Word Retrieval in the Wild: Establishing Preliminary Efficacy of Semantic Feature Ecological Momentary Intervention for Post-Stroke Aphasia

Anomia (i.e., impaired word retrieval) is one of the most common and debilitating deficits experienced by stroke survivors with aphasia, an acquired language disorder. Although many anomia treatments exist, treatment response is modest and inconsistent across studies, and the extent of generalization to everyday contexts-the ultimate goal of therapy-is largely unknown.

The critical gap between actual patient outcomes and the therapeutic potential of naming therapy is tied to treatment delivery limitations. Clinician-provided therapy requires hours of face time between a provider and patient, which limits the total dosage of therapy that can be delivered. Substantial evidence from cognitive psychology and emerging support from aphasia indicates distributed practice is best for long-term retention of treatment gains, yet standard aphasia therapy does not follow this model. Skill transference benefits from context, but most aphasia treatment remains decontextualized, with people with aphasia (PWA) completing sessions in clinic rooms outside of the real-world settings in which they live and communicate. In the current U.S. healthcare system, high-dosage, distributed and contextualized anomia treatment is infeasible.

In ecological momentary intervention (EMI), therapy is provided repeatedly throughout the day over time via prompts from a mobile device. Thus, EMI has inherent capacity for providing high-dosage, distributed, and contextualized therapy, but it is unused in stroke thus far. Research in PWA in EMI's sister method, ecological momentary assessment, does exist, including work by our team. In our work, 14 PWA and 14 controls completed a three-week smartwatch audio-based protocol in which they overtly named prompted pictures while going about daily life. We established feasibility of this method, including finding similar protocol compliance between PWA (77% response rate), controls (78%), and the extant EMA literature (79%). The objective of this phase 1 trial is to test a novel EMI based on our smartwatch naming protocol and semantic feature analysis therapy but with therapy trials temporally distributed throughout everyday life. We selected semantic feature analysis because it is one of the most effective and widely used anomia therapies, and it can be incorporated into our EMI system.

Twenty PWA will complete six weeks of EMI in their community. The EMI trial flow will include five steps: 1) a prompt alert and "Ready to name a picture?" cue, 2) a first naming attempt after a YES response, 3) the repetition of an auditory model, 4) semantic feature verification questions, and 5) a second naming attempt. To test the effects of contextualized training, trained items will be clustered into location categories (e.g., home, pharmacy), and we will use the smartwatch's location detection to match PWA's location and item category for ~50% of trials. All PWA will be scheduled to receive the same number of prompts (i.e., treatment trials) but typical variability in EMI compliance will induce natural variation between PWA in other treatment delivery factors (i.e., total dosage and trial spacing). With this paradigm, we will pursue the following aims:

Aim #1: Determine the extent to which semantic feature EMI improves naming of trained items and generalizes to untrained, related contexts. EMI efficacy will be determined by measuring a) change in trained item accuracy and b) generalization to untrained, semantically related items and functional communication. Consistent with spreading activation theory, we expect EMI to improve naming of trained items and untrained, semantically related items and result in modest (but significant) gains in functional communication.