In addition, critical high POCG confirmation rates were significantly greater when testing was performed in inpatient and emergency department locations ([chi square] test; P =.
Table 4 shows the frequency of critical POCG cases for which there was documentation that a primary caregiver was notified of results.
14-16) Because of the high volume of POCG testing and clinical impact in hospital and emergency department settings, it is important to understand factors that affect test performance in the critical value range.
When POCG devices are used as a waived test, confirmation of test performance such as linearity, accuracy, and precision, which is mandatory for moderately complex clinical laboratory testing, is not necessary as long as properly trained and qualified operators follow prescribed and approved procedures.
A key observation in this study was that only about half of the POCG measurements in the critical range were retested, and of these, about 19% of repeat measurements did not meet criteria to confirm results.
In addition, interpretative criteria to verify critical POCG results were defined in only about two-thirds of facilities.
During the study period, 139 cases of noncritical POCG measurements were identified in which venous blood was retested by a point-of-care blood gas instrument within 10 minutes.
This study addresses the accuracy of POCG measurements in the critical value range.
In addition, all health care personnel who performed POCG testing were trained and certified under the supervision of clinical laboratory technologists.
Most of the repeat tests that did not confirm the initial critical result were in the range of values that the POCG method was shown to measure accurately, based on comparison of values in that range to venous blood measured by another point-of-care device.
High critical POCG values were more common and were more often confirmed than low critical test results.
The Clarke error grid is another method for evaluating POCG methods designed to assess the clinical significance of method biases.