Reliability of Capillary Complete Blood Count in Children With Acute Gastroenteritis

Background: To assess the reliability of complete blood count (CBC) in the capillary blood of children with acute gastroenteritis (AGE), with a focus on leukocytes. Methods: This was a retrospective cross-sectional study. Complete blood count was compared between the capillary and venous blood in children admitted to a pediatric gastroenterology department with primary diagnosis of AGE (ICD-10 A09, A08.0, A08.2). Capillary blood was obtained in the emergency room and venous blood was sampled in the ward shortly thereafter during peripheral intravenous line placement. Results: One hundred and forty children were included. The mean (±SD) age and weight of patients were 3.0 ± 2.9 years and 16 ± 9 kg; 26% had leukocytosis. The mean difference between obtaining results of capillary and venous blood tests was 2 ± 1 h. Area under the receiver operating characteristic curve (AUC) for the identification of leukocytosis using the capillary blood was 0.98 (95% CI 0.96–1.0). The sensitivity and specificity were 86 and 98%, respectively (accuracy 95%). The positive and negative predictive values were 94 and 95%, respectively. The intraclass correlation coefficient revealed high concordance between capillary and venous CBC measurements (leukocyte count 0.94, hemoglobin 0.88, erythrocyte count 0.77, hematocrit 0.79, platelet count 0.90). Matched pairs comparisons revealed marginally higher erythrocytes (difference of medians: 0.2 T/L), hemoglobin (0.3 g/dL), hematocrit (1.0%), and platelets (9 G/L) in the capillary blood. Conclusion: Capillary CBC is useful in detecting leukocytosis in children with AGE.


INTRODUCTION
Acute gastroenteritis (AGE) is defined by ESPGHAN as a decrease in the consistency of stools and/or an increase in their frequency, with or without fever or vomiting. AGE remains a leading cause of pediatric hospitalizations, especially in children below 3 years of age (1).
Apart from clinical signs and symptoms of dehydration, physicians in emergency departments often rely on results of blood tests when making decisions on hospital admission. The tests may include electrolyte levels, blood gas (pH, bicarbonates, base excess) and also a complete blood count (CBC), which helps to rule out more severe conditions. In the ambulatory setting, the collection of capillary blood is also less invasive for the patient than venous blood sampling. However, dehydration, small fingers and the poor capillary network may cause technical difficulties in capillary blood sampling among pediatric patients and reduce diagnostic accuracy (2).
These limitations make the value of capillary CBC uncertain, especially in the emergency room in the course of differential diagnosis of vomiting, diarrhea, and/or fever in children. Therefore, the aim of this study was to assess the reliability of capillary CBC in children with AGE.

METHODS
This cross-sectional study was conducted in Karol Jonscher Clinical Hospital of Poznan University of Medical Sciences, Poznan, Poland. The data were retrospectively collected from electronic health records covering the years 2006-2015. The inclusion criteria comprised: (a) a primary diagnosis of AGE (codes according to ICD-10: A09, A08.0, and A08.2), (b) hematological and electrolyte measurements in the capillary blood in the emergency room and in the venous blood, which were obtained during venipuncture on admission to the hospital, and (c) the time between receiving the results of the two measurements shorter than 4 h. The same methodology was used in our previous study of capillary blood potassium levels (3). The study focused on AGE in order to provide a well-defined group of patients with high availability of the required data.
We used Statistica 12 software (Statsoft Inc., Tulsa, USA) to perform the statistical tests, calculate the areas under the curve (AUC) of the receiver operating characteristic, and draw Bland-Altman plots (4). Receiver operating characteristic curve calculations used binary outcome data obtained by comparing venous blood leukocyte count against the above-mentioned age-specific thresholds. The predictor variable was obtained by dividing capillary blood leukocyte concentration by upper limit of the norm for the age. The Youden's index was used to determine the optimal cut-off. The same was done for hemoglobin. The Kolmogorov-Smirnov test was used to test for normality. Forward stepwise regression was used to investigate the influence of age and base excess (as a proxy for dehydration) and other factors on the absolute error of capillary blood measurements.

