Establishment and Validation of a Predictive Model for Post-Treatment Anxiety Based on Patient Attributes and Pre-Treatment Anxiety Scores.

Abstract

In this study, we aim to establish and evaluate a predictive model for post-treatment anxiety state based on basic patient attributes and pre-treatment SAS scores, with the expectation that this model will guide clinical precision intervention. Data were collected from 606 patients with breast cancer who underwent surgery at our hospital between January 1, 2015 and December 30, 2018 and 144 newly diagnosed patients with breast cancer who were admitted between June 1, 2019 and December 30, 2019, for a total of 750 patients with breast cancer. The relationship between SAS_A scores and prognosis was verified by analyzing patient baseline characteristics, follow-up data, pre-treatment self-rating anxiety scale (SAS) scores, and SAS_A scores in follow-up period after the end of treatment. A risk prediction model was developed in view of the SAS_A scores, which was then screened, validated, and simplified by scoring, with a nomogram plotted. The SAS_A score can be utilized to differentiate prognosis. In K-M analysis, the high SAS_A score group had a significantly poorer progression-free survival rate than the low score group, p-value < 0.0001. Through model feature selection and clinical analysis, all variables were finally incorporated to establish a predictive model with a ROC AUC of 0.721 (0.637-0.805) for the validation set and external data, and an AUC of 0.810 (0.719-0.902) for external data, demonstrating good predictive performance. Calibration curves and probability distribution maps were constructed. DCA and CIC analyses demonstrated that model intervention could boost clinical benefits more effectively than intervention for all patients. Using a predictive model to guide clinical management for anxiety in breast cancer patients is feasible, but additional research is required.