the participant's last known address when calls failed as a means of contact. These efforts were needed, as this study dealt with individuals who had long-standing BHCs with varying degrees of desire to participate in mental health services, were in unstable living situations, were primarily uninsured, and were at the lowest socioeconomic stratum. All data were regularly tracked, and all attempts were made to contact individuals for visits. The data analysis described above, which used mixed effects regression, was able to handle data missing at random by imputing missing outcomes based on a random sampling of outcomes that were not missing. As stated above, if analyses demonstrated group differences, we examined the assumption of missing completely at random using the MCAR test, conducted analyses using all patients with an initial assessment, and conducted multiple imputations. For other analyses, we included patients with an initial assessment who had no further information as 1 group in the maximum likelihood analysis.
The CONSORT diagram appears in Figure 1. A total of 1377 individuals were screened for participation; of them, 465 signed informed consent. Between consent and first assessment, 139 were lost to follow-up and did not participate in any study assessments. This was due, in part, to individuals being recruited at the hospital and never coming to the TCC and, in part, to individuals being recruited at the TCC who could never be scheduled for research assessments. Of all patients who signed informed consent, 303 out of 465 (65.16%) missed at least 1 assessment. This includes people who missed all assessments. The difference between groups (EFC vs TAU) in the number of individuals who missed assessments at any time was not significant (χ2 (1) = 1.19; P = .28). Of those who got an initial assessment, 164 of 326 (50.31%) missed at least 1 of the follow-ups. The between-group differences in the number of individuals missing at least 1 of the follow-up assessments was not significant (χ2 (1) = 2.59; P = .11). In examining evaluable patients—that is, those with an initial assessment and at least 1 follow-up (n = 219)—the dropout rates did not differ between treatment groups following first assessment, with 16 out of 76 (21.05%) dropping out in TAU and 41 of 143 (28.67%) in EFC (χ2 (1) = 1.50; P = .22). Numbers for each time period appear in Table 1.
CONSORT Diagram.
Demographic and Clinical Characteristics of Participants by Treatment Condition (Includes All Patients With an Initial Assessment).
Table 1 presents the demographic and variable scores at each assessment, by treatment group, for those with a first assessment. There were no statistically significant treatment group differences in those with an initial assessment—with one exception: Individuals randomized to EFC were slightly younger than those randomized to TAU (t = 3.92; P = .0486). Participants dropping after the first assessment did not differ in demographics or symptom variables from those who continued participation and had at least 1 follow-up assessment (all P values > .20).
In EFC, 135 out of 219 individuals (62%) participated in SDM appointments with the coach at least once; a total of 273 visits were delivered. On average, individuals in EFC who participated in SDM had 28.8 minutes of SDM coaching (this does not include those who did not meet with an SDM coach). Problems in delivery of SDM included individuals who did not show up early enough to attend appointments or did not have time to keep their SDM appointments after seeing the doctor. The most common needs individuals expressed were for reassurance that the doctor wanted them involved in the decision-making process and for the opportunity to role-play difficult scenarios, so they could practice how to discuss these topics with their providers (eg, discussion of sexual side effects of medication, not wanting a specific medication).
The mixed-effects regression examining subjective QOL yielded significant effects of group at 3 months follow-up (F1,216 = 4.14; P = .04); and nonsignificant effects of time (F1,132 = 2.31; P = .13) and group by time (F1,132 = 0.32; P = .56; Table 2). An inspection of means indicates greater improvement of QOL in EFC than in TAU across the follow-up period. The group difference represents a small effect (Cohen d = 0.28). Figure 2 depicts means derived at specified time points by treatment group. To determine how much impact missing data had on this result, we ran a repeated-measures analysis of variance using all individuals with a first assessment regardless of whether they had follow-up data. Results of this analysis indicated that (1) there were no significant differences at initial assessment, suggesting that there is no evidence that participation in group vs individual intake had an impact on first assessment scores (F1,325 = 0.51; P = .47); and (2) group differences were statistically significant only at TCC discharge at 3 months postrandomization and not at follow-up 6 months after discharge from TCC (F1,325 = 4.39, P = .037; and F1,325 = 1.82, P = .179, respectively). In addition, we conducted the MCAR test.50 Results suggested that it is plausible that data are missing completely at random. This suggests that the analysis of covariance for mixed models is likely the more powerful test to use. Finally, due to missing data we conducted multiple imputations with n = 50 imputations and found a significant result (t = 2.03; P = .044) very close to that in the original analysis of covariance. The intervention groups differed on QOL, particularly right after discharge from TCC treatment. While the difference is numerically small, the SD of the variable is also small. It is possible that small changes in QOL may be noticeable to the individual.
Mixed-Effect Regression Estimates and CIs for Mean Differences Between Groups at Each Follow-up Visit and Mean Differences Between Visits for Each Group.
