The rapid spread of telehealth interventions, accelerated by the COVID-19 pandemic, has presented healthcare providers and researchers with a unique opportunity to save on delivery costs of health-related interventions, while potentially increasing reach and accessibility to marginalized groups of patients (Gajarawala and Pelkowski, 2021).

Several studies have now had the opportunity to evaluate the acceptability of telehealth in a range of clinical disciplines, across adult and pediatric populations, and in real-world settings (Tse et al., 2021; Cuthbert et al., 2022). For example, several models of psychological therapy have been delivered and evaluated utilizing telehealth and found to be acceptable, and importantly, addressing a health disparity gap by overcoming geographical barriers and supporting vulnerable remote populations (Cuthbert et al., 2022).

Some work has been conducted to compare efficacy of telehealth delivery vs. traditional face-to-face CBT-I (Gehrman et al., 2021). The effectiveness of CBT-I and adapted CBT-I on sleep outcomes among people with non-malignant chronic pain, in addition to the high levels of acceptability of telehealth interventions in behavioral medicine provide a strong rationale for developing an adapted CBT-I intervention, delivered via telehealth. Furthermore, the NG-193 guidelines (NICE, 2021) have made a key research recommendation to increase the evidence base for CBT-I in the treatment of chronic pain (Zambelli et al., 2022).

Given the predominant focus on feasibility and acceptability, we chose to use a pre-post single-arm trial design with three measurement points (baseline, immediate post-treatment, and one-month follow-up).

Adult women with musculoskeletal chronic pain (MSK; pain typically affecting bones, joints, ligaments, and tendons) were recruited over a two-month period June–July 2022 from an opt-in mailing list created during a previous UK-wide survey (Zambelli et al., 2022).

All participants provided voluntary informed consent prior to the initiation of the study. The study was approved by UCL’s Institute of Education Ethics Review Committee, reference: Z6364106/2022/02/136. This feasibility study was pre-registered via the Open Science Framework: https://osf.io/q5y9x.

Participants were included if they provided consent, were willing to comply with all study procedures, aged between 25 and 65 (working aged adults), were able to communicate in English, resided in the UK, self-reported chronic pain for more than 6 months, received a clinical diagnosis of non-malignant chronic pain affecting their: lower back, knee, hip, shoulder, neck, had a stable pain medication regime, had access to a computer or tablet with stable internet connection as well as a web camera. Finally, participants were included if had an insomnia score of above 8, indicated through screening procedures which were scored by the research team.

Exclusion criteria were: if participant was self-reporting medication or dosage change in the previous 1 month, had the presence of clinically assessed self-injury, current suicidal intent, self-reported borderline personality disorder or any form of schizophrenia, epilepsy, bipolar disorder or dementia spectrum disorders, additional intellectual disability or cognitive and/or genetic impairment that would prevent them from understanding the research protocol, self-reported history or current substance abuse, self-reported clinical diagnosis of unstable restless leg syndrome or period limb movement disorder, currently receiving any form of CBT, overnight shift work in the past 3 months, or pregnant / breastfeeding.

Recruitment took place between June and July 2022, and data collection was completed by November 2022. Figure 1 displays the study procedure flowchart.

The intervention consisted of six individual treatment sessions, including sleep hygiene, stimulus control (adapted for chronic pain), sleep restriction (adapted for chronic pain), as well as additional components targeted specifically to MSK pain, such as relaxation techniques, psychoeducation, and pain management in the context of insomnia. The traditional guidelines for sleep restriction and stimulus control were adapted to make consideration for the impact of chronic pain on an individual’s mobility and bedtime routine (Soong et al., 2021).

The adapted CBT-I intervention protocol was manualized and followed a strict treatment protocol, aiming to keep fidelity to the original aims of CBT-I: reducing beliefs and behaviors that perpetuate insomnia and improving overall sleep quality. The adapted intervention had an average duration of 50 min per session, bar the first session of 90 min. Supplementary File S1 provides an outline of the intervention content by session.

