Advancing equity in routine-care internet-delivered cognitive behaviour therapy: Representativeness, outcomes, engagement, and satisfaction☆☆☆ Authors: Alayna Gretton, Ram P Sapkota, Gladys O Orobosa, Blake F Dear, Nickolai Titov, Heather D Hadjistavropoulos Published: 2026 ABSTRACT AbstractBackgroundInternet-delivered cognitive behaviour therapy (ICBT) has potential to reduce access barriers, yet it remains unclear how well routine-care ICBT reaches underserved groups and whether symptom outcomes, engagement, and satisfaction are comparable across demographic subgroups.MethodsRoutine-care ICBT clinic data (N = 1640) collected in 2021 were compared with provincial census estimates from the 2021 Canadian census (N = 22,705) after harmonizing demographic categories. Longitudinal depressive and anxiety outcomes were analyzed using mixed-effects models with categorical time, testing subgroup main effects and subgroup-by-time interactions. Engagement (lesson completion) and satisfaction were analyzed using logistic regression models.ResultsSeveral groups had lower representation relative to census benchmarks, including men, clients with lower education, Indigenous and some racialized groups, newcomers, rural/small-town residents, whereas women, higher education, and White clients had higher representation. Depression and anxiety symptoms declined significantly over time across all demographic groups. Engagement was high (~82% completed ≥3 of 5 lessons). Older age predicted greater lesson completion. Satisfaction was high overall and with course materials, with relatively lower satisfaction among clients who were not partnered.ConclusionsAlthough some groups were underrepresented, consistent with face-to-face treatment, routine-care ICBT was associated with substantial symptom improvement across all demographic subgroups, alongside high engagement and satisfaction. Areas of lower representation highlight the need for targeted outreach and engagement supports, particularly to men, people with lower educational attainment, Indigenous and some racialized communities, newcomers, and people living outside large urban centres.Keywords: Internet-delivered cognitive behaviour therapy (ICBT), Ethnocultural, Diversity, Equity, Representativeness, Engagement Introduction Internet-delivered cognitive behavioural therapy (ICBT) has demonstrated effectiveness for a wide range of psychological disorders (Hadjistavropoulos et al., 2022; Käll et al., 2024), including depression and anxiety, and produces treatment effects equivalent to in-person CBT (Hedman-Lagerlöf et al., 2023). The past two decades have seen an increase in the development and use of ICBT in clinical research and practice (Andersson et al., 2019; Etzelmueller et al., 2020). Of note, some research suggests that the growth of digital mental health services accelerated during the early stages of the COVID-19 pandemic (Mahoney et al., 2021; Sorkin et al., 2021). In part, enthusiasm for ICBT has been driven by its potential to address barriers to mental health care – barriers that are more commonly experienced by marginalized groups (McGuire and Miranda, 2008). Compared to face-to-face CBT, ICBT has been proposed to lower service costs, reduce wait times, and increase flexibility in terms of geographical and transportation barriers and therapist availability, while facilitating discretion, thereby mitigating stigma-related barriers (Mental Health Commission of Canada, 2017; Webb et al., 2017). Some research evidence supports the conjecture that ICBT may circumvent barriers, and demonstrates effectiveness, acceptability, and engagement among marginalized populations. Research suggests, for example, ICBT may be delivering on its potential to increase access to mental health services for individuals with lower socioeconomic status (SES). In a cost-free, routine-care ICBT study, individuals from low- and middle-SES groups were overrepresented, with lack of affordable treatment alternatives cited as the primary motivation for seeking ICBT (Staples et al., 2025). Despite the financial constraints typically experienced by these groups, participants across SES levels demonstrated similar improvements in depression, anxiety, and distress, with large effect sizes, as well as similarly high levels of treatment satisfaction. Although the evidence is not specific to ICBT, a broader literature on internet-delivered interventions suggests that digital modalities can be used to extend the reach of mental health services to underserved populations. For example, mobile phone-delivered interventions have been successfully implemented in inner-city neighborhoods, rural settings, and among Indigenous peoples and other ethnoculturally diverse groups (García-Vázquez et al., 2016). Despite the above supporting evidence, some routine-care data indicate that use of therapist-assisted ICBT varies across demographic groups. For example, a study examining ICBT utilization over a six-year period (2013–2019) by our group in Saskatchewan, Canada found lower use among ethnic minority clients (approximately 10%) relative to rough population benchmarks (approximately 32%) (Hadjistavropoulos et al., 2022). Men are also consistently underrepresented among ICBT users, with women typically comprising the majority of clients (Hadjistavropoulos et al., 2022; van der Vaart et al., 2019). Other researchers have found patterns of use further vary by education level, age, marital status, and living arrangements (Fournier et al., 2025; van der Vaart et al., 2019). Importantly, these patterns do not imply that proportional representation is an inherent benchmark for service quality. Ideally, the benchmarks for such comparisons would use the demographic distribution of adults with depression and/or anxiety for whom ICBT may be appropriate; however, sufficiently detailed population-level estimates are not available for this group. These patterns underscore the need for empirical evaluation of who is reached in routine care and highlight the importance of examining structural factors and engagement processes that may shape real-world impact. Differences in utilization may reflect variation in need for services, awareness, preferences, and practical access constraints. Although imperfect, benchmarking service use against available population data can provide a descriptive indication of which groups are currently reached or underreached by ICBT. Persisting disparities in mental health service utilization have been attributed to a confluence of structural factors that disproportionately affect marginalized groups. Within the Canadian context in which the current study is situated, it has been argued that there is an “embeddedness of colonialism” within Canadian healthcare systems (Nelson and Wilson, 2017, p. 94). Healthcare delivery—and by extension, online healthcare delivery—may be incongruent with Indigenous patients' cultural values and practices, potentially contributing to inequitable access, engagement, and health outcomes (Davies, 2022; Pendse et al., 2022; Sekalala and Chatikobo, 2024). Despite the surge in information technologies and the availability of digital interventions, there remains a substantial “digital divide,” referring to “disparities in Information and Communications Technology access, usage, and outcomes” (Lythreatis et al., 2022, p. 1; Office of the Auditor General of Canada, 2023). Internet access and use among Canadians is not ubiquitous, and people living in rural, remote, and Indigenous communities experience particularly pronounced gaps (Carpenter-Song, 2020; Office of the Auditor General of Canada, 2023). Although approximately 90% of Canadian households have access to high-speed internet (50/10 Mbps), only 42.9% of households in Indigenous communities and 59.5% of households in rural and remote areas have access (Office of the Auditor General of Canada, 2023). In Saskatchewan, the 2021 provincial breakdown indicates substantial within-province disparities in access, with 99.02% of urban households having access to high-speed internet, compared to 48% of rural households and 10% of households on First Nations reserves (Office of the Auditor General of Canada, 2023). More recently, a federal backgrounder reported that 85.7% of Saskatchewan households currently have access to high-speed internet, with access projected to reach 90.2% by the end of 2026 (Innovation, Science and Economic Development Canada, 2026). Beyond representativeness and symptom outcomes, routine-care research also indicates that engagement and dropout are not uniform across clients and may reflect systematic differences in who benefits most from therapist-assisted ICBT. A routine-care study examining predictors of symptom change and dropout found that older age and lower intake distress were associated with greater likelihood of treatment completion, while higher baseline symptom severity, being off work on disability, and education level were associated with differences in symptom trajectories (Edmonds et al., 2018). This line of work underscores the importance of evaluating equity not only in clinical outcomes, but also in engagement processes (e.g., lesson completion) and patient experience (e.g., satisfaction), as disparities in these domains may constrain real-world impact even when average symptom reductions are strong. The question of generalizability remains largely unanswered in the existing literature. Only 2.1% of randomized clinical trials of interventions for depression report treatment effects for demographic subgroups (Polo et al., 2019). Similarly, a systematic review of 62 ICBT trials for depression found that 27% of studies reported race and 19% reported ethnicity (De Jesús-Romero et al., 2024). The limited studies that have examined outcomes for subgroups—primarily by race/ethnicity or socioeconomic status—have yielded mixed findings. Some studies suggest no significant differences in effectiveness between groups (e.g., Bakhti et al., 2025; Staples et al., 2025), whereas others have found slightly greater improvements among racial minority participants compared to White participants (Jonassaint et al., 2020). Another meta-analysis reported that younger clients and ethnic minority participants may experience fewer benefits from therapist-supported internet-delivered interventions in terms of treatment response and remission (Karyotaki et al., 2018). Clarifying whether diverse groups experience different treatment outcomes is essential for determining the breadth of ICBT's reach and for understanding