A Clinical Trial Investigating Telehealth and In-Person Social Communication Skills Training for People With Traumatic Brain Injury: Participant-Reported Communication Outcomes

Abstract

Objective: 

To investigate the efficacy of telehealth-based and in-person social communication skills training (TBIconneCT) for people with moderate to severe traumatic brain injury (TBI) based on outcomes reported by the survivor and a close communication partner.

Setting: 

Australia. Two telehealth dyads were located outside Australia.

Participants: 

Adults (n = 51) at least 6 months after moderate-severe TBI with social communication skills deficits, and their usual communication partners (family members, friends, or paid carers).

Design: 

Partially randomized controlled trial, with a telehealth intervention group, in-person intervention group, and a historical control group.

Main Measures: 

La Trobe Communication Questionnaire (LCQ) (total score, and number of items with perceived positive change). Both self- and other-reports.

Results: 

Trained participants had significantly more items with perceived positive change than did historical controls. A medium effect size in the sample was observed for improvements in total score reported by trained communication partners after treatment. Comparisons between telehealth and in-person groups found medium to large effect sizes in the sample, favoring the telehealth group on some LCQ variables.

Conclusions: 

Whether delivered via telehealth or in-person, social communication skills training led to perceived positive change in communication skills. It was unexpected that outcomes for the telehealth group were better than for the in-person group on some variables.

TRAINING COMMUNICATION PARTNERS is best practice in providing intervention for cognitive communication impairments after traumatic brain injury (TBI).¹ Providing communication partners with skills to enable effective conversations creates a more positive daily communication environment for people with a TBI. Training focuses on addressing negative patterns that communication partners may use in conversations with people with TBI, such as failing to follow up on the person's contributions, not giving enough information to support the person's comprehension, using questions designed to test the person's memory rather than asking for meaningful information, and querying the person's accuracy.² The aim of training is to facilitate collaborative interactions between people with TBI and communication partners.³ The TBI communication partner training program with the highest level of evidence is TBI Express.⁴ This program has been shown to improve the quality of conversations^5,6 and self-reported communication outcomes.⁷

TBI Express is an intensive program, involving 35 hours of intervention consisting of a combination of group sessions and dyad sessions (attended by both the person with TBI and their communication partner). This makes it difficult to implement TBI Express in full, given limitations on clinician time.⁸ Availability of families is a further barrier to accessing training, given factors of distance from rehabilitation services⁹ and competing time demands.¹⁰ The TBIconneCT program is a reduced-intensity version of TBI Express involving 15 hours of intervention over 10 sessions with the option for in-person or telehealth delivery. Each session involves the person with TBI and their communication partner attending together. TBIconneCT has shown positive outcomes using telehealth delivery with 2 participants in a single case experimental design study.¹¹ To further investigate the effectiveness of TBIconneCT, a clinical trial was conducted. This trial had 3 arms: in-person TBIconneCT training, telehealth TBIconneCT training, and a historical control group.

The current study reports on a secondary outcome measure from this previously reported trial of TBIconneCT.¹² The previous report addressed the primary outcome measure of conversational quality using the Adapted Measure of Support in Conversation (Reveal Competence scale)¹³ based on ratings from a blinded assessor. The secondary outcome measure reported in the current study evaluated the impact of the training on communication problems from the perspectives of the most important stakeholders: people with TBI and their usual communication partners.¹⁴ The La Trobe Communication Questionnaire (LCQ)¹⁵ is the secondary measure analyzed in the current article. Although level of insight will affect participants' reporting, the LCQ has been shown to be a valuable tool that is sensitive to change, and it has been recommended as a supplemental outcome measure in TBI research.¹⁶

The research questions for the current study were:

• (1) Is TBIconneCT training more efficacious than no training in improving LCQ outcomes as rated by people with TBI and their usual communication partners?

For the purposes of this question, the group of trained participants (formed by combining in-person and telehealth participants into a single group) was compared with the group of historical control participants.

• (2) What is the magnitude of any difference in LCQ outcomes between telehealth and in-person training?

The sample size of this trial was not large enough for a noninferiority design. This question is therefore exploratory in nature.

METHODS

Trial design

The trial protocol¹⁷ was registered at the Australian New Zealand Clinical Trials Registry (ACTRN12615001024538). There were no important changes to study methods after trial commencement. The project was approved by the Hunter New England Human Research Ethics Committee (12/06/20/4.04). The CONSORT statement¹⁸ was used to guide reporting of this trial.

Participants with TBI were recruited from brain injury units based in urban and rural areas of New South Wales, Australia. The project was also advertised through social media and brain injury support groups. Participants with TBI who consented to participate nominated a communication partner (family member, friend, or carer) who was invited to participate. As noted earlier, the clinical trial of TBIconneCT reported here had 2 intervention arms (in-person TBIconneCT training and telehealth TBIconneCT training) and a historical control group. Participants were assessed at baseline, posttreatment, and at a follow-up assessment.

