Telerehabilitation for mild traumatic brain injury patients: patients’ preferences in Aotearoa New Zealand

Study design

The study involved a cross-sectional survey of participants in New Zealand. The study was part of a larger project that explored patients’ and providers attitudes toward telerehabilitation for patients with mTBI. The DCE design and analysis experiment followed the International Society for Pharmacoeconomics and Outcomes Research (ISPOR) checklist on patient-preference methods.³⁵ Data are available from the authors by request.

Identification of attributes to be used in the DCE

As part of the wider study, 12 clinicians with expertise and operational experience delivering mTBI rehabilitation from a broad range of perspectives (ie professional backgrounds, experience of telerehabilitation, gender, culture, and geographical location) and five mTBI patients who were undergoing/had received rehabilitation within the past 12 months following a clinical diagnosis of mTBI participated in a series of semi-structured focus groups/interviews regarding the design of rehabilitation services for patients living with a mTBI. The participants were identified through a rehabilitation company, with contact details of interested patients forwarded to the research team. A purposive sampling framework was developed to ensure a broad range of experiences were represented with respect to age, gender, ethnicity, time since mTBI and geographical location. The research team used their clinical experience along with interview themes and literature findings to develop the dimensions for each attribute (see Table 1).

Sample method

DCE participants were recruited though the NZ TBI Network (https://tbin.aut.ac.nz/about-us). The network consisted of patients living with a mTBI and their friends and family members. All members of the network were contacted and asked if they would be interested in participating in the study. Those who responded positively were screened to ensure that they either had experienced a mTBI/concussion or were a family or friend of a person living with a mTBI. Participants were offered the opportunity to take the survey online or via a paper survey (provided along with a self-addressed, stamped envelope). Thus, the survey used a convenience sampling technique involving individuals who self-identified as either having a mTBI or being a friend or family member of a person living with a mTBI. All participants consented to be part of the study and the study received ethical approval from the New Zealand Ministry of Health’s Health and Disability Ethics Committee (NZ 21/CEN/237).

Questionnaire design

The fractional factorial design (created using the Bayesian design approach in SAS JMP³⁵) with the greatest efficiency involved eight versions of a 16-choice set and was found to be the most efficient design with 97% based on the geometric mean of the eigenvalues, 67% for the maximum standard error for prediction over the candidate set and 94% based on the algebraic mean of the eigenvalues.³⁶ The survey began with a detailed explanation of all the attributes and their respective levels, an example that involved the choice between two cars, and then 16 choices between two rehabilitation programs and one opt-out option. An example of the choice that participants were asked to complete is shown in Fig. 1. For each choice, the participants were asked to select their preferred program. After completing the 16 choice sets, participants were asked a series of questions, including demographic information and their preferences for rehabilitation services and estimated cost/time for each type of service.

Fig. 1.

Example of a choice set.

Model estimation

The DCE choices were analysed using conditional logit modelling (CLM).³⁷ The model was estimated in two ways: using each individual attribute (Categories Model) and using a linearised cost variable (Linear Model). The results were analysed for the entire sample and subsequently by whether the participant had a mTBI or was a friend or family member of someone who had a mTBI. The results from the DCE estimation provides an indication of the impact (positive or negative, as indicated by the sign of the coefficient (B) for each attribute) and the impact or importance of the attribute (the value of the coefficient). In addition, because participants were given the option of choosing no rehabilitation (‘Don’t like either programme’ option shown in Fig. 1), the analysis also provides an indication of whether the participant would prefer to not engage in rehabilitation (the ‘No Rehab’ option reported below).