how ICBT can be optimized to support equitable mental health outcomes across populations. The current study Prior research has generally examined demographic composition, selected demographic characteristics, or individual aspects of the ICBT care pathway. Consequently, limited evidence is available on who is reached by routine-care ICBT and whether clinical outcomes, engagement, and satisfaction are comparable across demographic subgroups once clients access care. The present study addresses this gap by benchmarking a routine-care ICBT sample against population-level data across a broad range of demographic characteristics while concurrently examining symptom outcomes, engagement, and satisfaction across subgroups. Using provincial census data as a benchmark, this study addressed three primary research questions: (1) To what extent does the demographic composition of ICBT clients differ from provincial census benchmarks? (2) Do clinical outcomes for depression and anxiety differ across demographic subgroups? and (3) Do engagement (lesson completion) and satisfaction differ across demographic subgroups? Consistent with evidence of disparities in mental health service utilization (Stephenson and Fournier, 2026), we hypothesized that some marginalized demographic groups would be underrepresented among ICBT users. Based on prior findings suggesting broadly comparable clinical benefits across demographic groups once care is accessed, we hypothesized that symptom outcomes would not differ substantially across subgroups. Given limited prior research, analyses of engagement and satisfaction were exploratory. Methods Design and ethics This study is a retrospective observational study utilizing data previously collected by the University of Regina's Online Therapy Unit (OTU; Regina, Canada). The OTU delivers ICBT to residents of Saskatchewan, Canada, free of cost through funding from the provincial government supporting service delivery (see Hadjistavropoulos and Peynenburg, 2025 for additional details). This study received ethics approval from the research ethics board at the University of Regina (ethics approval notice REB# 2019-197). Informed consent was gathered from participants during intake procedures prior to data collection. Participants and recruitment This study performed secondary analyses on data gathered by the OTU from clients seeking ICBT for depression and/or anxiety between January 4th, 2021, to December 23rd, 2021, to match population level data. Participants were made aware of the OTU's services through community and online advertising, and directed to the OTU's online platform for registration, informed consent and intake procedures (www.onlinetherapyunit.ca) (Fig. 1). For intake, participants completed an initial eligibility survey online via REDCap, which gathered demographic data and administered a battery of standardized questionnaires assessing medical background, mental health history, current symptoms and functioning, as required to administer care. Individuals who met the initial inclusion criteria progressed to a phone interview with a trained staff member to confirm their eligibility. Individuals eligible for participation in this study: 1) were residents of Saskatchewan, Canada, 2) were adults aged 18 or older, 3) reported elevated symptoms of depression and/or anxiety, 4) had reliable access to the internet and comfort navigating online platforms, 5) had the time to commit to treatment, and 6) provided informed consent. Exclusion criteria were: 1) being assessed during the phone interview as high risk of suicide by scoring >2 on item 9 of the Patient Health Questionnaire-9 (PHQ-9; Kroenke et al., 2001) or otherwise through clinical judgement, 2) having an unmanaged primary substance use disorder, 3) having an unmanaged severe psychological disorder (a bipolar mood disorder or psychotic disorder), 4) hospitalization for a mental health concern in the year prior, 5) being unable to provide an emergency medical contact if deemed necessary to coordinate care, 6) seeking help for a condition other than depression and/or anxiety, 7) reporting a physical health condition that would interfere with receiving treatment from the OTU, and 8) already receiving mental health care from another care provider or service (i.e., being currently enrolled in another OTU course, or receiving two or more therapy sessions per month from another service provider). Participant flowchart. Intervention The Wellbeing Course is a transdiagnostic ICBT intervention developed by the eCentreClinic (www.ecentreclinic.org,) at Macquarie University (Sydney, Australia) and tailored by the OTU to be delivered in a Canadian context (see Titov et al., 2015 for a description). The Wellbeing Course delivers standardized CBT content in an online modularized format. The Course is comprised of five core modules, referred to as “lessons”, released sequentially across an 8-week intervention period. Each lesson focuses on core CBT principles and skills including psychoeducation, thought challenging, activity scheduling, controlled breathing, graded exposure and relapse prevention. The intervention incorporates case stories from fictional characters to enhance the relatability of the course material and demonstrate skill application. The Course also offers downloadable homework worksheets, and supplemental resources focused on a variety of individual clinical concerns, such as additional readings on worry, beliefs, sleep, motivation and anger. Booster lessons were made available to clients 12 weeks after the start of the intervention (see Hadjistavropoulos et al., 2022). Therapist support The Course was offered with therapist support, via secure emails and/or phone calls. Participants were offered the choice of receiving “standard support”, which involved receiving weekly supportive messages and reflections from therapists over email, or “optional support” where participants reached out to therapists for support only as needed. Regardless of the level of support received, therapists were prompted to contact participants by phone if a rapid deterioration of clinical symptoms was observed (as assessed by weekly standardized symptom measures), or if participants were inactive on the online platform for a full week. Measures Demographic questionnaire Participants completed a survey querying psychosocial characteristics, medical, and mental health background, and current symptoms and functioning. Specifically, the demographic questionnaire included questions on age, gender, educational attainment, ethnicity, newcomer/immigrant status, previous mental health diagnoses, current medical provider, relationship/marital status, medication use, mental health service use, whether participants resided in an urban versus rural location, along with a variety of standardized measures assessing current mental health symptoms. The demographic questionnaire was administered once, during the intake process. Statistics Canada 2021 census of population data This study performed a secondary analysis of data from Statistics Canada's 2021 Census of Population (Statistics Canada, 2022; Statistics Canada, 2023a; Statistics Canada, 2024a). Saskatchewan data were filtered from the greater Canada-wide dataset to determine demographic characteristics of Saskatchewan residents, including age, gender, highest level of educational attainment, ethnicity, newcomer/immigrant status, urban versus rural locality, relationship status, employment status and income level. Demographic data reported from adults aged 18 to 85+ (as per Statistics Canada's reported age categories) were drawn from the 2021 Census Public Use Microdata File (PUMF) on Individuals (Statistics Canada, 2024a). This study also utilized data from Statistics Canada's Population and Dwelling Counts: Canada, Provinces and Territories, and Census Subdivisions (Municipalities; Statistics Canada, 2022), and the Census Program Data Viewer Data Visualization tool (Statistics Canada, 2024b); however, specific age categories were not reported in these resources, and these data reflected characteristics of Canadians of all ages. Depressive symptoms: Patient Health Questionnaire-9 (PHQ-9; Kroenke et al., 2001) The PHQ-9 contains 9 items assessing the frequency of depressive symptoms experienced in the past 2 weeks on a scale of 0 (not at all) to 3 (nearly every day). Total scores range from 0 to 27 (Kroenke et al., 2001), with scores of ≥10 indicating the probable presence of major depressive disorder (Manea et al., 2012). Reliable improvement was defined as a decrease of ≥6 points on the PHQ-9 (Gyani et al., 2013). Internal consistency in this study ranged from Cronbach's α = 0.84–0.87. Anxious symptoms: Generalized Anxiety Disorder 7-item (GAD-7; Spitzer et al., 2006) The GAD-7 is comprised of 7 items measuring the frequency of anxiety symptoms within the past 2 weeks. Respondents rate items on a scale of 0 (not at all) to 3 (nearly every day), for a total possible score range of 0–21, with scores ≥10 suggesting a probable diagnosis of generalized anxiety disorder. Reliable improvement was defined as a decrease of ≥6 points on the PHQ-9 or ≥ 4 points on the GAD-7 (Gyani et al., 2013). Internal consistency in this study ranged, Cronbach's α = 0.87–0.91. Engagement and satisfaction Treatment engagement was assessed using lesson completion indicators derived from the ICBT platform. Three engagement variables were used: (a) lessons completed (capped at five core lessons; values above five, reflecting booster completion, were recoded to five), (b) majority lesson completion (≥3 lessons vs. 0–2 lessons), and (c) completion of all core lessons (≥5 lessons vs. 0–4 lessons). Treatment satisfaction was assessed at posttreatment using a brief satisfaction questionnaire (Dear et al., 2011). Participants rated overall satisfaction with the program and satisfaction with the quality of the lessons and “Do It Yourself” guides on 5-point Likert scales (1 = very dissatisfied to 5 = very satisfied). For analysis, these were also summarized using endorsement ratings of satisfied/very satisfied. Participants also completed two yes/no items assessing whether they would recommend the course to a friend and whether the course was worth their time. Finally, participants rated perceived impact of the course on (a) confidence in managing symptoms and (b) motivation to seek additional treatment in the future if needed, each on a 5-point scale (1 = greatly reduced to 5 = greatly increased). Statistical analysis To examine the