Comprehensive prediction of urolithiasis based on clinical factors, blood chemistry and urinalysis: UROLITHIASIS score

Abstract

AbstractComprehensive prediction of urolithiasis using available factors obtained in the emergency department may aid in patient-centered diagnostic imaging decisions. This retrospective study analyzed the clinical factors, blood chemistry and urine parameters of patients who underwent nonenhanced urinary computed tomography for suspected urolithiasis. A scoring system was developed from a logistic regression model and was tested using the area under the curve (AUC). The prevalence of urolithiasis and important possible causes in the three risk subgroups were determined. Finally, the scoring model was validated. In the derivation cohort (n = 673), 566 patients were diagnosed with urolithiasis. Age > 35 years, history of urolithiasis, pain duration < 8 h, nausea/vomiting, costovertebral angle tenderness, serum creatinine ≥ 0.92 mg/dL, erythrocytes ≥ 10/high power field, no leukocytes ≤  + , and any crystalluria were retained in the final multivariable model and became part of the score. This scoring model demonstrated good discrimination (AUC 0.808 [95% CI, 0.776–0.837]). In the validation cohort (n = 336), the performance was similar (AUC 0.803 [95% CI, 0.756–0.844]), surpassing that of the STONE score (AUC 0.654 [95% CI, 0.601–0.705], P < 0.001). This scoring model successfully stratified patients according to the probability of urolithiasis. Further validation in various settings is needed.