Impact of computed tomography radiomics-based predictive model for immunotherapy efficacy in non-small cell lung cancer on predictive nursing outcomes

The aim of this study was to investigate the feasibility of predicting the efficacy of immunotherapy in non-small cell lung cancer (NSCLC) patients using computed tomography (CT) radiomics, and to evaluate the impact of predictive nursing interventions on the quality of life (QOL) and complication rates in these patients. A total of 68 NSCLC patients treated in hospital were included and randomly assigned into two groups: control group (routine care) and intervention group (predictive nursing intervention in addition to routine care), with 34 patients in each group. All patients underwent contrast-enhanced CT scanning, with the tumor region of interest (ROI) delineated. Intensity, morphological features (sphericity), and texture characteristics (gray-level co-occurrence matrix) were extracted. A predictive model for immunotherapy efficacy was established, and the performance of logistic regression (LR), support vector machine (SVM), K-nearest neighbor (KNN), and Boosting algorithms was compared in terms of accuracy, specificity, precision, recall, and F-score. Additionally, the time of drain removal, hospital stay, post-care QOL scores (including physical function, psychological function, social function, and material life), and complication rates were recorded for both groups. In the prediction model, the LR model performed better on multiple key indicators, with higher accuracy (0.787 vs. 0.771/0.752/0.754), specificity (0.525 vs. 0.352/0.256/0.236), precision (0.895 vs. 0.753/0.761/0.778), recall (0.889 vs. 0.762/0.858/0.793), and F1 score (0.892 vs. 0.757/0.807/0.785) than SVM, KNN, and AdaBoost algorithms. Regarding nursing interventions, the intervention group had significantly shorter drain removal time (52.39 ± 4.32 h) and hospital stay (11.82 ± 1.84 days) compared to the control group (67.93 ± 5.76 h and 17.92 ± 2.91 days, respectively), with statistically significant differences ( P < 0.05). Regarding nursing interventions, no significant difference was observed in the baseline QOL scores between the two patient groups at enrollment ( P > 0.05), indicating comparability. Post-care, the intervention group showed significantly higher scores in physical function, psychological function, social function, and material life quality compared to the control group ( P < 0.05). The incidence of complications in the intervention group (2.94 %) was lower than that in the control group (17.65 %), but the difference did not reach statistical significance ( P > 0.05). CT imaging omics technology has shown potential as an auxiliary tool for predicting the efficacy of immunotherapy in NSCLC patients. The predictive nursing model studied in parallel within the same cohort indicates that this prospective nursing approach can significantly improve patient outcomes. This integration approach highlights a promising “prediction to care” framework worth further exploration.