The population-based, small area analysis strategy
Over the past two decades, the Dartmouth Atlas Project (DAP) has developed a national strategy of providing continuous feedback of population-based information describing resource inputs, utilization and outcomes of care among the 3,436 hospital service areas and 306 hospital referral regions in the United States. The aggregation of up to 12 years of Medicare data in a single site has resulted in a rich source of information that can be used in evaluating the performance of the national, regional, and local health care systems. Much of the variation among areas in per capita resource inputs and utilization has proven to be unwarranted; it cannot be adequately explained on the basis of differences among regions in illness rates, patient preferences or the dictates of evidence-based medicine. Much of the variation relates to provider quality defects. In addition to variations in medical errors such as mortality following surgery, the DAP documents unwarranted variation in three categories of service: (1) systematic underuse of effective care such as beta-blockers after heart attack, or diabetic eye care; (2) misuse of preference-sensitive care such as discretionary surgery (as documented by striking variations among neighboring communities in rates of surgery); and (3) overuse of supply-sensitive care such as physician visits and hospitalization rates among chronically ill patients. The categories are important because the causes, as well as the remedies, of unwarranted variation differ according to category.

The problem of unwarranted variation has attracted wide attention from the press, policy makers and, increasingly, clinicians interested in quality improvement and health care reform. It has led to legislation promoting demonstration projects to deal with unwarranted variations among the Medicare population. Of particular importance to the debate over health care reform is the recently published DAP finding that populations living in regions with greater levels of spending and greater use of physician visits and hospitalizations do not experience better health care outcomes or better quality of care.

The importance of hospital-specific measures
While useful for calling attention to unwarranted variations (and for research examining the relationship between inputs, utilization and outcomes), small-area, community-based data lack the specificity required to characterize performance among different provider groups located in the same region. This is an important limitation because reform must ultimately be directed at changing the behavior of the administrative and clinical staffs of specific health care organizations. The DAP has adapted the small area analysis strategy to provide provider-specific performance measures. This methodology has been applied to evaluate variations among prominent academic medical centers; the results indicate striking variations in use of effective care, preference-sensitive care and supply-sensitive care among institutions with strong national reputations for excellence in geriatric care and the management of chronic illnesses including cancer, heart and pulmonary disease.

The next five years
The DAP was originally organized in anticipation of passage of the Clinton health plan as a means of identifying natural health care markets and describing their actuarial costs and utilization rates - information we thought essential for the implementation of managed care. The project was, from its inception, funded by the Robert Wood Johnson Foundation. When the proposal failed, we used the resources of the grant to develop our feedback strategies. While the value of the DAP has become widely recognized by leaders in the payer community, the provider community, policy makers and analysts, we are finding that the re-emergence of interest in health care reform has greatly increased the demand for unbiased population-based data.

Our goals for the next five years include:

1. Systematic feedback of small area and provider-specific information using the DAP website, hard copy and electronic editions of the Dartmouth Atlas, special reports and presentations by faculty.
2. Development of alliances and relationships that will create feedback channels to specific provider groups such as academic medical centers, providers interested in reform and quality improvement (e.g., the Institute for Improving Health Care's Pursuing Perfection project) and providers participating in CMS demonstration projects.
3. Development of relationships with foundations that want to use DAP data to plan and implement interventions to deal with problems in the quality and equity of care (for example, addressing racial disparities or implementing data-driven quality improvement projects).
4. Development of relationships with leaders in the quality movement in order to promote the expansion of performance measures to include preference-sensitive and supply-sensitive measures of efficiency and decision quality.
5. Development of relationships with the payer community in order to broaden the health purchasing agenda to include misuse of preference-sensitive and overuse of supply-sensitive care.
6. Continued interaction with members of Congress and Congressional staffers concerning the policy implications of variations.
7. Opening channels of communication with state governments, particularly those aspiring to leadership in improving health services.
8. Continued research into the development of new performance measures and the understanding and interpretation of unwarranted variation.
9. Continued efforts to promote the "real time" use of Medicare claims as a valuable tool in the management of disease and the evaluation of health care outcomes, including post-marketing surveillance.