Construction and Validation of a Risk Prediction Model for Cancer‐Related Cognitive Impairment in Lung Cancer Patients
Mengyuan Qiao, Li Luo, Hui ZhangABSTRACT
Background
Cancer‐related cognitive impairment (CRCI) is a major clinical challenge faced by lung cancer patients during or after treatment. Early identification of at‐risk populations by healthcare professionals is inadequate, and little is known about measures that can be taken to enhance their prevention. Existing systematic reviews and meta‐analyses have summarized common risk factors for CRCI in lung cancer patients, but integrated predictive models based on holistic theoretical frameworks remain scarce.
Aim
To construct a visual assessment tool for the identification of CRCI in lung cancer survivors based on the theory of unpleasant symptoms (TOUS), complementing existing predictive models with a multidimensional theoretical perspective.
Design
A prospective, observational, single‐centre study.
Methods
The present study was conducted in a major hospital in Urumqi, China, between October 2023 and July 2024. A total of 350 lung cancer survivors participated in this survey, which was divided into a training and validation group in a 7:3 ratio. Lasso regression and logistic regression analyses were employed to identify the risk factors for CRCI, construct a nomogram prediction model and test the prediction effect in the validation set. Model performance was evaluated using the area under the curve ( AUC ) and goodness‐of‐fit statistics, and the model was internally validated.
Results
A total of 350 lung cancer patients, comprising 245 in the training and 105 in validation groups, were included. Of these, 117 (33.4%) experienced CRCI. The predictive model identified significant predictors, including age, pathological stage, chemotherapy, post‐traumatic stress disorder (PTSD), depression and social support scores. At the 32.3% optimal cut‐off, the model had AUC values of 0.863 and 0.818 in the training and validation groups. Calibration plots demonstrated a strong correlation between predicted and observed rates, and decision curve analysis revealed optimal net benefit at threshold probabilities ranging from 10% to 80%.
Conclusions
The risk prediction model constructed in this study, based on TOUS, demonstrates satisfactory predictive ability superior to some existing models. It integrates physiological, psychological and environmental factors, serving as a valuable complementary tool for healthcare professionals in identifying high‐risk groups, particularly in clinical settings emphasizing holistic symptom management.