proportionate representation of demographic groups in ICBT, separate datasets from Statistics Canada and the OTU were compared on select demographic variables. To enable head-to-head comparisons between the OTU and census datasets, demographic variables were harmonized through recoding such that their categories/levels were operationally equivalent between the two datasets. Within each dataset, prevalence rates were calculated for several nominal demographic variables, including gender, educational attainment, ethnicity, newcomer status, relationship status, living alone, location type, employment status, and income level. For each demographic comparison, percentages and RQs were calculated using respondents with valid, classifiable data in the harmonized categories; missing, “not available,” or otherwise non-comparable responses were excluded from the comparison-specific denominator and reported descriptively where applicable. Means and Standard Deviations were calculated for a single continuous variable, age. The N-1 Chi-squared test (Campbell, 2007) was used to test for statistically significant differences between population-level and OTU demographic proportions using the MedCalc Comparison of Proportions Calculator Version 23.4.4 (MedCalc Software Ltd., 2025). The magnitude of the difference between proportions was quantified using the representation quotient (RQ), which signifies the representativeness of a study sample relative to a reference population, and has been used to quantify the proportionate representation of demographic groups within psychological interventions (e.g., Wrede et al., 2025). The RQ is computed by dividing the proportion of a subgroup within an observed sample by the proportion of the subgroup within a greater reference population. In this study, ICBT clients were treated as the observed sample and the census data served as the reference population. The RQ produces a statistic where 0.00 indicates no representation, 1.00 indicates proportionate representation, and values >1.00 indicate over-representation. For example, an RQ of 0.50 indicates representation at approximately half the population benchmark, whereas an RQ of 1.50 indicates representation approximately 50% above the population benchmark. Following Wrede et al. (2025), 95% confidence intervals for RQs were obtained by dividing the lower and upper bounds of the 95% confidence interval around the observed ICBT subgroup proportion by the corresponding census reference proportion. Longitudinal change in depressive and anxiety symptom severity was evaluated using linear mixed-effects models. Separate model series were estimated for PHQ-9 and GAD-7 symptom scores across 12 scheduled assessment occasions (baseline; weeks 1–8; week 12; week 16; week 20). Time was treated as a categorical factor to allow non-linear patterns of change and to obtain model-based means at each assessment occasion without imposing a functional form. All models included a random intercept for participant to account for between-person differences in overall symptom severity. To model within-person residual correlation, candidate covariance structures were compared, including first-order autoregressive (AR[1]), heterogeneous first-order autoregressive (ARH[1]), and spatial power. Model fit was evaluated using information criteria (AIC and BIC), and the best-fitting structure was selected for the primary analyses. Based on these comparisons, an ARH(1) structure was retained, allowing residual variability to differ across assessment occasions and modeling stronger correlations between observations closer in time and weaker correlations between observations farther apart. Models were estimated using maximum likelihood, and fixed effects were evaluated using Type III tests with denominator degrees of freedom as provided by SPSS (Version 28.0). Statistical tests were two-sided with α = 0.05. Pretreatment demographic and clinical variables were reviewed for completeness and suitability and were recoded into analysis subgroups. Gender was operationalized as woman versus man; other gender responses were excluded from gender-moderator models due to small cell sizes. Ethnicity was grouped as White, Indigenous (First Nations, Inuit, Métis), and other racialized (non-White, non-Indigenous), with “prefer not to answer” treated as missing. Education was collapsed into four categories (high school or less; some postsecondary/below bachelor's; bachelor's; above bachelor's). Relationship status was grouped as not partnered (single/dating), partnered (married/common-law), and previously partnered (divorced/separated/widowed). Location was categorized as rural/small town (1 to fewer than 7000 citizens), large town/medium city (7000 to 100,000), and large city (>100,000); responses not classifiable within these categories were treated as missing for location analyses. Employment was grouped as employed, not in the labor force (retired/student), not working due to health, and not working for other reasons; a binary employed versus not employed indicator was also used descriptively and in sensitivity analyses. Household income was collapsed into low, middle, and high brackets; “not available” was treated as missing, and given substantial missingness, education was prioritized as the primary socioeconomic covariate while income was examined secondarily. Living arrangement was modeled as living with someone (yes/no). To examine demographic differences in symptom levels and change over time, we estimated covariate-by-covariate model series for each outcome. All models included time, ethnicity, educational attainment, and mean-centered age. When ethnicity, education, or age was the focal characteristic, it was modeled as the focal predictor rather than solely as an adjustment variable. Models examining gender, relationship status, location, employment status, household income, or living arrangement additionally included the focal characteristic. For each focal characteristic, we first fit a main-effects model (time and the focal characteristic) and then added the characteristic × time interaction to test whether group differences varied across the treatment and follow-up period. When the interaction was not statistically significant, group differences were interpreted as generally consistent across the assessment period; when statistically significant, interpretation focused on differences at specific assessment points. For reporting, estimated marginal means with 95% confidence intervals were extracted at baseline, week 8 (post-treatment), and week 20 (follow-up) for each group. We summarized absolute change from baseline to week 8 and baseline to week 20 (Δ in points) and calculated percent change relative to baseline estimated marginal means. Where group contrasts were examined, Bonferroni-adjusted pairwise comparisons were used. As a threshold-based complement to the longitudinal symptom models, reliable improvement and reliable deterioration were examined from baseline to posttreatment (Week 8) and follow-up (Week 20). Reliable improvement was defined as a decrease of ≥6 points on the PHQ-9 or ≥ 4 points on the GAD-7 (Gyani et al., 2013) and was calculated among participants with clinically elevated symptoms at baseline (score ≥ 10 on the respective measure) who had paired baseline and relevant follow-up scores. This restriction was applied because participants with low baseline scores may not be able to achieve a decrease large enough to meet the reliable-change threshold. Reliable deterioration was defined as an increase of ≥6 points on the PHQ-9 or ≥ 4 points on the GAD-7 and was calculated among all participants with paired baseline and relevant follow-up scores, regardless of baseline symptom level. Reliable deterioration was defined as an increase of ≥6 points on the PHQ-9 or ≥ 4 points on the GAD-7 and was calculated among all participants with paired baseline and relevant follow-up scores, regardless of baseline symptom level. Reliable-change proportions are reported with 95% Wilson score confidence intervals (Brown et al., 2001). Subgroup estimates were summarized descriptively using the same categorical sociodemographic groupings used in the longitudinal analyses; no inferential comparisons of subgroup reliable-change rates were conducted. Engagement was operationalized using two binary indicators derived from the number of lessons completed: (a) majority lesson completion (≥3 lessons vs. 0–2 lessons) and (b) completion of all core lessons (≥5 lessons vs. 0–4 lessons). These outcomes were analyzed using generalized linear models with a binomial distribution and logit link (GENLIN). All models were estimated using maximum likelihood with Type III Wald tests. Each model simultaneously included mean-centered age, gender, education, ethnicity, relationship status, location, employment status, household income, and living with someone. Cases with missing data on any model variable were excluded listwise. Posttreatment satisfaction and perceived benefit were assessed using six items administered at posttreatment. Two satisfaction outcomes—overall satisfaction and satisfaction with materials —were analyzed using logistic regression with the same covariate set described above. The remaining satisfaction and perceived benefit outcomes (recommendation to a friend, perceived value, increased confidence to self-manage symptoms, and increased motivation to seek future treatment) showed substantial positive skew with small cell counts in the non-endorsement categories. To avoid unstable estimation and separation, these outcomes were summarized descriptively only. Results are reported as Wald χ2 tests and odds ratios (ORs) with 95% confidence intervals where applicable. Results Analyses were performed on data from N = 22,705 Saskatchewan residents drawn from Statistics Canada's 2021 Census Public Use Microdata File (Statistics Canada, 2024a) and N = 1640 clients who had completed screening for ICBT. Participants in ICBT were significantly younger (M = 37.41, SD = 13.86, range = 18–87) than adults aged 18 and older in the general Saskatchewan population (M = 40.85, SD = 22.89, range = 18–85+); t(24343) = 6.0069, p < .0001). Results for the representativeness of demographic subgroups in ICBT are presented below and summarized in Table 1, Table 2, Table 3, Table 4. Table 1Demographic comparisons: Saskatchewan 2021 Census population benchmarks versus the ICBT routine care sample.Patient characteristicsSaskatchewan (2021 Census Data)Wellbeing CourseRepresentation Quotient (RQ)95% CI