Variation inflation factor-based regression modeling of anthropometric measures and temporal-spatial performance: Modeling approach and implications for clinical utility. (January 2018)

• Record Type: Journal Article
• Authors: Long, Jason T.; Neogi, Smriti; Caldwell, Cailee M.; DeLange, Matthew P.
• Is Part Of: Clinical biomechanics. Volume 51(2018)
• Page Start: 51
• Page End: 57
• Publication Date: 2018-01
• Subjects: Gait -- Temporal-spatial parameters -- Modeling -- Variation Inflation Factor; Biomechanics -- Periodicals; Osteopathic medicine -- Periodicals; Biomechanics -- Periodicals; Osteopathic Medicine -- Periodicals; 612.76
• DOI: 10.1016/j.clinbiomech.2017.11.008
• Languages: English
• ISSNs: 0268-0033

Abstract

Abstract: Background: Understanding the relationship of underlying anthropometry to temporal-spatial performance is critical to appropriate assessment of patients with ambulatory dysfunction. The current body of literature has established the importance of limb length in this relationship. This study sought to re-examine these relationships in light of recent trends in body habitus and obesity, using Variation Inflation Factor analysis to optimize the model. Methods: Elementary school children (n = 452; ages 5–13) were tested during walking at a self-selected speed across an instrumented walkway. Temporal-spatial and anthropometric measures were compiled for all children. The relationship between temporal-spatial and anthropometric measures was assessed using regression modeling with Variation Inflation Factor optimization. Findings: Body weight was identified as a significant predictor of cycle duration, stride length, stance duration, and step width during initial modeling. However, it did not meet the constraints imposed during Variation Inflation Factor optimization and was removed from the final models. The final optimized models identified significant relationships between both temporal-spatial parameters of interest and other temporal-spatial measures, with the best fit identified for walking speed (R 2 = 0.6148). Interpretation: The use of the Variation Inflation Factor constraint during the regression modeling process ensured final models composed of truly independent predictor variables. The resulting models are highly robust and highlight the complex relationships between body structure, functional conditions, and walking performance. These models have value for routine clinical assessment of ambulatory dysfunction, and may provide a foundation for classifying temporal-spatial performance in the context of multiple contributing parameters. Highlights: Model rigor improves with Variation Inflation Factor to control multicollinearity. Weight is not a significant predictor for any measure after Variation Inflation Factor analysis. Final models may lead to more complex classification of temporal-spatial performance.