HSR&D Home » Research » PPO 13-121 – HSR&D Study
Identifying Innovations for Managing High-Cost Mental Health Patients
Daniel M. Blonigen, PhD MA
VA Palo Alto Health Care System, Palo Alto, CA
Palo Alto, CA
Funding Period: July 2014 - June 2015
Use of psychiatric emergency services - i.e., emergency departments [ED] and acute inpatient psychiatry - contributes substantially to the cost of care for VA mental health patients. Further, prior research outside the VA has shown that a small percentage of patients who utilize these services contribute a disproportionate share of the total cost ("high utilizers"). Despite their high cost, there have been no efforts in the VA to identify the patient characteristics of VA mental health patients who utilize these services repeatedly, or (2) formatively evaluate the care practices provided to mental health patients during and after their encounters in these emergency services. The current pilot project sought to obtain this information in order to identify the treatment needs of, and gaps in the quality of care for, this patient population and develop implementation projects to reduce the cost of VA mental health care.
Aim 1: Use VA administrative databases to identify the patient characteristics - i.e., demographic, psychosocial, clinical, service utilization, and medical - of high utilizers of psychiatric emergency services. To provide sufficient focus to the current study, psychiatric emergency services were defined in terms of ED visits with a mental health diagnosis.
Aim 2: Use the Consolidated Framework for Implementation Research (CFIR; Damschroder et al., 2009) to conduct a formative evaluation of care practices with high utilizers of psychiatric emergency services to identify barriers and facilitators to reducing this population's repeated utilization of emergency services.
A mixed-methods approach was employed to achieve the proposed aims. Aim 1: We used VA administrative databases to identify our study cohort for the patient-level analyses. Specifically, we selected all unique mental health patients treated in the VA in fiscal years (FYs) 2011-2012 (i.e., at least 2 or more outpatient visits with a mental health diagnosis, as defined by the International Classification of Diseases-9th Revision [ICD-9], or one inpatient stay with a mental health diagnosis), which identified 1,790,325 unique mental health patients. Out of this initial sample, we then selected those with at least one ED visit during FYs 2011-2012 that was associated with a mental health diagnosis (n=226,217; 13% of our original cohort). Ninety-five of these patients were ultimately excluded because their vital status records indicated a data of death prior to the study period. This resulted in a final cohort of 226,122 unique patients that were included in our Aim 1 analyses. Subsequently, we identified those who were "high utilizers" of emergency services (i.e., five or more ED visits during any one of the aforementioned fiscal years), and tested demographic, psychosocial, and clinical variables as predictors of membership into this high-utilizer group.
For the Aim 1 analyses, the primary data source was from the National Patient Care Database (NPCD) Medical SAS outpatient files in FYs 2011-2012, which provided visit dates, diagnoses, procedures, and other patient-level variables of interest. Non VA care which is captured via the Fee basis files and contains information on VA-paid care provided to Veterans seen at non-VA facilities was also used to capture additional veteran mental health patients that had used ED services in the community. The VHA Medical SAS inpatient Patient Treatment Files (PTF), which captures inpatient stays, was also used to obtain information for variables of interest. Patient-level data on demographic, psychosocial, and clinical characteristics was constructed from the NPCD SE files, PTF, and MCA outpatient and inpatient pharmacy files. For demographic variables, information that was missing in the SE files was obtained through the Vital Status database.