RQp values (χ2)AgeMean (SD)40.85 (22.89)37.41 (13.86)––p < .001Range18–85+18–87––n%n%GenderFemale11,48750.59124677.491.531.49, 1.57p < .001Male11,21849.4136222.510.460.42, 0.50p < .001Transgender/non-binary⁎Not reported in census data–221.34⁎–––Educational attainmentLess Than High School325014.44603.660.250.20, 0.32p < .001High School Diploma/752733.4431619.270.580.52, 0.64p < .001Trade Certificate/ University Diploma697030.9671743.721.411.34, 1.49p < .001Bachelor's Degree370816.4740524.71.51.38, 1.63p < .001University Professional Degree (e.g. MD)1100.49613.727.615.95, 9.72p < .001Graduate degree (e.g., MSc, PhD)9464.2814.941.180.95, 1.45p = .153EthnicityWhite/European15,75073.76138985.951.171.14, 1.19p <.001Southeast/East Asian11855.55332.040.370.26, 0.51p <.001Black2721.27150.930.730.44, 1.20p = .228First Nations/Inuit/Metis307914.421227.550.520.44, 0.62p <.001Latin American700.33110.682.081.16, 3.71p = .021West Asian/Middle Eastern960.4590.561.240.65, 2.35p =.537South Asian7153.35150.930.280.17, 0.46p <.001Other/multiple population groups1850.87221.361.571.04, 2.37p =.042Prefer not to answer/not available13535.96241.46–––Newcomer/immigrant (arrived 2016 or later)Yes8323.6690.550.150.08, 0.28p < .001Relationship and marital statusSingle, dating or never married565325.0355533.841.351.26, 1.44p < .001Married, common law or living with partner13,98561.9393557.010.920.88, 0.96p < .001Separated4822.13593.61.691.31, 2.16p = .001Divorced12915.72704.270.750.59, 0.94p = .014Widowed11725.19211.280.250.16, 0.38p < .001Living aloneYes347115.2933420.371.331.20, 1.46p < .001Note. Analyses were performed on data from the 2021 Census Public Use Microdata File (PUMF) (Statistics Canada, 2024a) and data gathered by the OTU during ICBT routine care. For gender comparisons, OTU percentages and representation quotients were calculated using participants in Census-comparable female and male categories (n = 1608); transgender/non-binary participants (n = 22) were reported descriptively but excluded from Census comparisons because comparable Census data were unavailable. Participants with missing gender data (n = 10) were excluded from gender-specific comparisons. For ethnicity comparisons, percentages and representation quotients were calculated using respondents with valid, classifiable ethnicity data (Census n = 21,352; OTU n = 1616); “Prefer not to answer/not available” responses were excluded from the comparison denominators and reported descriptively only.⁎Comparable Census data were not available; therefore, RQs, 95% CIs, and p values were not calculated. Demographic comparisons: Saskatchewan 2021 Census population benchmarks versus the ICBT routine care sample. Note. Analyses were performed on data from the 2021 Census Public Use Microdata File (PUMF) (Statistics Canada, 2024a) and data gathered by the OTU during ICBT routine care. For gender comparisons, OTU percentages and representation quotients were calculated using participants in Census-comparable female and male categories (n = 1608); transgender/non-binary participants (n = 22) were reported descriptively but excluded from Census comparisons because comparable Census data were unavailable. Participants with missing gender data (n = 10) were excluded from gender-specific comparisons. For ethnicity comparisons, percentages and representation quotients were calculated using respondents with valid, classifiable ethnicity data (Census n = 21,352; OTU n = 1616); “Prefer not to answer/not available” responses were excluded from the comparison denominators and reported descriptively only. Comparable Census data were not available; therefore, RQs, 95% CIs, and p values were not calculated. Table 2Geographic comparisons: Saskatchewan 2021 Census population benchmarks versus the ICBT routine care sample.Saskatchewan (2021 Census Data)Wellbeing CourseRepresentation Quotient (RQ)95% CI RQp valuesLocation typen%n%Small Town (<7000),456,08340.344427.190.680.62, 0.73p <.001Large Town to Medium City (7000–100,000)183,87716.223014.080.870.77, 0.98p = .02Large City (>100,000)492,54543.595958.731.351.29, 1.40p <.001Note. Analyses were performed on data from Population and dwelling counts: Canada, provinces and territories, and census subdivisions (municipalities) (Statistics Canada, 2022) and data gathered by the OTU during ICBT routine care. For location comparisons, OTU percentages and representation quotients were calculated using participants with classifiable location data (n = 1633); responses categorized as “Other” (n = 7) are reported descriptively but were excluded from Census comparisons. Geographic comparisons: Saskatchewan 2021 Census population benchmarks versus the ICBT routine care sample. Note. Analyses were performed on data from Population and dwelling counts: Canada, provinces and territories, and census subdivisions (municipalities) (Statistics Canada, 2022) and data gathered by the OTU during ICBT routine care. For location comparisons, OTU percentages and representation quotients were calculated using participants with classifiable location data (n = 1633); responses categorized as “Other” (n = 7) are reported descriptively but were excluded from Census comparisons. Table 3Income status comparisons: Saskatchewan 2021 Census population benchmarks versus the ICBT routine care sample.IncomeSaskatchewan (2021 Census Data)Wellbeing CourseRepresentation Quotient (RQ)95% CI RQp values (χ2)n%n%Less than $20,000164,28519.4422819.761.020.90, 1.14p = .788$20,000 to $39,999232,37027.5028324.520.890.80, 0.98p = .024$40,000 to $59,999177,04520.9525422.011.050.94, 1.17p = .378$60,000 to $79,999112,50013.3117314.991.130.98, 1.29p = .094$80,000 to $99,99969,5108.2313511.71.421.21, 1.66p < .001$100,000 and above89,22010.56817.020.660.54, 0.82p < .001Note. Analyses were performed on data from the Census Program Data Viewer (Statistics Canada, 2024b) and data gathered by the OTU during ICBT routine care. For income comparisons, OTU percentages and representation quotients were calculated using participants with valid income data (n = 1154); missing or “Not Available” income responses were excluded from the income-specific denominator. Income status comparisons: Saskatchewan 2021 Census population benchmarks versus the ICBT routine care sample. Note. Analyses were performed on data from the Census Program Data Viewer (Statistics Canada, 2024b) and data gathered by the OTU during ICBT routine care. For income comparisons, OTU percentages and representation quotients were calculated using participants with valid income data (n = 1154); missing or “Not Available” income responses were excluded from the income-specific denominator. Table 4Labour force status comparisons: Saskatchewan 2021 Census population benchmarks versus the ICBT routine care sample.Labour force statusSaskatchewan (2021 Census Data)Wellbeing CourseRepresentation Quotient (RQ)95% CI RQp values (χ2)n%n%Employed in paid work14,28360.1584151.280.850.81, 0.89p < .001Not employed in paid work946139.8579948.721.221.16, 1.28p < .001Not available602020.23⁎–––––Note. Analyses were performed on data from the 2021 Census Public Use Microdata File (PUMF) (Statistics Canada, 2024a) and data gathered by the OTU during ICBT routine care. For labour force status comparisons, Census percentages and representation quotients were calculated using respondents with valid employment data (n = 23,744); “Not Available” responses (n = 6020) were excluded from the comparison denominator.⁎The percentage for “Not Available” was calculated using the full Census sample (N = 29,764) and is reported descriptively only; no RQ, 95% CI, or p value was calculated. Labour force status comparisons: Saskatchewan 2021 Census population benchmarks versus the ICBT routine care sample. Note. Analyses were performed on data from the 2021 Census Public Use Microdata File (PUMF) (Statistics Canada, 2024a) and data gathered by the OTU during ICBT routine care. For labour force status comparisons, Census percentages and representation quotients were calculated using respondents with valid employment data (n = 23,744); “Not Available” responses (n = 6020) were excluded from the comparison denominator. The percentage for “Not Available” was calculated using the full Census sample (N = 29,764) and is reported descriptively only; no RQ, 95% CI, or p value was calculated. Gender Compared to the general Saskatchewan population, females had higher representation in ICBT (RQ = 1.53, 95% CI [1.49, 1.57]), while males had lower representation (RQ = 0.46, 95% CI [0.42, 0.50]). Comparisons for gender-minority groups could not be made because directly comparable categories were not available in the census dataset used for these analyses. See Table 1 for additional details. Highest educational attainment Individuals with postsecondary education generally had higher representation in ICBT relative to the Saskatchewan population benchmark. Specifically, higher representation was observed among those with a trade certificate, college diploma, or university education below a bachelor's degree (RQ = 1.41, 95% CI [1.34, 1.49]), those holding a bachelor's degree (RQ = 1.50, 95% CI [1.38, 1.63]), and those holding a university professional degree (RQ = 7.61, 95% CI [5.95, 9.72]). The exception was for individuals with a graduate degree (e.g., MSc, PhD), who were proportionately represented (RQ = 1.18, 95% CI [0.95, 1.45]). Individuals who had not obtained a high school diploma had lower representation (RQ = 0.25, 95% CI [0.20, 0.32]), as did those with a high school diploma as their highest level of educational attainment (RQ = 0.58, 95% CI [0.52, 0.64]) (see Table 1). Ethnicity Ethnocultural representation varied across groups. White/European clients had higher representation in ICBT relative to the Saskatchewan population benchmark (RQ = 1.17, 95% CI [1.14, 1.19]), as did Latin American clients (RQ = 2.08, 95% CI [1.16, 3.71]) and those identifying with multiple population groups or a group not listed (RQ = 1.57, 95% CI [1.04, 2.37]). For Black clients, the RQ was 0.73 (95% CI [0.44, 1.20]), and for West Asian/Middle Eastern clients, the RQ was 1.24 (95% CI [0.65, 2.35]). Lower representation was observed among Southeast/East Asian clients (RQ 0.37, 95% CI [0.26, 0.51]), Indigenous clients (RQ = 0.52, 95% CI [0.44, 0.62]), and South Asian clients (RQ = 0.28, 95% CI [0.17, 0.46]). Newcomers also had lower representation (RQ = 0.15, 95% CI [0.08, 0.28]) (see Table 1). Relationship status and living alone Relationship status and living arrangement also varied relative to