Objectives  This study evaluates the diagnostic accuracy of Haemoglobin Colour Scale (HCS), compared with clinical diagnosis, to detect anaemia and severe anaemia in preschool‐age children attending primary healthcare clinics in rural Zanzibar.

Methods  In all participants, haemoglobin (Hb) concentration was independently estimated by clinical examination for palmar pallor, HCS and HemoCue™. HemoCue was considered the reference method. Data collection was integrated into the usual health services and performed by local healthcare workers (HCWs). Sensitivity, specificity, positive and negative predictive values were calculated for HCS and clinical examination for palmar pallor. The limits of agreement between HCS and HemoCue, and inter‐observer variability for HCS, were also defined.

Results  A total of 799 children age 2–59 months were recruited to the study. The prevalence of anaemia (Hb < 11 g/dl) and severe anaemia (<5 g/dl) were 71% and 0.8% respectively. The sensitivity of HCS to detect anaemia was 33% [95% confidence interval (CI) 29–36] and specificity was 87% (83–91). The sensitivity of HCS to detect severe anaemia was 14% (95% CI 0–58) and specificity was 100% (99–100). The sensitivity of palmar pallor to detect anaemia was low, but superior to HCS (58%vs. 33%, P < 0.001); specificity was inferior to HCS (55%vs. 87%, P < 0.001). There was no evidence of a difference in either sensitivity (P > 0.1) or specificity (P > 0.1) between HCS and palmar pallor to detect severe anaemia.

Conclusions  Haemoglobin Colour Scale does not improve the capacity of HCWs to diagnose anaemia in this population. Accuracy is limited by considerable variability in the performances of test operators. However, optimizing the training protocol for those using the test may improve performance.