Evaluating three stuttering assessments through network analysis, random forests and cluster analysis

In stuttering, cognitive and behavioural variables interact in nonlinear fashion. These variables can be assessed by instruments which evaluate perceived impact of stuttering and stuttering severity. We applied three statistical methods in combination to the analysis of three assessment protocols to discover relationships within and between the tests to better understand variations in behavioural and social aspects of stuttering.

Scores from Stuttering Severity Index (SSI-IV), Overall Assessment of the Speaker's Experience of Stuttering scale (OASES), and Unhelpful Thoughts and Beliefs About Stuttering scale (UTBAS), collected from 26 participants were compared using three statistical methods: network analysis, random forests, and cluster analysis.

Network analysis demonstrated that SSI-IV only weakly interacts with a quality of life index (OASES) and a self-perception and belief systems index (UTBAS). Random forest analyses revealed the last two measures relate strongly to each other. The results from cluster analysis suggest a) a possible regrouping of OASES items and b) a possible use of one UTBAS scale instead of the three.

A combination of three statistical methods allowed us to evaluate the three assessments in more depth. The lack of interaction between the SSI-IV on the one hand, and OASES and UTBAS on the other, suggests that the network of the three commonly used stuttering assessments may be fractured in a non-productive way. A potential gap may exist for an assessment tool that would link behavioural and social aspects of stuttering.