BACKGROUND
Traumatic Brain Injury (TBI) is a leading cause of disability worldwide. TBI patients are characterized by a wide heterogeneity, one of the most significant barriers to effective therapeutic interventions. Cluster analysis (CA) to identify homogeneous subgroups based on performance in neuropsychological baseline tests has been extensively applied in state-of-the-art. Nevertheless, most analyzed samples are rarely larger than n = 100, different CA approaches and cluster validity indices (CVI) have been scarcely compared and not applied in web-based rehabilitation treatments.
OBJECTIVE
1) Apply state-of-the-art CVIs to different cluster strategies: hierarchical, partitional and model based.
2) Apply combined strategies of dimensionality reduction by means of principal components analysis and random forests and perform stability assessment to the final profiles
3) Characterize the identified profiles by means of demographic and clinically relevant variables
4) Study the external validity of the obtained clusters considering three relevant aspects of TBI rehabilitation: Glasgow Coma Scale (GCS), functional independence measure (FIM) and executions of web-based cognitive tasks.
METHODS
Different cluster strategies were executed with mclust, factoextra and cluster R packages. For combined strategies we used the factominer and random forest R packages. Stability analysis was performed with fpc R package. Between groups comparisons for external validation were performed using T-test, chi-square test or Mann-Whitney U test as appropriate. The period under study is August 2008 to July 2019
RESULTS
We analyzed 574 adult (mostly severe) TBI patients undergoing web-based rehabilitation. We identified and characterized 3 clusters with strong internal validation: 1) moderate attentional impairment and moderate disexecutive syndrome with mild memory impairment and normal spatio-temporal perception being 66% highly educated (P<.05)
2) severe disexecutive syndrome with severe attentional and memory impairments and normal spatio-temporal perception being 49.2% highly educated (P<.05)
3) very severe cognitive impairment, being 45.2% highly educated (P<.05).
We externally validated them with severity of injury (P<.006), and functional independence assessments: cognitive (P< .001), motor(P<.001) and total(P<.001). We mapped 151,763 web based cognitive rehabilitation tasks performed by the 574 participants during the whole period (all cognitive functions) to the 3 obtained clusters (P<.001) and confirmed the identified patterns.
Stability analysis indicates that Cluster 1 and 2 were respectively rated as 0.6 and 0.75 therefore they are indeed measuring a pattern and Cluster 3 was rated as highly stable.
CONCLUSIONS
CA in web-based cognitive rehabilitation treatments allows for ii)identifying and characterizing strong patterns of response to neuropsychological tests, iii) externally validating the obtained clusters using important aspects of TBI rehabilitation (severity or functional independence in daily activities) iv) tailoring cognitive web based tasks executed in the web platform to the identified profiles providing clinicians a tool for treatments’ personalization not addressed in previous traditional CA. iv) extension of a similar approach to other medical conditions
CLINICALTRIAL