Which model best assesses gait in healthy elderly? A confirmatory factor analysis of existing conceptual gait models

Abstract

Background: Various conceptual gait models have been created with exploratory factor analysis to assess gait performance in healthy individuals or patients with orthopaedic or neurological diseases. However, a direct comparison between these models, to determine which is best for assessing gait in healthy elderly subjects, has never been performed. Confirmatory factor analysis is a statistical technique which allows to compare the structure of these models and to evaluate their validity and reliability.

Research question: Which of the current models for evaluating gait shows the best construct, convergent and discriminant validity and reliability when replicated in a sample of healthy elderly subjects?

Methods: 92 healthy elderly subjects (aged 73.3 ± 6.8 years) were enrolled. Participants were instructed to walk on a baropodometric walkway; gait variables were then extracted and analyzed according to 8 different gait models (published between 2007 and 2019). Correlation between variables of each model were investigated. The number of factors to include for each model was assessed with different criteria of principal component analysis. The construct validity of the gait models was assessed in terms of goodness of fit indexes through confirmatory factor analysis. Convergent and discriminant validity and reliability of the models were also assessed.

Results: Of the models considered, only our model previously created for patients with Parkinson's Disease reached convergence, with an excellent fit in all indexes (χ² = 18.34, df = 13, p = 0.15, RMSEA = 0.06 (95%CI = 0.00-0.13), CFI = 0.98, TLI = 0.96, SRMR = 0.05). The model showed an acceptable convergent validity. The three factors resulted to be mutually independent (correlations among factors <0.85) and reliable. Most of the other seven considered models presented a high number (12.5-65%) of highly inter-related variables and more factors than necessary.

Significance: Future studies can use our gait model as a framework for clinical practice or research.

Keywords: Factor analysis; Gait; Healthy subjects; Model; Older adults.