Provider profiling: advancing to episodes of care

Judging by the interest expressed by managed care organizations, provider profiling has arrived. Surveys indicate that most organizations have adopted, or plan to adopt in the near future, a means to describe, provider practice patterns. A further vote of confidence came from providers: In 1994, the American College of Physicians, the largest national specialty organization, issued a position paper supporting provider profiling and questioning the value of other approaches to utilization management, such as preauthorization of individual services. Also, an article and an editorial in the New England Journal of Medicine cautiously supported the concept of profiling. Provider profiling has great promise as a means to promote cost-effective care without the limitations of case-by-case preauthorization. The combination of a sophisticated episode of care methodology and a set of validated practice benchmarks offers the opportunity to perform true clinical profiling and to supply providers with data to review and alter practice patterns.

What is provider profiling? A reasonable definition is description of practice patterns on a disease-specific basis, comparison of these patterns to an appropriate peer group, and also comparison patterns, to validated benchmarks. By examining practice patterns, profiling seeks to promote cost-effective care. This goal relies on two assumptions:

* Cost-effective strategies can be identified for specific illnesses.

* There is substantial variance among providers around such strategies.

Evidence supports both assumptions. Although definitive evidence for "best practice" is available for relatively few conditions, consensus evidence and clinical data support at least a limited number of most effective strategies for many common illnesses. Second, academic researchers have documented extensive practice variations without clear foundation in demographics or patient characteristics.

Provider profiling can be thought of as a means to reduce variation in practice patterns - not as a means to promote the one correct pattern for a given illness, because limitations of evidence and patient variation make this unlikely, but rather to promote the range of patterns that appears cost-effective. Provider profiling in this sense is educational. The approach seeks to establish a feedback loop for providers, describing their current practice patterns over patient populations and relating these patterns to the practice of their peers, and to establish benchmarks.

The value of this approach can be enhanced by analysis of episodes of care. An episode serves as a clinical unit," especially for acute and subacute care. If both cost and clinical issues are to be considered, comparison of episodes is preferable to analyzing care over a time interval. The latter approach does not differentiate patients who may have only one episode in an interval from those with multiple episodes.

Creation of episodes of care using claims or encounter data is a substantial methodological challenge. Not only must the beginning and end of the episode be identified, but also related diagnoses must be linked. For example, if a patient is initially seen for chest pain, the diagnosis of angina is made, and then a myocardial infarction occurs, the methodology must link these services into a single episode, rather than breaking them up into multiple episodes.

How can this be accomplished? With regard to identifying the start of an episode, in Health Payment Review's EpisodeProfiler this is based on the first appearance of services for a diagnosis. The termination of the episode and, by extension, the beginning of another episode of the same illness are based on a "clean" period, an interval in which no services for the diagnosis occur. This methodology is superior to the establishment of a fixed interval for defining episode duration. The latter approach, unless the interval is quite long, excludes extended episodes that may be of particular interest for analysis of resource use.

With regard to linking diagnoses, EpisodeProfiler divides all diagnoses into 553 categories, Episode Treatment Groups (ETGs). Analogous to hospital-based DRGs but incorporating care at any site, ETGs consist of groups of primary diagnoses that involve similar clinical management. Distinct ETGs within the same primary illness can occur on the basis of comorbidities, complications, or defining surgeries. For example, there are multiple ETGs for bronchitis based on age and on the presence of specific comorbidities and multiple ETGs for asthma based on the presence of complications. This approach is critical in allowing appropriate comparisons; if all asthma was in a single ETG, variation in patient condition might well overwhelm variation in practice patterns.

A considerable number of other methodological issues are beyond the scope of this discussion. Suffice it to say that the methodology must be sophisticated enough to address issues raised by providers, who will demand (appropriately) that profiling be conducted accurately. In addition, profiling should incorporate sophisticated case-mix adjustment methods. That is, when entire practices are compared, the illness burden of the population must be considered. EpisodeProfiler uses ETGs to evaluate case mix. This approach has advantages over other, regression-based methods, that link unrelated diagnoses based largely on resource use. Unlike resource-derived systems, a clinically based system such as EpisodeProfiler can be used to case-mix adjust within a specific clinical area (e.g., renal disease) or a specific specialty (e.g., urology).

Lawrence G. Miller, MD, MPH, is Vice President, Clinical Affairs, Health Payment Review, Boston, Mass.

COPYRIGHT 1995 American College of Physician Executives
COPYRIGHT 2004 Gale Group