Rapid urine tests, such as microscopy, for bacteria and white cells, and dipsticks, for leucocyte esterase and nitrites, are often used in children that are unwell to guide early diagnosis and treatment of urinary tract infection. We aimed to establish whether these tests were sufficiently sensitive to avoid urine culture in children with negative results and to compare the accuracy of dipsticks with microscopy. Medline, Embase, and reference lists were searched. Studies were included if urine culture results were compared with rapid tests in children. Data were analysed to obtain absolute and relative accuracy estimates. Data from 95 studies in 95 703 children were analysed. Summary estimates for sensitivity and specificity for microscopy for Gram-stained bacteria were 91% (95% CI 80-96) and 96% (92-98), for unstained bacteria were 88% (75-94) and 92% (84-96), for urine white cells were 74% (67-80) and 86% (82-90), for leucocyte esterase or nitrite positive dipstick were 88% (82-91) and 79% (69-87), and for nitrite-only positive dipstick were 49% (41-57) and 98% (96-99). Microscopy for bacteria with Gram stain had higher accuracy than other laboratory tests with relative diagnostic odds ratio compared with bacteria without Gram stain of 8.7 (95% CI 1.8-41.1), white cells of 14.5 (4.7-44.4), and nitrite of 22.0 (0.7-746.3). Microscopy for white cells should not be used for the diagnosis of urinary tract infection because its accuracy is no better than that of dipstick, laboratory facilities are needed, and results are delayed. Rapid tests are negative in around 10% of children with a urinary tract infection and cannot replace urine culture. If resources allow, microscopy with Gram stain should be the single rapid test used.