population benchmarks. Individuals who were single or dating had higher representation in ICBT (RQ = 1.35, 95% CI [1.26, 1.44]), as did those who were separated (RQ = 1.69, 95% CI [1.31, 2.16]) and those living alone (RQ = 1.33, 95% CI [1.20, 1.46]). Lower representation was observed among those who were married or common law (RQ = 0.92, 95% CI [0.88, 0.96]), divorced (RQ = 0.75, 95% CI [0.59, 0.94]), or widowed (RQ = 0.25, 95% CI [0.16, 0.38]) (see Table 1). Urban and rural location type Geographic representation also varied relative to population benchmarks. Individuals living in large cities had higher representation in ICBT (RQ = 1.35, 95% CI [1.29, 1.40]). Lower representation was observed among those living in small towns, villages, or rural areas (RQ = 0.68, 95% CI [0.62, 0.73]) and among those living in large towns to medium cities (RQ = 0.87, 95% CI [0.77, 0.98]) (see Table 2). Income and employment Among participants with valid income data (n = 1154), representation varied across income categories relative to population benchmarks. Representation was comparable to the population benchmark among individuals earning less than $20,000 (RQ = 1.02, 95% CI [0.90, 1.14], p = .788), $40,000–$59,999 (RQ = 1.05, 95% CI [0.94, 1.17], p = .378), and $60,000–$79,999 (RQ = 1.13, 95% CI [0.98, 1.29], p = .094). Individuals earning $20,000–$39,999 had lower representation (RQ = 0.89, 95% CI [0.80, 0.98], p = .024), whereas those earning $80,000–$99,999 had higher representation (RQ = 1.42, 95% CI [1.21, 1.66], p < .001). Individuals earning $100,000 or more also had lower representation (RQ = 0.66, 95% CI [0.54, 0.82], p < .001). Using Census respondents with valid labour force status data (n = 23,744), individuals employed in paid work had lower representation in the ICBT sample (RQ = 0.85, 95% CI [0.81, 0.89], p < .001), whereas individuals not employed in paid work had higher representation (RQ = 1.22, 95% CI [1.16, 1.28], p < .001). See Table 3, Table 4 for additional details. See Table 3, Table 4 for additional details. Clinical outcomes PHQ-9 subgroups Depressive symptoms declined over time across all models, Fs(11, 1275–1535) = 54.25–202.14, ps < 0.001 (Table 5). Older age was associated with lower PHQ-9 scores, F(1, 1295.28) = 25.75, p < .001, and the age-by-time term was not significant (p = .099). Ethnicity was associated with PHQ-9 scores, F(2, 1367.54) = 3.77, p = .023; Indigenous clients reported higher scores than White clients (mean difference = 1.53, p = .019), whereas other racialized clients did not differ from White clients (p > .05), and the ethnicity-by-time term was not significant (p = .275). Educational attainment showed a small main effect, F(3, 1313.55) = 3.86, p = .009, and an education-by-time interaction, F(33, 1469.55) = 1.55, p = .025. Baseline differences favored higher education (ps = 0.040 to <0.001), and these differences persisted at posttreatment for above-bachelor's education versus high school or less and below bachelor's/some postsecondary education (ps = 0.039 and 0.006), but no education differences remained at Week 20 after correction (ps = 1.000). Relationship status was associated with PHQ-9 levels, F(2, 1321.24) = 14.42, p < .001, with no interaction with time (p = .294); the previously partnered group was higher than the partnered group at baseline (mean difference = 2.79, p < .001), and the partnered group remained lower than both not partnered and previously partnered at Week 8 (mean differences = 1.22 and 2.41, ps = 0.003 and < 0.001) and Week 20 (mean differences = 1.38 and 2.54, ps = 0.004 and < 0.001). Location was associated with overall PHQ-9 scores and change over time, with a location main effect F(2, 1330.37) = 5.35, p = .005, and location-by-time F(22, 1343.11) = 1.77, p = .015. Large-city clients reported higher PHQ-9 scores than rural clients at Week 8 (mean difference = 1.17, p = .004) and Week 20 (mean difference = 1.15, p = .019), with no baseline differences after correction (ps > 0.05). Gender showed no main effect (p = .787); the gender-by-time term was at the 0.05 threshold (p = .050), with a Week 20 difference favoring men (mean difference = −0.95, p = .036) but no differences at baseline or Week 8 (ps > 0.05). Employment status showed a level effect, F(3, 1321.86) = 17.74, p < .001, with no interaction with time (p = .195); clients not working due to health had higher PHQ-9 scores than employed and not-in-labor-force clients at baseline (ps < 0.001), Week 8 (ps ≤ 0.004), and Week 20 (ps = 0.037 and 0.007). Household income showed a significant effect, F(2, 940.98) = 9.45, p < .001, with no interaction with time (p = .331); low income exceeded middle income (mean difference = 1.391, SE = 0.349, p < .001, 95% CI [0.553, 2.229]) and high income (mean difference = 1.789, SE = 0.446, p < .001, 95% CI [0.720, 2.858]), whereas middle and high income did not differ (p = 1.000). Living with someone was associated with PHQ-9 scores, F(1, 1319.88) = 12.79, p < .001, with no interaction with time (p = .957). In summary, depression decreased over time across groups. Some between group differences persisted across time, reflecting a higher burden of depressive symptoms for certain subgroups. Most subgroup-by-time interactions were not statistically significant, indicating no clear evidence that symptom change over time differed across most demographic characteristics. See Table 5 for additional information. Table 5Model-adjusted PHQ-9 estimated marginal means at pretreatment, posttreatment (week 8), and 3-month follow-up (week 20) by sociodemographic subgroup.PHQ subgroupPre mean (95% CI)W8 post mean (95% CI)W20 (≈3 mo) mean (95% CI)Δ Pre→W8 (points)Δ Pre→W20 (points)%Δ Pre→W8%Δ Pre→W20Gender Man13.51 (12.68, 14.35)7.66 (6.86, 8.47)5.91 (5.01, 6.81)−5.85−7.6−43.3−56.2 Woman13.77 (13.19, 14.36)7.45 (6.87, 8.02)6.86 (6.25, 7.47)−6.32−6.91−45.9−50.2Age (−1 SD / mean / +1 SD) Younger (−1 SD)14.27 (13.64, 14.90)8.14 (7.51, 8.77)7.50 (6.82, 8.18)−6.13−6.77−43−47.4 Mean age (0 SD)13.75 (13.19, 14.30)7.55 (7.00, 8.10)6.74 (6.17, 7.32)−6.2−7.01−45.1−51 Older (+1 SD)13.23 (12.58, 13.88)6.96 (6.33, 7.59)5.99 (5.32, 6.66)−6.27−7.24−47.4−54.7Ethnicity White13.11 (12.74, 13.49)6.89 (6.53, 7.25)6.21 (5.80, 6.61)−6.22−6.9−47.4−52.6 Indigenous14.55 (13.21, 15.89)8.85 (7.53, 10.17)6.15 (4.57, 7.72)−5.7−8.4−39.2−57.7 Other racialized13.93 (12.58, 15.28)7.80 (6.47, 9.11)6.11 (4.53, 7.70)−6.13−7.82−44−56.1Education High school or less14.29 (13.44, 15.14)8.02 (7.19, 8.85)6.91 (5.98, 7.85)−6.27−7.38−43.9−51.6 Below bachelor / some postsecondary14.66 (14.01, 15.32)8.15 (7.51, 8.79)7.13 (6.43, 7.82)−6.51−7.53−44.4−51.4 Bachelor's degree12.95 (12.18, 13.73)7.23 (6.48, 7.97)6.38 (5.57, 7.20)−5.72−6.57−44.2−50.7 Above bachelor's degree12.07 (10.95, 13.18)6.34 (5.30, 7.39)6.87 (5.66, 8.08)−5.73−5.2−47.5−43.1Employment Employed (paid work)13.32 (12.68, 13.96)7.17 (6.54, 7.79)6.45 (5.78, 7.13)−6.15−6.87−46.2−51.5 Not in labor force (household/retired/student)12.66 (11.92, 13.40)6.81 (6.10, 7.52)6.02 (5.23, 6.82)−5.85−6.64−46.2−52.4 Not working due to health16.60 (15.52, 17.68)10.09 (9.04, 11.13)8.12 (6.97, 9.26)−6.51−8.48−39.2−51.1 Not working (unemployed/other reasons)14.85 (13.64, 16.05)7.55 (6.37, 8.72)7.40 (6.05, 8.76)−7.3−7.45−49.2−50.2Household income Low (<$40,000)14.80 (14.03, 15.56)8.57 (7.81, 9.34)7.57 (6.78, 8.35)−6.23−7.23−42.1−48.9 Middle ($40,000–$79,999)13.41 (12.66, 14.15)7.18 (6.45, 7.92)6.18 (5.41, 6.94)−6.23−7.23−46.4−53.9 High (≥$80,000)13.01 (12.10, 13.92)6.78 (5.88, 7.69)5.78 (4.85, 6.71)−6.23−7.23−47.8−55.6Relationship status Not partnered (single/dating)14.05 (13.29, 14.81)8.14 (7.39, 8.90)7.41 (6.59, 8.24)−5.91−6.64−42.1−47.3 Partnered (married/common law/cohabiting)13.17 (12.56, 13.78)6.92 (6.33, 7.52)6.03 (5.40, 6.67)−6.25−7.14−47.5−54.2 Previously partnered (separated/divorced/widowed)15.96 (14.81, 17.11)9.33 (8.22, 10.44)8.57 (7.33, 9.81)−6.63−7.39−41.5−46.3Location Rural13.25 (12.46, 14.03)6.60 (5.84, 7.37)5.90 (5.07, 6.74)−6.65−7.35−50.2−55.5 Large town to medium city14.06 (13.10, 15.02)7.72 (6.77, 8.66)6.11 (5.02, 7.21)−6.34−7.95−45.1−56.5 Large city13.70 (13.09, 14.31)7.77 (7.18, 8.37)7.05 (6.41, 7.69)−5.93−6.65−43.3−48.5Living with someone No14.66 (13.80, 15.52)8.49 (7.66, 9.33)7.59 (6.68, 8.51)−6.17−7.07−42.1−48.2 Yes13.46 (12.88, 14.04)7.26 (6.70, 7.83)6.45 (5.84, 7.05)−6.2−7.01−46−52.1Note. Values are estimated marginal means (EMMEANS) from linear mixed-effects models with random intercepts and a heterogeneous first-order autoregressive residual covariance structure (ARH[1]). Pre = baseline; W8 = post-treatment; W20 = approximately 3-month follow-up. Each subgroup block is based on the corresponding subgroup model (see Statistical Analysis). Models were adjusted for mean-centered age, ethnicity, and educational attainment, except when one of these variables was the focal characteristic, in which case it was entered as the focal predictor rather than solely as an adjustment covariate; therefore, estimates should not be interpreted as originating from a single combined model across all subgroup blocks. Model-adjusted PHQ-9 estimated marginal means at pretreatment, posttreatment (week 8), and 3-month follow-up (week 20) by sociodemographic subgroup. Note. Values are estimated marginal means (EMMEANS) from linear mixed-effects models with random intercepts and a heterogeneous first-order autoregressive residual covariance structure (ARH[1]). Pre = baseline; W8 = post-treatment; W20 = approximately 3-month follow-up. Each subgroup block is based on the corresponding subgroup model (see Statistical Analysis). Models were adjusted for mean-centered age, ethnicity, and educational attainment, except when one of these variables was the focal characteristic, in which case it was entered as the focal predictor rather than solely as an adjustment covariate; therefore, estimates should not be interpreted as originating from a single combined model across all subgroup blocks. GAD-7 subgroups Anxiety symptoms declined over time across all models, Fs(11, 1256–1880) = 51.98–211.30, ps < 0.001, and were lower at older ages (e.g., F(1, 1247.97) = 39.17, p < .001; Table 6). Ethnicity was associated with GAD-7 levels, F(2, 1354.01) = 5.68, p = .004, and the ethnicity-by-time term was significant, F(22, 1772.97) = 1.61, p = .037. Bonferroni-adjusted contrasts indicated higher scores for Indigenous relative to White clients at pretreatment (mean difference = 1.560, SE = 0.641, p = .045, 95% CI [0.023, 3.096]) and at several treatment timepoints (Weeks 3–8: mean differences = 1.652–2.738, ps = 0.028 to <0.001), with no ethnic-group differences evident at Week 20 after correction (ps = 1.000). Indigenous clients also reported higher scores than other racialized clients at Week 3 (mean difference = 3.031, SE = 0.894, p = .002, 95% CI [0.889, 5.173]) and Week 4 (mean difference = 2.161, SE = 0.872, p = .040, 95% CI [0.071, 4.251]). Education was not associated with overall GAD-7 level (ps = 0.249–0.369), although the education-by-time interaction reached significance, F(33, 1939.89) = 1.45, p = .049; Bonferroni-adjusted education-group contrasts were not significant (ps = 1.000). Relationship status was associated with GAD-7 levels, F(2, 1264.32) = 5.48, p = .004, with no interaction with time (p = .292). The previously partnered group reported higher scores than the partnered group (mean difference = 1.342, SE = 0.406, p = .003, 95% CI [0.368, 2.315]) and not partnered group (mean difference = 1.284, SE = 0.462, p = .017, 95% CI [0.177, 2.391]), whereas partnered and not partnered groups did not differ (p = 1.000). Location was not associated with GAD-7 (ps = 0.173–0.206) and did not interact with time (p = .225). In the gender model, neither the main effect nor the gender-by-time interaction was significant (ps = 0.617 and 0.399). Employment status was associated with GAD-7 scores, F(3, 1267.43) = 6.34, p < .001, with no interaction with time (p = .473). Clients not working due to health reported higher scores than employed and not-in-labor-force clients (mean differences = 1.635 and 1.729, ps < 0.001), while the contrast versus unemployed/other was not significant after adjustment (p = .072). Household income showed a small main effect, F(2, 893.30) = 4.07, p = .017, with no interaction with time (p = .228); low income exceeded middle income on the GAD-7 (mean difference = 0.799, SE = 0.314, p = .033), whereas low versus high (p = .061) and middle versus high (p = 1.000) were not significant. Living with someone was not associated with GAD-7 and did not interact with time (ps = 0.440 and 0.514). In summary, anxiety symptoms diminished over time for all subgroups, and between group differences reflected a higher anxiety burden for certain demographic groups. Most subgroup-by-time interactions were not statistically significant, indicating no clear evidence that symptom change over time differed across most demographic characteristics. See Table 6 for additional information. Table 6Model-adjusted GAD-7 estimated marginal means at pretreatment, posttreatment (week 8), and 3-month follow-up (week 20) by sociodemographic subgroup.GAD subgroupPre mean (95% CI)W8 post mean (95% CI)W20 mean (95% CI)Δ Pre→W8 (points)Δ Pre→W20 (points)%Δ Pre→W8%Δ Pre→W20GenderMan12.98 (12.22, 13.74)6.86 (6.11, 7.61)5.83 (5.00, 6.66)−6.12−7.15−47.1−55.1Woman13.05 (12.52, 13.58)6.99 (6.46, 7.52)6.36 (5.81, 6.92)−6.06−6.68−46.4−51.2Age (−1 SD / mean / +1 SD)Younger (−1 SD)13.89 (13.32, 14.47)7.70 (7.12, 8.28)7.12 (6.50, 7.75)−6.19−6.77−44.6−48.7Mean age (0)13.06 (12.56, 13.57)7.03 (6.53, 7.53)6.32 (5.79, 6.85)−6.04−6.75−46.2−51.6Older (+1 SD)12.24 (11.64, 12.83)6.35 (5.77, 6.94)5.51 (4.90, 6.12)−5.88−6.72−48.1−55.0EthnicityWhite12.38 (12.03, 12.72)6.29 (5.96, 6.63)5.70 (5.32, 6.08)−6.08−6.68−49.2−54.0Indigenous (First Nations/Inuit/Metis)13.94 (12.71, 15.16)8.10 (6.85, 9.35)6.26 (4.78, 7.75)−5.84−7.68−41.9−55.1Other racialized / non-White non-Indigenous12.94 (11.70, 14.17)7.59 (6.33, 8.86)5.48 (4.00, 6.95)−5.35−7.46−41.3−57.7EducationHigh school or less13.29 (12.51, 14.06)7.36 (6.59, 8.13)6.46 (5.59, 7.32)−5.92−6.83−44.6−51.4Below bachelor / some postsecondary13.47 (12.88, 14.07)7.49 (6.89, 8.08)6.55 (5.91, 7.19)−5.99−6.93−44.4−51.4Bachelor's degree12.76 (12.05, 13.47)6.75 (6.05, 7.44)6.01 (5.25, 6.76)−6.01−6.75−47.1−52.9Above bachelor's degree12.45 (11.43, 13.46)6.11 (5.13, 7.09)6.57 (5.44, 7.69)−6.33−5.88−50.9−47.3EmploymentEmployed (paid work)12.84 (12.29, 13.39)6.81 (6.26, 7.36)6.08 (5.50, 6.65)−6.03−6.76−47.0−52.7Not in labor force (household/retired/student)12.74 (12.13, 13.36)6.71 (6.11, 7.32)5.98 (5.35, 6.61)−6.03−6.76−47.3−53.1Not working due to health14.47 (13.66, 15.29)8.44 (7.63, 9.26)7.71 (6.88, 8.54)−6.03−6.76−41.7−46.7Not working (unemployed/other reasons)13.11 (12.20, 14.02)7.08 (6.17, 7.99)6.35 (5.42, 7.27)−6.03−6.76−46.0−51.6Household incomeLow (<$40,000)13.62 (12.92, 14.31)7.41 (6.72, 8.11)6.71 (6.00, 7.42)−6.2−6.9−45.6−50.7Middle ($40,000–$79,999)12.82 (12.14, 13.49)6.62 (5.95, 7.28)5.91 (5.22, 6.60)−6.2−6.9−48.4−53.9High (≥$80,000)12.68 (11.86, 13.51)6.48 (5.67, 7.30)5.78 (4.94, 6.62)−6.2−6.9−48.9−54.4RelationshipNot partnered (single/dating)12.92 (12.28, 13.55)6.89 (6.25, 7.52)6.16 (5.51, 6.81)−6.03−6.76−46.7−52.3Partnered (married/common law/living with partner)12.86 (12.33, 13.39)6.83 (6.30, 7.36)6.10 (5.54, 6.66)−6.03−6.76−46.9−52.6Previously partnered (separated/divorced/widowed)14.20 (13.33, 15.07)8.17 (7.30, 9.04)7.44 (6.56, 8.33)−6.03−6.76−42.5−47.6LocationRural12.70 (12.05, 13.35)6.67 (6.02, 7.31)5.94 (5.27, 6.60)−6.03−6.76−47.5−53.2Large town to medium city13.14 (12.40, 13.89)7.11 (6.36, 7.85)6.38 (5.62, 7.14)−6.03−6.76−45.9−51.5Large city13.16 (12.64, 13.69)7.13 (6.60, 7.66)6.40 (5.85, 6.95)−6.03−6.76−45.8−51.4Living with someoneNo13.04 (12.26, 13.83)7.16 (6.38, 7.94)6.19 (5.35, 7.04)−5.88−6.85−45.1−52.5Yes13.00 (12.48, 13.53)6.94 (6.41, 7.46)6.27 (5.72, 6.82)−6.07−6.74−46.7−51.8Note. Values are estimated marginal means (EMMEANS) from linear mixed-effects models with random intercepts and a heterogeneous first-order autoregressive residual covariance structure (ARH[1]). Pre = baseline; W8 = post-treatment; W20 = approximately 3-month follow-up. Each subgroup block is based on the corresponding subgroup model (see Statistical Analysis). Models were adjusted for mean-centered age, ethnicity, and educational attainment, except when one of these variables was the focal characteristic, in which case it was entered as the focal predictor rather than solely as an adjustment covariate; therefore, estimates should not be interpreted as originating from a single combined model across all subgroup blocks. Model-adjusted GAD-7 estimated marginal means at pretreatment, posttreatment (week 8), and 3-month follow-up (week 20) by sociodemographic subgroup. Note. Values are estimated marginal means (EMMEANS) from linear mixed-effects models with random intercepts and a heterogeneous first-order autoregressive residual covariance structure (ARH[1]). Pre = baseline; W8 = post-treatment; W20 = approximately 3-month follow-up. Each subgroup block is based on the corresponding subgroup model (see Statistical Analysis). Models were adjusted for mean-centered age, ethnicity, and educational attainment, except when one of these variables was the focal characteristic, in which case it was entered as the focal predictor rather than solely as an adjustment covariate; therefore, estimates should not be interpreted as originating from a single combined model across all subgroup blocks. Reliable change Among participants with clinically elevated baseline PHQ-9 scores and paired Week 8 data, 430/648 (66.4%, 95% CI [62.6, 69.9]) showed reliable improvement; at Week 20, 319/437 (73.0%, 95% CI [68.6, 76.9]) showed reliable improvement. Reliable deterioration on the PHQ-9 was uncommon, occurring in 15/958 participants (1.6%, 95% CI [1.0, 2.6]) at Week 8 and 9/632 (1.4%, 95% CI [0.8, 2.7]) at Week 20. For the GAD-7, 521/655 participants (79.5%, 95% CI [76.3, 82.5]) with clinically elevated baseline symptoms and paired Week 8 data showed reliable improvement, as did 346/427 (81.0%, 95% CI [77.0, 84.5]) at Week 20. Reliable deterioration on the GAD-7 occurred in 21/954 participants (2.2%, 95% CI [1.4, 3.3]) at Week 8 and 19/630 (3.0%, 95% CI [1.9, 4.7]) at Week 20. Subgroup reliable-change estimates are presented descriptively in Supplementary Table S1; no inferential comparisons of subgroup rates were conducted. Engagement and satisfaction Among eligible participants (n = 907 included; 69.7%), multivariable logistic regression models examined demographic correlates of lesson completion. Most participants completed at least three lessons (81.6%). The model predicting majority completion (≥3 lessons vs 0–2) showed an overall improvement in fit relative to an intercept-only model, χ2(17) = 39.58, p = .001. In this adjusted model, older age was associated with higher odds of completing ≥3 lessons, Wald χ2(1) = 8.13, p = .004, OR = 1.03, 95% CI [1.01, 1.05]. Household income was also associated with majority lesson completion, Wald χ2(2) = 7.11, p = .029; participants in the lowest income category showed lower odds of completing ≥3 lessons relative to the highest income group, although this pairwise contrast did not reach conventional significance (OR = 0.58, p = .072). No significant associations were observed for gender, education, ethnicity, relationship status, location, employment status, or living with someone (all ps > 0.05). A second adjusted model predicting completion of all core lessons (≥5 lessons vs 0–4) also improved fit relative to an intercept-only model, χ2(17) = 55.07, p < .001. Age was the only statistically significant predictor, with older participants more likely to complete all core lessons, Wald χ2(1) = 31.06, p < .001, OR = 1.04, 95% CI [1.03, 1.06]. Relationship status showed an omnibus effect that did not reach statistical significance, Wald χ2(2) = 4.53, p = .104, although previously partnered participants had higher odds of completing all core lessons compared to partnered participants (OR = 1.79, p = .035). No other demographic variables were significantly associated with completion of all core lessons in this model (all ps > 0.05). Satisfaction ratings were generally high. Most respondents endorsed overall satisfaction (82.7%); in the adjusted logistic regression model, relationship status was associated with overall satisfaction, Wald χ2(2) = 7.33, p = .026, with not-partnered participants reporting lower odds of overall satisfaction than partnered participants (OR = 0.38, 95% CI [0.15, 0.96], p = .040). No significant effects were observed for age, gender, education, ethnicity, location, income, employment status, or living with someone (all ps > 0.05). Satisfaction with materials (lessons and guides) was also high (88.5%); inferential analyses were attempted using the same adjusted logistic regression framework, but no demographic predictors showed robust or consistent associations after adjustment (all ps > 0.05). The remaining posttreatment satisfaction outcomes were highly