Subjective Quality of Life Over Time by Treatment.
Regarding engagement in mental health services following treatment at the TCC, our primary measure of engagement, we examined data from the first call survey following TCC discharge. Approximately 61% of all participants were engaged in mental health services following discharge from the TCC. These numbers were nearly identical for both treatment groups, and the difference was not statistically significant (χ2 (1) = 0.45; P > .50). We also examined data including all that from individuals for whom no follow-up data on engagement were available in a 2 (EFC vs TAU) by 3 (engaged, not engaged, no information) χ2 analysis. The result did not change the conclusions, showing no significant associations between group and engagement category (χ2 (3) = 1.04; P = .59).
As a secondary measure of treatment engagement, we compared the proportion of scheduled TCC visits that were kept while in treatment at TCC by group (no data on people who dropped before TCC). The results revealed that, in TAU, 80.4% (SD, 27.5%) of all appointments were kept; in EFC, 85.4% (SD, 22.7%) were kept. This difference did not reach statistical significance, but there was a strong trend in 1-tailed testing, which was more in line with the hypothesized results (z = −1.6129; P = .054).
For the subset of participants recruited from inpatient units, we examined the likelihood of presenting to TCC intake by group. Participants scheduled for individual intakes in TAU attended intake in nearly identical proportion to those scheduled for Access group intake (χ2 (1) = 0.11; P > .73).
First assessment scores on the CPS are presented in Figure 3. Results indicated that about 26% of participants wanted mostly active roles in the decision-making process and less active roles for the input of prescribers. Roughly 22% wanted a primarily passive role, with the doctor having the largest say in care decisions. Most individuals, about 52%, wanted a primarily collaborative process with a role fairly equal to that of prescribers. Examined another way, these data suggest that 91% of our participants in transitional care wanted to have at least some say in decisions about their treatment.
Shared Decision-Making Preference at Initial Assessment.
We examined changes in the CPS from first assessment to end of treatment at the TCC (discharge assessment) only by categorizing individuals into 2 groups: (1) unilateral decision-making (wanting to make all or most decisions themselves or wanting the doctor to make all or most decisions); or (2) collaborative decision-making (categories in which both the provider and the consumer had a say in decisions). In the χ2 analysis (group by assessment period), there were no significant differences from first assessment to follow-up regarding these preferences (χ2 (1) = 0.03; P > .78). Moreover, using general linear models for repeated measures, we found no group differences over time by treatment (χ2 = 0.07; P =.75).
In addition, we were interested in whether consumers and providers would view communication during visits more similarly on the MPI over time and by treatment group. The mixed-effects regression model examining mean difference on the MPI between consumer and provider on ratings of in-session SDM parameters indicated a nonsignificant effect of group (F1,272 = 0.41; P = .52), a significant effect of time (F6,437 = 3.70; P = .002), and a nonsignificant group by time interaction (F6,437 = 0.27; P = .55). An inspection of means shows that, in both groups, consumers' and providers' ratings of how the session went became more similar over time (see Figure 4).
Matched-Pairs Difference Score Over Time by Treatment Group.
We investigated changes in symptomatology over time by treatment group as an important background against which to interpret other study results. Results of mixed-effects regression on the BPRS total score indicated no significant effect of group (F1,216 = 0.12; P = .72) or group by time interaction (F1,333 = 0.04; P = .84), but a significant effect of time (F1,133 = 3.95; P = .48). Inspection of means indicates that participants improved in terms of symptomatology over time while in treatment at the TCC, regardless of group. We ran no further analyses on this variable because we found no significant group differences.
Individuals at the TCC rated satisfaction as very good. In TAU, overall treatment satisfaction was 6.4 (SD, 0.87), and in SDM, it was 6.1 (SD, 1.0). The results of a Wilcoxon rank sum test indicated that this difference was statistically significant (z = 2.85; P < .004).
Of individuals, 96% in EFC and 90% in TAU reported intent/desire to keep their post-TCC appointments for mental health care (Fisher exact test not significant, P = .19).
In examining crisis visits, hospitalizations, arrests, and incarcerations in the 6 months after discharge from the TCC, we counted the number of these unwanted events and examined differences between groups as a function of days covered using the log of exposure days. Individuals in EFC had a total of 118 events with a log exposure of 9.3, and those in TAU had a total of 76 events with a log exposure of 8.9. This difference was not statistically significant (χ2 (1) = 0.001; P = .9507).
