Wong Carlos King Ho, Cheung Prudence Wing Hang, Samartzis Dino, Luk Keith Dip-Kei, Cheung Kenneth M C, Lam Cindy Lo Kuen, Cheung Jason Pui Yin
Department of Family Medicine and Primary Care, The University of Hong Kong, Hong Kong SAR, China.
Department of Orthopaedics and Traumatology, The University of Hong Kong, Pokfulam, Hong Kong, SAR, China.
PLoS One. 2017 Apr 17;12(4):e0175847. doi: 10.1371/journal.pone.0175847. eCollection 2017.
This is a prospective study to establish prediction models that map the refined Scoliosis Research Society 22-item (SRS-22r) onto EuroQoL-5 dimension 5-level (EQ-5D-5L) utility scores in adolescent idiopathic scoliosis (AIS) patients. Comparison of treatment outcomes in AIS can be determined by cost-utility analysis. However, the mainstay spine-specific health-related quality of life outcome measure, the SRS-22r questionnaire does not provide utility assessment. In this study, AIS patients were prospectively recruited to complete both the EQ-5D-5L and SRS-22r questionnaires by trained interviewers. Ordinary least squares regression was undertaken to develop mapping models, which the validity and robustness were assessed by using the 10-fold cross-validation procedure. EQ-5D-5L utility scores were regressed on demographics, Cobb angle, curve types, treatment modalities, and five domains of the SRS-22r questionnaire. Three models were developed using stepwise selection method. EQ-5D-5L scores were regressed on 1) main effects of SRS-22r subscale scores, 2) as per 1 plus squared and interaction terms, and 3) as per 2 plus demographic and clinical characteristics. Model goodness-of-fit was assessed using R-square, adjusted R-square, and information criteria; whereas the predictive performance was evaluated using root mean square error (RMSE), mean absolute error (MAE), and the proportion of absolute error within the threshold of 0.05 and 0.10. A total of 227 AIS patients with mean age of 15.6 years were recruited. The EQ-5D-5L scores were predicted by four domains of SRS-22r (main effects of 'Function', 'Pain', 'Appearance' and 'Mental Health', and squared term of 'Function' and 'Pain'), and Cobb angle in Model 3 with the best goodness-of-fit (R-square/adjusted R-square: 62.1%/60.9%). Three models demonstrated an acceptance predictive performance in error analysis applying 10-fold cross-validation to three models where RMSE and MAE were between 0.063-0.065 and between 0.039-0.044, respectively. Model 3 was therefore recommended out of three mapping models established in this paper. To our knowledge, this is the first study to map a spine-specific health-related quality of life measure onto EQ-5D-5L for AIS patients. With the consideration and incorporation of demographic and clinical characteristics, over 60% variance explained by mapping model 3 enabled the satisfactory prediction of EQ-5D-5L utility scores from existing SRS-22r data for health economic appraisal of different treatment options.
这是一项前瞻性研究，旨在建立预测模型，将改良版脊柱侧凸研究学会22项问卷（SRS - 22r）与青少年特发性脊柱侧凸（AIS）患者的欧洲五维健康量表5级（EQ - 5D - 5L）效用评分进行映射。AIS治疗结果的比较可通过成本效用分析来确定。然而，脊柱特异性健康相关生活质量的主要结局指标，即SRS - 22r问卷并未提供效用评估。在本研究中，前瞻性招募AIS患者，由经过培训的访谈员完成EQ - 5D - 5L和SRS - 22r问卷。采用普通最小二乘法回归来建立映射模型，并通过10折交叉验证程序评估其有效性和稳健性。EQ - 5D - 5L效用评分与人口统计学特征、Cobb角、曲线类型、治疗方式以及SRS - 22r问卷的五个领域进行回归分析。使用逐步选择法开发了三个模型。EQ - 5D - 5L评分分别基于以下因素进行回归分析：1）SRS - 22r子量表评分的主效应；2）如1加上平方项和交互项；3）如2加上人口统计学和临床特征。使用决定系数（R方）、调整后的决定系数和信息准则评估模型的拟合优度；而预测性能则使用均方根误差（RMSE）、平均绝对误差（MAE）以及绝对误差在0.05和0.10阈值内的比例进行评估。共招募了227例平均年龄为15.6岁的AIS患者。在拟合优度最佳的模型3中，EQ - 5D - 5L评分由SRS - 22r的四个领域（“功能”“疼痛”“外观”和“心理健康”的主效应以及“功能”和“疼痛”的平方项）和Cobb角预测（决定系数/调整后的决定系数：62.1%/60.9%）。在对三个模型应用10折交叉验证的误差分析中，三个模型均表现出可接受的预测性能，其中RMSE和MAE分别在0.063 - 0.065和0.039 - 0.044之间。因此，在本文建立的三个映射模型中推荐模型3。据我们所知，这是第一项针对AIS患者将脊柱特异性健康相关生活质量指标映射到EQ - 5D - 5L的研究。通过考虑并纳入人口统计学和临床特征，映射模型3解释了超过60%的方差，从而能够根据现有的SRS - 22r数据对EQ - 5D - 5L效用评分进行令人满意的预测，用于不同治疗方案的健康经济评估。