positively skewed and are reported descriptively: 96.3% indicated they would recommend the course to a friend, 95.8% indicated the course was worth their time, 83.3% reported increased or greatly increased confidence in managing symptoms, and 78.3% reported increased or greatly increased motivation to seek future treatment if needed. Because non-endorsement responses were uncommon (approximately 11–22% across items), subgroup inferential analyses were not conducted for these outcomes. Discussion This study aimed to (a) compare the representativeness of a routine-care ICBT sample against population-level benchmarks and (b) evaluate whether clinical outcomes, engagement and satisfaction are comparable across demographic subgroups once clients access care. This study addressed an important gap, given that routine-care settings can differ from trials in uptake, adherence, and patient characteristics, even when clinical outcomes remain strong overall (Etzelmueller et al., 2020). Access On average, routine-care ICBT clients were younger than adults in the general Saskatchewan population, although the age range indicated uptake across adulthood. This pattern is consistent with evidence that mood and anxiety disorders are common in working-age adults and that digital modalities may be particularly appealing or accessible for some younger help-seekers (Dorow et al., 2018; Hybels et al., 2022; Statistics Canada, 2023b). A clear and consistent finding was the lower representation of men, alongside higher representation of women—patterns that have been repeatedly observed in ICBT services (Sapkota et al., 2023b; van der Vaart et al., 2019). This likely reflects multiple influences, including sex differences in the prevalence of depression and anxiety (Statistics Canada, 2023b; Steel et al., 2014) and gendered help-seeking norms and stigma-related concerns that may reduce men's likelihood of accessing care even when services are available (Affleck et al., 2018). Alternatively, the observed underrepresentation of men in ICBT may reflect differences in preferred help-seeking pathways rather than an absence of help-seeking per se; that is, men may be accessing support through other routes while ICBT remains less frequently used. Clarifying which pathways men use, and what barriers or preferences shape these choices, is an important direction for future work. Educational attainment also showed a pronounced gradient: clients with postsecondary credentials had higher representation, while those with high-school-only education or less had lower representation. This is consistent with evidence that educational and socioeconomic factors shape service use and barriers to care (Steele et al., 2007), even though lower educational attainment is often associated with greater mental health burden (Niemeyer et al., 2019). Underrepresentation of lower-education groups may reflect differences in awareness of ICBT, referral pathways, digital access/literacy, or preference for other modalities, and it highlights the importance of implementation efforts that reduce informational and practical barriers to entering care. Ethnocultural representation was heterogeneous rather than uniform. White clients had higher representation, which is consistent with reports of White overrepresentation in ICBT settings and trials relative to population benchmarks (De Jesús-Romero et al., 2024; Hadjistavropoulos et al., 2022). At the same time, representation was close to census benchmarks for some groups (e.g., Black and West Asian), whereas other groups had lower representation, including Southeast/East Asian, South Asian, and Indigenous clients. Lower participation among some ethnocultural groups is consistent with evidence of barriers such as acculturation, stigma, and uncertainty about how to access services (Han and Pong, 2015; Yang et al., 2020), and with Canadian data showing lower in-person service use among some groups even when mental health need is present (Chiu et al., 2018). For Indigenous peoples, underrepresentation is particularly concerning given evidence of unmet healthcare need (Goetz et al., 2023; Statistics Canada, 2024c) and longstanding concerns regarding culturally incongruent care in systems shaped by colonialism (Nelson and Wilson, 2017). Newcomers also had lower representation, consistent with evidence that recent immigrants may underutilize mental health services due to barriers related to language constraints, system navigation and cultural fit, with variation by country of origin (Durbin et al., 2015; McDonald and Kennedy, 2004; Stephenson and Fournier, 2026). In this context, the OTU work emphasizing culturally responsive, patient-oriented service improvement and evaluation of adapted models is directly relevant to improving equitable reach in routine care (Sapkota et al., 2023b). Geographic patterns likewise suggested that online delivery alone does not eliminate access disparities: large-city residents had higher representation, whereas small-town and rural residents had lower representation compared to census benchmarks. This finding is consistent with ongoing challenges in rural and remote mental health service delivery and the potential constraints posed by digital infrastructure and connectivity gaps, even as internet-delivered care is often promoted as a solution for geography-related barriers (Mental Health Commission of Canada, 2020; Office of the Auditor General of Canada, 2023). More broadly, disparities in mental health care service use by sociodemographic characteristics have been documented in Canada, underscoring the importance of examining reach in routine-care delivery models (Stephenson and Fournier, 2026). Finally, patterns in relationship status and living arrangement suggested differential reach by social context, with higher representation of single/dating, separated, and living-alone clients. These patterns may reflect associations between relationship disruption or social isolation and mental health need and help-seeking (Shankar, 2023; Whisman et al., 2022), although these mechanisms should be tested directly. Income and employment patterns likewise indicated that access was not evenly distributed across socioeconomic strata. Although ICBT is often positioned as scalable and cost-effective (Kaehlke et al., 2022; Mental Health Commission of Canada, 2017), socioeconomic factors may still influence access to and uptake of digital mental health services (Williamson et al., 2019; Keast et al., 2025). Income representation showed a mixed rather than linear pattern, with representation comparable to population benchmarks across several income categories. In contrast, individuals employed in paid work had lower representation, whereas those not employed in paid work had higher representation relative to the population benchmark. . Together, these representativeness findings underscore the value of pairing population benchmarking with targeted implementation strategies—such as community outreach, culturally responsive adaptation, and attention to digital and informational barriers—to improve equitable reach in routine-care ICBT. Clinical outcomes Beyond representativeness, depressive and anxiety symptoms declined substantially during treatment and follow-up across demographic subgroups, indicating that therapist-assisted ICBT delivered in routine care is associated with meaningful clinical improvement once clients access care. For most demographic characteristics, subgroup-by-time interactions were not statistically significant, with differences primarily observed in overall symptom burden rather than in rates of change. Specifically, older age was consistently associated with lower symptom levels. Higher symptom burden was observed among certain groups, including Indigenous clients during pre-treatment and treatment, and individuals not working due to health across all assessment periods. Some subgroup differences, particularly those involving education and ethnicity, were no longer evident at follow-up. Other differences, including those associated with relationship status, location and employment status remained evident at Week 20. The threshold-based analyses provided a complementary individual-level perspective: reliable improvement was common among participants with clinically elevated baseline symptoms, whereas reliable deterioration was uncommon at both Weeks 8 and 20. Taken together, these findings indicate substantial symptom reductions across diverse client subgroups who accessed routine-care ICBT, while also highlighting that inequities in symptom burden can persist across treatment and follow-up (e.g., higher symptom levels among Indigenous clients, and among those not working due to health). In this respect, the outcome findings align with the service reach findings: some groups with lower representation in access (e.g., Indigenous clients) are also groups for whom elevated symptom burden is observed among those who do enroll, underscoring the importance of both improving access and potentially optimizing culturally informed, context-responsive delivery. Related OTU work has similarly emphasized that symptom improvements can be strong in routine care among people of diverse backgrounds while engagement and experience remain essential equity-relevant outcomes to monitor and optimize (Edmonds et al., 2018; Sapkota et al., 2023a; Sapkota et al., 2024). Future research could extend these subgroup analyses by examining combinations of pretreatment characteristics using multivariable prediction approaches to identify individuals who may experience different treatment outcomes (e.g., Wallert et al., 2022). Future research could formally examine subgroup differences in reliable change and treatment response using adequately powered, prespecified analyses. Engagement and satisfaction Engagement analyses further clarify where disparities may arise in the care pathway. In this sample, completion was high, with most participants completing at least three lessons (81.6%). These engagement findings suggest that younger participants complete fewer lessons, a pattern consistent with OTU routine-care evidence that completion tends to be higher among older clients (Edmonds et al., 2018). Satisfaction overall with the course was high (82.7%). Participants who were not partnered reported lower satisfaction than those who were partnered, whereas other demographic characteristics showed limited or no association with overall satisfaction among respondents. Satisfaction with materials was also high (88.5%). Most