Mapping the EORTC QLQ-C30 and QLQ-LC13 to the SF-6D utility index in patients with lung cancer using machine learning and traditional regression methods.

Background: Preference-based measures of health-related quality of life (HRQoL), such as the Short Form Six-Dimension (SF-6D) is essential for health economic evaluations. However, these measures are rarely included in clinical trials for lung cancer. This study aims to develop mapping algorithms to predict SF-6D health utility scores from the European Organization for Research and Treatment of Cancer Quality of Life Questionnaire Core (EORTC QLQ-C30) and Quality of Life Questionnaire-Lung Cancer 13 (QLQ-LC13).

Method: The study sample comprised a Chinese population with lung cancer (n = 625). Traditional regression techniques, including Ordinary Least Squares regression, Generalized Linear Model, as well as machine learning techniques, such as Gradient Boosting Tree, Support Vector Regression, Ridge Regression are used. Five-fold cross-validation was performed. The performance metrics used to evaluate the models including R, root mean square error (RMSE),mean absolute error (MAE) and mean absolute percentage error (MAPE) were used to screen the optimal model.

Results: The mean and median of SF-6D health utility values were 0.774 (SD = 0.154) and 7.795, respectively. The model with the best mapping performance was the Ridge regression model Five-fold cross-validation (CV) results show that the Ridge regression model has the best mapping performance, the final prediction indexes are R = 0.753, RMSE = 0.074, MAE = 0.057, MAPE = 8.169%.

Conclusions: This study developed an optimized mapping algorithm to predict the utility index from the QLQ-C30 QLQ-LC13 to the SF-6D. This algorithm offers provides an effective alternative for estimating SF-6D estimation when the preference-based health utility values are unavailable.