Clinical-immunological phenotyping and score-based automated diagnostic support for chronic nonspecific upper urinary tract infections: a combined prospective-retrospective diagnostic study

Abstract Background Chronic nonspecific upper urinary tract infections (CNUTIs), including chronic pyelonephritis, are characterized by recurrent inflammation, persistent urinary abnormalities, and variable responses to conventional diagnostic and therapeutic approaches. Standard clinical and laboratory methods may not fully reflect the immunological heterogeneity underlying disease persistence and recurrence. This study developed and internally validated an integrated clinical-immunological phenotyping framework for diagnostic stratification of patients with CNUTIs. Methods The study included 187 participants: 167 patients with CNUTIs and 20 apparently healthy controls. Clinical indicators, routine laboratory parameters, and immunological biomarkers were analyzed, including neutrophil phagocytic activity, phagocytosis completion index, CD16 + CD56+ natural killer cells, CD4/CD8 ratio, serum IgA and IgG, circulating immune complexes, IL-6, IL-8, IL-10, urinary secretory IgA, and urinary neopterin. Multivariate logistic analysis identified predictors associated with phenotype formation. The Clinical and Immunological Phenotype Assessment Scale (CIPAS) was developed, and a score-based automated diagnostic support system (CIPAS-AI) was used to automate phenotype classification. Internal validation used bootstrap resampling with 1,000 iterations to assess the stability of the predefined classification logic under sampling variability; it was not used for model training or cutoff optimization. Results Three clinical-immunological phenotypes were identified: antibacterial defense insufficiency, excessive inflammatory activation, and immune regulation insufficiency. CIPAS showed the highest diagnostic performance, with an AUC of 0.893, sensitivity of 85.6%, and specificity of 82.3%, outperforming the conventional clinical and laboratory approach (AUC = 0.712; sensitivity = 68.4%; specificity = 65.7%). CIPAS-AI also demonstrated high diagnostic performance (AUC = 0.861; sensitivity = 82.4%; specificity = 79.5%), only slightly lower than the full CIPAS scale. Bootstrap validation confirmed stable phenotype classification: antibacterial defense insufficiency was detected in 348 iterations (34.8%), excessive inflammatory activation in 371 iterations (37.1%), and immune regulation insufficiency in 281 iterations (28.1%), with low coefficients of variation. Conclusions Integrated clinical-immunological phenotyping using CIPAS improves diagnostic differentiation among patients with CNUTIs by identifying immune-response patterns beyond conventional clinical and laboratory assessment. CIPAS-AI provides a practical score-based diagnostic support tool for standardized preliminary phenotype classification, although prospective external validation is required before routine clinical implementation.