For the 2 intervention arms of the trial, we recruited 23 participants with TBI within 2 hours' drive from the university and 13 participants with TBI beyond 2 hours' drive. This resulted in a total of 36 intervention participants. The 13 participants outside Sydney were allocated to telehealth training. The 23 participants within Sydney were randomized to telehealth or in-person training at a ratio of 1:3 (6 telehealth: 17 in-person) using a blocked randomization approach, with blocks of 4. Sequences were generated using computer software by an independent university staff member, who was outside the study team. Clinicians enrolling participants were blinded to the allocation sequence for participants within Sydney. Allocations were stored in sealed opaque envelopes and opened after participants completed initial assessments. For participants outside Sydney, clinicians were not blinded to training mode when enrolling participants, as the clinicians were aware of the participants' location and the allocation protocol.

Rather than allocating available participants to a third no-treatment arm, a historical control group was used to increase the study's power. An earlier study involved recruiting 44 outpatients with moderate-severe TBI and their communication partners from brain injury rehabilitation units in Sydney, Australia.⁷ These participants were allocated to 3 groups based on their availability to attend training: (1) joint social communication skills training where the person with TBI and communication partner attended together, (2) solo social communication skills training where only the person with TBI attended, and (3) delayed treatment control. For the present study, we accessed the data from the no-treatment period for the 15 participants in this historical control group, creating a total sample of 51 participants.

Participants

Eligibility criteria for participants with TBI recruited to the training arms were (1) a moderate-severe TBI occurring at least 6 months previously, defined as a score on the Glasgow Coma Scale of 9 to 12 (moderate) or 8 or less (severe) and/or a period of posttraumatic amnesia of 1 to 24 hours (moderate) or more than 24 hours (severe)¹⁹; (2) discharged from hospital; (3) significant social communication deficits identified by brain injury service staff based on observation in the context of usual clinical interactions, or for self-referring participants, at least one aspect of social communication rated as inappropriate by a speech pathologist using a pragmatic protocol²⁰ during a 15-minute conversation; (4) have an available communication partner; (5) aged 18 to 70 years; (6) have a home computer with internet connection; and (7) proficient in English. Exclusion criteria were (a) severe aphasia (ie, unable to produce sentence-level responses in a conversation); (b) severe amnesia preventing participants from providing informed consent; (c) severe dysarthria significantly affecting intelligibility to an unfamiliar listener; (d) drug or alcohol addiction, or active psychosis as determined by report from the referring clinician, self or family member; and (e) co-occurring degenerative neurological disorder or multiple TBIs. The inclusion criteria for communication partners were that they were already known to the person with TBI (eg, spouse, family member, or paid carer) and that they interacted with the person at least once a week. The only exclusion criterion for communication partners was history of a severe brain injury. The historical control participants were recruited using the same eligibility criteria, with the exception that participants were required to have sustained a moderate-severe TBI at least 9 months previously (rather than 6 months, as was the case for the 2 TBIconneCT training arms). A lower criterion for months after injury was used for the training arms to broaden the pool of potential participants to maximize recruitment. Participants were permitted to continue usual multidisciplinary rehabilitation during the study; however, any concurrent speech pathology intervention was not permitted to target social communication skills or communication partner training. Adherence to this criterion was confirmed through liaison with the rehabilitation team speech pathologist.

Case history and baseline functional evaluation tools

Several measures were taken at baseline to examine the comparability of the 3 samples of participants and to determine whether there was any evidence that statistical adjustments for baseline differences between groups might enhance the analysis. A structured case history form was used to collect demographic information on age, years of education, and number of months after injury. Participants' location was categorized as major city, inner regional, outer regional, remote, or very remote by inputting addresses into a search tool.²¹ This tool is based on the Australian Statistical Geographical Standard Remoteness Area (ASGS-RA) scheme, which classifies locations based on measures of population and access to services.²² Two activity-based measures were administered to describe the participants' everyday level of functioning. First, participants completed Task 4 of the Functional Assessment of Verbal Reasoning and Executive Strategies (FAVRES).²³ The FAVRES is an assessment of cognitive communication skills evaluating the ability to participate in functional communication contexts across 4 assessment tasks. Of these 4 tasks, FAVRES Task 4 has been most strongly correlated with attention, speed of processing, memory, and executive function scores.²⁴ This task was used with special permission for administration over videoconferencing from the publisher, CCD Publishing. Second, the Sydney Psychosocial Reintegration Scale (SPRS)²⁵ was used as a global measure of psychosocial functioning.

