Physician compliance with clinical practice guidelines: A multi-level analysis

Unwanted variation in the health care of patients can lead to costly, unnecessary care, the underuse of needed services, and the exposure of patients to additional risk. The medical profession developed clinical practice guidelines in the early twentieth century to address this variation. Previous research has shown that unwanted variation is related to geographic location, to patient characteristics, and to physician characteristics.;The proportion of patients receiving care compliant with guidelines ranges from eight% (upper respiratory infections in children) to 68% (lipid-lowering medication for coronary artery disease). Unlike previous studies, this study found that patients with Medicaid insurance are more likely than patients with private insurance to receive care compliant with guidelines, and elderly patients are more likely to receive compliant care than patients in other age groups. Male physicians and specialists are more likely to provide care compliant with guidelines. Of all characteristics, the patient's diagnosis was most important. Patients with coronary artery disease, for example, are 87% more likely to receive the medications required by guidelines for their care than patients with asthma. Physicians who follow guidelines in one patient are similarly likely to follow guidelines in patients with the same diagnosis, but not in patients with other diagnoses.;For both the physician-level and patient-level coefficients, the coefficients are smaller than logit regression analysis performed on the same data. This methodological insight reveals that failing to account for the context in which health care is provided (the nesting of patients within physicians) causes inaccuracies in explanatory models. Not accounting for the ecological fallacy by using only single-level analytical models will lead researchers and policymakers to overestimate the impact of both patient and physician characteristics on compliance with guidelines in providing care.;This study uses a multi-level model appropriate for a dichotomous dependent variable for an analysis of patients nested within physicians. The results demonstrate that the variation between patients nested within physicians is responsible for more than ten times the variation in compliance with clinical practice guidelines than the variation between physicians. Few physician variables are significant in predicting compliance with guidelines.