RESULTS
Out of 1,080 children with the primary diagnosis of acute gastroenteritis, 140 met the inclusion criteria (90 with A09, 40 with A08.0; age range 0-16 years). The group characteristics, including the mean ± SD values of basic hematological parameters, are presented in Table 1.
Matched pairs analysis revealed a small yet consistent bias of capillary hemoglobin, RBC count, and hematocrit toward higher values ( Table 2). The capillary measurements' absolute  error was low ( Table 2). The agreement of measurements in the capillary and the venous blood is illustrated in Bland-Altman plots (Figure 1; Table 3). AUC for identification of leukocytosis using the measurements in the capillary blood was 0.98 (95% CI 0.96-1.0). The sensitivity and specificity were 86 and 98%, respectively (accuracy 95%). The positive and negative predictive values were 94 and 95%, respectively.
AUC for detecting hemoglobin levels corresponding to anemia-not for diagnosing anemia since the patients were dehydrated-was 0.86 (95% CI 0.68-1.00). The sensitivity was 71% and the specificity 98%, accuracy 97%, positive predictive value 71%, negative predictive value 98%. In seven children, venous hemoglobin levels corresponding to the diagnosis of anemia were found. Five of these were also identified by the capillary blood analysis. However, two false-positive read-outs were also provided by capillary blood analyzers.
Venous hematocrit did not correlate with an absolute error of any of the hematological measurements. Of the five analyzed hematological variables, only RBC count weakly correlated with the difference in time between receiving the results of assessments performed in the venous and the capillary blood (Spearman's ρ = 0.19, p = 0.02). However, forward stepwise regression did not confirm any important, independent relationships between the relative or absolute error of capillary CBC measurements and the aforementioned time difference, age, sex, pH, base excess of extracellular fluid, or urea. A few weak associations (R 2 < 0.20) were found using regression: the relative error of capillary hematocrit associated with body weight, hemoglobin correlated with body weight and age, and absolute capillary white blood cell (WBC) count error related to height.
Measurements were grouped time-wise to identify potential systematic bias related to hypothetical changes in sampling or

DISCUSSION
This is the first study to assess the reliability of capillary CBC in the setting of AGE. Capillary blood, as well as venous blood measurement, were performed in a specialized medical laboratory, using automated hematology analyzers. This pragmatic study had a sizeable convenience sample, employed a statistical approach suitable for determining diagnostic value, and attempted compensation for potential confounders (2). Overall, the differences between the hemoglobin concentration, hematocrit, WBC, and RBC count in the venous and the capillary blood were practically insignificant.
A recently published study comprehensively reviewed the value of capillary vs. venous blood in the assessment of anemia and called for further research (2). Although the results were inconclusive, most of the identified studies revealed a higher hemoglobin concentration in the capillary blood compared with the venous blood. Our results confirm a marginally higher hemoglobin in capillary vs. venous samples. Capillary CBC has a lower sensitivity in the diagnosis of anemia and should not be used to this aim, at least unless its limitations are thoughtfully considered (6)(7)(8)(9).
Multiple biological factors influencing the accurateness of capillary blood hemoglobin concentration measurement in AGE were found. This study revealed that body weight and age slightly affected accuracy of capillary hemoglobin and hematocrit measurements. Previous research pointed toward the dependence of the value of capillary CBC on sex, which could be mediated by the vasoconstrictive effect of androgens. As in males capillaries contain less RBCs, hence less hemoglobin (10). This study found no effects of sex on obtained differences between capillary and venous CBC in AGE.
The method of capillary blood sample collection also significantly affects results. Hemoglobin concentration in pooled sample drops is higher than in the single drop of capillary blood (6). Therefore, the pooled sample method (also used herein) more precisely reflects the actual hemoglobin. This study indicates that capillary CBC may be more precise in the estimation of hemoglobin concentration among younger children and those with lower body weight, despite increased technical difficulties in sampling capillary blood. Similar results were obtained previously (11).
The main aim of this retrospective work was to determine the reliability of capillary CBC in AGE, which results in dehydration. No relationships were found between base excess corrected for the extracellular fluid (BE-ECF) and the accuracy of capillary CBC. BE-ECF can only estimate the level of dehydration, which is challenging to assess correctly even using complex clinical questionnaires.
In summary, despite being prone to biological, technical and environmental confounders, capillary CBC is useful in detecting leukocytosis in children with AGE (21).

DATA AVAILABILITY STATEMENT
The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.

ETHICS STATEMENT
The study was approved by the Bioethical Committee at Poznan University of Medical Sciences (541/15). Written informed consent from the participants' legal guardian/next of kin was not required to participate in this study in accordance with the national legislation and the institutional requirements.