A number of patient-level variables were extracted from the above-noted VA administrative databases and examined as predictors of membership into the high-utilizer group. These variables were selected based on conceptual or empirical support as being a correlate of frequent psychiatric emergency utilization. For the sake of parsimony, the details on how some variables were constructed (e.g., specific stop codes) were omitted from this report, but are available upon request from the PI: (1) sociodemographic variables (age, gender, race, ethnicity, marital status; period of service [OEF/OIF vs. not], geographic area [highly-urban, urban, highly-rural, and rural], driving time to the closest VA facility, and patient's copay status [yes/no]); (2) psychosocial variables (homeless [yes/no] during the study period, housing stability [number of times a patient's zip code changed during the study period], and criminal-justice involvement [yes/no] during the study period; (3) clinical variables (ICD-9 diagnoses of psychotic-spectrum disorders [e.g., schizophrenia]; bipolar disorders, depression, acute stress disorders, major depressive disorders [MDD], anxiety disorders, posttraumatic stress disorders [PTSD], substance use disorders [SUD], adjustment disorders, personality disorders, dissociative disorders, conduct/impulse control disorders, somatoform disorders, attention deficit//hyperkinetic disorders, suicidal ideation and homicidal ideation. Separate dichotomous variables were created [yes/no] for each diagnosis. The diagnoses were coded present only if associated with at least two outpatient or fee-basis mental health encounters in a fiscal year, an approach that has been shown to improve the positive predictive value of mental health diagnoses in VA administrative data); (4) service utilization (detoxification from alcohol and/or other drugs; total number of mental health encounters during the study period; use of opiates (long-acting [yes/no], short-acting [yes/no], and tramadol only [yes/no]); medication possession ratios for antipsychotics among patients with a psychotic-spectrum diagnosis; and number of missed mental health appointments during the study period; (5) chronic conditions (total number of chronic medical conditions, identified by the Health Economics Resource Center and the Program Evaluation Resource Center). Subsequent to our construction of the patient-level database, we conducted univariate ANOVAs to identify which patient-level predictor variables were significantly different between the high-utilizer and non-high utilizer groups in our cohort and obtain descriptive data on prevalence of patient-level predictors in these sub-groups.
Aim 2: To address Aim 2, we conducted interviews with a range of administrators (e.g., ACOSs for Mental Health; ED Directors) and front-line providers (ED social workers, care coordinators, psychiatrists, psychologists, and nurses) from psychiatric emergency services in the VA. Specifically, we randomly selected 10 VA facilities from both the bottom quartile (LOW) and top quartile (HIGH) in terms of ratio of high utilizers of psychiatric emergency services to total number of unique mental health patients treated in FY2012. To ensure that interviews were conducted with a range of facilities, within each quartile, we sought to interview at least one key informant from each of five Facility Complexity Levels. All potential participants were contacted via email and volunteered to participate in the 45-minute interview. All participants were phone consented prior to the audio-taped interview. A total of 31 key informants (14 LOW, 17 HIGH) were interviewed from Jan-June 2015 on issues of continuity of care (i.e., care coordination; discharge and continuing care planning) using CFIR domains related to context of care and characteristics of the individuals involved in the delivery of care.
Interviews were de-identified, transcribed by a non-VA contractor, cleaned, and entered into ATLAS ti. v.7, and then coded by the Project Coordinator. To identify themes in the textual data related to barriers and facilitators to reducing repeated utilization of psychiatric emergency services, we utilized a modified version of Neal and colleagues' (2015) approach to rapid analysis of field notes. During each interview, as well as immediately after, the Project Coordinator took detailed notes using a template to document themes reported by the respondent related to facilitators and barriers to reducing frequent utilization of psychiatric emergency services among VA mental health patients. Six domains were specified a priori to organize the identified themes into barriers and facilitators at the patient-, provider-, and systems-level. These domains were operationalized in a codebook that was developed jointly by the Project Coordinator and PI. Once the interviews notes were completed, the Project Coordinator copied the notes into a matrix in order to compare and contrast the themes within each domain (columns) and across participants (rows). The matrix was organized into the aforementioned domains and a brief summary was used in each cell of the matrix to describe the theme. After the Project Coordinator created the matrix, the PI and project coordinator independently reviewed the matrix and interview notes, and then each created a list of meta-themes. The PI and Project Coordinator then meet to compare their independent list of themes and then engaged in a consensus process to categorize the meta-themes into barriers and facilitators at the patient-, provider-, and system level.
To date, we have two main sets of findings, which correspond to the specific aims of the project. Manuscripts are currently in preparation; however, some analyses are ongoing (e.g., multivariate modeling), and therefore the findings reported below are preliminary.
