This is a pilot study to first measure the cross-sectional relationship between variations
in physicians� vignette scores and aggregated scores of individual physicians� patient
health outcomes; and second to (longitudinally) determine whether feedback of vignette
scores improves physicians� clinical performance as measured by vignettes
1. Study Design: Physicians will complete computerized vignettes for four conditions �
diabetes, coronary artery diseases (CAD), chronic obstructive pulmonary disease (COPD),
and depression. We will collect retrospective outcomes data and develop composite
outcome measures for on two conditions, Diabetes and CAD. For the longitudinal
analysis, only vignette data will be collected and fed back to providers. Feedback
will consist of specific data on vignette outcome scores for the individual physicians
and for the sites overall.
2. Site Selection: Primary care clinics at 2 VAMCs
3. Study Population and Sampling: We will enroll primary care physicians at 2 VAMCs. 30
consenting physicians will be prospectively randomized into two groups. One group will
receive feedback of their vignette scores, and the other group will serve as control,
receiving no feedback.
4. Variables and Measurement Instruments: Computerized vignettes measuring clinical
practice completed by the physicians for diabetes, CAD, COPD, and Depression and a
composite health outcome measures from the medical records of these physicians�
patients with diabetes and CAD.
5. Data Collection Strategy and Timeline: Vignettes will be administered to all physicians
at baseline, with feedback of scores 3 months later and readministration of vignettes 9
months thereafter to measure the trend in improvement. The composite outcome data will
only be collected at baseline only.
6. Data Analysis: The statistical analysis will compare the effects within the context of
an analysis of covariance (ANCOVA) model. The analyte is the quality of care
physicians give to patients with four common conditions. The relationship between
vignette scores and patient outcomes will be modeled accounting for clustering effects.
The prospective experimental design will be used to quantify possible differences
between the intervention and control groups. The data will be analyzed using a
three-way crossed, one-way nested ANCOVA model where the covariate is the baseline
vignette score. This model can be used to look at case effects, by domain, level of
training, and by site.
Physician at SF VAMC with panel of primary care patients