Abstract
Population-based surveys matched by time but using different methodologies for determining hemoglobin (Hb) concentration have shown inconsistencies in estimating anemia prevalence. This study aimed to estimate measurement errors in Hb quantification in HemoCue 201+ using venous blood (VB) and capillary blood both drops (DCB) and pools (PCB), and compare the results against those of a reference method (VB analyzed in hematology analyzers based on the cyanmethemoglobin method). Children (n = 49), adult females (n = 50), and older adults (n = 50) were randomly allocated to donate VB (4 mL) and either DCB (three drops) or PCB (350 µL). Results in HemoCue were analyzed through Bland Altman and Lyn's concordance against Hb concentration by the reference method. A positive average bias (systematic error) was found for the HemoCue (0.31 g/dL) using the same VB samples. This value was then subtracted from all readings carried out in the device. After this adjustment, DCB still produced a positive bias (0.42 ± 0.81 g/dL), and the variation of single results was ±1.6 g/dL (95% CI). PCB and VB performed similarly; the average bias was negligible (-0.02 ± 0.36 and 0.00 ± 0.33 g/dL, respectively) and the variation of the results (95% CI) was ±0.7 g/dL or lower. Lyn's concordance values were 0.86, 0.96, and 0.98 for DCB, PCB, and VB, respectively. Random variation using DCB is too large to approximate the true Hb values, and therefore DCB should be discontinued for diagnosing anemia both in individuals and in populations.