The system was developed by the University of Bern along with the speech and language therapist of the University Hospital of Bern (Inselspital). The system contains five main parts (Figure 1): (1) The patient interface that allows the patient to train according to the therapy plan. (2) The therapist interface that allows the therapist to assign new exercises to the patient, create new patient accounts, and supervise the progress of the patient. The patient and therapist interfaces run on tablet computers (iPad) programmed in Object-C. (3) The Web page to create new exercises independent of the app. (4) The NoSQL (Not Only Structured Query Language) database to store patient and exercise data, and the Binary Large Objects storage to store all images and videos of the exercises. The database is located behind a secure firewall on a managed server and can only be accessed by authorized personnel (ie, administrator of the system). (5) The open-source server served as a backend service.

In collaboration with speech and language therapists, 10 different exercise types were defined to meet all linguistic modalities (Table 1). More than 30,000 exercises in German were implemented, and new exercises can be added online by the exercise creator. The exercises are split into training units (deck), consisting of about 25 tasks.

The exercises consist of three main elements (ie, fixed, response, and supportive elements). Fixed elements build the structure of exercises and cannot be moved. Response elements are needed to complete the task and can be moved to fulfill the task. The number of fixed and response elements can be varied and are either videos, images, written language, placeholders, or audio tracks. Supportive elements (ie, videos, images, and audio tracks) act as an aid to solve the tasks and can thus be used to fine tune task difficulty. For example, in the sentence completion task in Figure 2, “kalt” (cold) and “brandheiss” (boiling) are response elements, whereas “Der” (The), “Winter” (winter), “ist” (is), and “.” are fixed elements. The video of a speech and language therapist spelling the correct word acts as a supportive element.

Based on the broad spectrum of impaired language functions, the difficulty of an exercise differs in patients, and thus, categorization into one difficulty measure is not possible and must be adapted individually. Therefore, a hierarchical system that is structured according to characteristics (ie, based on linguist rules) and difficulty within an exercise type (eg, phonology, morphology, syntax, semantics, and pragmatics) was implemented. Detailed information of each exercise type and the hierarchical structure can be found in the Multimedia Appendices 1 and 2.

The therapist has to first create an account for the patient. Depending on the type and severity of the language impairment, the therapist assigns tasks and the corresponding decks to the patient. The tasks and decks are downloaded automatically onto the patient’s tablet computer. Each time the patient trains using the tablet computer, results are sent to the patient’s database. Feedback about the correctness of response is given automatically after the exercise is completed. In parallel, the therapist is able to change the exercise types and decks remotely, according to the needs of the patient. Furthermore, the therapist can monitor the patient’s progress in real time. The workflow is presented in Figure 3.

The study was divided into two phases: the development phase followed by evaluation of the system by patients and therapists. The study was carried out in accordance with the current version of the Declaration of Helsinki and approved by the Ethics Committee of the Canton of Bern, Switzerland. The participants (patients and therapists) were recruited via the University Hospital of Bern. Prior to participation, written informed consent was obtained, and procedures related to the study were explained to the participants. The main inclusion criteria for patients were age >18 years, diagnosis of aphasia, and a minimal level of cognitive function to handle a tablet computer and to understand the task and questionnaire. Patients with hemiparesis were not excluded from the study. For therapists, neither inclusion nor exclusion criteria were set. The study was conducted in the general ward (ie, neurorehabilitation) at the University Hospital Bern, and questionnaires were completed during the patient’s stay.

To assess patients’ and therapists’ opinion, attitude, and perception of the system, the well-established usability scale (System Usability Scale) [35] was used. The System Usability Scale is based on 10 questions and has a 5-point scale (1=strongly disagree, 5=strongly agree). To measure the patient’s subjective experience, enjoyment, and stress experienced, an adapted selection of 14 items of the intrinsic motivation inventory questionnaire was used (Textbox 1) [36]. The adapted Intrinsic Motivation Inventory questionnaire encompassed the dimensions of Feasibility, Interest/Enjoyment, Value/Usefulness, and Tension/Pressure based on a 7-point scale (0=strongly disagree, 6=strongly agree). For patients’ understanding, the questionnaires were explained and filled out with the therapists. The assignment of exercises based on diagnostic tests and the assessment of the system was performed by the same therapist. The intrinsic motivation inventory questionnaire was added later to the study and was therefore not assessed in all patients.

