Purpose The aim of this study was to establish a model that would enable healthcare providers to use routine follow-up measures of peritoneal dialysis to predict frailty in those patients. Design A cross-sectional design with Logistic regression and XGBoost machine learning algorithms analysis. Methods One hundred and twenty-three cases of peritoneal dialysis patients who underwent regular follow-up at our center were included in this study. We use the FRAIL scale to confirm the frailty of the patients. Clinical and Laboratory data were obtained from the peritoneal dialysis registration system. Factors associated with patient Frailty were identified through regularized logistic regression and validated using an XGBoost model. The final selected variables were in-cluded in the unregularized Logistic Regression to construct the model Findings A total of 123 patients were reviewed in this study, with an average age of 61.58 years, and the median dialysis Duration was 38.5(18.07,60.53) months. 39 patients (31.71%) were female, 54 PD patients (43.9%) were classified as frail. Age, Ferritin, and TCH are the top three im-portant features labeled by the XGBoost. The results are consistent with the regularized logistic regression. Conclusions In this study, age, total cholesterol, and ferritin are the most important features associated with the frailty in peritoneal dialysis patients. This model can be used to predict frailty status and help health monitoring of peritoneal dialysis patients. Clinical Evidence Logistic regression and XGBoost machine learning algorithms can be used to construct a predictive model of frailty in peritoneal dialysis patients. The model could provide doctors with an objective tool to find frailty in peritoneal dialysis patients. As the data is obtained from routine examinations, the prediction model will not bring additional burden to the work of doctors or nurses.

The rapid adoption of generative artificial intelligence (AI) tools in higher education is transforming how students engage with undergraduate mathematics, raising concerns about learning and assessment validity. This study examines the impact of AI accessibility across a two-semester, multi-course dataset including Business Calculus, Linear Algebra, and Calculus III. By comparing unproctored homework and proctored exam performance, we analyze how student learning behaviors shift in AI-accessible environments, particularly through epistemic off-loading of mathematical work. Guided by a sociocognitive framework, we employ complementary measures -- performance gaps, homework-exam correlations, and Wasserstein distance -- to characterize divergence between practice and mastery. Results reveal a growing integrity gap as course content shifts from procedural to conceptual and spatially intensive mathematics. In both Business Calculus and Linear Algebra, differences in homework format (online versus hand-written, TA-graded) do not yield substantively different performance patterns, indicating that paper-based homework is not inherently more resistant to AI-mediated offloading. While homework retains partial predictive validity in procedural courses, upper-division courses exhibit a collapse in alignment between homework and exams, indicating that unproctored assessments increasingly reflect synthetic fluency rather than internalized understanding. These findings highlight the need to rethink assessment practices in the AI era.