This was a randomized control trial utilizing a waitlist control group (WLCG). The immediate group completed study assessments and activities in the following order: baseline, intervention, 1-month, 6-month, 1-year. The WLCG completed studies in the following order: baseline, 6-month-pre-training, intervention, 1-month, 6-month. For the WLCG, 6-month-pre-training occurred at the same time as the 6-month assessment for the immediate group and consisted of the same assessments.

Participants were enrolled at four Spinal Cord Injury Model Systems Centers: Midwest Regional SCI Care System, Northern New Jersey SCI System, South Florida SCI System, and University of Pittsburgh Model Center on SCI.

Separate trainings were held for MWC and power wheelchair (PWC) users at each site. Participants were randomized to either an immediate or WLCG using permutated blocks of 2 or 4 based on level of injury (paraplegia, tetraplegia) and years since injury (<1 year, ≥1 year). Group assignment codes were generated before the start of the study for a 1:1 allocation ratio by the study biostastician. Allocation was concealed with study members at individual sites contacting the study coordinator at the lead site before baseline testing to receive the group assignment. WLCG participants were told that they were scheduled in the next available intervention in 6 months.

Institutional Review Board approval was obtained at each site and participants provided written informed consent for participation.

A convenience sample was recruited between March 2015 and January 2017 through advertisements, research registries, and word of mouth. Eligiblity criteria were age 18 to 75 years, wheelchair users secondary to a non-progressive SCI, used wheelchairs as their primary means of mobility (>50% of weekly mobility), resided in the community, and had Mini-Mental State Examination scores ≥23. Individuals who were not able to complete maintenance themselves were asked to bring a caregiver or assistant who could assist with maintenance at home or were provided with a volunteer unfamiliar with wheelchair maintenance who they could direct to do the maintenance under their guidance.

Hansen et al., ¹⁰ reported that the number of wheelchair accidents decreased by 100% after a maintenance intervention while there was no decrease in number of accidents for control subjects. By taking a conservative approach and assuming a medium effect size, we would have over 90% power to detect significant differences with a total sample size of 140 participants. Enrollment concluded after recruitment goals were met.

At each site participants attended two classes, each 120 minutes in length. Two make-up classes were offered as necessary. Class sizes ranged from 4-7 wheelchair users to reflect likely group training sizes in a clinical setting. Each class was taught by two trainers who received training from content experts, as described in a previous publication.¹¹ Class 1 included an introduction (5 minutes), overview of objectives and relevance (15 minutes), instructional video on how to take care of a wheelchair at home (5 minutes), didactic and demonstration of caring for a wheelchair at home (60 minutes), and a summary (5 minutes). Class 2 consisted of hands-on wheelchair maintenance guided by a checklist (90 minutes), and a summary and discussion (25 minutes). Trainers at each site completed a 1-day training in-person train-the-training with 2 content experts. Intervention fidelity was addressed through the use of a reference manual and annotated PowerPoint presentation and consistent lesson plans, hands-on activity checklists, maintenance cards, and educational videos across sites. Links to these materials can be found here – [url removed for blinding]. Participants received a printed or digital copy of the slide presentation to enable note-taking. To maximize attendance at both classes, participants received their toolkit and checklist at the second class. A copy of the checklist and list of items in the toolkit can be found in the Supplementary Material.

The following data were collected at baseline: age, sex, body mass index, race, highest level of formal education, occupation, income, SCI level (tetraplegia [C8 and higher] vs paraplegia [below C8]), length of time using a wheelchair, and age of wheelchair used most often.

The Wheelchair Maintenance Training Questionnaire (WMT-Q) evaluates the knowledge of wheelchair maintenance and frequency of wheelchair maintenance performance of wheelchair users.¹¹ The MWC or PWC version of the WMT-Q were used accordingly.