Intervention

Each participant in telehealth and in-person intervention modes received 10 training sessions of approximately 1.5 hours' duration, as guided by the TBIconneCT manual. The content of the manual was modified from the TBI Express manual.⁴ Content was identical across training modes, with telehealth participants receiving Skype-based sessions and in-person participants receiving home visits from the clinician. TBIconneCT, like TBI Express, is based on behavioral, cognitive, and educational theory. It aims to increase positive and decrease negative communication behaviors of people with TBI and their communication partners. The treatment components or “ingredients” of TBIconneCT are consistent with those of TBI Express and other social communication skills interventions.^26,27 The core processes were to (a) reflect on positive and/or negative communication experiences since last session, (b) discuss completion of home practice tasks, (c) replay at least one recorded conversation, (d) discuss aspects of the conversation with the participants, (e) set home practice tasks in collaboration with participants, and (f) provide participants with a session summary page. Activities included education about communication processes, identification of personally relevant strategies, and practice within conversational tasks. Treatment ingredients included repeated trials, clinical modeling, feedback, rehearsal and role-play, strategy instruction, self-monitoring, self-correction, and education (see Table, Supplementary Digital Content 1, available at: https://links.lww.com/JHTR/A333, which uses the TIDieR²⁸ framework to describe the intervention). Participants in the historical control group continued with any usual multidisciplinary rehabilitation and did not receive any social communication skills intervention or communication partner training during the study period.

Outcome measure

This article reports on data from the LCQ,²⁹ a secondary outcome measure in the TBIconneCT trial. The full list of outcome measures is available in the trial protocol.¹⁷ The LCQ measures frequency of communication problems of the person with TBI, from the perspective of the person with TBI (referred to as self-report) and a usual communication partner such as a family member, friend, or carer (referred to as other-report).²⁹ The questionnaire comprises 30 items about social communication problems occurring after brain injury. Respondents are asked to consider all the communication situations encountered in daily life (eg, family, social, and work situations) and report current frequency of difficulty. An example item is, “When talking to others does your family member/friend leave out important details?” Respondents report the frequency of each problem on a 4-point scale (never or rarely, sometimes, often, and usually or always). Individual item scores are summed to produce the LCQ total score. On postintervention and follow-up forms for each LCQ item, respondents additionally reported on their perception of any change in frequency for each item since last questionnaire administration (no change, behavior happens more often, or behavior happens less often). Participants reported on their perceptions of change without referring to scores provided at the previous questionnaire administration. This data therefore reflects participants' subjective impressions of progress, or lack of progress since the previous assessment. The current study focused on 2 variables: (a) LCQ total score (self- and other-report), ranging from 30 to 120 (higher scores represent greater perceived difficulty); and (b) number of items with perceived positive change (ie, reduced frequency) (self- and other-report), ranging from 0 to 30 (higher scores represent greater perceived progress).

The LCQ total score is widely used as an outcome measure in TBI intervention research,^6,7,30–36 but the LCQ number of items with perceived positive change is used less frequently.^7,11 Both variables were included in the clinical trial of TBI Express⁷ (the program from which TBIconneCT was adapted) and therefore both were included in this trial. The LCQ total score has high internal consistency (Cronbach α: self-report = 0.91, other-report = 0.92) and acceptable test-retest coefficients (self-report: r = 0.81, other-report: r = 0.87).²⁹ The LCQ is also reliable for administration over videoconferencing.³⁷ The LCQ total score summarizes the overall degree of change. Large change in a few items, or smaller change in many items, can produce the same total degree of change. Number of items with perceived positive change is an index of the more abstract concept “breadth of change,” in areas sampled by the LCQ, so it is treated as a continuous variable rather than a count.

The sample size for the clinical trial (calculated on the primary outcome measure¹⁷) is adequate for the LCQ data. Based on other-reported LCQ data from a previous trial,⁷ which provided an estimated effect size of d = 1.21 for comparisons between trained and control participants, a sample size of 12 participants in each group provides 80% power to detect differences between 2 groups with α of 0.05.³⁸ The sample size of 36 trained participants and 15 control participants is therefore adequate for addressing research question 1. The sample size was insufficient for a noninferiority design comparing outcomes between the in-person and telehealth training groups. Research question 2 is therefore exploratory in nature and involves analysis of effect sizes in the sample, using P values to determine whether there are any significant differences using the sample size available.

The LCQ was administered to both participants with TBI and communication partners at the initial assessment, at 1 to 3 weeks after completion of the intervention phase and at a follow-up assessment. There were 2 participants with TBI (1 in the in-person group and 1 in the telehealth group) who did not demonstrate adequate comprehension of the LCQ questions at the initial assessment, and therefore only data from their communication partners are reported. The follow-up assessment for participants in the telehealth and in-person training groups occurred 3 months after the postassessment, whereas the follow-up assessment for the historical control group occurred 6 months after the postassessment. Given that all participants were in the chronic stage of recovery, we do not expect this difference in timing to have a marked impact on the results. A 3-month follow-up time point was used for participants in the training groups to contain data collection within the constraints of the project timeline. This was necessary given anticipated slow rates of participant recruitment and the staggered nature of enrolling individual participants to the training arms. It was also anticipated that use of a 3-month time point would increase the retention of participants at follow-up. The 6-month follow-up time point had been more feasible for the historical control participants, as this data had been collected as part of a larger multisite project with recruitment of participants in planned waves.