Marginal probabilities

Marginal analysis was conducted in two ways. First, the willingness to pay (WTP) for each attribute was calculated by dividing the coefficient (B) each of the attributes by the coefficient for the linearised cost (Linear Model). The WTP estimates the relative importance of each of the attribute in monetary terms. This was done for the entire sample and separately for two groups: patients living with a mTBI (Had a mTBI) and their friends and family members (Friends and Family). The second type of marginal analysis explored several policy scenarios, including: (a) having online resources available; (b) the threat of the spread of COVID-19 in the community; and (c) how changes in external conditions – namely increased risk of contracting COVID-19 – influenced the choice between online and clinic-based sessions. In addition, the analysis examines two policy options that could be used to increase the choice of rehabilitation after a mTBI; (a) lowering the cost of the telerehabilitation sessions but subsidising internet connectivity; and (b) reimbursing for wages lost as a result of attending (either virtually or at a clinic) the rehabilitation session. Each analysis reports the probabilities of each of the five options – quick (30 min) telerehabilitation, long (60 min) telerehabilitation, quick clinic visit, long clinic visit, or no visit under four scenarios relative to a ‘base case’ that was formed using the participants reported perceptions of the cost and time associated with each mode of delivery. The travel time and session cost information were combined with an estimate of the income that would be lost to form a total estimated cost of each mode of rehabilitation.

Respondent demographics

The study sample included 68 participants with an average age of 41 years (see Table 2). Overall, 44% of the sample (n = 30) reported having a mTBI. The average time since their first mTBI was 5 years, and the average number of mTBIs was 2.6 events. A majority (83%) of the respondents had undergone rehabilitation following their mTBI. There was a high degree of connectivity for the participants (all reported having some access to the internet).

Discrete choice results: entire sample

The results from the conditional logit analysis are shown in Table 3. The first model (labelled Categories Model in Table 3) provides the coefficient results for each attribute included in the experiment (the reference variable for each attribute are indicated). As predicted, all cost results had negative slopes with increasing degrees of magnitude. The second model (Linear Model) reports the results with cost linearised for the entire sample. The results suggest that there was a general preference for therapy compared to not having therapy, as suggested by the significant negative coefficient for the No Rehab (the Neither option) in the analysis. The results also suggest that, all else being equal, participants preferred long clinic appointments and long telerehabilitation appointments to a quick clinic appointment. There was a strong preference for having a therapist that you know and like (as evidenced by the willingness to pay $107.55). Overall, there was little demand for online resources, either with or without clinician support. There was also a significant preference for not having significant COVID-19 spread and for a long telerehabilitation appointment if there was significant spread in the community.

Discrete choice results: mTBI and family or friend

Results from the conditional logit analysis are shown for the two groups – those who reported having a mTBI and those whose friends or family had a mTBI (Table 3). Both groups favoured rehabilitation over no rehabilitation and preferred a therapist that they liked over a new therapist each week. Those who had a mTBI showed a preference for a long clinic or long telerehabilitation session over both the short clinic and the short telerehabilitation sessions whereas the family or friends group showed no preference for a short versus long rehabilitation session.

Marginal analysis

Table 4 shows the participants’ estimated costs of the travel that would be required to attend a long or short clinic visit, the perceived cost of a clinic visit and an online session, and the preference for therapist and resource materials. The travel time and session cost information were combined with an estimate of the income that would be lost to form a total estimated cost for each mode of rehabilitation, ranging from $77.32 for a long session in a clinic to $28.44 for a short telerehabilitation session. The results from the linear DCEs reported in Table 3 and the preferences and estimated costs shown in Table 4 were used as the base cost for the marginal analysis.

The marginal analysis shown in Table 5 shows the marginal probabilities for a choice between the five options – no visit, short clinic visit, long clinic visit, short telerehabilitation visit, and long telerehabilitation visit – for the policy scenarios described above. The first rows show the comparison between all five options, while the last two rows show the probabilities for a long clinic vs. rehabilitation visit with and without the inclusion of the lost wages (work costs). The results suggest that for the base case analysis, the probability of participants choosing a long telerehabilitation visit was 42% and a short telerehabilitation visit was 23%, meaning a combined probability of 65% of the respondents choosing a telerehabilitation visit compared with a probability of 33% choosing a clinic visit and 2% choosing to have no rehabilitation. The probability of choosing a telerehabilitation clinic visit increased when there was significant COVID-19 spread (78%) and if internet costs were subsidised (78%). In contrast, if respondents were compensated for the time required for the visit (either clinic based or telerehabilitation), the probability of choosing a clinic visit increased from 33% (base case) to 41% (no work cost), whereas the probability of a telerehabilitation visit decreased from 65% (base case) to 58% (no work cost). Those with a mTBI were relatively unaffected about the risk of contracting COVID-19.