Variation among laboratories and point-of-care (POC) devices in International Normalized Ratio (INR) determinations for individual patients in the same community was studied. For each patient, blood samples obtained from a single venipuncture followed by a finger stick were tested at three laboratories and with two POC devices. A nonparametric analysis of variance and regression analysis were used to compare INRs across sources. The potential clinical impact of differences between pairs of INR readings (i.e., whether a difference was likely to result in a change in therapy) was evaluated. Complete data were obtained and analyzed for 41 patients. Of the five sources of INRs, results from two laboratories and one POC device were significantly similar. Results from one laboratory and one POC device were significantly different from the other results. Variation between INR readings from two different sources might have resulted in a change in therapy 10-30% of the time. Variations occurred in INR readings among three laboratories and two POC devices and could have led to a substantial number of inappropriate changes in therapy.