ABSTRACT
BACKGROUND: Anemia is an important public health problem, and accurate estimates may inform policy and programs. Although hemoglobin (Hb) assessment of venous blood via automated hematology analyzers (AHAs) is recommended, most population-based surveys estimate anemia prevalence based on analysis of capillary blood via portable hemoglobinometers. OBJECTIVES: We aimed to evaluate screening methods for hemoglobin and anemia assessment using paired venous samples. METHODS: Participants were women 15-40 y who were not pregnant or lactating. Paired venous whole blood samples (n = 896) were analyzed for hemoglobin (Hb) via portable hemoglobinometer (HemoCue 301) and Coulter Counter AHA. Anemia and severe anemia were defined as Hb <12.0g/dL and <8.0 g/dL, respectively. Bland-Altman methods were used to assess the level of agreement for Hb results (mean difference, SD of differences, limits of agreement). Diagnostic accuracy parameters (sensitivity, specificity, positive predictive value, negative predictive value, accuracy) were calculated to evaluate HemoCue performance compared to the AHA reference, overall and by sociodemographic, nutritional, and metabolic characteristics. RESULTS: The estimated anemia prevalence was significantly lower via HemoCue vs. AHA (36.3% compared with 41.6%; P value < 0.0001). The HemoCue had 84.4% accuracy for anemia screening and 98.8% for severe anemia, compared to the AHA reference. The HemoCue had 74.8% sensitivity and 91.2% specificity, compared to AHA. HemoCue sensitivity was higher in women with iron deficiency [serum ferritin (SF) <15.0 µg/L: 81.6% compared with SF ≥15.0 µg/L: 41.3%], and lower in women with metabolic risk factors, including overweight [BMI ≥25.0 kg/m2: 63.9% vs. BMI <25.0 kg/m2: 78.8%], or elevated CRP (>1.0 mg/L: 67.2% vs. ≤1.0 mg/L: 82.9%), trunk fat (>35%: 62.7% vs. ≤35%: 80.1%), or whole-body fat (>35%: 63.9% vs. ≤35%: 80.3%). CONCLUSIONS: Findings suggest that women with anemia may be incorrectly identified as not anemic via portable hemoglobinometer, and anemia prevalence may be underestimated at the population level.This study was registered at clinicaltrials.gov as NCT04048330.