Development of the patient-reported outcomes (PRO) based survival prediction model for survivors who underwent lung cancer surgery.

Abstract

e23084 Background: To construct a prognostic model of 5-year survival among disease-free survivors who underwent lung cancer surgery using socio-clinical and patient-reported outcomes (PRO), and to compare its predictive performance with that of a traditional model based on known clinical variables. Methods: Data on 809 survivors who underwent lung cancer surgery between 2001 and 2006 in two Korean tertiary teaching hospitals were used. The training data set was utilized to generate the prediction model and the remaining 20% was employed as a testing set to estimate the model’s accuracy. Three Cox proportional hazard regression models were constructed and compared that of 5-year survival prediction ability through the evaluation of their performance in terms of discriminative and calibration ability. The three models were constructed with: 1) only clinical and socio-demographic variables, 2) only PROs, and 3) variables from model 1 and 2 considered altogether. The performance of each 5-year survival prediction model was evaluated using C-statistics and Hosmer-Lemeshow-type χ2-statistical analyses. Results: From the validation set, the C-statistics for the model 1, 2, and 3 were 0.70 (95% confidence interval [CI], 0.67−0.73), 0.77 (95% CI, 0.74−0.80), and 0.81 (95% CI, 0.78−0.84), respectively. In this study, model 3 (including PRO and other variables together) showed the highest discriminative and calibration ability compared to others. Conclusions: The findings suggested that the PRO included model in addition to clinical and socio-demographic variables, is more accurate in the survival prediction of lung cancer survivors than models constructed with only well-known socio-clinical variables.