Multiple Regression Analysis of Factors Affecting Health-Related Quality of Life in Adult Spinal Deformity

BackgroundPrevious studies demonstrated the adult spinal deformity (ASD) population is heterogeneous. Multiple parameters may affect health-related quality of life (HRQL).AimTo understand the ranking of parameters affecting HRQL in ASD using multiple regression analysis.Patients and MethodsA total of 483 patients enrolled in a prospective multicenter ASD database from the population. Multiple regression analysis was performed for Scoliosis Research Society–22 (SRS-22) and Oswestry Disability Index (ODI) separately. Initially proposed primary variables of diagnosis (highest correlation), age, lordosis gap (L gap), and coronal curve location were regressed for each response variable (SRS-22 and ODI) univariately. Age and L gap could not be used together because of high colinearity. Coronal curve location was removed owing to an insignificant correlation. Two initial models were considered per response, consisting of diagnosis and age in one and diagnosis and L gap in the other. The rest of the potentially predictive variables were introduced in these models one at a time. Final models were evaluated using stepwise automatic model selection.ResultsFor ODI, body mass index (BMI), gender, and sagittal and spinopelvic parameters were in the basic model but only BMI and gender in the model with L gap and only gender in the model with age were highly predictive. For SRS-22, a large number of parameters were in the basic model but BMI, gender, coronal balance, lordosis curve, and sagittal vertical axis in the model with L gap and only gender in the model with age were highly predictive. Coronal curve location was not significantly predictive in any model.ConclusionsThese findings reiterate the importance of patient diagnosis, age, and/or the amount of lordosis as the most important factors affecting HRQL in ASD. Gender, BMI, and sagittal vertical axis appear to be consistently important co-variables whereas coronal balance and magnitude of L curves may also be important in SRS-22. These may aid in better understanding the problem in ASD and may be useful in future classifications.