Optimizing Task Engagement in Computer-Based Aphasia Treatment

Veterans Health Administration, Decatur PA

Investigators

Linked publications & trials

Abstract

There are over a million people with aphasia (PWA) living in the United States. Recent estimates suggest that VHA outpatient clinics see 2000 new cases of aphasia each year, indicating that approximately 100,000 Veterans are currently living with the condition. Aphasia has a lasting negative impact on quality of life, and it is important to find cost-effective ways to improve treatment outcomes for Veterans with this disorder. The proposed study seeks to improve aphasia treatment by addressing maladaptive speed/accuracy trade-offs in PWA performance. It will employ a computational model (the diffusion model; Ratcliff, 1978), which uses accuracy and response time data to derive estimates of individual components of the decision process. These components include a measure of processing efficiency, and a measure of response threshold setting. Response thresholds have been shown to account for speed-accuracy tradeoffs in performance and is thought to reflect an individual's level of ?response caution? in avoiding mistakes. In preliminary work using the diffusion model, PWA were found to set more cautious response thresholds than matched controls in a lexical decision task, leading to much longer processing times overall. However, when provided with online speed and accuracy feedback, PWA responded 22-28% faster with only a 1-3% loss in accuracy. While this suggests PWA set their response thresholds mal-adaptively, it also shows that targeted feedback can improve their threshold setting. This finding is important for aphasia rehabilitation because it suggests that response threshold miscalibration is an unnecessary and addressable source of slowed performance. In treatment, this could reduce the number of trials per session, affecting dosage. More generally, unnecessary slowing could negatively affect communication in day-to-day life. These considerations motivate a novel treatment, designed to improve language performance by targeting response threshold optimality. Preliminary results showed that the diffusion model can be successfully used to study speed-accuracy tradeoffs in aphasia in tasks requiring key press. Therefore, the first part of this project will extend the model to word production tasks in order to increase its applicability to naming therapy. It is predicted that variants of the standard diffusion model will adequately fit single word production data by demonstrating standard patterns of parameter sensitivity and inter-parameter correlations between and across tasks. Models will be tested using corpora data from the English Lexicon Project and an existing PWA confrontation naming dataset. Successful results of this aim will support the use of these models for assessment and feedback purposes in naming. The second part of this project will develop and test the efficacy of a computer-based treatment specifically designed to improve response threshold optimality. This novel treatment will provide both lexical- semantic stimulation and response optimality training by combining picture naming and semantic feature verification with adaptive response feedback. Model-derived optimality feedback will be provided for feature verification decisions, and also for naming decisions if the models in part 1 are successfully extended to naming. Ten PWA will be tested in a single-subject experimental design with multiple baselines across participants. Lexical-semantic stimulation is predicted to produce stable improvements on trained and semantically related untrained items, and response optimality training is predicted to generalize by producing stable improvements response threshold optimality in probe tasks. Better levels of response threshold optimality are also predicted to correlate with better treatment effect sizes, due to increased treatment dosage. If successful, this project will demonstrate a novel way to improve speed of processing in aphasia, and contribute to the development of more effective computer-based treatments. This will improve services for Veterans in need by making available adaptive patient-specific language rehabilitation in cost-effective ways.