Comparing visual and automated urine dipstick analysis in a general practice population

Abstract

Introduction Urinary symptoms constitute the primary reason for female patients to consult their general practitioner. The urinary dipstick test serves as a cornerstone for diagnosing urinary tract infections (UTIs), yet traditional visual interpretation may be subject to variability. Automated devices for dipstick urinalysis are routinely used as alternatives, yet the evidence regarding their accuracy remains limited. Therefore we aimed to compare concordance between visual and automated urinary dipstick interpretation and determine their test characteristics for the prediction of bacteriuria.Material and methods We conducted a prospective validation study including urine samples originating from adult patients in general practice that were sent to the Maastricht Medical Centre + for urinary culture. Urinary dipstick tests were performed on each sample, which were interpreted visually and automatically. We calculated Cohen’s κ and percentage agreement and used 2 × 2 tables to calculate test characteristics.Results We included 302 urine samples. Visual and automated analysis showed almost perfect agreement (κ = 0.82 and κ = 0.86, respectively) for both nitrite and leukocyte esterase, but moderate agreement for erythrocytes (κ = 0.51). Interpretation of clinically relevant (nitrite and/or leukocyte esterase positive) samples showed almost perfect agreement (κ = 0.88). Urinary dipsticks show similar test characteristics with urinary culture as gold standard, with sensitivities of 0.92 and 0.91 and specificities of 0.37 and 0.41 for visual and automated interpretation respectively.Conclusion Automated and visual dipstick analysis show near perfect agreement and perform similarly in predicting bacteriuria. However, automated analysis requires maintenance and occasionally measurement errors can occur.

Keywords

• Urinary tract infections
• point-of-care testing
• general practice
• primary health care
• diagnostic equipment