Each assessment for participants in the telehealth and in-person training groups was conducted over Skype via an interview format. Respondents were mailed a copy of the questionnaire to refer to during the interview. Each assessment for participants in the historical control group was conducted face-to-face in an interview format, with respondents referring to a provided copy of the questionnaire during the interview. Some participants preferred to self-complete the questionnaire. In these cases, the assessor reviewed the completed questionnaire together with the respondent to ensure completion and comprehension of all items.

We aimed for assessors to be blinded to the intervention to which participants were allocated; however, there was unblinding for some assessments. At initial assessment, clinicians were blinded for all Sydney participants as allocation occurred after this assessment. Clinicians conducting initial assessments for all historical control group participants were also blinded.³⁹ However, clinicians conducting initial assessments for out-of-Sydney participants were aware of the participants' location and the protocol for allocation to groups, and therefore were unblinded. Postassessments and follow-up assessments were completed by an independent clinician (ie, a clinician who had not delivered the training). For postassessments and follow-up assessments, assessing clinicians were inadvertently unblinded for 16% (7/45) of participants at postassessment and 16% (6/38) of participants at follow-up due to participants revealing cues about their location or training mode. It was not possible to blind research participants or clinicians providing the intervention.

Data analysis

Given that allocation to groups was not fully randomized, demographic variables and LCQ scores at baseline were compared to determine whether groups were similar. Comparisons involving categorical variables were conducted using Fisher exact tests. Continuous variables were first tested for normality and then compared using 1-way analysis of variance (ANOVA) for variables with normal distributions and Kruskal-Wallis H tests for variables with nonnormal distributions.

The primary research questions were analyzed with pairs of planned orthogonal contrasts, testing self-report LCQ mean change scores from baseline to posttreatment, and posttreatment to follow-up and, similarly, other-report LCQ mean change scores. The first contrast of each pair compared treated participants with control participants, and so addressed the first research question. The second contrast of each pair addressed the second research question by comparing mean change between in-person to telehealth treatment participants. For the second research question, type II error rates were likely to be high, so the emphasis of the analysis was on the effect sizes in the sample means rather than significance. Effect sizes were calculated as d, in the form mean 1 – mean 2 divided by the pooled standard deviation. Cohen's guidelines for effect size were used, where d = 0.2 is “small”, 0.5 “medium,” and 0.8 “large.”⁴⁰ Supplementary analyses using the mean of number of items with perceived positive change at posttreatment and follow-up also employed the same pattern of orthogonal contrasts and effect sizes as analysis of the primary measures. Contrasts keep the type I error rate at .05 for each contrast pair. Across the set of contrast pairs, no Bonferroni adjustment was made. Readers should accordingly exercise judgment in interpreting P values.

Where variables had a nonnormal distribution, bootstrapping using 2000 samples and a 95% confidence interval was used to calculate significance. For all analyses of variables with normal distributions, data were initially inspected to identify any outliers. Influence of outliers was tested by conducting analyses with and without outliers. Exclusion of outliers did not change significance of any group comparisons (ie, P values < .05 remained < .05, values > .05 remained > .05) and therefore all outliers were retained.

RESULTS

Allocation and retention of participants

Figure 1 illustrates the allocation process. Data collection occurred from January 2015 to May 2018 and ceased after the sample size had been reached. No harms or unintended effects of the intervention were reported to the researchers or via the human research ethics committee.

Figure 1.: Allocation flow diagram.

A total of 16/19 telehealth participants and 15/17 in-person participants completed training and at least some postassessment. There were 14/15 control participants who completed the postassessment. All dropouts during intervention from both training groups (telehealth: n = 3, in-person: n = 2) were due to lack of availability. One dyad allocated to in-person training moved overseas after 4 training sessions and so completed remaining sessions via telehealth. This participant's data were analyzed with the in-person group as per intention-to-treat analysis.

At postassessment, there was missing data from one communication partner within the in-person group, who partially completed the postassessment but did not have time to complete the LCQ. One control participant and his communication partner did not complete postassessment due to illness. At follow-up, all dropouts were due to participants declining to participate or being uncontactable.

Preliminary analysis: Participant baseline characteristics

Demographic variables and scores on functional evaluation tools for the 3 groups are reported in Table 1. SPRS scores indicated that this cohort of participants generally had low levels of participation across the areas of work and leisure, relationships, and living skills. Similarly, FAVRES scores showed reduced cognitive communication skills. However, a few participants scored at the upper range on specific domains.

