Enhancing the Classification of Aphasia: A Statistical Analysis Using Connected Speech.
Davida Fromm, Joel Greenhouse, Mitchell Pudil and 2 others
PMID 36457942WHAT IT FOUND
Analyzing connected speech from 168 people with aphasia, two simple measures, total words in personal narratives and closed-class words in storytelling, accurately sorted participants into seven distinct groups.
These groups often cut across traditional WAB-R diagnoses, suggesting that fluency ratings alone can misclassify severity.
Key findings
01A decision tree using only two variables (total words in free speech and closed-class words in Cinderella storytelling) achieved high accuracy (0.86 on testing set) in assigning participants to seven clusters identified by unsupervised clustering.
02Traditional WAB-R aphasia types did not map neatly to the new clusters; for example, 35 participants diagnosed with Broca's aphasia were distributed across five different clusters, though mostly concentrated in two.
03Standardized test scores (naming, repetition) were not useful in distinguishing the discourse-based clusters, whereas discourse measures like amount of output and grammatical content were key.
STILL TO COME
How it was doneWhat they foundWhat it means for SLPs
Read the rest of this summary
You get three full summaries a month, free, and we do not ask for a card. Search, the TL;DRs and your library stay unlimited either way.
What it does not show
The clusters were derived from a specific set of discourse tasks (AphasiaBank protocol), so the decision tree thresholds may not generalize to other assessment materials. Interpretation of the clusters relied on subjective auditory-perceptual judgments by a single researcher who was not blinded to the group assignments. The study used a subset of 168 participants from a larger database, requiring complete data on 221 variables, which may introduce selection bias. The analysis is cross-sectional; it does not track how these clusters change over time or in response to treatment.
Declared interests
None declared.
The easy way to misread this
Do not assume that the seven new clusters replace current diagnostic labels. The authors state that the intention was not to recommend new aphasia subtypes but to illustrate how discourse analysis can reveal heterogeneity within traditional categories. The decision tree cutoffs (e.g., 309 words) are specific to the tasks used and are not universally prescriptive for other clinical settings.