Aim 1: Out of our total cohort of unique mental health patients who had at least one ED visit with a mental health diagnosis from FY 2011-2012 (n=226,122), 20,706 (9%) patients had five or more ED visits during the study period and were defined as "high utilizers" in our sample. The majority of patients were not married during the study period (67%), and were non-Hispanic (88%) white (68%) males (90%) with a mean age of 51 years. Most of the patients resided in an urban region (66%) and were relatively close to a VA facility (Median driving time was 14 minutes). Marked differences with respect to the above demographics were not seen between high utilizers and non-high utilizers. Most patients were Veterans of prior wars (85% not OEF/OIF). A little less than half of the patients had housing instability (45%), while 31% had been homeless at one point within the study period. Only 7% of the patients had legal problems. There were higher frequencies of patients with housing instability (moved 4+ times (16% vs 3%), homelessness (65% vs 28%), and legal problems (15% vs 6%) among the high utilizers. There was a higher usage of opiates in the high-utilizer group (65% vs. 55%) with the short-acting opioids used most frequently. Only a minority of patients had received detoxification services (16%). However, out of the patients that had detoxed within the study period, most had been high utilizers of the ED (46% vs 13%). The median number of outpatient mental health visits during the stud period was 14, though significantly higher for the higher utilizers (44 visits) than the non-high utilizers (12 visits). The most frequent mental health diagnosis in this cohort was MDD (65%) followed by major alcohol use disorders (47%) and major anxiety disorders (45%). PTSD was diagnosed in 41% of the cohort. These conditions were found at higher frequencies in the high utilizer group, especially for the alcohol/drug disorders. Major psychotic disorders were nearly twice as prevalent in the high utilizer group. Among the chronic medical conditions, the most common conditions were hypertension (56%), lower back pain (46%), injury (41%), acid-related diseases (33%), and arthritis (28%). Each of these conditions was seen at a higher frequency in the high-utilizer group.
Aim 2: Qualitative analyses revealed a number of robust themes, in terms of barriers and facilitators (at the patient-, provider-, and system level) of reducing repeated psychiatric emergency service utilization. Our analyses did not suggest any marked differences in terms of either barriers or facilitators by quartile (LOW vs. HIGH), facility-level complexity, or type of key informant (administrator vs. front-line provider): Barriers: (patient-level) non-compliance with treatment recommendations, particularly engaging in SUD specialty care; and difficulty tracking and following up with patients due to homelessness/transiency; (provider-level) negative perceptions or "stigma" attached to high utilizers by providers leading to challenges in aligning with patients; and lack of communication and coordination between medical and mental health professionals in the ED; (system-level) lack of intensive case management services for mental health patients with long-term substance abuse problems; and insufficient and poorly coordinated discharge planning, particularly for homeless veterans; Facilitators (patient-level) no robust themes were identified in this domain; (provider-level) a recovery-oriented approach to care, focused on relationship building with patients and identifying their life goals beyond symptom management; (system-level) interdisciplinary care-coordination teams within EDs; intensive case management within EDs, including identifying patients who are at risk of high utilization; peer mentors to facilitate engagement with outpatient services; and psychiatric observation beds within EDs.
Knowledge gained from the proposed project informed the development of a recently submitted Partnered Evaluation Center proposal focused on the implementation of peer-based interventions to increase engagement in SUD and mental health care among homeless veterans with dual-diagnoses. Specifically, our findings related to the significance and challenges of addressing homelessness among high utilizers informed the decision to include a specific aim in that proposal to examine facilitators and barriers of VA clinical staff's use of the recently developed "Hot Spotter" program. This program, developed within the Office of Homelessness, includes algorithms and dashboards to inform care decisions on interventions that target Veterans at especially high risk of homelessness. In addition, as we finalize the manuscripts based on this work and present the findings to other VA researchers and providers in the coming months, we will help to inform the emerging national efforts of PACT intensive case management teams focused on the care of high-risk/high-need patients, nearly half of whom have a mental health diagnosis.
External Links for this Project
NIH ReporterGrant Number: I21HX001338-01A1
Dimensions for VADimensions for VA is a web-based tool available to VA staff that enables detailed searches of published research and research projects.
If you have VA-Intranet access, click here for more information vaww.hsrd.research.va.gov/dimensions/
VA staff not currently on the VA network can access Dimensions by registering for an account using their VA email address. Search Dimensions for this project
DRA: Mental, Cognitive and Behavioral Disorders
MeSH Terms: none