Cost analysis in the Department of Veterans Affairs: consensus and future directions.

OBJECTIVES: In 1997, the Management Decision and Research Center of the Department of Veterans Affairs convened cost experts and health economists in a working meeting. Its goal was to provide consensus guidelines for conducting cost analyses in managed care systems, such as VA, that do not have encounter-level cost data or that do not prepare itemized patient bills. The impetus for the meeting was that too often computer-based cost data were proposed or used in studies that were inappropriate for the question being addressed. There was also a sense that often great effort was being expended by VA health economists "reinventing the wheel" in developing new cost components for each study. METHODS: A group of 45 VA and non-VA health economists, health researchers, and policy-makers attended a 2 day working meeting organized around a series of case vignettes to identify areas of consensus, controversy, and gaps in knowledge. RESULTS: Consensus emerged in the following four areas: (1) Cost Methods. A "hybrid model" was identified as the current standard of cost analysis in VA and entails mixing "micro-costing" primary data collection and "gross-cost" computer-based methods to reflect resource-use variations that are essential to the research question. (2) Cost Infrastructure. VA is developing a new, but unevaluated, costing system that could allow for computer-based cost analyses at much finer levels of detail than is currently possible. (3) Data Quality. Ongoing data validation of existing and developing cost databases is needed, especially concerning interfacility variation. (4) Dissemination. A new cost data center was recommended to provide training, information dissemination, and coordination. CONCLUSIONS: Consensus was reached about the hybrid model as the current paradigm for cost analysis in systems like VA.

More or less?: Methods to compare VA and non-VA health care costs.

OBJECTIVE: To examine past comparisons of the costs of the Veterans Health Administration (VA) and of non-VA providers to determine lessons and data requirements for future cost comparisons, particularly those assessing VA efficiency and to determine whether VA should purchase care from non-VA providers. CONCEPTUAL FRAMEWORK: Over the past two decades, researchers have tried to establish how VA costs compare to those of non-VA health care delivery systems. Existing studies of overall acute care costs address one of two distinct questions: How do VA costs compare to costs in private sector hospitals? and Would it cost more to have VA patients treated in nonfederal hospitals? For both questions, the major factors underlying differences in health care costs are variations in outputs, input prices, and levels of efficiency. Health care cost comparisons across systems must also wrestle with accounting differences. CONCLUSIONS: That review finds no convincing evidence that VA has been significantly more or less efficient than nonfederal hospitals in delivering care. However, VA costs do appear to have been significantly lower than fee-for-service charges that the federal government might have to pay if veterans were treated in private sector hospitals for the same diagnoses. Future comparisons of costs in the era of managed care will require better diagnostic and population data to control for observable and unobservable case-mix differences. They should also include measures of the quality of outcomes. Finally, consistent accounting practices, particularly in the treatment of capital costs, are needed.

Do individuals with substance abuse diagnoses incur higher charges than individuals with other chronic conditions?

Concerns about high costs have led to limits on the services covered by most insurance plans for substance abuse treatment. But, the commonly used comparison group for cost analyses, all enrollees in a health-care plan, may not be appropriate because addiction is a chronic condition. Therefore, to determine whether substance abusers incur higher charges than patients with other serious chronic conditions, we used health insurance information for employees and dependents over 3 years (1989 to 1991) for two firms with a total of almost 40,000 employees to do alternate comparisons. We compared average annual charges for patients with the following diagnoses: substance abuse, substance abuse with mental illness, arthritis, asthma, and diabetes. Patients who undergo treatment for abusing alcohol, drugs, or both often (but not always) incur higher charges than people with other chronic conditions. Clear differences in average charges emerge between patients with and without mental health claims.

Private sector coverage of people with dual diagnoses.

In general, people with dual diagnoses account for a significant proportion of both the mental health and substance abuse populations. Most published information on dual diagnosis comes from research on selected treatment programs that are largely funded from public sources. This analysis uses private health insurance claims and eligibility files for 1989 to 1991 for three large firms to identify individuals with both substance abuse and mental health claims and to examine their characteristics, charges, and utilization. More than half of people with dual diagnoses incurred significant charges over three years in both mental health and substance abuse. These individuals with high mental health charges were more likely to be male than were patients with mental health claims alone; they were less likely to be male than were patients with claims for substance abuse and no mental health services. They were also significantly younger than were patients with substance abuse or mental health utilization only for two of the firms. The average charges for people with dual diagnoses were higher than those for patients with substance abuse or mental health claims only.

