A novel comprehensive predictive model for obstructive pyonephrosis patients with upper urinary tract stones.

BACKGROUND

Calculous pyonephrosis tended not to be accurately diagnosed before operations. It is mostly confirmed during percutaneous nephrolithotripsy or percutaneous nephrostomy. We aimed to evaluate the risk factors for predicting obstructive pyonephrosis patients with upper urinary tract stones.

METHODS

Clinical data of 322 patients with upper urinary tract stones and obstructive hydronephrosis were retrospectively searched and analyzed in our study. The patients were divided into two groups; pyonephrosis and non-pyonephrosis groups. Both disease related factors and infection-associated indicators were analyzed. Univariate and multivariate logistic analyses were performed on preoperative variables. Accordingly, ROC curves were drawn, and a novel comprehensive model was constructed to predict the pyonephrosis.

OUTCOMES

Compared to the non-pyonephrosis group, patients in the pyonephrosis group showed statistical differences in sex, urinary tract infection (UTI) within 3 months, stone density, computerized tomography (CT) value of hydronephrosis, serum creatinine, hydronephrosis, contralateral kidney severe hydronephrosis or atrophy, preoperative white blood cells, neutrophils, serum C-reactive protein, urine leukocyte, nitrite, and urine culture revealed statistical difference (<0.05). Univariate analysis showed that there were significant differences for sex, UTI history, degree of hydronephrosis, contralateral severe hydronephrosis or atrophy, serum creatinine, and CT value of hydronephrosis (<0.001). Multivariate analysis demonstrated several independent risk factors for pyonephrosis, including degree of hydronephrosis (=0.02), CT value of hydronephrosis (=0.001), urine leukocyte (=0.002), urine culture (=0.001) and blood neutrophils (=0.009). Based on these risk factors, we constructed a novel comprehensive model and confirmed it was an effective method to predict pyonephrosis (AUC, 0.970). Bootstrapped calibration curves showed no untoward deviation in both training and validation dataset (mean absolute error of 0.027, 0.036).

CONCLUSIONS

Hydronephrosis, CT value of hydronephrosis, blood neutrophils, urine leukocyte, and urine culture were independent risk factors to predict pyonephrosis. The novel comprehensive model was found to be an effective method to predict pyonephrosis and needed to be further confirmed in prospective studies.