Use of a clustering method to describe the clinical profiles of older fallers: the value of a multidisciplinary consultation

Objectives The population of older adults is particularly heterogeneous with regard to frailty and the risk of falling, the two of which are linked. We conducted an exploratory, analysis (with no preconceived ideas) of data collected during multidisciplinary falls consultations (MFCs), to identify people with similar profiles. Materials and methods We performed an observational, multicentre study of older patients (aged 75 and over) having been evaluated in an MFC. We excluded adults with a Mini Mental State Examination score < 14/30, an activities of daily living score < 4/6, or an unstable medical condition. Each participant underwent a clinical interview, impedancemetry, and a physical activity assessment (a questionnaire, and use of an activity tracker on 5 consecutive days). The K-means method and ascending hierarchical clustering were used to identify clusters of people with common characteristics. Results Of the 106 participants, the median [IQR] mean number of falls in the previous 6 months was 1 [2]. Three functional clusters were identified: (i) fallers with poor mobility, difficulty getting up off the ground after a fall, and using a mobility aid for walking; (ii) an intermediate sedentary group with a gait speed of similar to 0.6 m s(-1), and (iii) active people with a timed up and go test time below 15 s and a gait speed above 0.8 m s(-1). Conclusions The population of older fallers referred for an MFC is heterogeneous. The presence of certain clinical characteristics enabled the definition of three patient clusters, which might help physicians to determine the most appropriate care objectives and pathways.

Automated summary

This study aimed to identify similar profiles among older adults referred for multidisciplinary falls consultations (MFCs) and determine their frailty and risk of falling. The study used clinical interviews, impedancemetry, and physical activity assessment to evaluate the participants, and employed K-means and ascending hierarchical clustering to identify three clusters of patients. The results provide guidance for physicians in determining appropriate care objectives and pathways.