Analysis was conducted using R (R Foundation for Statistical Computing, Vienna, Austria), whereas for the intrinsic motivation inventory, a nonparametric Wilcoxon signed rank test was used, which accounts for a small sample size. For the subscale Interest/Enjoyment, Value/Usefulness, and Feasibility, a score>3 (ie, mean of the score scale) and for Pressure/Tension, a score<3 was regarded as positive and significant. To analyze the System Usability Scale, a nonparametric Wilcoxon signed rank test was used if the score was significantly above 70 (Good) or 85 (Excellent) [37].

In this study, we developed a multimodal telerehabilitation system to train the linguistic modalities in patients with aphasia. In line with our hypothesis, the system is simple to use, highly adaptable to the patient’s need, and highly accepted. It ran stably and was appreciated by patients as well as therapists.

The first main finding is that the questionnaire response in terms of usability was rated as excellent, and thus, the developed system is well accepted by patients.

Training with the Bern Aphasia App was rated as enjoyable, which confirms that the personalized content and difficulty of exercises could be adapted by therapists to the needs of patients. The slight increased pressure and tension to train indicate that the exercises were challenging but not frustrating. Overall, the usability and motivation indicate that the design for the patient interface is clear and consistent and thus offers the possibility of training independently. The intrinsic motivation score was consistent with that in the literature, whereas the usability score was higher [38,39].

The second main finding was that the new system allows therapists to adapt and monitor the training of the patient remotely. The usability for the therapist interface was rated lower than that for the patients’ interface but still considered good. One reason for the lower usability score might be the need for more functionality in the therapist interface compared to the patient interface and thus the need for more time for familiarization. Another reason for the lower usability score in the therapist interface could be that the patient interface was better designed and thus more adapted to their needs.

Due to the study design of a feasibility study, it remains unclear whether this result can be generalized to all patients with aphasia and therapists and whether evaluation can be transferred from a clinical to a home setting. An additional limitation of this study is the small sample size of patients and therapists and that the therapist who assigned exercises and conducted the study was the same person.

Furthermore, when using the app, patients must be connected to the internet. In rare cases, the internet connection was too slow, or patients lost internet connection and had trouble reconnecting. Therefore, it is crucial that future telerehabilitation apps can be used offline while exercising and that internet connection is only needed to synchronize the app (ie, loading new exercises and sending statistical reports to the therapist).

In the next step, we will investigate the effect of the Bern Aphasia App in a randomized multicenter clinical trial at patients’ homes. Positive results in clinical trials could have a great socioeconomic impact in addition to increased quality of life of the affected patients. With tablet-based apps like the Bern Aphasia App, both patients and therapists can benefit from an intuitive, cost-efficient, touch-based reliable product that fits well with the current trend of moving health treatment from hospital to home. We suggest standard linguistic tests (eg, Boston naming [40], Token test [41], and Amsterdam-Nijmegen everyday language [42]) prior to and after the intervention as a primary outcome and follow-up measurement as well as questionnaires about motivation and quality of life as secondary outcomes to determine the actual improvement caused by the therapy exercises.

The exercises in the hierarchical language structure are usually created by recommendations and heuristics, which is highly time consuming. Therefore, future research should focus on automated exercise creation based on artificial intelligent algorithms to ease the burden of therapists.

Based on the questionnaire scores, the developed system is well accepted and simple to use for patients and therapists. The tablet computer–based app and the hierarchical language exercise structure offer patients with different forms of aphasia a possible chance to train with different doses and intensities independently at home. Overall, the novel system has potential for treatment of patients with aphasia as a supplement to face-to-face therapy.