TABLE 1 - Comparison between the 3 groups on demographic and baseline variables

	In-person (n = 17)	Telehealth (n = 19)	Control (n = 15)	H (Fa)	df	P
Demographic variables
Age, median (range), y	54 (20-68)	42 (19-66)	36 (19-68)	5.50	2	.06
Gender, male/female, n	13/4	17/2	13/2	1.22b	n/a	.63
Education, mean (SD), y	14.4 (2.7)	13.8 (3.2)	12.7 (3.2)	1.17a	2,48	.32
TPI, median (range), mo	12 (6-574)	53 (6-342)	91 (24-276)	7.00	2	.03
PTA, median (range), d	42 (10-98)	46 (1-183)	40 (6-182)	0.43	2	.81
CP age, median (range), y	43 (20-78)	57 (27-67)	57 (21-79)	0.97	2	.62
CP gender, male/female, n	2/15	3/16	3/12	0.54b	n/a	.89
CP education, median (range), y	14 (10-26)	13 (10-17)	12 (10-16)	2.62	2	.27
Communication partner relationship to person with TBI
Family member, n	14	18	10	4.85b	n/a	.24
Friend, n	1	0	1
Paid support worker, n	2	1	4
Participant location (ASGS-RA)
Major city, n	16c	7	14	n/a	n/a	n/a
Inner regional, n	1	8	1
Outer regional, n	0	3	0
Overseas, n	0	1	0
Functional evaluation tools
SPRS Work & Leisure, median (range)	7 (1-12)	10 (1-14)	11 (1-14)	2.40	2d	.30
SPRS Relationships, mean (SD)	8.9 (4.5)	9.3 (3.1)	8.7 (3.5)	0.10a	2,43	.90
SPRS Living Skills, median (range)	10 (3-16)	13 (1-15)	11 (3-16)	0.98	2d	.61
FAVRES Accuracy SS, median (range)	41 (1-106)	41 (1-106)	42 (1-106)	0.22	2e	.90
FAVRES Rationale SS, median (range)	44 (16-107)	71 (16-107)	44 (16-71)	0.83	2e	.66
FAVRES Time SS, median (range)	102 (62-113)	95 (59-120)	101 (84-117)	0.74	2f	.69
FAVRES Reasoning score, mean (SD)	16.7 (5.4)	15.4 (5.5)	11.8 (7.7)	1.65a	2,31	.21
Outcome measure: baseline scores
LCQ self-report, mean (SD)	61.3 (19.6)	58.7 (15.8)	57.2 (16.8)	0.22a	2,46	.80
LCQ other-report, mean (SD)	64.4 (12.5)	64.6 (16.6)	57.7 (15.7)	1.07a	2,48	.35

Abbreviations: ASGS-RA, Australian Statistical Geography Standard—Remoteness Area; CP, everyday communication partner; FAVRES, Functional Assessment of Verbal Reasoning and Executive Strategies; LCQ, La Trobe Communication Questionnaire; n/a, not applicable; PTA, posttraumatic amnesia; SPRS, Sydney Psychosocial Rating Scale; SS, standard score; TBI, traumatic brain injury; TPI, time postinjury.

^aF-statistic reported.

^bFisher exact test statistic reported.

^cOne of these participants moved overseas during the training program.

^dn = 15 for in-person group, n = 16 for telehealth group, n = 15 for control group.

^en = 12 for in-person group, n = 13 for telehealth group, n = 8 for control group for these variables.

^fn = 11 for in-person group, n = 13 for telehealth group, n = 8 for control group for this variable.

Groups were statistically similar on all demographic and functional evaluation variables, except for a significant difference between groups in time postinjury (P = .030). Additionally, it was noted that in-person participants were older than telehealth and control participants. The difference did not reach statistical significance (P = .06) but the effect size was medium (d = 0.64). For time postinjury, pairwise comparisons were performed with a Bonferroni correction for multiple comparisons, with the adjusted P value reported. This post hoc analysis showed a statistically significant difference in time postinjury between the in-person and control groups (P = .029), but not between any other group combinations. To investigate whether months postinjury influenced LCQ scores, multiple regression models were run to predict LCQ variables (change scores and number of items with perceived positive change, both self- and other-reports) using 2 predictors: months postinjury, and intervention versus control (using nonorthogonal simple contrast coding, with separate models for in-person vs control, and telehealth vs control, applying a Bonferroni correction). Bootstrapping using 2000 samples and a 95% confidence interval was used due to the small sample size. Months postinjury was not a significant predictor of any LCQ variables, and therefore was not used as a covariate when analyzing treatment effects.

Baseline LCQ scores are also reported in Table 1. One-way ANOVAs were conducted with no significant differences between groups identified. Additionally, there were no significant differences in baseline scores between participants retained in the study versus dropouts at the postassessment time point, and between participants retained in the study versus dropouts at the follow-up assessment time point.