The WMT-Q was completed electronically using Qualtrics Survey Software. Each version of the WMT-Q has three domains: capacity, performance and knowledge. Capacity questions (n=20 MWC, n=21 PWC) asked, “Do you (or your caregiver) know how to do the maintenance skill?” with response options of “yes”, “no” or ”not possible.” If the participant reported the capacity to accomplish a maintenance skill, they were then asked about performance, “How often do you actually do this skill during your daily life?”, with response options of daily, weekly, monthly, quarterly, yearly, or never. The Knowledge domain (n=15 MWC, n=18 PWC; 13 common question) consisted of multiple choice and true/false questions related to the importance of maintenance practice such as, “How many times more likely are you to sustain an injury if you do not maintain your wheelchair?” with response options of “No increased likelihood, two times more likely, five times more likely, ten times more likely”; both question types included “I don’t know” as a response to discourage guessing. Scoring for each domain of the WMT-Q was as follows: Capacity=ΣTasks answered YesTotal Possible − Not Possible∗100 Performance=ΣTasks completedcapacity Denominator∗100 Knowledge=ΣCorrect answersTotal Possible Answers∗100

Participants repeated the WMT-Q at 1-month and 6-months after training. In addition, the immediate group completed a 1-year follow-up while the WLCG completed a 6-month pre-training follow-up (identical to the 6-month follow-up). Participants received 3-month reminders by either phone call, text message or email to remind them of upcoming training or follow-up.

Participants included in data analysis completed at least 1 follow-up assessment time point. Generalized Estimating Equations (GEE) were used to examine between and within group differences in WMT-Q scores. A sensitivity analysis was competed for WMT-Q between group differences using intention-to-treat analysis, specifically last observation carried forward, to evaluate the effects of missing data for participants missing follow-up data. For consistency with previous publications on disparities in wheelchair performance,^1–3 the following variables were recoded: race was dichotomized to White and minority (Black, other); level of education was recoded to high school or less, 2-4 years of college, and graduate degree; occupation was dichotomized to working, student, or on-the-job training and retired, homemaker, unemployed or other. A dichotomous variable was created for responsiveness to training (improved vs stayed the same/decreased) for each WMT-Q domain to indicate an improvement in scores at 1-month-post training. Chi-square and independent t-tests were used to investigate factors associated with responsiveness to training and between group differences (immediate vs. WLCG) at baseline for the following demographic characteristics: wheelchair type, age, years since injury, years of wheelchair use, race, gender, level of injury, level of education, and occupation. All statistical analyses were completed using IBM SPSS version 26.

Of the 147 individuals enrolled in this study, there were 80 MWC users and 67 PWC users. Participation in each study time point for those who did and did not complete the intervention are found in Figure 1. Completion rates at baseline, 1-month-post, and 6-month-post were 70%, 55%, and 51% respectively. 96% of immediate and 70% of WLCG completed the assessment immediately before training (baseline and 6-month-pre-training, respectively). Baseline testing was completed by 97% and 93% of MWC and PWC participants, respectively.

Fifty-six percent of participants had paraplegia with an average (SD) of 14.1 (11.3) years of wheelchair use. Additional demographic characteristics can be found in Table 1. The only between group difference based was a lower percentage of minorities in the immediate group compared to the WLCG. Comparing those who did complete all assessments to those who did not, there were no significant differences between groups (Supplementary Material). Further, sensitivity analyses demonstrated that the results for the full sample were similar in magnitude, direction and extent of statistical significance to those for the 147 participants who met inclusion for data analysis (Supplementary Material).

Following the intervention, there were no differences between the immediate and WLCG for MWC users between baseline and subsequent time points for capacity (χ² (1)=0.25 p=.621, 95% CI (−16.44, 9.806)), performance (χ² (1)=0.25 p=.617, 95% CI (−15.88, 9.414)), or knowledge (χ² (1)=0.5 p=.479, 95% CI (−8.29, 3.888)),). There were also no differences for PWC users for capacity (χ² (1)=0.16 p=.69, 95% CI (−8.782, 5.81)), performance (χ² (1)=0.054 p=. 816, 95% CI (−16.44, 9.806)), or knowledge (χ² (1)=1.007, p=.316, 95% CI (−8.259, 6.502)) between baseline and subsequent time points following the intervention.

There were statistically significant gains in MWC WMT-Q capacity scores, χ² (4)=25.95 p<.001, 95% CI (4.819,15.608) (Figure 2a). Following the training, MWC WMT-Q capacity scores were significantly higher at 1-month (36%), 6-months (42%), and 1 year (38%), compared to baseline scores (26%). All pairwise comparisons were significant at the p<.001 level. MWC WMT-Q capacity for the WLCG 6-month-pre-training time point did not differ from baseline (29%).

