Background. Self-managed digital therapies are increasingly used in post-stroke aphasia rehabilitation, yet the clinical meaning of in-app performance gains and the causal impact of practice dosage remain unclear. Real-world analyses suggest that higher weekly practice time is associated with greater improvements in app-based domain scores, but prior work has relied largely on in-app metrics and conventional observational models without explicit causal inference. Objectives. To (1) quantify the association between changes in digital therapy domain scores and changes in standardized aphasia and functional communication outcomes, and (2) estimate the causal effect of weekly practice dosage on both in-app and clinical outcomes using longitudinal causal modeling. Methods. We conducted a prospective observational cohort study of adults with post-stroke aphasia who used a self-managed tablet-based therapy program for 24 weeks. Standardized assessments (Western Aphasia Battery–Revised Aphasia Quotient [WAB-R AQ], Boston Naming Test [BNT], and Stroke and Aphasia Quality of Life Scale [SAQOL-39]) were administered at baseline, 12 weeks, and 24 weeks. In-app “domain scores” summarizing task difficulty across language and cognitive domains were computed weekly, and weekly practice dosage was defined as total minutes in active therapy. We fit linear mixed-effects models for longitudinal change and applied marginal structural models with inverse probability of treatment weighting to estimate the causal effect of sustained weekly dosage categories (0–15, 15–40, \(>\)40 minutes/week) on outcomes. Results. Participants demonstrated clinically meaningful improvements in both WAB-R AQ and domain scores over 24 weeks. Greater gains in domain scores were moderately correlated with improvements in WAB-R AQ, BNT, and functional communication, suggesting that in-app performance can serve as a proxy for real-world recovery. In weighted marginal structural models, maintaining 15–40 or \(>\)40 minutes of weekly practice produced larger average improvements in WAB-R AQ and domain scores than \(<\)15 minutes/week, with evidence of diminishing returns beyond approximately 60 minutes/week. Conclusions. In a real-world digital aphasia therapy setting, improvements in domain scores were meaningfully linked to standardized language and quality-of-life outcomes, supporting their use as clinically relevant progress indicators. Causal analyses suggest that sustaining at least 15–40 minutes of self-managed practice per week yields substantially greater recovery than very low practice, while much higher doses confer only modest additional benefit, providing an empirically grounded basis for dosage recommendations in self-managed digital aphasia therapy.
Background. Self-managed digital therapies are increasingly used in post-stroke aphasia rehabilitation, yet the clinical meaning of in-app performance gains and the causal impact of practice dosage remain unclear. Real-world analyses suggest that higher weekly practice time is associated with greater improvements in app-based domain scores, but prior work has relied largely on in-app metrics and conventional observational models without explicit causal inference. Objectives. To (1) quantify the association between changes in digital therapy domain scores and changes in standardized aphasia and functional communication outcomes, and (2) estimate the causal effect of weekly practice dosage on both in-app and clinical outcomes using longitudinal causal modeling. Methods. We conducted a prospective observational cohort study of adults with post-stroke aphasia who used a self-managed tablet-based therapy program for 24 weeks. Standardized assessments (Western Aphasia Battery–Revised Aphasia Quotient [WAB-R AQ], Boston Naming Test [BNT], and Stroke and Aphasia Quality of Life Scale [SAQOL-39]) were administered at baseline, 12 weeks, and 24 weeks. In-app “domain scores” summarizing task difficulty across language and cognitive domains were computed weekly, and weekly practice dosage was defined as total minutes in active therapy. We fit linear mixed-effects models for longitudinal change and applied marginal structural models with inverse probability of treatment weighting to estimate the causal effect of sustained weekly dosage categories (0–15, 15–40, \(>\)40 minutes/week) on outcomes. Results. Participants demonstrated clinically meaningful improvements in both WAB-R AQ and domain scores over 24 weeks. Greater gains in domain scores were moderately correlated with improvements in WAB-R AQ, BNT, and functional communication, suggesting that in-app performance can serve as a proxy for real-world recovery. In weighted marginal structural models, maintaining 15–40 or \(>\)40 minutes of weekly practice produced larger average improvements in WAB-R AQ and domain scores than \(<\)15 minutes/week, with evidence of diminishing returns beyond approximately 60 minutes/week. Conclusions. In a real-world digital aphasia therapy setting, improvements in domain scores were meaningfully linked to standardized language and quality-of-life outcomes, supporting their use as clinically relevant progress indicators. Causal analyses suggest that sustaining at least 15–40 minutes of self-managed practice per week yields substantially greater recovery than very low practice, while much higher doses confer only modest additional benefit, providing an empirically grounded basis for dosage recommendations in self-managed digital aphasia therapy.
