Description

We propose to constitute a prospective cohort of subjects carrying chronic disease. The objective is to explore all the different components of mobility (balance, muscle force, body composition, walking ability, metabolism, etc.), to assess their impact on the functional capacity of individuals and to identify their potential interactions. Processing this data could ultimately allow the development of a model to determine a composite standard profile of "dysmobility" in patients with chronic disease.

In the present protocol, parameters will be measured on five occasions (at inclusion, at 6 months from the inclusion day, at 1,2 and 5 year(s) from the inclusion day). All tests will be performed on the same day. New assessments will be done on the same principle as the initial evaluation.

Statistical analyses will be carried out using Stata software (version 13, StataCorp, College Station, USA). Data will be described by frequencies and percentages for categorical variables and by means and standard deviation (or median and interquartile range if data are not normally distributed) for continuous variables. The normality of continuous data will be assessed graphically and using the Shapiro-Wilk test. The main analysis will consist in determining patient profiles regarding their mobility. Clustering-type approaches (supervised or not) will be proposed: k-means, vector machine support, machine learning, factor analysis. For example, factor analyses on mixed data will allow the main components of mobility to be characterized. They can be followed by an ascending hierarchical classification in order to determine homogeneous groups of patients.

These groups will be described and compared on the main criteria evaluated using standard statistical tests: chi2 test (or Fisher's exact test if applicable) for the categorical criteria and using an analysis of variance (or Kruskal-Wallis if data are not normally distributed) for continuous criteria. The main analysis will be broken down more specifically for each of the pathologies considered. In addition, the sensitivity to change will be assessed in each pathology and for each assessment criterion. The search for factors related to the evolution of the different criteria will be carried out using usual tests and mixed multivariate models (logistics for categorical / linear criteria for continuous criteria) considering the subject as a random effect, and adjusting on the time criteria (inclusion / follow-up) and the criteria highlighted in the univariate analyses and in light of the elements reported in the literature. A sensitivity analysis will be proposed in order to study the statistical nature of the missing data and to propose, if necessary, the most suitable method of imputing the data: multiple imputation, maximum bias or LOCF in the case of longitudinal data.