Individuals in EFC saw greater improvement in a subjective measure of QOL than did those in TAU; the effect size was 0.28, a small effect clinically. It makes sense that getting into treatment rapidly and helping people communicate better with their provider using an SDM coach could improve how people think about the quality of their life during treatment in transitional care. This may be related to issues such as locus of control or the feeling of empowerment. Improving empowerment through SDM is consistent with preferences expressed by individuals in transition from inpatient settings to care in the community.13 This difference in QOL could not be attributed to differential symptom change within the 2 treatment groups. In other words, QOL differentially improved, but symptoms did not. Both groups improved over time in terms of level of symptomatology, and we found no treatment group differences.
It is important to note that, while treatment satisfaction at the TCC in general was >6 on a 7-point scale, individuals in SDM had slightly lower treatment satisfaction. It may be that focusing on SDM and on how to get what they wanted out of visits with a prescriber created slightly higher expectations of the visit and more awareness of dissatisfaction when the provider did not engage as desired.
We found no differences in engagement for individuals who received EFC vs TAU on our primary measure of engagement: attending treatment following discharge from the TCC. Other engagement variables also did not differ significantly among groups. Lack of group differences may be because our TAU at the TCC focuses on engagement and providing a warm hand-off to follow-up services.13 Typically, we did not discharge individuals until they had an appointment scheduled with a prescriber (ie, just intake with no follow-up was not sufficient). We found it interesting that virtually identical numbers of individuals showed up for group and individual intake when study recruitment took place at the hospitals. This suggests that a group intake appointment may not be a disincentive for people to attend an initial follow-up visit after hospitalization, and the group process allows more rapid scheduling, accommodating more individuals. Access group could be an important model for overburdened services and should be further investigated. Results show that patients attending a group vs individual intake had similar symptom levels and QOL.
Initial preferences for SDM varied considerably among individuals; however, even in this population in postacute transitional care, the majority of those in treatment wanted to be either equally involved or have the primary say in decisions about their care. Additional research on ways to make this participation possible is important. Almost no research has focused on tailoring SDM coaching to the type of decision-making preferred by individuals with BHCs.
About two-thirds of our participants used SDM when available, but session times had to accommodate people's busy schedules and so occurred either before or after visits, for between 15 and 60 minutes, depending on such factors as time, transportation, and content of session. The dose of SDM in the study was thus quite low, with patients who did receive SDM spending only about a half hour, on average.
It is also interesting that consumers and providers had more discrepant views about their dyadic communication during treatment sessions early in the course of treatment and that they tended to view sessions more similarly after more time and more contact with each other. This is possibly a consequence of getting to know each other and becoming more comfortable with the other person's communication style. Alternatively, this could be due to the attrition of the sample over time. People who like their providers and feel communication is improving may be more likely to stay in treatment. Recall that the goal of the clinic was to transition individuals to long-term care in the community when space opened for them. Therefore, individuals received as few visits as needed to provide care until transition could take place.
Overall, results with EFC did not significantly differ from those with TAU in our TCC—except with respect to subjective QOL. The TCC is a state-of-the art treatment program focused on engagement. The program provides a level of evidence-based TAU care and outreach that is in many ways superior to the standard of care in typical community settings. This likely made detecting differences between our TAU and EFC very difficult.
Empirical studies have shown that involving patients in the treatment decision-making process may lead to increased satisfaction with treatment. In addition, studies suggest that SDM leads to lower decision conflict, as patients are better informed, do a better job of following through on treatment recommendations, and see improved outcomes, including medical markers such as blood pressure and blood sugar; however, evidence on this topic is not consistent.51-59 Regarding QOL specifically, one study found weak evidence suggesting that SDM could improve QOL in cancer patients.60 A retrospective study by Andersen and colleagues demonstrated that in survivors of breast cancer, perceived involvement in decision-making was correlated with health-related QOL.61 We were unable to find any studies of SDM that examined QOL as an outcome among patients with BHCs. The results of the current study are likely to be useful for providers of transitional care for individuals with significant BHCs who have been recently discharged from hospitals or emergency services. Previous research has reported that as many as 40% of individuals with BHCs do not attend any outpatient visits in the 30 days following discharge.1 Studies show that problems in continuity of care are associated with poorer outcomes and higher health care costs.2 Engaging individuals in treatment posthospitalization is important for producing the best outcomes and controlling the costs associated with frequent hospitalization, which are estimated to total more than $2 billion annually.62
EFC could be utilized in outpatient mental health settings. The group intake process makes sense as a method of getting people rapidly into treatment during this critical transitional phase. It is no more expensive to complete a group intake, and it did not make a difference in terms of symptoms or QOL. The group component is particularly applicable for underfunded service areas with limited capacity for taking new patients immediately after discharge. Certainly, SDM would be applicable in multiple transitional services, but more needs to be understood about the individuals' desires for SDM and how best to encourage this behavior.
