Application Of Mapping In Ankylosing Spondylitis Patients’ Health-Related Quality Of Life

Ankylosing Spondylitis(AS)is a kind of autoimmune disease that can seriously affect patients’ health-related quality of life.The measurement of health utilities is the key and basis of Health Technology Assessment,but the direct measurements and the indirect measurements both have limitations,especially the indirect measurements are not sensitive to the health status of patients.Therefore,the "mapping" was used in this study to construct a mapping model based on the AS-specific scale to obtain more accurate health utilities.Hitherto,frequency theoretical models and Bayesian networks have been used in the mapping as a kind of frontier approach.This study aimed to construct mapping models of ASQoL,BASFI,and BASDAI to EQ-5D-5L and SF-6D(SF-12)based on Chinese population preference,and select the optimal models through a variety of goodness of fit measurement indexes and obtain the health utilities of patients with AS.From December 2019 to October 2020,243 AS inpatients in Qilu Hospital of Shandong University were surveyed by questionnaire.The patients filled in the above five scales and social demographic information.In frequency theoretical mapping models,five-fold crossvalidation was used to divide the research objects into construction set and verification set.Six frequency theoretical models,including OLS,Tobit,CLAD,GLM,OLOGIT,and MLOGIT,were adopted.In the Bayesian Network,the parent nodes and child nodes of the starting scale and the target scale were searched to obtain the structure and conditional probability of the Bayesian Network.To construct a mapping model for patients with AS.Five goodness of fit measurement indexes including R2,MAE,RMSE,AIC,and BIC were selected to judge the prediction ability of models to select the optimal model.A total of 240 patients were included in the study,with an average age of 36.199±9.55 years.45.8%of the patients had a disease course of more than 6 years,indicating that the sample patients were generally young and had a long disease course.Kruskal-Wallis test results showed that there was no statistically significant difference between the social-demographic information of the constructed set and the validation set.The health utility values based on EQ5D-5L and SF-6D(SF-12)scales were 0.73910.268 and 0.572±0.068,respectively.ASQoL,BASDAI,and BASFI scores were 6.949±5.815,3.604±2.373,and 3.198±2.612,respectively.Compared with SF-6D(SF-12)utility value,the EQ-5D-5L utility value was more strongly correlated with ASQoL,BASDAI,and BASFI scores,with all the correlation coefficients above 0.8.The Bayesian Network of BASDAI and BASFI to SF-6D(SF-12)dimensions was a single line structure.The EQ-5D-5L dimensions of mobility and usual activities were related to most BASDAI/BASFI items.The Bayesian Network from ASQoL items to EQ-5D-5L anxiety/depression dimension was the most complex,involving 7 child nodes.The number of physiological and psychological items on the ASQoL scale was balanced.Except for the social function dimension in SF-6D(SF-12),the Bayesian Network of other dimensions is not a single line structure.On the whole,among the mapping models based on frequency theory,OLS and MLOGIT have the most goodness of fit indicators,which were the optimal fitting models of direct mapping and indirect mapping respectively.Combined with Monte Carlo Simulation results and scatter plot,Bayesian Network shows better prediction performance than OLS and MLOGIT,which can be promoted in mapping.After the modification of mapping models for the health utilities of AS patients,the health utility value of AS patients without serious complications could be determined in the range of 0.730-0.750.