Bethesda Hyatt
Bethesda Maryland
December 13-14, 2007

NIDA Organizer:
Sarah Q. Duffy, Ph.D.

Meeting Purpose and Intent:

Cost function analysis is a method used by economists to examine the costs of production in a particular industry. Derived from the Neoclassical Economic Theory of the Firm, it relates firm-level data on costs to data on output levels, input prices, and other factors and can be used to determine whether or not an industry exhibits economies of scale, how factors such as client mix affect costs, and whether or not specific firms produce their outputs efficiently.

Although cost function analysis has been used extensively to examine many other industries, it has been used only rarely to examine substance abuse treatment costs. Knowledge from carefully conducted cost function studies is critical to the development of reimbursement, licensing, and performance management policies and practices designed to promote the efficient provision of high-quality treatment services.

The meeting had the following goals:

• Review existing research on cost function analysis in substance abuse treatment and other health care settings.
• Describe the salient characteristics of the current substance abuse treatment system that would inform cost function analysis.
• Identify how cost function analysis could be used to inform policymakers and others responsible for funding or otherwise promoting high-quality, efficient substance abuse treatment for all those who need it.
• Identify gaps in theoretical and applied research, methods and data collection that may currently hinder the ability of cost function analysis to answer relevant research questions.

Brief Discussion of Meeting Outcome:

Participants presented the state of the science revealing the following:

• Cost function analysis relates a firm's costs to its output and input prices in a way that is consistent with the Neoclassical Economic Theory of the Firm. Under certain assumptions, many of which are testable, the cost function serves as a dual to, and can adequately describe, the underlying production process. Efficiency analysis is a related technique that can be used to measure several types of inefficiency in production compared to best practices, decompose inefficiency into its constituents, and explain the sources of inefficiency.
• Although there have been scores of cost function analyses examining other health care settings and other industries, only three published studies exist on cost function analysis of substance abuse treatment facilities.
• These three studies suggest that there are significant economies of scale in both the outpatient drug-free and methadone treatment modalities, but that client and facility characteristics have not uniformly explained variations in the observed costs. However, these studies were not fully informed by economic theory nor did they incorporate emerging practices in conducting these analyses for other health care settings such as hospitals and nursing homes. Furthermore each of the three studies also used different data sources, dependent variables, functional forms, model specifications and measures of output further limiting the strength of the results
• Substance abuse researchers are ahead of many other fields in developing and refining structured instruments, such as the Drug Abuse Treatment Cost Analysis Program, Substance Abuse Services Cost Analysis Program, and Treatment Cost Assessment Tool, and other methods for costing treatment and collecting accurate cost data including both accounting and economic costs. Many of these instruments are capable of providing useable information to programs, both aiding in the provision of services and increasing provider participation in data collection. However, data collection efforts mostly have been piecemeal.

Participants described the following research, policy and oversight questions that might be informed by cost function analysis:

• What are the relationships between costs and quality of care, and between cost and casemix?
• What are reasonable reimbursement rates for substance abuse treatment and how do they differ by modality (e.g. outpatient, intensive outpatient) and setting (e.g. primary care vs specialty care) and for specialized populations (pregnant women, adolescents, SSI clients, clients from different cultures)?
• Are there economies of scale in the production of treatment for difficult-to-treat clients, for example SSI clients and adolescents? Are they sufficient to consider policies designed to promote system consolidation, given that many facilities are relatively small? Is it possible to achieve economies of scale while maintaining access to treatment in rural areas? What are the tradeoffs?
• What is the optimal size of group therapy groups?
• How does technological change affect costs?
• What client and facility-level characteristics need to be adjusted for in performance measurement and management systems that evaluate treatment costs?
• When should health plans contract for services rather than provide them internally?

Participants identified some of the key technical issues that need to be considered when designing cost function analyses:

