Construction and internal assessment of an exploratory risk prediction model for psychological distress in brain tumor patients: a cross-sectional study

This study aimed to identify determinants of psychological distress (PD) in patients with brain tumors, determine relevant risk factors, and develop an exploratory nomogram-based predictive model with internal assessment via bootstrapping. A total of 321 brain tumor patients admitted to the Department of Neurosurgery at a tertiary-grade A cancer hospital between April 2023 and February 2025 were recruited via convenience sampling. Univariate and multivariate logistic regression analyses were conducted to identify factors associated with PD. Patients were categorized into low distress thermometer (DT < 4) and high distress thermometer (DT ≥ 4) groups based on symptom-related PD levels. A nomogram prediction model was developed using the rms package in R (version 4.3.1), and model performance was assessed through ROC analysis, calibration curves, and the Hosmer–Lemeshow goodness-of-fit test. The final prediction model included eight variables: marital status, education level, monthly income, payment method, disease duration, financial toxicity (COST), self-efficacy (GSES), and medical coping mode (MCMQ). The model exhibited good apparent discriminative ability (AUC = 0.840, 95% CI: 0.795–0.885, P < 0.05), with a sensitivity of 0.867 and specificity of 0.678 at the optimal cutoff of 0.513. Internal validation via 1000 bootstrap resamples yielded an optimism-corrected AUC of 0.823 (95% CI: 0.774–0.872), a bootstrapped calibration slope of 0.942 (95% CI: 0.831–1.053), and an optimism-corrected Brier score of 0.187, indicating minimal overfitting. Calibration curve analysis showed the calibration line closely approximated the ideal 45° line, and the Hosmer–Lemeshow test revealed no significant difference between predicted and observed PD incidence (P = 0.371), confirming adequate goodness-of-fit. The exploratory nomogram developed in this study exhibits acceptable internal performance for PD risk stratification in brain tumor patients. As a single-center, cross-sectional model, it requires further internal validation (e.g., cross-validation) and external validation across diverse populations to confirm generalizability. It provides a preliminary tool for clinicians to identify high-risk patients and implement targeted nursing interventions.