Veterans’ Use of Mental Health Services

Principal Investigator: Carole Siegel, Ph.D.; Co- Investigators: Shang Lin, Ph.D., Joseph Wanderling, M.S., Ellen Weissman, M.D. (Department of Veterans Affairs)
        

• To examine the types, amounts and costs of mental health services that veterans receive from the Department of Veterans� Affairs (VA) and from other public mental health systems; and
• To explore explanatory reasons for the differential use of these systems.

Rationale: Veterans obtain mental health services from both the Department of Veteran Affairs (VA) system of health care and/or the non-federal mental health system. This project was initiated in response to concerns by the New York State Office of Mental Health (OMH) and VA policy makers about veterans' service utilization patterns for each system. The types, amounts and costs of mental health services that veterans receive from the Department of Veterans' Affairs (VA) and from other sectors are hypothesized to vary depending on Federal and State policies that determine mental health spending. As spending priorities and coverage guidelines change, it is important to be able to assess the cross-system impact.

Phase I of the study proposed, using data for the study period 1988 to 1997 in New York State (NYS), to:

• Describe the veterans who are users of
• mental health services in the VA system
• mental health services provided, funded or certified by the NYSOMH;
• Estimate on a county level the number of persons using either or both systems for mental health services; 
• Develop an area level explanatory model of differential sector use in terms of area descriptors and VA and OMH service and funding policies.

In Phase II the study's focus was to:

• Determine why veterans select and use either or both systems, through direct interviews with veterans who are mental health service users.

Phase I of this study received funding from the VA through a funded VISN-3 application for a Mental Illness Research and Education Clinical Center (MIRECC): Maximizing Recovery for Veterans with SMI by Bringing Research into Practice, Larry Siever, M.D., Director.

The protocols of five similar "overlap" studies were reviewed in order to ensure the uniqueness and relevance of the present study. Prior work has been conducted by one of the investigators (RR) on the factors that affect access to VA mental health services (Rosenheck & Phil, 1997). Another investigator (CS) has conducted work on estimating the size of a population based on a cross-sectional survey and will be applied to data from the Patient Characteristics Survey of NYS, a primary data source for this study (Siegel et al., 1997; Laska et al., 1996, 1988).

RESEARCH ACTIVITIES AND RESULTS

Method: The study population is New York State veterans who were users of federal or non-federal mental health services in alternate years between 1988 and 1999. These veterans will be described in terms of the proportion of the total number of veterans they represent, their demographic and clinical characteristics and the types and costs of mental health services received in the year. The study year coincides with the OMH Patient Characteristic Survey time frame. The Laska-Meisner-Siegel (Laska et al., 1988) methodology will be used to estimate from the PCS cross-sectional sample, the annual number of veterans within broadly specified client subgroups receiving services of a given type. Annual costs will be based on the utilization data of the one week sample imputed up to the annual level, where programmatic cost data are obtained from the OMH Consolidated Cost Report Form. For VA mental health service users client level data on service use and characteristics and cost data are available from VA administrative files.

For county level estimates of the number of persons using both systems: These numbers will be obtained on a county level (for each of the 62 counties in New York State) for client subgroups based on ethnicity, age and broad diagnostic categories and for each of the study years. No will be annualized using the LMS method (Laska, et. al., 1966). The estimate of Notv will be based on the birthdate methodology (Larsen, 1994; Banks, et. al., 1996), merging the list of one-week OMH users with the annual list of VA users.

We then estimate the number of veterans who use only OMH services in the year, the number using both systems, and the number making use of either system.

An explanatory model of sector usage in terms of county level factors:

The vector of rates (To/T, Tv/T, Tov/T) is a measure of differential sector use. The aim is to examine county-level factors expected to affect these rates. A logic model of influence will guide the selection of variables hypothesized to impact on sector and overlap usage. These will include time-dependent variables of service availability and expenditures of each of the sectors, endogenous county characteristics reflecting economic and social characteristics, and policy impact variables. The user characteristics are also hypothesized to affect usage including location of inpatient services measured by the average distance of VA county residents to the closest State hospital, and to the closest VA hospital; VA and non-VA mental health expenditure variables; VA and non-VA service availability as measured by the density of non-federal mental health providers in an area; and policy variables chosen to reflect managed care introduction and penetration. Candidates for county factors include rural/urbanicity and socioeconomic conditions. Since we will have overlap counts for ethnic/ age groups, and broad diagnostic groupings, we will be able to use these data in the modeling process described below. In addition new methodology for subset analysis under development in the Methods Core (Project: II-2) may also be applied.

A polychotomous weighted logistic regression model will be used to examine the explanatory value of the factors accounting for the variation in VA and OMH usage rates. Even though the rates are on a county level, to take account of different population sizes in the counties, the unit of analysis will be, in effect, a veteran and, in the same county each will be described by the same county factors. The dependent variable will be a three category variable indicating whether the veteran receives services only from the VA; only from OMH, or from both systems.

We will also be able to examine sector usage in terms of models run for each of the study years. Significant variables in each year's models as well as the similarities and dissimilarities in the set of significant explanatory variables across years will provide insight into policy impact on sector usage. An additional time series model will be fit in which all observations over time are simultaneously used in a generalized linear model with a logistic link function. This type of model will allow us globally to test hypotheses concerning the impact of policy variables. Thus, for example, to examine the hypothesis that the penetration of Medicaid managed care will increase VA sector usage, we would expect the rate of VA use to be positively (and significantly) related to the increase in the penetration of managed care. A caveat of analysis is that these rates are based on the usage in a two-week sample. Missing are the sporadic users. However, if we assume that sporadic users are uniformly distributed across counties, the relational analyses that are performed should not be biased.

Results:   

Population Counts by Year

p---VA                      �---OMH Annualized                     l---Matches

Service use patterns were generated for 1995, 1997 and 1999. The refined birth date methodology was used to estimate overlap usage in each system and to improve the precision of the overlap (matches) estimates.

In the VA, male usage slightly increases over time while in OMH, there is a slight downward trend. For females in the VA, usage doubles from 1995 to 1997, while in OMH, there is a considerable decrement in usage. Whereas in 1995, OMH usage by females was almost 2.5 times higher than that in the VA, by 1999 there is similar usage. Overlap usage rates are approximately 4.5% for males whereas for females they trend upward over time from 2 to 5%. The elderly are less likely to use multiple systems and approximately 4 times more likely to use the VA than other mental health providers (in comparison to 2 times as likely for those less than 65).

There is considerable variation in usage rates over counties. Explanatory county level variables for greater VA usage than OMH usage, based on a multinomial regression include having a VA hospital (1995, 1997), lower median income (1999), being rural (1997,1999) or upstate urban (1999) and to some extent lower expenditures for non-VA services per capita (1999).

Use of both systems in comparison to the VA only was predicted by lower proportions of the population >64 (all years), rural (1995,1997), having a VA hospital (1999), greater OMH expenditures per population (1999), and greater percents nonwhite (1995,1997), greater expenditures of non-VA services per capita (1995,1999), lower median income (1999) and a greater % nonwhite and % vets over 64 (1999). A relative cost weighting methodology was used to compare costs of each system. In all years non-VA funds for inpatient services were greater than VA funds and per persons costs were also higher. Total outpatient costs were also higher based on non-VA funding in 1995 but essentially equal to VA costs in subsequent years. There was a slight increase in the number of persons seeking outpatient services in the VA over time and a slight decrease in those using non-VA dollars for these services. The number of vets that received community support services through the VA over time steadily increased as did expenditures for this service. Non-VA funded community supports to vets decreased somewhat over time. In reporting these data no statistical tests are made as the sample is total.

In September 2002, we presented our findings at a VA MIRECC retreat. Queries from the audience placed into question the validity of the comparisons being made between VA and non-VA services, specifically with regard to residential and ER services. Subsequently we discovered that residential data other than inpatient days had not been included in our original VA data file and that some VA Community Support Program services had been erroneously identified as residential services. We also learned that ER services for mental health treatment are provided by the VA but not specifically coded as mental health visits. The data sets and their analyses were re-analyzed taking these newly learned facts into account. The re-analyzed data are incorporated in the findings described above.

As spending priorities and coverage guidelines change, it is important to be able to assess the cross-system impact in terms of utilization patterns.

Results will be written up for presentation and publication.

 Inclusion of Gender and Minority Groups*:  
* Only data that are currently available filled in. 

Top

HOME

Research Cores: Negotiating Lives in Communities | Methods for MH Services Research | Systems Integration |� Promoting Recovery |
Research Topics: Homelessness | Instrument Development | Managed Care | Mental Illness & Chemical Abuse
Methods to Improve Services Research | Monitoring and Improving Service Delivery Systems
Multi-Cultural Issues | Recovery | Treatment Innovations