Long-term changes in technology acceptance of a robotic system in stroke treatment: a pilot study

Robot-assisted arm therapy in stroke treatment

Robot-assisted therapy has become a widely used therapeutic approach in stroke rehabilitation. A variety of clinical studies showed that patients benefit in recovery of upper limb from training with robotic systems [5], [6], [7]. The effectiveness can be explained by the degree of intensity and the number of repetitions per therapy session [7]. Most of the systems use an assist-as-needed strategy [6]. Patients are supported to such an extent that they can perform the desired exercise with maximum use of their remaining voluntary motor skills [6]. The main goal is the relearning of movements through exercise [6]. The use of multimodal feedback as well as training in virtual realities is attributed an increasingly central role [6], [7]. Studies showed that augmented feedback influence motor relearning [8] and mixed reality-based interventions have an effect on functions after stroke [9].

In the project “BeMobil – Regain Mobility and Motivity” the therapy robotic device Bi-Manu-Track (BMT) has been further developed for use in telerehabilitation environment. In order to prove a concept an experimental prototype Bi-Manu-Interact (BMI) by a Rehabilitation Robotics group, at the institute Industrial Automation Technology (IAT) of the TU Berlin. The BMI was tested in an experimental setup that is described below: Therapist and stroke patient sit opposite to each other and interact through the BMI. Each person receives visual feedback on a monitor about the quality of the movement [10].

Technology acceptance

In the mid-1990s, it was recognised that in order to improve user satisfaction and acceptance, other features of the technology had to be taken into account. These go beyond the classical functional application context and are related to the user’s attitudes, needs, desires and abilities [11], [12]. In addition, Marcellini and colleagues (2000) refer to socio-demographic factors such as age, gender, level of education, health status and social conditions as predisposing variables in the context of technology acceptance, since they influence the way the technology is used [13].

One model that includes those factors is the Technology Acceptance Model (TAM), which was first developed by Davis and colleagues in 1989 [14]. The TAM is dedicated to the prediction of behavioural intention and the use of technology. A meta-analysis conducted by King and He (2006) with 88 studies on the application of the TAM in different areas (e.g. Internet, e-commerce) showed that it is a valid, robust and widely used model [15]. The authors of the TAM assume that the actual use of a technology is determined by the perceived benefit and ease of use [14]. Previous findings of technology acceptance confirm that with an increased experience in the use of a device, the influence on the perceived benefit and the intention to use decrease. This is due to the advantages and disadvantages of a technology are perceived in a more differentiated way [16].

Research question and hypotheses

Based on the presented research findings the present study focusses on a detailed analysis of TAM and UX of the BMI. Therefore, the central research question is “Are there differences in TAM, UX and the willingness to use the technology at home between the first (T0) and after the 10th (T1) training session with the BMI?”

The following hypotheses are derived: H1: T1-ratings in terms of UX show higher values than T0-ratings. H2: T1-ratings in terms of TAM show higher values than T0-ratings. H3: The willingness to use the BMI at home is higher at T1 than at T0.

Study design and sampling

To answer the research question, a study design with a repeated measurement was applied, looking for differences for the variables UX, TAM and potential of home use within the same group of participants. Inclusion criteria for the study were adequate cognitive and language skills to participate in an interview. These skills were subjectively assessed by therapists on site. Exclusion criteria were the following: severe language and speech disorders, neglect, limitations in cognitive performance and decreased awareness. Nine patients (n=4 females) aged 33–78 years (M=60.22, SD=13.17) participated in the study. They already received different therapeutic treatments as a part of their stroke rehabilitation at T0 was for one up to 32 weeks (M=11.33, SD=9.15). The Barthel Index (BI), a test for evaluation of independence in everyday life in a range of 0–100 points (scored with 0, 5 or 10 points). The participant had an average of 58 points (SD=13.46) and were rated as “need of help”. The Fugl-Meyer Score (FMA), a test for evaluation of motor skills in the upper extremities in a range of 0–66 points (scored with 0, 1 or 2 points). The participant had an averaged at 54 points (SD=16.72) and the performances were rated as “moderate” impairments.

Questionnaire and data analysis

The questionnaire applied in the study used questions with a structured, standardized, closed structure. Questions were answered by dichotomous answer options or on five point Likert scales (1=strongly disagree, 2=disagree, 3=undecided, 4=agree and 5=strongly agree). The items were divided into three areas: UX (e.g. concentration, etc.), TAM (e.g. perceived use, etc.) and potential use at home.

Non-parametric methods were used to analyse the differences between measurements. This procedure was chosen because of small sample size, lack of normal distribution and ordinal response scale. Two-sample paired Wilcoxon signed-rank tests were performed. Because of the small sample size (n<10), one-sided exact tests were used. All statistical tests were performed on a significance level of α=0.05. Cohen’s d was used as effect size.

User experience with BMI

In order to assess the experience of stroke patients trained with the BMI, the subjects were asked to answer various questions after using the training device. Table 1 lists medians, p-values and effect sizes of various UX factors during training with the BMI and the visual feedback in comparison of T0–T1. According to the medians, it can be seen that above-average positive values were given at both measurement points in the variables of concentration, safety, well-being and motivation during the training as well as all variables for visual feedbacks. At the same time, both measurement points show that fatigue, anxiety, under- and overstrain, boredom and pain were given as rather little or not at all. The analysis shows no statistically significant differences between T0 and T1 in terms of UX (p>0.05). However, small to large effect sizes can be seen in most variables, which imply a practical relevance of those differences.

Acceptance of the BMI

The ratings of the four TAM dimensions were assessed on a five-level Likert scale. Table 2 shows the descriptive data, p-values and effect sizes of the measurement points T0 and T1. The medians show that positive values were given in all four dimensions at both measuring points. The variable “intended use” remains constant in the ratings. The ratings for the scales “perceived use”, “user-friendliness” and “social norm” decreased marginally at the second measurement point compared to the first. The analysis shows no statistically significant differences between T0 and T1 in all TAM dimensions (p>0.05). But, small to medium effect sizes were found in all dimensions, which implies a practical relevance of these differences.

Potential for home use of the BMI

The patients were asked whether they could imagine using the BMI at home. The answer was given dichotomously with “yes” or “no”. Figure 1 shows that the willingness to use the BMI at home increased at T1 compared to T0. At T0 four patients could imagine use the BMI at home. At T1 the willingness increased to six affirmations. The data show a migration of two patients who did not show willingness at T0. After 10 training sessions (T1) they could imagine to use the BMI at home.

Figure 1:

Descriptive data of willingness to use the Bi-Manu-Interact (BMI) at home and migration of patients T0 vs. T1 (n=9).