clients reported that they would recommend the course to others (96.3%), they believed it was worth their time (95.8%), reported increased confidence to manage symptoms (83.3%) and reported increased motivation to seek future treatment (78.3%). High satisfaction and strong acceptability in routine care is consistent with OTU benchmarking work examining satisfaction, engagement, and outcomes in an adapted ICBT program for people of diverse ethnocultural groups (Sapkota et al., 2024). Strengths and limitations This study has several limitations and strengths. This study was conducted in a single Canadian Province, and these findings may not generalize to other regions. A further limitation is that census benchmarks reflect the general population rather than the subset of Saskatchewan adults with depression and/or anxiety who might plausibly be candidates for ICBT. The representation of gender-minority groups could not be compared because the Census Public Use Microdata File used in the present analyses did not provide directly comparable transgender and non-binary categories; these data are sometimes aggregated to protect confidentiality (Statistics Canada, 2023a). The representation of gender-minority groups (e.g., non-binary, trans) in ICBT could not be examined, due to a lack of available Canadian census data; the Census Canada dataset did not report gender minority status, as very low population prevalence rates could compromise confidentiality (Statistics Canada, 2023a). As a second limitation, census data for location type (e.g., rural, small town) contained data from both adults and children, while OTU data were from adults exclusively. Therefore, precise head-to-head (i.e., adult-to-adult) comparisons of proportions were not possible, and while our results suggest that individuals living outside large cities had lower representation in our sample, the validity of these comparisons may be limited, precluding firm conclusions. Other ICBT routine care studies report varied results; some report a higher concentration of clients living outside major cities (Titov et al., 2018) while others indicate a greater proportion of users in larger cities (Mehta et al., 2022). Other limitations pertain to the subgroup outcome and engagement/satisfaction analyses. Because this was an uncontrolled observational study, symptom reductions over time cannot be attributed causally to ICBT. In addition, nonsignificant subgroup-by-time interactions should not be interpreted as evidence of equivalent outcomes across groups, particularly because subgroup sizes, and therefore precision and statistical power, varied across analyses. Engagement models were conducted on eligible cases with complete data on model predictors, resulting in 69.7% included and 30.3% excluded, which may introduce selection bias if excluded participants differed systematically on unmeasured characteristics. Several satisfaction outcomes were highly skewed and had substantial missingness, limiting the interpretability of subgroup modeling; as a result, inferential subgroup analyses were restricted to indicators with sufficient variability and data completeness (e.g., overall satisfaction and satisfaction with materials) and the remaining items were reported descriptively. The study examined multiple subgroup models and multiple comparisons; although this approach supports a broad equity-focused description, some time-specific differences (e.g., differences emerging at later follow-up) should be interpreted cautiously and ideally replicated. Finally, engagement was operationalized primarily as lesson completion; future work may benefit from including additional engagement indicators (e.g., homework engagement), given evidence that homework engagement predicts symptom change in routine-care ICBT (Sapkota et al., 2023a). A strength of this study is the specificity of comparisons made between population-level census data and ICBT data; by recoding and collapsing demographic variable categories/levels, it was possible to operationalize demographic characteristics in an equivalent manner across the two datasets, enabling head-to-head comparisons. This is the first study, to our knowledge, that has examined the representativeness of an ICBT sample with this degree of specificity, and across such a wide range of diverse groups. Although ICBT studies have reported proportions of minoritized groups in ICBT, few have compared sample proportions to a reference population to examine proportionate representation. Finally, this study examined the representation of demographic subgroups in a routine care setting, permitting a naturalistic snapshot of demographic representation in ICBT across diverse outcomes. It can be argued that this naturalistic recruitment method provides a more accurate picture of ICBT's reach and fulfillment of its potential to circumvent healthcare barriers, compared to convenience sampling methods. Clinical implications Symptoms improved across all examined groups, and treatment satisfaction was high. Compared to demographic benchmarks from the greater population, certain groups had lower representation. Strategies could be implemented to increase representativeness, for example by targeting certain communities to promote awareness of ICBT and uptake. However, ICBT researchers and clinicians may have fewer opportunities to address other barriers, such as self-directed stigma, acculturation, lower awareness of healthcare resources, a lower need for mental health services due to a healthy immigration effect, and a technological divide that exists in Canada for remote and Indigenous communities. Furthermore, psychological disorder prevalences are not evenly stratified among demographic subgroups within the general population (Statistics Canada, 2023b), and differences in representation may reflect differences in need (e.g., greater help seeking may arise from a higher need for anxiety and depression treatment in women versus men). Additionally, ICBT may not be the most appropriate treatment option for certain marginalized groups facing serious psychosocial challenges such as limited resources, housing instability, or severe symptoms, as their difficulties may require in-person interventions and supports (e.g., case management, psychiatric management, intensive, face-to-face services). ICBT may not align with forms of mental health support preferred by minority groups, such as affiliating with friends and family, or culturally specific healing practices. Nevertheless, among individuals in our sample who sought and accessed ICBT, symptom improvements were observed across subgroups, reinforcing the potential value of ICBT as a low-cost, low-barrier treatment option. Conclusions and future directions The key contribution of this study was the joint examination of representativeness, outcomes, engagement, and satisfaction of diverse demographic subgroups in routine-care ICBT using population benchmarks. Although clinical trials have demonstrated the strong efficacy of ICBT, and the considerable potential of ICBT to reduce mental healthcare disparities, ICBT in routine care is not reaching several demographic groups in proportion to their representation in the population. Groups with lower representation included men, individuals with lower levels of educational attainment, Indigenous, Southeast/East Asians, South Asians, and newcomer populations, individuals living in rural communities and small towns, and persons who are married, common law, or divorced. Income-related representation was mixed, with lower representation in the $20,000–$39,999 and $100,000-or-more categories. Importantly, among individuals from these groups who did access care, symptom improvements, engagement, and satisfaction were generally strong across subgroups. We encourage ICBT researchers to continue examining the representation of demographic subgroups in routine-care ICBT relative to population benchmarks across regions, to better understand gaps in reach. Researchers and clinicians may also consider targeted outreach, culturally informed and context-responsive adaptations and enhanced supports to extend the reach and benefits of ICBT to underserved groups and communities. Ethics statement The study was approved by the Research Ethics Board of the University of Regina (2019-197), and all participants provided informed consent. Declaration of Generative AI and AI-assisted technologies in the writing process During the preparation of this work RPS used ChatGPT-5.2 to generate SPSS syntax and to improve grammar and phrasing. HDH used ChatGPT-5.2 to review consistency of the reported results and references. Following the use of this tool, the authors reviewed and edited the content as needed and take full responsibility for the content of the publication. Funding We did not receive any funding specific to this study. The Online Therapy Unit is funded by the Saskatchewan Ministry of Health. N.T. and B.F.D. are funded by the Australian Government to operate the national MindSpot Clinic. Funders had no involvement in the study design, collection, analysis, or interpretation of the data. Declaration of competing interest The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper. Acknowledgements The authors wish to acknowledge the patients, screeners, therapists, research staff, research associates, students, and web developers associated with the Online Therapy Unit at the University of Regina. Footnotes This article is part of a Special issue entitled: ‘Advancing Equity’ published in Internet Interventions. An earlier version of this paper was accepted for presentation at The International Society for Research on Internet Interventions (ISRII) 13th Scientific Meeting. However, the paper was not presented, as travel was not recommended. Supplementary data to this article can be found online at https://doi.org/10.1016/j.invent.2026.100998. Supplementary data Supplementary Table S1 Reliable improvement and reliable deterioration at posttreatment and follow-up by sociodemographic subgroup Data availability The data were collected during routine care and are only available from the Online Therapy Unit (OTU@uregina.ca) upon reasonable request. References Associated Data Supplementary Materials Supplementary Table S1 Reliable improvement and reliable deterioration at posttreatment and follow-up by sociodemographic subgroup Data Availability Statement The data were collected during routine care and are only available from the Online Therapy Unit (OTU@uregina.ca) upon reasonable request.