Interobserver Reliability of Urine Sediment Interpretation

Abstract

Urine sediment interpretation is frequently used in the evaluation of patients with kidney disease. There has been no systematic evaluation of the reliability of this diagnostic maneuver. Digital photographs of urine sediment images were acquired from 165 consecutive patients being evaluated by the nephrology consultation service at a tertiary care hospital. Urine sediment images of 100 patients were randomly selected; 86 patients had images that were deemed to be of sufficient quality, and one image per patient was chosen for inclusion in an internet-based questionnaire. For each image, the presence or absence of 14 potential urinary structures was ascertained. Ten nephrologists (senior readers [n = 3]: >10 yr of experience; intermediate readers [n = 3]: 1 to 10 yr of experience; and junior readers [n = 4]: first year of practice) completed the questionnaire. For each urinary structure, we measured the rate of complete agreement among the readers as well as the kappa statistic as a marker of agreement beyond chance. Unanimous agreement was highest (79.1%) regarding the presence of broad and fatty casts and poorest (31.4%) for the identification of dysmorphic red blood cells and white blood cells. Interobserver agreement was best for squamous epithelial cells (kappa = 0.54) and hyaline casts (kappa = 0.52) and worst for transitional epithelial cells (kappa = 0.14) and fatty casts (kappa = 0.06). When assessed within strata of physician experience, interobserver agreement was not associated with seniority. Nephrologists achieved slight to moderate agreement in the identification of structures that are commonly observed in the urine sediment.