This application proposes to study referrals from primary care physicians to specialist physicians in a managed care organization that enrolls over half of the population of Rochester, NY. The six specific aims are the examination of: 1) variations in referral rates among primary care physicians; 2) variations in referral rates among primary care physicians across diagnostic groups and over time; 3) the relationship between physician and practice characteristics and referral rates; 4) the relationship between referral rates and costs, 5) the relationship between referral rates and avoidable hospitalizations, and 6) the relationship between referral rates and patient health status and satisfaction. Data will be incorporated from multiple sources.
Cl aims data from the managed care organization will be used to identify primary care physicians and calculate referral rates; to examine variations in referral rates among primary care physicians, across diagnostic groups, and over time (Aims 1 and 2); and to quantify test-ordering intensity for Aim 3, calculate costs for Aim 4, and identify avoidable hospitalizations for Aim 5. A random sample of 150 primary care physicians will be surveyed about personal information, characteristics of their practice, and various attitudes including satisfaction, discomfort with uncertainty, risk aversion, malpractice concern, and psychosocial orientation. These characteristics will be used to address Aim 3 and may be included as control variables in the multivariate analyses of Aim 6. Fifty consecutive managed care enrollees of one hundred primary care physicians will be surveyed. Physicians whose patients will be surveyed will consist of the 25 primary care physicians with the lowest referral rates, the 25 with the highest referral rates, and half of those with intermediate referral rates. Patients will be asked about their personal information, health status, satisfaction, medical skepticism, and characteristics of their relationship with their physician. These characteristics will be used to address Aims 3 and 6. All analyses will control for case-mix using Ambulatory Care Groups (ACG) and Clinical Complexity Indices (CCI). Analyses incorporating variables derived from the patient survey will employ the statistical package SUDAAN (SUrvey DAta ANalysis) to account for the nested sample design.