• The analysis must be designed with the ultimate goal in mind. If, for example, the goal of a cost function analysis is to promote the optimal allocation of resources, the analysis must adequately capture the complexity of substance abuse treatment to be credible to the policy-makers, practitioners, clients and other stakeholders, and likely must be tailored to the specific context under study. Also, it is important to keep in mind that efficiency likely is but one of many stakeholder objectives.
• Variables need to be chosen carefully. There are many kinds of output measures that can be used including final outcomes (e.g. restoration of patients' health), intermediate outputs (e.g. discharges) or services provided (counseling sessions). Output measures should be consistent with what the organization is meant to accomplish, and must be adequately adjusted for heterogeneity in the client population. For example, the number of clients who completed treatment might be appropriate, but it might not adequately value contributions to recovery of incomplete treatment episodes. When considering input prices, it important to use the market wage while accounting for the possibility that higher quality may require better-paid personnel.
• Careful consideration needs to be given to which type of cost function is estimated, for example average vs. total (variable) or long- vs. short-run. Although it is somewhat easier to obtain an estimate of economies of scale from average cost functions, they cannot be used in multiproduct cost functions when firms produce different subsets of products. By definition, "short-run" applies to the period during which capital equipment (or some other input) is fixed, and "long-run" refers to a time horizon long enough so that capital is variable (e.g. the firm can purchase more). However, given that future capital purchases may depend on current demand, capital may be endogenous. Several methods may be used to determine which type of function to estimate (e.g. Hausman (1978) specification tests for the endogeneity of capital, running the model both ways and seeing which provides the most sensible results, or testing for conditions implied by economic theory).
• Attention also needs to be paid to functional form (e.g. log-log, Cobb Douglas, translog, etc.). Each of these design factors imposes certain assumptions that must either be tested or acknowledged when interpreting the results of the analysis. Appropriately-specified flexible functional forms can accommodate many of the unique characteristics of substance abuse treatment (as it does for other industries). Using such forms is likely to be far more appropriate than the use of simple linear or hybrid forms.
• Studies comparing substance abuse treatment facilities on other outcomes and that rely on more sophistocated measures of casemix reveal that failing to account for differences in casemix can mask important differences in performance. A claims-data based method for estimating the severity of an episode of illness as a function of the number of claims with drug abuse and related diagnoses over a given time period and the costs of each encounter is available which might be appropriate for some purposes, although methods that rely on the receipt of services to generate severity information may be especially prone to gaming.

Participants made the following suggestions for future research:

• Investigate how best to conceptualize and measure casemix for the purpose of comparing costs across substance abuse treatment facilities. What are the relevant aspects of the client and the addictive disorder that would lead to differences in the cost of treatment, and how are they best measured? How can casemix measures be designed so that they do not promote over- or under-treatment? How can the required data be collected in a standardized way that can be accomplished in a busy treatment practice, given that most current administrative data systems in substance abuse treatment are not well suited for this purpose?
• Conceptualize, develop, and test quality measures for inclusion in cost function and other related analysis.
• Investigate the properties of a variety of output measures to help determine which are most informative for specific purposes. Studies also are needed to understand the extent to which output is endogenous and, if it is, how to model it appropriately.
• Develop estimates of the cost of high-quality care (e.g. care that conforms to NIDA treatment principles).
• Investigate economies of scale and scope, as well as the increment in costs that may be necessary to establish access in rural areas and for clients needing special programs (e.g. those with co-occurring disorders). Also, what are the relationships between scale, other aspects of technical efficiency, and outcomes?
• Consider panel data approaches, which have been show to have advantages in the hospital literature.
• Conduct efficiency analyses of the substance abuse treatment industry.
• Consider the role of management styles and interventions on costs and cost functions.
• Consider production function and complementarity analyses as additional ways to investigate the economics of treatment.
• Develop models of the system of care to better reflect the idea that some clients require multiple episodes of care each encompassing multiple modalities to achieve recovery, clients often need both core and comprehensive services, and that the same client might be appropriately treated in different settings, such as inpatient hospital vs. residential, or specialty vs. primary care. These models may also offer insights for optimizing systems of care for other conditions requiring service coordination and a spectrum of treatment options.
• Investigate how best to present information from these analyses to decision makers.
• Address gaps in data collection incrementally, as opposed to undertaking a nationwide data collection effort, until other issues (especially casemix and quality measurement) are further investigated. For example, it may be possible to use data from closed delivery networks with good information systems to inform output measurement and casemix and quality measurement and adjustment mechanisms, which could then be applied in a larger scale effort. Other approaches may be to conduct targeted data collection efforts to complement existing data collection.
• Data collection needs to be further standardized and streamlined, and additional opportunities for providing timely information to providers that help them respond to specific changes in the field (such as mergers and labor unionization) are needed to improve the quality of the data and feasibility of data collection. There is little routine administrative collection of cost data for substance abuse treatment, suggesting that primary data collection will be needed to advance the field for the foreseeable future.
• Research is needed to determine the costs of ancillary and wraparound services, and to measure administrative and indirect costs and the costs of information technology.

Brief Description of Resulting Publications:

None, although a journal supplement may be pursued.

Participant list [PDF Format, 40kb]
Agenda [PDF Format, 36kb]