OBJECTIVE: To determine whether the HemoCue haemoglobin value measured in fingertip skin puncture blood corresponds to the reference value measured in venous blood. DESIGN: Prospective. METHOD: In two health centres, patients' blood was first drawn from a fingertip skin puncture and the haemoglobin was measured with the HemoCue method (HemoCue B-Haemoglobin AB photometer, Angelholm, Sweden). The same patients were sent to the regional laboratory for laboratory haemoglobin determination. Agreement between the two haemoglobin values was assessed using the method of Bland and Altman. RESULTS: Both haemoglobin measurements were carried out in 58 patients. The mean HemoCue haemoglobin was 8.0 mmol/l (95% CI: 7.6-8.4) and the mean venous haemoglobin was 8.2 mmol/l (95% CI: 7.9-8.6). Of all values, 2 were above the level of agreement of 2 SD and 17 values were above the level of 1 SD. The sensitivity of the HemoCue measurement was 81% (95% CI: 62-100) and the specificity 95% (95% CI: 88-100). In the population investigated, with a prevalence of anaemia of 28%, the predictive value of a positive HemoCue result was 87% and of a negative result 93%. CONCLUSION: According to the test characteristics, the HemoCue is a good device for haemoglobin determination. However, in several cases there is a significant difference between the haemoglobin measured with the HemoCue method and the laboratory haemoglobin value. If a reliable haemoglobin value is needed, a laboratory venous haemoglobin assessment is preferred.