PWC users also demonstrated a significant increase in capacity scores, χ² (4)=41.76p<.001, 95% CI (23.698,36.372) (Figure 2a). Correspondingly, PWC WMT-Q capacity scores were significantly higher at 1-month (89%), 6-months (86%), and 1-year (85%), following the training, compared to baseline scores (59%). All pairwise comparisons were significant at the p<.001 level. PWC WMT-Q capacity for the WLCG 6-month-pre-training time point did not differ from baseline (57%).

There were statistically significant gains in MWC performance scores, χ² (4)=29.78 p<.001, 95% CI (6.363,17.293) (Figure 2b). Following the training, participant average scores were significantly higher at 1-month (35%), 6-months (40%), and 1-year (37%), compared to baseline scores (24%). All pairwise comparisons were significant at the p<.001 level. MWC WMT-Q performance for the WLCG 6-month-pre-training time point did not differ from baseline (26%).

Similarly, PWC users demonstrated a significant increase in performance scores, χ² (4)=41.76p<.001, 95% CI (24.462,36.882) (Figure 2b). WMT-Q performance scores were significantly higher at 1-month (85%), 6-months (84%), and 1 year (82%), compared to baseline (55%). All pairwise comparisons were significant at the p<.001 level. PWC WMT-Q performance for the WLCG 6-month-pre-training time point did not differ from baseline (52%).

Knowledge scores did not change for MWC users over time, χ² (1)=6.76 p=.149, 95% CI (−1.474,3.356). For PWC users, knowledge scores increased significantly, χ² (4)=28.94 p<.001, 95% CI (5.233,11.426) (Figure 2c). Following the training, participant average scores for PWC knowledge were significantly higher at 1-month (28%, p<.001) and 6-months (23%, p=0.028), compared to baseline scores (20%). These changes were not retained at 1-year-post-training (23%). PWC WMT-Q knowledge for the WLCG 6-month-pre-training did not differ from baseline (19%).

The percentage of participants who improved in domain score at 1-month follow-up were as follows for MWC and PWC users: capacity (86.00% and 90.91%) performance (86.36% and 88.33%) and knowledge (47.06% and 65.67%). No participant demographic characteristics were associated with responsiveness for any of the WMT-Q domains; however, there were differences based on baseline WMT-Q scores (Figure 3). MWC users who responded to training had lower baseline scores for capacity (t=4.36, p<.001), performance (t=5.37, p<.001), and knowledge (t=4.42, p<.001). PWC users who responded to training had lower baseline scores for knowledge (t=4.82, p<.001).

Participants were a convenience sample across 4 SCI Model Systems Centers and may not be representative of the general SCI population. Individuals were aware that the study targeted wheelchair maintenance and may have been motivated to learn the techniques. Sacrificing this aspect of blinding was important as a previous study on wheelchair skills highlighted the limitations of a double blind study design, such as enrolling individuals with high baseline skill or lack of interest in the intervention.¹⁷ Future work should investigate additional effects of maintenance training including the incidence of repairs, consequences, quality of life, and participation. Future studies may also consider comparing 1:1 training to group training to identify potential benefits of either type of training. Further, secondary to the limitations of in-person training as highlighted by the 2020 COVTD-19 global pandemic, future studies may also investigate how maintenance training could be translated to a virtual setting.

Missing data secondary to failure to complete assessment was an issue. Only 70% completed at least two assessments and the intervention and thus were eligible for inclusion in analysis. Decreasing completion rates with follow-up (baseline vs 1-month vs 6-months) suggests the rate of failure to complete assessments may be partly related to the duration of study enrollment. Intervention arm assigned may also be a factor as pre-training assessment was higher for the immediate group. Wheelchair type does not appear to be a factor as there were no differences in baseline completion between groups. Although WLCG designs are attractive alternatives to active control groups delivered conjoint with the intervention, they may increase risk that participants may not complete “pre-intervention” (i.e. “post-waitlist”) assessment.