Analysis of treatment effects

Analysis of LCQ total scores: Pre- versus posttreatment

LCQ scores for individual participants are provided as online content (see Table, Supplementary Digital Content 2, available at: https://links.lww.com/JHTR/A334). Table 2 shows pretreatment and posttreatment LCQ total scores for both self-report and other-report versions. The first research question was assessed with the first planned orthogonal contrast, which showed no significant differences between trained and control participants for self-report and a negligible effect size in the sample (P = .82, d = 0.07). The second research question was assessed with the second orthogonal contrast, comparing in-person and telehealth participants on pre-/postchange scores for self-report. In keeping with the second question's focus on effect sizes, it found an effect size in the sample in the “small” range by Cohen's guidelines (d = 0.11, P = .79). The first orthogonal contrast using pre-/postchange scores for other-report showed no significant differences between trained and control participants; however, the effect size in the sample was medium (d = 0.60, P = .10), with trained participants reporting greater improvement than control participants. There was also a medium effect size in the sample for comparison between in-person and telehealth participants on other-report scores, with telehealth participants reporting greater improvement than in-person participants, but the difference was not significant (d = 0.70, P = .07).

TABLE 2 - Pretreatment versus posttreatment self and other LCQ total scores

	Pretreatment			Posttreatment			Treatment effect Question 1 (trained vs control)				Treatment effect Question 2 (IP vs TH)
	IP n = 14	TH n = 16	CTRL n = 14	IP n = 14	TH n = 16	CTRL n = 14	t	df	P	d	t	df	P	d
LCQ self total, mean (SD)	61.1 (20.8)	58.3a (16.1)	57.4 (17.4)	59.8 (20.5)	56.1a (12.5)	56.3 (13.9)	−0.24	40	.82	0.07	−0.27	40	.79	0.11
LCQ other total, mean (SD)	62.1 (12.3)	65.5 (15.1)	56.1 (14.9)	58.8 (13.5)	55.9 (12.2)	54.6 (11.7)	−1.72b	25.0	.10	0.60	−1.95b	24.7	.07	0.70

Abbreviations: CTRL, control; IP, in-person; LCQ, La Trobe Communication Questionnaire; TH, telehealth.

^an = 15 for LCQ self-data for telehealth group.

^bWelch's t-test used due to unequal variances.

Analysis of LCQ total scores: Posttreatment versus follow-up

We examined any further change in LCQ scores between posttreatment and follow-up assessments. Table 3 shows posttreatment and follow-up LCQ total scores for both self-report and other-report versions. The first orthogonal contrast testing posttreatment to follow-up change scores for self-report showed no significant differences between trained and control participants (P = .36, d = 0.32). Comparison between in-person and telehealth participants on self-report scores had a medium effect size in the sample, with telehealth participants reporting greater improvement than in-person participants, but the difference was not significant (d = 0.58, P = .19). Contrast tests using posttreatment to follow-up change scores for other-report showed no significant differences between trained and control participants, but a close to medium effect size in the sample (d = .49, P = .17), with trained participants reporting a greater reduction in frequency of communication problems at follow-up than control participants. Comparison between in-person and telehealth participants showed a negligible effect size in the sample (d = 0.03, P = .95).

TABLE 3 - Posttreatment versus follow-up self and other LCQ total scores

	Posttreatment			Follow-up			Treatment effect Question 1 (trained vs control)				Treatment effect Question 2 (IP vs TH)
	IP n = 10	TH n = 11	CTRL n = 12	IP n = 10	TH n = 11	CTRL n = 12	t	df	P	d	t	df	P	d
LCQ self total, mean (SD)	60.1 (16.3)	57.3 (9.3)	53.8 (13.1)	60.9 (16.7)	53.5 (11.8)	49.6 (11.9)	0.93	30	.36	0.32	−1.34	30	.19	0.58
LCQ other total, mean (SD)	61.4 (14.5)	52.4 (8.7)	53.5 (12.3)	59.7 (14.3)	50.9 (10.3)	55.3 (12.7)	−1.43a	22.3	.17	0.49	0.06a	13.7	.95	0.03

Abbreviations: CTRL, control; IP, in-person; LCQ, La Trobe Communication Questionnaire; TH, telehealth.

^aWelch's t-test used due to unequal variances.

Supplementary analyses: Number of items with positive change

Table 4 shows the number of LCQ items that participants with TBI and communication partners perceived as improved at postassessment and follow-up. For postassessment data, contrast tests showed that trained participants reported significantly more items with perceived positive change than did control participants for both self- and other-reports, with very large effect sizes in the sample (both P < .001, d = 1.93 and 2.43, respectively). Comparisons between in-person and telehealth participants found small effect sizes in the sample with no significant differences between groups for both self- (d = 0.17, P = .65) and other-reports (d = 0.23, P = .55). For follow-up data, contrast tests showed that trained participants reported significantly more items with perceived positive change than did control participants, again with very large effect sizes in the sample, for both self- (P = .001, d = 2.07) and other-reports (P < .001, d = 2.37). Comparisons between in-person and telehealth participants on follow-up data found a large effect size in the sample favoring the telehealth group for self-report, but this difference was not significant (d = 0.90, P = .10). For other-report, there was a negligible effect size in the sample and no significant difference (d = 0.07, P = .89).

