Agreement between visual and automated UniScept API readings.

The UniScept API system was evaluated for agreement of visual versus automated readings of both its identification panels and its antimicrobial susceptibility panels. The biochemical responses of 340 oxidase-negative and oxidase-positive fermentative bacterial cultures were read both visually and automatically in the UniScept API 20E system. Automated and visual readings agreed with 99.3% of the biochemicals. Of the 45 tests that disagreed, the tests for indole and citrate were most often in disagreement. A total of 470 fermentative and nonfermentative cultures were used in the UniScept MIC system to compare visual and automated readings of susceptibility results with 17 antimicrobial agents. Agreement within +/- 1 dilution occurred with 94.1% of the enteric fermenters and with 91.7% of the other cultures. Comparison of visual and automated readings resulted in very major discrepancies in 0.95% of the readings, with the largest percentage of discrepancies associated with glucose nonfermenters (1.8%). It was felt that an automated reading is an acceptable alternative to a visual reading of the biochemicals but that 0.95% was just within the acceptable range of the 1% allowable very major discrepancies in the automated reading of susceptibilities.