Assessment of Interobserver Reliability of Nephrologist Examination of Urine Sediment
Palsson R, Colona MR, Hoenig MP, Lundquist AL, Novak JE, Perazella MA, and Waikar SS. Assessment of Interobserver Reliability of Nephrologist Examination of Urine Sediment. JAMA Netw Open 2020; 3(8):e2013959.
JAMA Netw Open
Importance: Urine sediment microscopy is commonly performed during the evaluation of kidney disease. Interobserver reliability of nephrologists' urine sediment examination has not been well studied.
Objective: Assess interobserver reliability of the urine sediment examination.
Design, Setting, and Participants: In this diagnostic test study, urine samples were prospectively collected from a convenience sample of adult patients from an academic hospital in the United States undergoing kidney biopsy from July 11, 2018, to March 20, 2019. Digital images and videos of urine sediment findings were captured using a bright-field microscope. These images and videos along with urine dipstick results were incorporated in online surveys and sent to expert nephrologists at 15 US teaching hospitals. They were asked to identify individual sediment findings and the most likely underlying disease process.
Exposures: Urine dipstick results and urine sediment images from patients undergoing native kidney biopsy.
Main Outcomes and Measures: Interobserver reliability of urine sediment microscopy findings estimated by overall percent agreement and Fleiss κ coefficients. Secondary outcomes included concordance of diagnoses suspected by nephrologists with corresponding kidney biopsy results.
Results: In total, 10 surveys from 10 patients containing 76 study questions on individual features were sent to 21 nephrologists, 14 (67%) of whom completed them all. Their combined 1064 responses were analyzed. Overall percent agreement for casts was an estimated 59% (95% CI, 50%-69%), κ = 0.52 (95% CI, 0.42-0.62). For other sediment findings, overall percent agreement was an estimated 69% (95% CI, 61%-77%), κ = 0.65 (95% CI, 0.56-0.73). The κ estimates ranged from 0.13 (95% CI, 0.10-0.17) for mixed cellular casts to 0.90 (95% CI, 0.87-0.94) for squamous epithelial cells.
Conclusions and Relevance: In this study, substantial variability occurred in the interpretation of urine sediment findings, even among expert nephrologists. Educational or technological innovations may help improve the urine sediment as a diagnostic tool.