Our results include several major findings: (a) pooled capillary blood samples, in our hands, show acceptable performance in Hb measurement using HemoCue 201+ when compared with venous blood samples; (b) drops of capillary blood showed the largest Hb measurement variations and inaccuracy than pooled capillary or venous blood measurements using HemoCue 201+; (c) variation of results with drops of capillary blood is random, and therefore not correctable; (d) the HemoCue 201+ device in our study showed a positive bias (that is a systematic error) which can be corrected by adjusting the Hb values of the same blood samples (reference samples with hemoglobin concentration within the range of hemoglobin concentrations of 8-16 g/dL) analyzed in HemoCue and by a laboratory reference method; and (e) the HemoCue measurements (considering the measurement error from personnel) contributed to low variation in the Hb measurement system performance as expected (7%), with the main source of variability being the individual heterogeneity of the same blood samples, especially in drops of capillary blood. Variation from measurement error was highest in children, which reflects the challenge of obtaining good quality blood samples in this population group.
Results of hemoglobin concentration values as estimated using drops of capillary blood in the HemoCue 201+ varied widely from the true values calculated using venous blood, such that they could not be considered reliable. In addition, the higher intra-variability in subsequent drops of capillary blood is additional evidence that this is not a suitable source for measuring hemoglobin concentration as the error can be as high as 20-30% of the true value. Bond  documented a coefficient variation 3.4 times for Hb for subsequent drops than in venous blood measured in a hematology analyzer.
Our results coincide with Conway et al , who discourage the finger prick for drops of capillary blood to estimate Hb values due to larger variation than when using pooled capillary blood samples and, logically, much higher than using venous blood. Bond et al  also concluded that hemoglobin measurement using several drops of blood may be necessary to reduce variation, but this is neither practical nor economically sensible. In our study, two measurements from capillary drops did not reduce the variation. Our results suggest that drops of capillary blood introduce considerable misinterpretation of the anemia status of individuals (under- or overestimation) and populations (overestimation of the prevalence, due to the use of a threshold value to estimate anemia prevalence).
Pooled capillary blood produced a slightly higher difference in Hb concentration of »0.1 g/dL than venous blood, but which we considered small enough to not necessitate corrections. The high concordance of Hb from pooled capillary versus venous blood found in our study highlights that pooled capillary blood may be considered an acceptable alternative for estimating Hb using a portable photometer in population-based field studies when the collection of venous blood is challenging. Dasi et al  reported a similar finding: a concordance of 0.97 from Hb from pooled capillary blood measured by a portable autoanalyzer versus venous blood measured using a clinical method. The reliability of Hb measurement from pooled capillary blood data may be explained by the low variability due to collection of a greater amount of blood as compared to single drops of blood. Nevertheless, pooled capillary blood variability might be affected by other factors, such as insufficient training and standardization of personnel who collect the blood sample (as discussed below), time duration in collecting the sample, and postural effect and a defective mixed sample . In the Hb distribution in pooled samples, we found that six values were out of the expected range, which may suggest coagulation of blood samples during collection due to possible failure to adequately mix the blood with the anticoagulant. In order to narrow the ratio of EDTA to blood, our study collected 350-500 µL of blood. Previous data analyzing Hb from venous blood with EDTA versus heparin did not find statistically significant differences in Hb determination (Méndez Gómez-Humarán I, in process of publication).
One study performed in Guatemala and Honduras showed HemoCue bias with the opposite tendency to that observed in our study, finding a systematic and negative bias (-0.13g/dL, p=0.01) in Hb measurement with HemoCue 201+ models. Nevertheless, a positive bias (+0.24 g/dL, p<0.001) was found with the HemoCue 301. Therefore, the bias may depend on the apparatus and not the model, so the performance of each apparatus should be verified before use and its performance examined periodically. Calibration to manufacturer standards is insufficient, as it checks optical conditions of the apparatus but not accuracy of measurement against blood samples . The latter is especially important if the apparatus is intended for use in population-based surveys where small changes may have important implications for the measured distribution of hemoglobin concentration in the population. Systematic bias can be easily corrected, either by adjusting the results of the apparatus using blood standards analyzing them with a reference method and in each portable photometer or by collecting additional blood samples in a subsample of the study population for analysis with a clinical hemocounter.
Our findings may explain the discrepancy in estimations of anemia prevalence among surveys [14–18] when verification of the equipment is uncommon, and the blood samples are frequently drops of capillary blood. Recently, Hruschka et al  documented wider discrepancies in Hb distribution and substantial differences (from 9 to 30%) in anemia estimation from two nationally representative survey matched by time and country when using the HemoCue device as compared to other methodologies. The DHS used capillary drop blood, while the BRINDA project used venous blood. The median difference in anemia prevalence was consistently lower in BRINDA surveys than in DHS surveys (+19% in children and +9% in women).
