Frontiers in Psychology (May 2015)
Using big-data to validate theories of rehabilitation in aphasia
Abstract
Introduction. While the evidence for efficacy of rehabilitation of language disorders is fairly robust and conclusive (Allen, et al., 2012; Brady, et al., 2012; Kelly, Brady, & Enderby, 2010; Cherney, et al., 2008), a major limitation identified by these reviews is that the sample size of patients in each of the interventions have been modest (5-20 patients). As technology moves our field forward, we can now collect and analyze larger sets of data to validate theories of rehabilitation. As a first step, we report data from a recently completed study examining the effectiveness of software platform (Constant Therapy) to deliver, monitor and analyze treatment for individuals with aphasia (Des Roches et al., 2015). Methods. Fifty one individuals with language and cognitive deficits were administered standardized tests (Western Aphasia Battery, Boston Naming Test, Pyramids and Palm Trees, and Cognitive Linguistic Quick Test) prior to initiation and following completion of therapy. Forty-two experimental patients used the iPad-based therapy once a week with the clinician and up to six days a week for home practice. Nine control patients practiced therapy on the iPad once per week with the clinician only. Thirty-eight therapy tasks were divided into language and cognitive activities that were developed (Des Roches et al., 2015), 28 of these tasks included buttons that revealed a hint to assist the patient answer the item. The assigned therapy tasks were tailored to that individual’s language and cognitive impairment profile based on an initial baseline assessment. Each task was practiced until accuracy on task reached 100% on multiple occasions at which point that task was replaced with the task at the next level of difficulty. The 51 patients each completed a 10 week program leading to total of 3327 therapy sessions across patients. Analysis and Results: Mixed regression models showed that both the experimental and control groups improved but experimental participants showed more significant changes on the therapy tasks and standardized tests than control patients. Additionally, the more severe patients benefited from the simpler language and cognitive tasks whereas the less severe patients improved on tasks that required a combination of language/cognitive processing. Correlations between starting severity scores and the amount of change on standardized tests showed that participants with lower initial scores showed more improvement on the standardized tests than participants with higher initial scores. In addition, because participants were able to self-administer the hints for individual items, the relationship between levels of hint use and participants’ corresponding accuracy showed that hint use often negatively predicted performance. Correlations between frequency of hint use and participant language/cognitive impairment profiles showed significant negative relationships between hint use and performance on standardized tests indicating that the more severe the participants the more often they utilized hints to complete the therapy tasks. Conclusions: Analysis at the patient, therapy task and item level allow an evaluation of the efficacy of rehabilitation and individual responsiveness to rehabilitation. In addition, preliminary results from a larger data set collected anonymously from over 2,000 independent patient users is also presented.
Keywords