Characteristics of private-sector managed care for mental health and substance abuse treatment.

OBJECTIVE: This study examined diversity during the late 1980s in managed care programs for mental health, alcohol abuse, and drug abuse to identify ways in which research can generate more meaningful data on the effectiveness of utilization review programs. METHODS: Telephone interviews were conducted with representatives of utilization review programs for employee health insurance plans in 31 firms that employed 2.1 million people in 1990. Questions addressed qualifications of personnel, clinical criteria to authorize care, integration with employee assistance plans, penalties for not complying with utilization review procedures, outpatient review, and carve out of mental health and substance abuse review. RESULTS: Large variations in utilization review programs were found. Programs employed a range of review personnel and used a variety of clinical criteria to authorize care. More than two-thirds did not carve out mental health and substance abuse review from medical-surgical review. Some firms' employee assistance plans were integrated with utilization review programs, while others remained unintegrated. Penalties for not following program procedures varied widely, as did review of outpatient services. CONCLUSIONS: Because of trends toward even more diversity in utilization review programs in the 1990s, research that identifies the specific features of managed care programs that hold most promise for controlling costs while maintaining quality of care will increasingly be needed.

Measuring quality of care: fundamental information from administrative datasets.

Under proposals for national health insurance reform in the USA, employers and purchasing cooperatives will have to measure the quality of health care services. Their need for data systems upon which to base their decisions has stimulated dramatic innovation and rapid change in how health care information is collected, integrated from multiple sources, and reported. To make administrative data useful for quality measurement, careful attention must be given to information about: medical care utilization; patient characteristics; provider characteristics; and health plans. In this paper, we describe the extent to which this information is included in existing administrative datasets. We then suggest how planned datasets should be designed so they can be used to assess the quality of health care.

How well do Americans understand their health coverage?

Data from two surveys are used in this DataWatch to explore Americans' understanding of their health insurance. First, data from a national survey of consumers are used to examine if people with private health insurance correctly report their coverage for six services. Second, information from an evaluation of a pilot project of subsidized insurance in New York is used to investigate how well newly insured persons understand their coverage. Based on these surveys, almost all privately insured people understand the basic elements of their insurance plans but underestimate their coverage for mental health, substance abuse, and prescription drug benefits and overestimate their coverage for long-term care. People who are newly insured in physician networks or health maintenance organizations seem uncertain about what services their plan covers and restrictions on their choice of hospitals.

Can practice guidelines reduce the number and costs of malpractice claims?

Practice guidelines have the potential to reduce the number of malpractice cases and the costs of settling them. However, for practice guidelines to exert any influence, they must be assumed to be (1) developed for conditions or procedures that frequently lead to events for which negligence claims are filed; (2) widely accepted in the medical profession; (3) fully integrated into clinical practice; and (4) straightforward and readily interpreted in a litigation setting. Because the validity of each of these assumptions can be questioned, the idea that inserting practice guidelines into the existing litigation process will generate large savings in the near future is overly optimistic.

Rural hospital wages.

Average fiscal year 1982 wages from 2,302 rural American hospitals were used to test for a gradient descending from hospitals in counties adjacent to metropolitan areas to those not adjacent. Considerable variation in the ratios of adjacent to nonadjacent averages existed. No statistically significant difference was found, however. Of greater importance in explaining relative wages within States were occupational mix, mix of part-time and full-time workers, case mix, presence of medical residencies, and location in a high-rent county within the State. Medicare already adjusts payments for only two of these variables.

Hospital wage gradients within U.S. urban areas.

Average wages from 2,275 general hospitals in metropolitan areas across the U.S. were used to test for a wage gradient descending from hospitals in the central city through those in urbanized and finally non-urbanized areas of each county-defined metropolitan area. Significant gradients were found in MSAs of all sizes. Urbanized-area wages were 3 to 6 percent lower than those in central cities of the same metropolitan area. Non-urbanized suburban wages were 10 to 12 percent lower than those in central cities. The explanations for the gradients differ somewhat between large and small areas. For example, while the relative mix of high-wage and low-wage occupations in each hospital is a significant explanatory variable for wages in all metropolitan area sizes, the relative use of part-time workers is not significant in those metropolitan areas with fewer than 250,000 people. Relative crime in each hospital's city is highly significant in explaining relative wages only for areas with populations of more than one million.