Comparison of Hemoglobin Measurements by 3 Point-of-Care Devices With Standard Laboratory Values and Reliability...

We compared the accuracy of 3 point-of-care testing (POCT) devices with central laboratory measurements and the extent to which between-method disagreements could influence decisions to transfuse blood.

Hemoglobin concentrations [Hb] were measured in 58 adult patients undergoing cardiothoracic surgery using 2 Ilex GEM Premier 3500 blood gas analyzers (BG_A and BG_B) and a HemoCue Hb-201 device (HemoCue). Measurements were compared with our central laboratory's Siemens Advia 2120 flow cytometry system (laboratory [Hb] [Lab[Hb]]), regarded as the gold standard. We considered that between-method [Hb] differences exceeding 10% in the [Hb] range 6-10 g/dL would likely erroneously influence erythrocyte transfusion decisions.

The 70 Lab[Hb] measurements ranged from 5.8 to 16.7 g/dL, of which 25 (36%) were <10.0 g/dL. Measurements by all 4 devices numbered 57. Mean POCT measurements did not differ significantly (P > .99). Results of the Bland-Altman analyses revealed statistically significant bias, with predominant underestimations by all 3 POCTs predominating. HemoCue upper and lower limits of agreement (LOA) were narrower, and the 95% confidence intervals (95% CIs) of the LOAs did not overlap with those of BG_A and BG_B. Similarly, a narrow mountain plot demonstrated greater precision for the HemoCue. Comparing BG_A with BG_B revealed no bias and narrow LOA. Error grid analysis within the [Hb] range 6-10 g/dL revealed that 5.3% of HemoCue measurements were beyond the permissible 10.0% error zone in contrast to 19.0% and 16.0% of the blood gas measurements. Possible inappropriate transfusion decisions based on POCT values generally erred toward unnecessary transfusions. Calculations of Cohen κ statistic indicated better chance-corrected agreement between HemoCue and Lab[Hb] regarding erythrocyte transfusions than the blood gas analyzers.

All 3 POCT devices underestimated the Lab[Hb] and cannot be used interchangeably with standard laboratory measurements. BG_A and BG_B can be considered to be acceptably interchangeable with each other. Whereas the HemoCue had little bias and good precision, the blood gas analyzers revealed large bias and poor precision. We conclude that the tested HemoCue provides more reliable measurements, especially within the critical 6-10 g/dL range, with reduced potential for transfusion errors. Decisions regarding erythrocyte transfusions should also be considered in the light of clinical findings.