We developed a colour tint scale to use as an aid in the clinical assessment of anaemia by measuring conjunctival pallor. The objectives of this study were to evaluate the accuracy and agreement among observers in detecting anaemia in three sequential phases with incremental information using clinical pallor of different anatomical sites, subsequently adding subjects' medical history for physical symptoms and the colour scale. After training in the application of these three sequential assessments, 12 primary health workers were assigned to independently examine 198 anaemic and 254 non-anaemic pregnant women while blind to the true anaemic status. Their assessments in each phase were then compared with the anaemic status based on haemoglobin level, measured using HemoCue, taken as the gold standard, to determine sensitivity and specificity, and agreements among observers in detecting anaemia were calculated. In the three sequential phases of assessment the sensitivities were 73.8, 78.3, 82.9% and specificities 76.0, 84.7 and 90.9%, respectively. In each subsequent step, the improvements in both the sensitivity and specificity were statistically significant [P(chi2(McNemar)) < 0.01]. Kappa statistics for agreement among 12 observers for assessing anaemia in the sequential phases were 0.50, 0.71 and 0.82, respectively. The Spearman rank correlation coefficient between haemoglobin level and the colour scale reading was 0.68 (P < 0.001). Taking medical history and incorporating a simple colour tint scale with examination of pallor improved the sensitivity, specificity and agreement for detection of anaemia by health workers.