TABLE 4 - Number of LCQ items with perceived positive change at posttreatment and follow-up

	LCQ Number of items with positive change, median (range)			Treatment effect Question 1 (trained vs control)				Treatment effect Question 2 (IP vs TH)
	IP	TH	CTRL	t	df	P	d	t	df	P	d
Posttreatment	n = 14	n = 16	n = 14
Self-report	13.5 (0-28)	13.5 (6-28)a	1 (0-12)	−5.49	39	<.001	1.93	−0.50	39	.65	0.17
Other-report	15 (0-25)	15 (2-28)	0.5 (0-4)	−9.06b	32.15	<.001	2.43	−0.61b	27.50	.55	0.23
Follow-up	n = 10	n = 10	n = 12
Self-report	9.5 (0-25)c	16.5 (8-30)	0.5 (0-8)	−6.05b	16.27	.001	2.07	−1.85b	12.29	.10	0.90
Other-report	14 (0-27)	13.5 (3-26)	0 (0-7)	−7.35b	22.10	<.001	2.37	0.15b	17.60	.89	0.07

Abbreviations: CTRL, control; IP, in-person; LCQ, La Trobe Communication Questionnaire; TH, telehealth.

^aTwo participants with TBI did not complete change scores, so n = 14 for self-report.

^bWelch's t-test used due to unequal variances.

^c Two participants with TBI did not complete change scores, so n = 8 for self-report.

DISCUSSION

Participants who underwent TBIconneCT communication partner training perceived improvements in their conversations after participating in training, although this was not consistently supported across all measures. The first research question was whether training (regardless of mode) was more efficacious than no training. At posttreatment, comparison between trained and control participants found no treatment effect on LCQ total score for self-report; however, a medium effect size in the sample was found for LCQ total score for other-report, with trained communication partners tending to report greater improvement than did controls. For change from post-treatment to follow-up, there was no significant difference between the trained and control participants, for both self-report and other-report. It was noted that trained communication partners reported some reduction in frequency of communication problems at follow-up with an effect size in the sample close to the medium range. These findings are commensurate with those from analysis of LCQ total scores in the TBI Express clinical trial,⁷ in which treatment effects were found for other-report but not self-report.

The LCQ variable, which was most responsive to intervention, was the number of items with positive change perceived by participants. Both trained participants with TBI and trained communication partners perceived significantly more LCQ items with positive change at posttreatment compared with historical control participants. At their follow-up assessment (at 3 months), trained participants again reported significantly more items with positive change than did the control group at their follow-up assessment (at 6 months). This suggested the trajectory of improvement was perceived by trained participants to continue over the 3 months after training. The bias of this data, which is based on participant perception of improvement, is acknowledged. It may also be difficult for participants to accurately recall extent of change in communication problems over a period of several months. However, it was useful to confirm that these findings align with those from the TBI Express clinical trial for this LCQ variable.⁷ In both studies, trained participants perceived improvements in communication both over the course of the intervention and in the following months.

The second research question related to comparing the efficacy of in-person training to telehealth-based training. This question was exploratory due to the small sample size. There were some medium to large effect sizes in the sample in favor of the telehealth participants noted on some variables. At posttreatment, there was a medium effect size favoring telehealth participants for LCQ total score (other-report). At follow-up, there was a medium effect size for telehealth participants for LCQ total score (self-report), and a large effect size for number of items with perceived positive change (self-report). This indicates participants with TBI who received telehealth training reported a more positive trajectory of improvement over time, compared with participants with TBI who received in-person training. While there were no significant differences in outcomes on the LCQ between in-person and telehealth participants, the obtained effect sizes in this sample may be useful for sample size calculations in any future noninferiority trial comparisons.

It is unclear why the telehealth group tended to have more positive outcomes on the LCQ than the in-person group. One possible reason is that families found using a novel delivery format more engaging. Another factor is that the telehealth group was composed mostly of regional participants, whereas the in-person group was composed mostly of participants living in major cities. While group comparisons on demographic and baseline characteristics did not reveal any factors which needed to be adjusted for in the analyses, there may be other unmeasured variables relating to the difference in context between the groups, which may have influenced the training outcomes. To further investigate potential reasons for the findings on effect sizes in this sample, which favored the telehealth group, the experience of participating in the training will be explored through analysis of qualitative interviews in a separate investigation.

The finding that participants with TBI and their communication partners viewed their conversations as improved after training complements the analysis of the primary outcome measure (reported elsewhere), which found that blinded ratings of conversations between people with TBI and their trained communication partners significantly improved after training.¹² Measurement of communication outcomes after TBI requires a multifaceted approach,^16,41 and the LCQ data provide important information on the perspectives of people with TBI and their communication partners on problems in conversation and progress over time. Although this data is not from an independent or blinded source, it is nonetheless valuable given that the participants with TBI and their communication partners are the most direct informants on whether the training has had a meaningful effect on their lives. They are the people who experience the impact of communication problems on their day-to-day interactions, and so their perceptions are important. The data from this trial demonstrate that people with TBI and their communication partners had a sense of making progress with conversation skills over the course of training, which is a positive outcome.

