Objective To determine the role of inter–observer error and the influence of training upon dipstick urine analysis.

Design A two phase observational and training study

Methods Five standard solutions of serum albumin were used to test the accuracy of midwives and nursing auxiliaries involved in dipstick urine analysis at a maternity hospital. The standard solutions were chosen such that they should have resulted in negative (n= 2) and positive (n= 3) dipstick test results, respectively.

Setting A teaching maternity hospital and academic department of obstetrics and gynaecology.

Participants Twenty midwives, 20 nursing auxiliaries and nine laboratory technicians.

Results For the two nonproteinuric solutions, a higher false positive rate was observed for nursing auxiliaries(40% and 55%), compared with midwives (5% and 30%) (P= 0.020 and P= 0.20, respectively). Before training, laboratory technicians recorded high false positive rates (67% and 89%), but after training these were reduced to 0% and 22% (P= 0.25 and P= 0.023, respectively). Both nursing auxiliaries and midwives recorded false negative rates of between 10% and 45% for the three proteinuric solutions.

Conclusions Observer error may be reduced by assigning midwives to urine dipstick analysis or by the implementation of directed training. Classification of pre-eclampsia or other hypertensive diseases of pregnancy on the basis of the presence and degree of proteinuria should be confirmed with a 24–hour quantitative protein collection.