Abstract
OBJECTIVE: To analyze the pathogen distribution of older type 2 diabetes mellitus (T2DM) patients with urinary tract infection (UTI) and to develop and validate the feasibility of a nomogram risk prediction model for older T2DM patients with UTI. METHODS: We retrospectively analyzed clinical data from older patients with T2DM admitted to the Department of Endocrinology of The First People's Hospital of Lianyungang City from December 2023 to December 2024. Random number sequences were generated using R software, and all patients were assigned to the modeling cohort and validation cohorts in a ratio of 7:3 through simple random sampling. We compared baseline features between training and validation sets using appropriate statistical tests. We employed univariate and multivariate logistic regression models to identify factors independently associated with UTI in older patients with T2DM. Subsequently, we developed a nomogram prediction model, which was then validated using the validation group. Four methods of statistics, namely, the Hosmer-Lemeshow goodness-of-fit test (H-L test), subjects' work characteristic curve and area under the curve (AUC), calibration curve, and clinical decision curve (DCA), were comprehensively applied to evaluate the model's fit, discrimination, calibration, and clinical utility. RESULTS: Among 521 older patients with T2DM, 82 developed UTI, with an incidence of 15.74%. Logistic regression analyses identified female (odds ratio [OR] = 2.53, 95% confidence interval [CI]: 1.25-5.12, P = 0.011), HbA1c (OR = 1.35, 95% CI: 1.15-1.59, P = 0.001), SGLT-2i in previous year (OR = 2.29, 95% CI: 1.19-4.41, P = 0.013), indwelling urinary catheter (OR = 3.04, 95% CI: 1.02-9.23, P = 0.048), and UTI in previous year (OR = 5.22, 95% CI: 2.22-12.25, P = 0.001) as independent risk factors for older T2DM patients with UTI. The AUC was 0.764 (95% CI: 0.683-0.898) in the training group and 0.779 (95% CI: 0.703-0.855) in the validation group. The H-L test results for the training (χ² = 9.834, P = 0.277) and validation (χ² = 5.432, P = 0.711) groups indicated good goodness-of-fit for the prediction model, with the internal validation curve demonstrating satisfactory performance. CONCLUSION: Establishing this simple, intuitive nomogram prediction model enables the early identification of risk factors in older T2DM patients with UTI. Based on readily accessible clinical variables, this model facilitates early risk stratification and promotes preventive interventions in a timely manner, ultimately reducing UTI incidence in clinical practice. Furthermore, the identified pathogen distribution and antimicrobial susceptibility profiles directly support targeted UTI treatment in this high-risk population, further enhancing the clinical utility of the predictive model and potentially improving patient outcomes.