Original articlesUse of grade membership analysis to profile the practice styles of individual physicians in the management of acute low back pain
Introduction
The extensive literature on practice variations has focused on two issues: the need to understand differences between the practice styles of physicians, and the need to foster evidence-based medicine. Medical services increasingly need to provide experimental evidence of clinical benefit in order to obtain the broad support of health professionals, regulatory agencies, and payors [1]. However, providers of services may have personal styles of practice that do not reflect current medical standards. Efforts have therefore been directed towards profiling health professionals and measuring their performance in order to better understand why practice varies and to improve the quality of care 2, 3.
The individual physician–patient encounter is a unique and complex situation shaped by the patient's symptoms and expectations, the physician's knowledge, attitudes, and training, and many other environmental factors. Several approaches have been used to try to describe the practice patterns of physicians. They include stratification according to diagnosis-related groups 4, 5, the development of appropriateness scores for single medical procedures 6, 7 or operations 8, 9, 10, the use of standardized patients 11, 12 or written case scenarios 13, 14, and the time-honored method of individual peer assessment [15].
Different approaches describe physician performance in different ways. For example, measures of appropriateness of care are generally used when the focus is on quality of care, whereas measures of cost or the intensity of resource use are preferred when describing the economic consequences of healthcare. However, such one-dimensional assessments of performance cannot be expected to fully reflect the complexity and multidimensional nature of the clinical encounter. A similar challenge in the field of continuing medical education is often met by describing performance as a combination of several dimensions 16, 17, 18, 19. For example, when assessing the ability of medical residents to conduct an objective structured clinical examination, medical expertise might be summarized in terms of ability to interact with the patient, conduct an adequate examination, gather diagnostic evidence, and formulate an appropriate plan of therapeutic action. The development of these types of measures, however, is subjective, time consuming, and complex, and involves a consensus of opinion from many experts.
The ideal measure would be less dependent on opinion and more descriptive of the multidimensionality of management techniques. Grade of Membership analysis (GoM) is a multivariable statistical technique suited for this type of problem 20, 21, 22. It simultaneously estimates several characteristic types of practice style and estimates the grade of membership of each physician within each style. For example, a style might be described as “highly investigative” or “minimal diagnostic,” and the grade of membership coefficient for each would tell the physician the extent to which his/her behavior is identified by those terms.
The present article reports the application of GoM to responses to a mailed survey of Ontario physicians concerning the management of low back pain (LBP) [23]. Analysis of the survey and identification of practice styles were guided by the 1994 Agency for Health Care Policy and Research (AHCPR) practice guidelines on the management of acute LBP [24].
Section snippets
Grade of membership analysis
GoM is similar to cluster analysis in that it identifies sets of variables related to each other in a characteristic way [25]. The relationships identified can then be used to make inferences about the underlying structure or nature of the world represented by the variables. For example, physicians with a very investigative and resource-intensive practice style might all cluster around a certain behavior type. Cluster analysis can: (1) identify the physician most representative of that behavior
Extreme profiles
The initial function of GoM is to identify a set of extreme profiles. To this end, the analysis was executed with partitions into 3, 4, and 5 extreme profiles (Table 2). The Log Likelihood value increased with an increasing number of parameters in the model, and the Akaike information criterion reached its maximum with the four-profile partition, suggesting that the partition into four extreme profiles would be the one to retain. However, each partition was also interpreted from a conceptual
Discussion
The present study demonstrates that GoM can be successfully applied to profile the practice styles of individual physicians. The database used consisted of responses to a survey of Ontario FPs and GPs concerning their management of acute LBP, including three hypothetical patient scenarios. GoM was shown to identify four characteristic and unique management profiles, and to provide a detailed description of the practice style of each individual physician surveyed, thereby potentially enabling
References (30)
- et al.
Written case simulationsdo they predict physicians' behavior?
J Clin Epidemiol
(1990) - et al.
Use of written cases to study factors associated with regional variations in referral rates
J Clin Epidemiol
(1991) - et al.
Mathematical typologyA grade of membership technique for obtaining disease definition
Comput Biomed Res
(1978) A new approach to teaching the practice of medicine
JAMA
(1992)- et al.
Containing costs while improving quality of carethe role of profiling and practice guidelines
Annu Rev Public Health
(1993) - et al.
Realizing the potential of practice pattern profiling
Inquiry
(1992) Profiling physician practice patterns using diagnostic episode clusters
Med Care
(1995)- et al.
Physician profiling. An analysis of inpatient practice patterns in Florida and Oregon
N Engl J Med
(1994) - et al.
The appropriateness of use of percutaneous transluminal coronary angioplasty in New York State
JAMA
(1993) - et al.
The appropriateness of use of coronary angiography in New York State
JAMA
(1993)