Interpretation of self-reported data after TBI can be complex due to awareness deficits¹⁴ and underreporting and overreporting of communication problems.⁴² Self-report can change over time, with reported difficulties in conversation and social behavior increasing in frequency from 2 to 10 years postinjury.⁴³ This may be due to increased awareness with neurological recovery, or in response to problems resuming preinjury roles. Given that the TBIconneCT program involves people with TBI self-monitoring conversations and attempting complex communication tasks, it is possible some individuals reported an increased frequency of communication problems at postintervention due to increased awareness rather than increased difficulties. This creates challenges in interpreting averaged group data, which combines participants with preserved insight and those who lack insight. The issue of awareness may also influence communication partner scores, as partners may increase their own awareness of communication problems during training. Further analyses of scores on the different factors of the LCQ,⁴⁴ individual participant data, or qualitative feedback may be illustrative of differing patterns in LCQ responses.

Limitations

While the sample size for the trial was adequate for comparing trained participants against the historical group, it was not large enough for a noninferiority study comparing in-person and telehealth delivery of the intervention. The trial was not adequately powered for detecting differences between in-person and telehealth groups. However, the medium to large effect sizes noted for this sample in comparisons between in-person and telehealth groups warrant further investigation in future research.

This self-reported data also has an inherent bias due to social desirability effects, particularly when participants indicate their perceptions of items with positive change. While we addressed this through use of an independent assessor (ie, not the clinician who delivered training), participants may have had a vested interest in reporting positive outcomes. Unblinding of assessors during some interviews may have also influenced the data. The LCQ is a well-established outcome measure¹⁶; however, the number of items with perceived positive change has not been used frequently in previous research. Given outcomes on this variable had been reported for the TBI Express intervention,⁷ it was relevant to report outcomes on the same variable for the TBIconneCT trial, so that comparisons between the original and modified versions of this intervention could be made. It is noted that this data provides limited information on clinical significance. It is unclear whether a report of positive change on a specific item represents a negligible, small, or large improvement in level of functioning for that individual. Further research is required to establish whether the number of items with perceived positive change is a valid and reliable outcome measure.

The difference in timing for the follow-up data is acknowledged as an issue. The in-person and telehealth participants completed follow-up assessment after 3 months, whereas the historical control group completed follow-up assessment after 6 months. The difference in timing of the follow-up assessment was related to feasibility concerns about the timeline of the study and participant retention. It was expected this difference would have limited impact on findings due to all participants being in the chronic stage of recovery. However, this did introduce a further source of variation to comparisons between the follow-up data of trained versus control participants. The key findings that were noted relating to this follow-up data were that (a) trained communication partners reported a (nonsignificant) reduction in frequency of communication problems at follow-up compared with the control group, and (b) trained participants reported significantly more items with perceived positive change at follow-up than control participants. It is unclear how these findings would differ if trained participants had completed their follow-up at 6 months: whether the trajectory of positive change would continue as skills are consolidated, or whether there would be loss of improvements over this longer period. The comparisons between trained and control participants using the follow-up data must therefore be interpreted with caution.

We also noted a statistically significant difference in months postinjury between the in-person and control groups. This may have been related to the difference in the inclusion criterion for time postinjury, with trained participants eligible from 6 months postinjury, and historical control participants eligible from 9 months postinjury. It is possible that the difference in time postinjury could have influenced outcomes. However, multiple regression models run to predict LCQ variables demonstrated that months postinjury was not predictive of any LCQ outcomes. This suggests that the difference in months postinjury had little effect. Beyond the key variables compared across groups, there may also have been other unmeasured factors, which differed between groups. Notably, data were not collected regarding the amount of external multidisciplinary rehabilitation received by participants during the study period. Difference between groups regarding the extent of participation in other rehabilitation services is a potential confounding influence.

Clinical implications

The findings support providing social communication skills training to people with TBI and their communication partners in either in-person or telehealth delivery mode. The Lancet Neurology Commission on TBI identified improving access to postacute care to address disparities between regions as a key priority,⁴⁵ and telehealth may be one approach to meeting this need. This study provides evidence to support the clinical use of the TBIconneCT program via telehealth to deliver training to families regardless of location. The activity-based measures, the SPRS and FAVRES, indicated that the cohort of participants had reduced levels of psychosocial functioning and cognitive communication skills, yet still achieved positive outcomes from participating in the TBIconneCT intervention. The findings also reinforce the usefulness of the LCQ as an outcome measure in clinical practice, although further research is needed to support the use of number of items with perceived positive change as an outcome measure.

CONCLUSION

Training using the TBIconneCT program led to improvements in perceived communication ability from the perspectives of people with TBI and their communication partners. The potential of telehealth service delivery is supported by the fact that comparisons between the in-person and telehealth participants in this sample on some variables had medium to large effect sizes favoring the telehealth group. The positive findings from the primary outcome measure (independently rated data)¹² combined with findings on this secondary participant-rated outcome measure provide a broad view of the effects of TBIconneCT conversation skills training.