The intervention sessions were delivered via Microsoft Teams. Participants were encouraged to keep their webcam on during each session with the research therapist. If the participant did not wish to have the webcam on, this would not exclude or disadvantage the participant. However, having the webcam on was viewed as closer to traditional face-to-face sessions and provided the research therapist with possible indications of anxiety signs. Participants did not have to register for an account to join the sessions. All sessions were manualized and delivered by ZZ, who received CBT-I practitioner training prior to commencement. Supervision was provided in a case conference format by LH, a BACP-accredited psychologist and trained CBT-I practitioner.

Participants were working-aged women aged between 38 and 65 (M = 50.76 years, SD = 8.03 years). Participants had received diagnosis of MSK pain in secondary care. Most participants reported their pain duration as more than 10 years (n = 6) and were taking medication as part of their pain management regime (n = 7). Participant characteristics are displayed in Table 1.

Feasibility was assessed using frequency analysis which examined the study recruitment rate calculated by dividing the number of participants who completed the study, by the number of participants enrolled. Compliance rates were examined by reporting the completion rates of the study procedure, follow-up completion rate, and rescheduling rates. These rates were expressed as percentages or whole numbers due to the small sample size.

To assess within-group treatment effects on primary outcomes of sleep, and subsequently, secondary outcomes related to pain severity, pain interference, depression, and anxiety a repeated measures ANOVA model was used to generate parameter estimates to determine the significance of changes across time points at baseline, post-intervention, and one-month follow-up for each outcome of interest. The data were found to have met assumptions for normality and sphericity before carrying out repeated measures analyses utilizing the Shapiro–Wilk test.

Effect size using Partial Eta Squared was reported (small d = 0.2, medium d = 0.5, and large d = 0.8; Cohen, 1988). Data were analyzed on IBM SPSS v. 28.

There were 27 participants who came forward and were pre-screened for their eligibility. Of these, 10 participants (37% recruitment/enrolment rate) were enrolled in the study after the pre-screening and full-screening procedure as per criteria outlined here. Five participants did not show up for pre-screening, four were deemed ineligible during pre-screening, and three did not respond to scheduling requests for the full-screening call. Of the 15 participants who attended the full-screening call, three were excluded due to medical history disclosures, and two did not respond to follow-up emails. One participant was later withdrawn due to noncompliance and referred to a sleep clinic for a suspected sleep disorder unrelated to insomnia (see Figure 2 consort flowchart).

One participant was excluded from the study after three sessions due to suspected paradoxical insomnia which caused an inability to comply with the study procedures as per exclusion criteria. This participant completed three further adapted sessions with no sleep restriction component, retaining only information about relaxation/ sleep hygiene techniques.

Of the remaining nine participants, 100% completed the six intervention sessions, and weekly sleep diaries, and completed the follow-up measures and post intervention and 1 month after session six.

Four sessions out of 36 intervention sessions (11%) among the nine final participants were rescheduled due to illness (n = 1), annual leave (n = 2) and scheduling conflicts (n = 1).

All nine participants who completed the study were able to complete the participant satisfaction survey, part of the post-intervention online questionnaire. All nine participants (100%) were “very satisfied” with the overall intervention, on a scale of 0 = very unsatisfied, to 4 = very satisfied. Furthermore, all nine participants (100%) reported they would be “very likely” to recommend the adapted CBT-I intervention to an adult with chronic pain and concomitant insomnia on a scale of 0 = not likely at all, to 4 = very likely.

Participants were asked how far they agreed with seven statements on a scale of 0 = do not agree at all, to 4 = completely agree. All participants agreed that the therapist was able to communicate in a clear way throughout the intervention, trusted the therapist throughout the intervention, the therapist was knowledgeable about sleep and chronic pain conditions, and importantly, participants felt more empowered to manage their sleep problems in the present and the future.

With regards to whether the intervention met the participants’ expectations, seven participants completely agreed with this statement, one participant “mostly agreed,” and one participant “neither agreed nor disagreed.” Eight participants completely agreed that the intervention sessions had been tailored to participants’ individual sleep goals, and one participant “mostly agreed.” Finally, eight participants completely agreed that they themselves felt more knowledgeable about sleep problems and how they interact with chronic pain.

