CASE: iReadMore: optimization and personalization of a reading therapy app for patients with central alexia
Alexander Leff, Ashley Peacock, NeuroDiversity
- To optimize the usability and clinical effectiveness of iReadMore in the context of a ‘rollout’ trial
- To develop patient-specific therapy algorithms
- To develop a predictive algorithm from completed patients’ data that will provide estimates of likely effect sizes that incoming patients can reasonably expect to achieve
- Background and rationale
iReadMore is a reading therapy app for adult aphasic patients. It contains diagnostic, therapeutic (practice-based) and outcome components.
Central alexia is the commonest type of acquired reading disorder in adults. It is usually encountered as part of a generalized language disorder known as aphasia. Stroke is the leading cause of aphasia but head-injury, brain tumours and dementia also figure prominently. In the UK alone, there are more than 350,000 people with aphasia; around 2/3rds of these will have central alexia. While standard speech and language therapy (SALT) improves outcomes in patients with aphasia, the ‘dose’ needed to achieve this (>100 hours) is ~10 times more than the average aphasic patient current receives on the NHS.
We have recently completed a phase II trial of iReadMore therapy in a group of patients with central alexia. The results (which are in revision with the journal Brain) demonstrate that the therapy works well on trained items. After an average of 35 hours therapy, patients improved their reading accuracy by 9%. The standardized effect size was “large” (Cohen’s d = 1.38) Reaction times also significantly improved.
We now plan to ‘rollout’ iReadMore therapy for suitable patients to use who have internet access. We wish to optimize the therapy and will do so by testing a series of hypotheses, all aimed at improving the usability and/or clinical effectiveness of iReadMore. We have ethical approval in place.
- Plan of investigation
The main method will be split-testing (users are randomized to one of two versions of iReadMore to see which performs best). The main areas we wish to optimize for iReadMore are:
2) Adaptive interface: while most patients in the phase II study enjoyed using iReadMore, there was no formal analysis of each user’s interactions with the graphical user interface. By collecting information on when and where the user touches parts of the assessment and therapy screens, we can move certain items on the screen to a better location. The main outcomes will be evidence of engagement and patient reported outcome measures
3) Tailoring therapy: the current version has fixed therapy items. Given that each patient has to work hard to improve their performance, it makes sense to offer them a choice on which words to work on. We will test this using a within-subject design. The main outcomes being reading impairment and patient reported outcome measures (e.g. satisfaction over the range of training words available)
4) Predicting individual patient’s responses to therapy: with a large pool of users, we should be able to make predictions for incoming patients by modelling their baseline assessments and demographic data (e.g. time since stroke) and comparing this with similar users who have completed the therapy
iReadMore webpage: http://www.ucl.ac.uk/aphasialab/apps/ireadmore.html
Woodhead ZVJ, Ong Y-H & Leff AP
Web-based therapy for hemianopic alexia is syndrome-specific
BMJ Innovations 2015; 1(3): 88-95
Ong Y-H, Jacquin-Courtois S, Gorgoraptis N, Bays PM, Husain M & Leff AP.
Eye-Search: a web-based therapy that improves visual search in hemianopia
Annals of Clinical and Translational Neurology 2014; 2(1): 74-78