Mahmood Yousefi1, Azin Nahvijou2, Ali Akbari Sari3, Hosein Ameri4. 1. Iranian Center of Excellence in Health Management, School of Management and Medical Informatics, Health Economics Department, Tabriz University of Medical Sciences, Tabriz, Iran. 2. Cancer Research Center, Cancer Institute, Tehran University of Medical Sciences, Tehran, Iran. 3. Department of Health Management and Economics, School of Public Health, Tehran University of Medical Sciences, Tehran, Iran. 4. Health Policy and Management Research Center, Department of Health Services Management, School of Public Health, Shahid Sadoughi University of Medical Sciences, Yazd, Iran. Electronic address: Hamery7@yahoo.com.
Abstract
OBJECTIVES: Many studies have mapped the QLQ-C30 onto the EQ-5D or the SF-6D utilities; however, these studies were limited to developed countries. So this study aimed to map QLQ-C30 onto the SF-6D version 2 (SF-6D-v2) and EQ-5D-5L using the data collected from colorectal and breast cancer patients in a developing country. METHODS: A cross-sectional data set of 668 inpatient and outpatient cancer patients was gathered from 4 teaching hospitals of cancer treatment in Tehran and Yazd from May 2017 to November 2018. The ordinary least squares (OLS) and censored least absolute deviations (CLAD) models were applied to estimate the utility values of both EQ-5D-5L and SF-6D-V2 using the QLQ-C30. Predicted R2 and adjusted R2 were used to evaluate the goodness of fit of the models. Moreover, the predictive performance of 2 models was assessed through estimating the mean absolute error (MAE), root mean square error (RMSE), intraclass correlation coefficients (ICC), and Spearman's rho. The 10-fold cross-validation method was also applied for validation of models. RESULTS: The OLS Model E4 was the best-performing model for EQ-5D-5L (Adj R2 = 71.7%, Pred R2 = 71.15%, MAE = 0.0770, RMSE = 0.1026), and the OLS Model S4 performed best for SF-6D-V2 (Adj R2 = 74.64%, Pred R2 = 73.86%, MAE = 0.0465, RMSE = 0.0621). CONCLUSION: The OLS Model E4 for EQ-5D-5L and the OLS Model S4 for SF-6D-V2 were the best models for policy makers to have more accurate evaluation of the healthcare interventions when the data are gathered through non-preference-based instruments.
OBJECTIVES: Many studies have mapped the QLQ-C30 onto the EQ-5D or the SF-6D utilities; however, these studies were limited to developed countries. So this study aimed to map QLQ-C30 onto the SF-6D version 2 (SF-6D-v2) and EQ-5D-5L using the data collected from colorectal and breast cancerpatients in a developing country. METHODS: A cross-sectional data set of 668 inpatient and outpatientcancerpatients was gathered from 4 teaching hospitals of cancer treatment in Tehran and Yazd from May 2017 to November 2018. The ordinary least squares (OLS) and censored least absolute deviations (CLAD) models were applied to estimate the utility values of both EQ-5D-5L and SF-6D-V2 using the QLQ-C30. Predicted R2 and adjusted R2 were used to evaluate the goodness of fit of the models. Moreover, the predictive performance of 2 models was assessed through estimating the mean absolute error (MAE), root mean square error (RMSE), intraclass correlation coefficients (ICC), and Spearman's rho. The 10-fold cross-validation method was also applied for validation of models. RESULTS: The OLS Model E4 was the best-performing model for EQ-5D-5L (Adj R2 = 71.7%, Pred R2 = 71.15%, MAE = 0.0770, RMSE = 0.1026), and the OLS Model S4 performed best for SF-6D-V2 (Adj R2 = 74.64%, Pred R2 = 73.86%, MAE = 0.0465, RMSE = 0.0621). CONCLUSION: The OLS Model E4 for EQ-5D-5L and the OLS Model S4 for SF-6D-V2 were the best models for policy makers to have more accurate evaluation of the healthcare interventions when the data are gathered through non-preference-based instruments.