The majority (n = 7) of participants began treatment with a combination of onset and maintenance insomnia. One participant began treatment with onset insomnia only, and one participant began treatment with early morning awakenings. Supplementary File S2 displays the sleep profiles for each of the nine participants who completed the study and attended all six sessions of CBT-I. Figure 3 demonstrates the mean ISI scores between baseline and follow-up which began above threshold for Moderate insomnia and fell to within “no insomnia” range at follow-up. Figure 4 demonstrates changes in individual SOL between baseline and follow-up. Figure 5 shows the increases in mean SE between baseline and follow-up.

Repeated measures ANOVAs were conducted to determine whether there were statistically significant differences in insomnia severity, SOL, WASO, TST, and sleep efficiency over the course of the intervention. Table 2 displays the descriptive statistics across the primary sleep outcomes of interest at baseline, post-intervention, and one-month follow-up.

The first model yielded statistically significant changes in insomnia severity as measured through the ISI, F(2, 16) = 45.13, p < 0.001, partial η² = 0.849, with insomnia severity scores decreasing from 18 ± 5.57 at baseline to 5.58 ± 1.72 at post-intervention and to 6.56 ± 5.74 at one-month follow-up. Post hoc analysis with a Bonferroni adjustment revealed that insomnia severity was statistically significantly decreased from baseline to post-intervention [12.89 (95% CI, 7.08–17.98), p < 0.001], and from baseline to one-month follow-up [11.44 (95% CI, 6.84–16.05), p < 0.001], but not from post-intervention to one-month follow-up [−1.44 (95% CI, −5.07 to 2.18), p = 0.793].

The second model yielded statistically significant changes in sleep onset latency, measured through the sleep diary contents, over time, F(2, 16) = 7.29, p = 0.006, partial η² = 0.477, with sleep onset latency decreasing from an average of 45.36 ± 29.54 min at baseline to 17.68 ± 16.08 min at post-intervention and to 14.13 ± 5.74 min at a one-month follow-up. Post hoc analysis with a Bonferroni adjustment revealed that sleep onset latency was statistically significantly decreased from baseline to one-month follow-up [31.23 (95% CI, 6.11–61.85) minutes, p = 0.046], but not from baseline to post-intervention [26.68 (95% CI, −4.76 to 60.11) minutes, p = 0.099], nor from post-intervention to one-month follow-up [3.56 (95% CI, −10.54 to 17.65) minutes, p = 1.00].

The third model yielded statistically significant changes in wake after sleep onset, measured through the sleep diary contents, over time, F(2, 16) = 3.64, p = 0.50, partial η² = 0.313, with wake after sleep onset decreasing from an average of 45.09 ± 36.90 min at baseline to 17.02 ± 13.83 min at post-intervention and to 21.86 ± 18.59 min at one-month follow-up. Post hoc analysis with a Bonferroni adjustment revealed that sleep onset latency was not statistically significantly decreased from baseline to post-intervention [28.07 (95% CI, −10.12 to 66.27) minutes, p = 0.17], nor from baseline to one-month follow-up [23.23 (95% CI, −16.81 to 63.28) minutes, p = 0.36], nor from post-intervention to one-month follow-up [−4.84 (95% CI, −22.63 to 12.95) minutes, p = 1.00].

The fourth model yielded non-statistically significant changes in total sleep time, measured through the sleep diary contents, over time, F(2, 16) = 1.79, p = 0.199. Post hoc analyses were not consulted.

Finally, the last model yielded statistically significant changes in sleep efficiency, measured through the sleep diary contents, over time, F(2, 16) = 5.41, p = 0.02, partial η² = 0.404, with sleep efficiency increasing from an average of 75 ± 15.30 per cent at baseline to 89.81 ± 7.30 per cent at post-intervention and to 87.81 ± 8.31 per cent at one-month follow-up. Post hoc analysis with a Bonferroni adjustment revealed that sleep efficiency was not statistically significantly decreased from baseline to post-intervention [−13.88 (95% CI, −30.97 to 3.21) per cent, p = 0.12], nor from baseline to 1-month follow-up [−11.88 (95% CI, −26.79 to 3.04) per cent, p = 0.13], nor from post-intervention to one-month follow-up [2.00 (95% CI, −5.26 to 9.27) per cent, p = 1.00].