Some studies have documented higher Hb values (in the range of 0.3 to 0.8 g/dL) from HemoCue 201+ when using a hemocounter [24,25], while others have not found a bias in the equipment. Our data came from a field setting scenario, where participants were interviewed at home in conditions outside clinical settings, and without humidity and temperature control, which are factors that might affect the stability of the HemoCue cuvettes and electronic performance of the HemoCue 201+ device [26,27]. A previous study done in San Luis Potosí, Mexico using HemoCue 201+ reported failures in completing Hb readings in locations with high temperatures (>35ºC) (Berenice Gaona, personal communication, data not published).
In summary, it is important to emphasize that the main reason for errors in the determination of hemoglobin using HemoCues is not the bias of the apparatus but the large and random variation if using single drops of blood obtained by finger prick. This measurement error was highest in children (17%) as compared to women (3%) and older adults (7.3%). Since this error is not amenable to correction, it may impact the anemia prevalence estimates, with the greatest affect in estimations for children.
Different sources of error may affect the accuracy and precision of the Hb measurements . Well-trained and standardized personnel are a crucial factor which minimizes potential error in Hb determination in field studies. In the present study, personnel were instructed to puncture only once (in the finger for capillary blood or the arm for venous blood) to avoid participant exclusion, especially due to the refusal of mothers to allow their children to participate. Our success rate in collecting pooled capillary blood was very high (97.3%) in all samples in comparison with a study done in Honduras , where non-trained health or medical personnel were involved. In that study, pooled capillary blood samples could not be collected in a high proportion of children (38/46, 83%) or some women (3/24 = 12%). We attribute the success rate of our study, both in pooled capillary and venous blood samples in children and adults, to the time invested (around two weeks) in training, sensitization, and standardization of our personnel for collecting and handling blood samples in field settings. These are important skills that must be assured for the reliable human measurement of Hb, independently of the method. The assumption that drops of capillary blood are suitable for estimating Hb concentration is incorrect. Our results confirm previous recommendations to discourage the use of single drops of capillary blood collected by finger prick for Hb measurement [11,12,22].
Other sources of error may include at the time of loading blood into the HemoCue micro-cuvettes, which increases the magnitude of error using these devices . This may be even worse if unexperienced personnel are involved.
Here, it is important to note that our results differ from one previous validation study using the HemoCue (B-Hemoglobin model) done in a hospital in Mexico . The results of that study found a non-statistically significant bias from drops of capillary blood in both children and adults (-0.14±1.05 and -0.12±0.87 p=0.025, respectively), as compared to pooled capillary (-0.29 ±0.78 in adults) and venous (-0.47±0.53 in adults) blood. Nevertheless, larger variation from drops of capillary blood, particularly in children, was consistent with our study . Notably, the previous study was performed in clinal setting (hospital), while our study was performed in a field setting where external factors may have affected methods.
The following should also be considered when interpreting our results and comparing them to conflicting data reported in population-based surveys: we used high-flow BD-type lancets, which are characterized by a higher number of bezels (pentapoint) and allow high blood flux. Most national surveys use low-flow lancets that cause a shallow wound (one bezel), likely with the aim for less invasive methods. However, the one bezel lancet may restrict the flux of red blood cells and therefore increase inaccuracy and variation in hemoglobin concentrations estimated using single drops of capillary blood. Our study did not examine the effects of using low-flow lancets.
To our knowledge, this is the first study to evaluate the results of hemoglobin concentration measurements using the HemoCue in a field setting scenario with different blood sample sources: venous blood, pooled capillary blood, and drops of capillary blood. Furthermore, it is common that the performance of HemoCues is not verified using venous blood and a reference methodology; our results demonstrate the importance of estimating the bias of each HemoCue apparatus against venous blood which is further analyzed using a clinical hemocounter.
Finally, it is important to highlight that our validation study was done in a field setting, and therefore had marked discrepancies from those studies done in a clinical setting. Nonetheless, our main findings related to drop-to-drop variability as well as pooled capillary blood variation were consistent with previous studies in clinical settings. Therefore, verification and adjustment of the results of HemoCues should be done before their use in the field. This verification should also be periodically confirmed to maximize the reliability of the results .
Determination of anemia prevalence in both individuals and in populations requires the use of reliable procedures and methodologies, because producing erroneous results may cause serious negative consequences, and waste of resources and time, that using only clinical indicators.