Work

The influence of hemoglobin measurement error on accuracy of population anemia surveillance: results from mathematical simulations

Year: 2025

Type: article

Abstract: BACKGROUND: Anemia is of public health concern among females of reproductive age and young children. Population anemia surveys are challenging because of inaccuracy and imprecision in common field-friendly assessment methods. However, there is uncertainty about how much error can be tolerated for effective surveillance. OBJECTIVES: Our objective was to quantify the influence of random and systematic errors in hemoglobin (Hb) measurement on prevalence bias under varying population characteristics using mathematical simulations. METHODS: We conducted a literature review to identify parameters of interest: means and standard deviations (SDs) in capillary and venous blood, the mean bias (capillary mean - venous mean), and the SD inflation ratio (capillary SD/venous SD). We generated reference Hb distributions with varying means (105-135 g/L) and SDs (10-16 g/L), assuming measurement using venous blood on an autoanalyzer. We then generated capillary blood comparison distributions with mean biases ranging from -5 to +11 g/L and SD inflation ratios ranging from 0.5 to 2.0. Using target prevalence accuracies of ±5 and ±10 percentage points (pp), we identified maximum tolerable mean biases and SD inflation ratios. RESULTS: Among 92 contrasts reported in 16 studies, 90% of mean biases were between -6.6 g/L and 10.7 g/L, and 90% of SD inflation ratios were between 0.79 and 1.54. In simulations, the prevalence bias was heavily influenced by the mean Hb bias, but less by the SD inflation ratio. To achieve the target prevalence accuracy, the mean bias would need to be within ±1 g/L (for 5 pp) or ±2.5 g/L (for 10 pp) and SD inflation ratio between 0.85 and 1.20 (for 5 pp) or 0.7-1.5 (for 10 pp). CONCLUSIONS: Random error (SD inflation ratio) has less influence than systematic error (mean bias) on anemia prevalence estimates. Efforts to reduce mean bias during data collection or through statistical adjustment may have a greater impact on anemia prevalence assessment.

Source: American Journal of Clinical Nutrition

DOI: 10.1016/j.ajcnut.2025.101185

Authors: Christine P Stewart, Charles D Arnold, Sika M Kumordzie, Reina Engle-Stone

Institution: University of California, Davis

Language: English

Topic: Clinical Laboratory Practices and Quality Control

Subfield: Physiology

Field: Medicine

Domain: Health Sciences

SDG: Good health and well-being

Open Access status: hybrid

Estimated APC paid: $5,200