Repeated measures ANOVAs were conducted to determine whether there were statistically significant differences in secondary clinical outcomes of interest: pain severity, pain interference, anxiety, and depression scores over the course of the intervention and follow-up period. The data were assessed for normality utilizing the Shapiro–Wilk test and all variables were found to be normally distributed (p > 0.05). Table 3 displays the descriptive statistics for all secondary quality of life outcomes of interest at baseline, post-intervention, and one-month follow-up.

The first model yielded non-statistically significant changes in pain severity, measured through the brief pain inventory, over time, F(2, 16) = 0.38, p = 0.689. Post hoc analyses were not consulted due to the non-significant result.

The second model yielded non-statistically significant changes in pain interference, measured through the brief pain inventory, over time, F(2, 16) = 0.64, p = 0.541. Post hoc analyses were not consulted due to the non-significant result.

The third model yielded statistically significant changes in anxiety, measured through the hospital anxiety and depression scale (HADS), over time, F(2, 16) = 4.87, p = 0.022, partial η² = 0.378, with anxiety scores decreasing from 10.22 ± 5.43 at baseline to 6.22 ± 3.67 at post-intervention and to 7.00 ± 4.30 at one-month follow-up. Post hoc analysis with a Bonferroni adjustment revealed that anxiety scores were not statistically significantly decreased from baseline to post-intervention [4.00 (95% CI, −1.75 to 8.18), p = 0.061], nor from baseline to one-month follow-up [3.22 (95% CI, −1.85 to 7.18), p = 0.26], nor from post-intervention to one-month follow-up [−7.88 (95% CI, −3.47 to 1.91), p = 1.00].

The fourth and final model yielded statistically significant changes in depression, measured through the hospital anxiety and depression scale, over time, F(2, 16) = 7.87, p = 0.004, partial η² = 0.496, with depression scores decreasing from 10.00 ± 3.16 at baseline to 5.56 ± 3.64at post-intervention and to 6.56 ± 4.10 at one-month follow-up. Post hoc analysis with a Bonferroni adjustment revealed that depression scores were statistically significantly decreased from baseline to post-intervention [4.44 (95% CI, 0.61–8.28), p = 0.04], but not from baseline to one-month follow-up [3.44 (95% CI, −0.29 to 7.18), p = 0.71], nor from post-intervention to one-month follow-up [−1.00 (95% CI, −4.01 to 2.02), p = 1.00].

This study adopted a single-arm design which makes it difficult to infer whether any observed improvements in sleep and mood outcomes were due to the intervention, other care provisions the individuals may have received beyond the study (private care, NHS), or simply the passing of time. Secondly, this study aimed to assess feasibility and acceptability as its primary aim, therefore a small target sample of 10 female participants with chronic MSK pain was sufficient, however, looking to the future, it would be beneficial to recruit a larger and perhaps more diverse chronic pain sample to enable sub-group analyses across different demographics, pain conditions, and pain complexity, as these factors have been shown to as independent risk factors for sleep disturbance and may impact the trajectory of a CBT-I treatment on sleep outcomes (Rollman and Lautenbacher, 2001; Smith et al., 2001). In order to validate the findings from this study further, a larger study adopting a randomized control design should be conducted.

Finally, the telehealth delivery may pose a limitation to individual who are digitally excluded from participation, e.g., without access to internet, or adequate internet-connected devices with which to participate in such interventions. However, telehealth may be used to serve individuals who are unable to participate in in-person/on-site treatment due to reasons such as mobility impairments, occupational or caring responsibilities, or personal preference.

Within NHS services, talking therapies for chronic pain and insomnia are typically offered as distinct interventions, necessitating separate referrals. Moreover, when sleep is addressed within a pain management program, evidence indicates that the sleep component is not personalized, consisting primarily of sleep hygiene recommendations, if at all (Cunningham et al., 2011; Burton and Shaw, 2015). Our adapted CBT-I intervention aimed to address sleep problems in the context of living and managing chronic MSK pain with preliminary evidence that it may ameliorate outcomes in sleep and mood which may lead to fewer healthcare interactions and address high-cost barriers to care. Furthermore, the telehealth delivery of the intervention may help to address other barriers known to prevent targeted care provision, such as a lack of trained practitioners in certain geographical areas and increased accessibility for individuals living remotely (Gajarawala and Pelkowski, 2021).

The findings from this feasibility study can provide clinical and policy implications, therefore we have listed some recommendations for consideration: