Community Support

the Evaluation Center@HSRI Toolkit

Estimating Per Unit Treatment Costs for Mental Health and
Substance Abuse Programs

V. DEVELOPING A FORMAT FOR SUMMARIZING SERVICE USE AND COSTS (STEP 3)>

1. Linking Service Use Data With Per Unit Cost Estimates

To provide an example of how to summarize client treatment costs, assume that you have collected data on service use over a one-year period. After the data on services used for each client are collected and all the necessary per unit costs have been estimated, set up a table in the format illustrated in Table 9. Worksheet software is very useful at this stage. It is not hard to see that this format permits the data to be summarized in a variety of ways, most importantly, the total cost to treat each individual. In the next section we will expand on the uses of these data and how they might be used in a cost evaluation.

Computerized spreadsheets are very useful, but the size needed to carry out the work is not easily reproduced on the printed page. A more realistic picture of the many different services provided a single individual over one year is summarized below. We provide it in its entirety to remind readers that the number of services available to consumers covers a lot of territory and spreadsheets need to be organized accordingly.

For example, in an 18 month study, ID#001 used 35 hospital days in two different hospitals, 13 case-management visits, 39 group therapy visits, 12 medication visits and 2 ER visits. The consumer lived for 9 months in supported housing, for 1 month in a shelter and for 8 months in a market apartment.

20 psychiatric hospital days @$597   $11,940
15 state hospital days@$605   $9,075
13 case-management visits @$78.50   $1,020
39 group therapy visits (1 hour) @$6.68   $261
12 medication visits @23.59   $283
2 ER visits @$98.00   $196

For housing costs
9 months, supported housing @$715   $6,435
1 month, shelter @$124   $124
8 months, shared market apartment@650   $2,600
8 months utilities@$30   $120

Sometimes it may be wise to group certain types of services together rather than report each separately to improve generalizability of the cost data. This was the case in our study of formerly homeless and mentally ill adults in Boston. Participants in this study were admitted to different hospitals with very different per diem rates, based on the area of town where they lived. After the per diems were calculated for each hospital, the per diem costs by private psychiatric hospitals, general hospitals and state hospitals were aggregated to yield the average per diem cost for each type of institution. (Before aggregating hospitals into these three categories, the per diems for individual hospitals ranged from $432 to $1,007). This cost averaging was done because clients were assigned randomly to houses within different catchment areas. The particular hospital a client went to was determined, at least in part, by the catchment area in which s/he lived. If hospital-specific per diems had been used, treatment cost for a client would be significantly influenced by the catchment area of the house to which the client happened to be assigned. This could have biased the comparison of costs between housing types. Each episode length of stay was then multiplied by the appropriate hospital-type average per diem.

Once the cost of treatment, by service type, is available for each person in a study, it is not hard to imagine the next steps: summarizing costs by service types, by client types or by provider types. In the Table 10 below, we provide a sample table that summarizes the data (not the data from the above worksheet) by general service use types and by state agency payer. This illustrates how service units, when converted to expenditures, allows the display of expenditures in simple descriptive summaries that reveal a lot about the clients in the study.

In Table 10, we provide four types of summary statistics: (1) the total number of users, (2) the total expenditures for a particular type of treatment (such as inpatient admissions), (3) the mean user expenditure for that treatment type, and (4) the mean expenditure for all clients (the sample mean) in the study, not just those who used the service. The TOTAL row provides information about everyone (19,194) studied. In this row, the last two columns are the same because the denominator (19,194) for users and total sample are the same.

Costs can also be categorized by payers, such as the cost to a particular state agency, the cost to state government, cost to federal government, and the cost to tax payers (which would be a sum of federal, state and local expenditures).

2. Summarizing Services

Common temporal categories are:

The month - as in the cost of x service per month
The episode - as in the cost of x service per acute hospital episode of schizophrenia

Common person categories are:

The individual - as in the cost of x service per person per month
Age groups - as in the monthly cost of x service for persons between the ages of 19 and 64
Disorder groups - as in the monthly cost of x service for persons with schizophrenia

Common categories for aggregations of persons are:

The program - as in the cost of community support program services per person per month
The health plan - as in the mean cost of services provided by a particular agency or health plan
The system - as in the annual cost of all public mental health services for persons with schizophrenia.

3. Grouping Services

Cost evaluations are rarely done for a single discrete audience. Reporting cost evaluations is helpful to both administrators and policy-makers, but language that is particular to a single health plan or program may make information sharing difficult. Once services have been defined, it may be possible to group some services with different names together as being approximately equivalent. It is particularly appropriate if these services are known to have similar unit costs.⁷

In addition to making cost information more accessible to a wider audience, it has the advantage of forcing one to think through the cost implications of programs and improves the conceptual clarity of the study. Grouping conceptually similar services, is also a way of summarizing the data for services that only a small number of persons receive. For example, we have often found that a small percentage of study subjects has a particular treatment type that no one else has but a small N makes it difficult or impossible to analyze separately. However, by dropping these data out of the study, valuable information would be lost.

However, if the service use data can be aggregated up into a particular group of services, the data will not be lost. This does not mean cost all the data in a particular group using only one cost per unit, rather that the cost of all services that fall within a group can be reported together.

Services can be grouped in any way that makes sense within the context of the study, but we have found that a six level grouping process works well. We use these groupings to provide a structure for how the study is conceptualized, for organizing tables when we present the findings (see Section 4) and for making decisions about the level of effort needed to determine the scope of a particular for the study. The levels we use are:

Inpatient
Residential
Outpatient Treatment
Psychosocial Rehabilitation
Other Support Services
General Medical Treatment

Table of Contents | Previous | Next