Using a group intake format such as that used in EFC can be easily implemented with current staff in outpatient settings. If 10 people come to intake, they can be seen by 2 to 3 staff members in a 1- to 1.5-hour process. This arrangement is similar to 10 individual half-hour appointments, which would keep 2 people busy for 2.5 hours. The process allows staff members to provide appointments based on the needs of each person. This makes better use of scarce prescriber resources. SDM coaching in many states can likely be billed as skill building even if it is provided by a certified peer specialist. The TAC-Review model, while developed for those with BHCs, could conceivably be transferrable to other types of consumer-provider interactions. TAC-Review is particularly applicable to those situations in which there is no clear “best” treatment and many choices with variable side effect profiles exist.63 This is the case in the treatment of mental health issues as well as of many physical conditions, including various cancers, diabetes, allergy, and hypertension.64-66 In fact, participants in our study who had comorbidities stated that SDM coaching, received as part of mental health treatment, improved their use of SDM with non-mental health providers.
We did not have a large enough sample to investigate comparative effectiveness of TAU and EFC in subpopulations defined by specific risk factors and comorbidities.
We did identify some limitations in this study of individuals with BHCs who had just been hospitalized or discharged from emergency services. We conducted the study in only one geographic area, so caution is warranted when applying the model to other locations. While planned, the randomization of individuals before one component of the intervention (group vs individual intake) did not allow for a true baseline. It is possible that participation in the intake process could have influenced the first assessment; however, we found no differences between groups on outcome variables at the initial assessment. It would be difficult for other investigators to replicate the study findings because a true intention-to-treat analysis could not be reported, given that patients who had been randomized were lost to follow-up and therefore not assessed. We had difficulty following up on this sample of individuals with BHCs, posthospitalization or post-ED use, for multiple reasons, including homelessness and transient telephone service.
The high dropout rates may have impacted the findings. Many participants dropped out before the first assessment. Missing data on follow-up assessments could have increased the risk of bias in the comparative estimates for outcome variables. We tried multiple statistical approaches to examine the impact of missing data where we found significant group differences in our proposed analyses. The high level of standard care in the university-run TCC was likely not comparable to that in traditional community settings. We did not examine the potential mechanisms of the EFC intervention, such as empowerment, which may have influenced QOL in the EFC group.
Systematic reviewers have found in meta-analysis that modified intent-to-treat analyses, such as the kind we performed, produce larger intervention effects than do true intent-to-treat analyses.67 It may be that the apparent benefit of EFC on QOL may be partially explained by the extensive loss to follow-up and the modified intent-to-treat analysis of available cases. Finally, the benefit of SDM skills may be larger if patients have an opportunity to practice those skills immediately. In the current study, some coaching sessions occurred after patients' visits with the doctor, and patients may not have learned the skills as effectively. Active learning and deliberate practice are strong determinants of retention. To mitigate this issue, we did provide written handouts to help with retention.
Future research will be needed to determine the value of EFC, using a sample of patients not in transition, where randomization could more properly occur after a baseline assessment. In addition, it will be important to examine the impact of group intake on resources and to tailor SDM level of involvement to the preferences of the people in treatment. If larger studies with randomization before any assessment indicate the need for follow-up, then researchers might need to study the implementation of access intake group and SDM coaching in EFC.
We sought to examine the impact of EFC, which included a group intake designed to lead to rapid access to services and an SDM coach vs usual care in a TCC treating individuals with BHCs immediately following discharge from hospital and EDs. Overall, results with EFC did not differ significantly from those with TAU—except regarding subjective QOL.
EFC that includes group intake and SDM coaching may improve QOL for individuals with BHCs who are transitioning from hospital to community care. This is a particularly relevant model in systems where there is a provider shortage. While our findings can inform practice, their limitations include that we conducted the study in a limited geographic area and that follow-up of individuals was very difficult, leading to a high rate of dropout and missing outcome data.
Research reported in this report was [partially] funded through a Patient-Centered Outcomes Research Institute® (PCORI®) Award (#IH-1304-6506) Further information available at: https://www.pcori.org/research-results/2013/helping-patients-mental-illness-engage-their-transitional-care
Velligan DI, Fredrick M, Sierra C, et al. (2019). Helping Patients with Mental Illness Engage in Their Transitional Care. Patient-Centered Outcomes Research Institute (PCORI). https://doi.org/10.25302/8.2019.IH.13046506
The [views, statements, opinions] presented in this report are solely the responsibility of the author(s) and do not necessarily represent the views of the Patient-Centered Outcomes Research Institute® (PCORI®), its Board of Governors or Methodology Committee.
This book is distributed under the terms of the Creative Commons Attribution-NonCommercial-NoDerivs License which permits noncommercial use and distribution provided the original author(s) and source are credited. (See https://creativecommons.org/licenses/by-nc-nd/4.0/
Your browsing activity is empty.
Activity recording is turned off.
See more...
