This study used a retrospective cohort design to examine service use patterns of children and adolescents in receipt of ambulatory mental health services between April 1, 2019, and March 31, 2023, using administrative health data collected by the IWK Mental Health and Addictions (MHA) program. The IWK MHA program provides family-centered mental health and addictions care for children and adolescents up to their 19th birthday in Nova Scotia, Canada. Outpatient (ambulatory) services offered are in 3 Community Mental Health and Addictions clinics and in some schools and other community locations in the Halifax region. Elsewhere provincially, outpatient child and adolescent services are offered by Nova Scotia Health. Ambulatory services are provided using the Choice and Partnership Approach (CAPA). Briefly, CAPA is a client-oriented service model that incorporates the active collaboration of clients, families, and clinicians in care planning and evidence-based treatment, demand and capacity management guided by lean thinking, clinical skill matching to client needs, and goal-based outcome monitoring [29]. “Choice” appointments represent the initial clinical contact, in which clients, families, and clinicians jointly decide upon clinical needs and treatment options [29]. Clients going on to treatment do so in “Partnership,” during which most intervention work occurs with a clinician with skills matched to the client’s clinical needs [29]. Clients may exit and return to care based on their needs.

Administrative health data sources included Meditech registration and Community Wide Scheduling databases held at IWK Health. Abstracted data included client and service information of children and youth who accessed IWK MHA services between April 1, 2018, and January 31, 2024, including client demographic information, clinic location, outpatient appointment dates, visit type (first “Choice” or treatment “Partnership”), modality (virtual or in-person), acuity at intake, and presenting concerns for approximately 11,000 unique clients.

In our analysis, study ID, a randomly assigned number, was used to identify unique individuals, while protecting confidentiality. Age was categorized into 3 groups representing clinically meaningful client populations: 0‐11 years, 12‐18 years, and 19+ years. Although the IWK MHA program typically provides services to clients younger than 19 years, it is not uncommon that those older than 19 years receive services. This is most common due to those who are transitioning to adult services being seen for bridging while this transition occurs. Additionally, those who are near the end of a treatment episode may be seen beyond their 19th birthday for completion of care, and some specialty services may continue to see clients due to lack of comparable care in adult services. Sex (biological) was classified as male or female, as available in the administrative data. Presenting concern was based on the primary issue identified at the initial Partnership visit. Presenting concerns are based upon Diagnostic and Statistical Manual of Mental Disorders (Fifth Edition) (DSM-V) diagnostic codes and are assigned by the Choice clinician and updated by the Partnership clinician as needed. The categories of presenting concern included in the model were substance use disorders, mood disorders (mood and bipolar), anxiety disorders, personality disorders, eating and other feeding disorders, neurodevelopmental disorders (attention-deficit/hyperactivity disorder and autism spectrum disorder), neurocognitive disorders (cognitive and developmental delay), trans health, to be determined, other mental health disorders (disruptive behavior, forensic risk, obsessive-compulsive disorder, Tourette syndrome, trauma, and psychosis), and “not available.” Urgent triage stream was assigned to episodes for which the first Partnership visit occurred within the urgent triage stream and was used to distinguish from episodes occurring within the nonurgent triage stream. The triage stream (urgent or nonurgent) differs in terms of initial wait time benchmarks (eg, 7 or 28 days) and is assigned based on clinical triage that reflects an assessment of the urgency of need for care. The distinction between Core and Core and Specific Partnership was made based on whether the episode consisted solely of Core Partnership visits or included both Core and Specific Partnership visits. In the CAPA model, the majority of treatment occurs within the Core pathway, with Specific Partnership work consisting of specialized assessment or treatment added as necessary to meet client needs and goals of treatment. It was anticipated that those with both Core and Specific Partnership visits would have longer episodes of care or have more frequent visits within an episode of care than those with solely Core Partnership care. Clinic referred to the location where the client was seen during the initial Partnership visit within an episode of care, which included the 3 outpatient clinics in the Halifax region (Halifax, Dartmouth, and Sackville) and School Mental Health. The time period of a given episode was assigned based on the fiscal year (2019/20, 2020/21, 2021/22, and 2022/23) in which more than 50% of the episode occurred. Acuity at presentation to Community Mental Health and Addictions services is measured using the HEADS-ED (Home, Education & Employment, Activities & Peers, Drugs & Alcohol, Suicidality, Emotions, Behaviors, Thought Disturbance, and Discharge or Current Resources) screening tool during intake at the IWK Central Referral service. The HEADS-ED is a validated mental health screening tool used to obtain a psychosocial history and guide decision-making regarding client disposition. A higher overall HEADS-ED score indicated a need for more intensive or immediate mental health services [30,31].

We used a multilevel mixed-effects negative binomial model to examine the association between treatment modality and treatment duration in numbers of visits. This model was chosen due to the hierarchical structure of the data, with episodes of care nested within individual study IDs and the overdispersed count nature of the outcome. Additionally, because the mean treatment duration in visits differed from the variance, a negative binomial model was selected over a Poisson model. An unadjusted model was used to estimate the relationship between treatment modality and treatment duration, and an adjusted model included fixed effects such as age, sex, presenting concern, urgent stream, Specific Partnership, and clinic. We treated the study ID as a random effect. To identify confounders, we compared the measure of association between treatment modality and treatment duration both before and after adjusting for a potential confounding factor, one at a time. If the difference was greater than 10%, we concluded that confounding was present and adjusted for these variables in our analysis [32]. To identify effect measure modification, we calculated stratum-specific measures of association. If the stratum-specific measures were different from one another, the variable was identified as an effect modifier, and the stratum-specific measures of association were reported [33]. As we expected real temporal differences in treatment duration by modality over the observation period, we treated time period (fiscal years 2019/20, 2020/21, 2021/22, and 2022/23) as an effect modifier and reported stratified results accordingly. We excluded outliers; therefore, we restricted the model to episodes shorter than 25 visits, which corresponds to the 95th percentile of the visit distribution. Episodes exceeding 25 visits may differ in client and clinical characteristics and may be more likely to be classified as hybrid due to their extended duration and increased likelihood of exposure to both modalities.

As a sensitivity analysis to assess the impact of treatment modality on treatment duration in days, a time-to-event analysis was used. We included client and system characteristics with statistically significant associations and retained modalities that met the proportional hazards assumption. The client characteristics included in the time-to-event analyses were sex and presenting concern. We stratified the results by time period.

The IWK Health Research Ethics Board approved the overarching research project, including this study (“Our Virtual Reality: Rapidly Responding to Changing Mental Health Needs Among Children and Adolescents,” project number 1026770). Informed consent was not required for this secondary analysis of deidentified administrative data. Members of the research team did not have access to any identifying information for visits captured in the administrative databases.

In this retrospective cohort study of administrative health data in a large child and adolescent ambulatory mental health service, we studied duration of treatment by episodes of care for virtual and in-person treatment modalities following the rapid implementation of virtual care during the COVID-19 pandemic. When considering episodes of care in terms of the number of visits, the observation that “virtual care takes longer” appeared to be true in the first year following the onset of the pandemic. The duration of treatment was longest during period 2 (fiscal year 2020/21) in terms of numbers of visits per episode of care, attenuating during the following 2 fiscal year periods of observation. Episodes of care delivered virtually in 2020/21 involved nearly 60% more visits than those delivered in person. However, by 2022/23, virtual care episodes involved 18% fewer visits than in-person episodes. When looking at the treatment duration in numbers of days per episode of care, these patterns were not significantly different.

While we accounted for key client and system factors in our analysis, there could be other explanations for our findings that are not available in administrative data. The period between April 1, 2020, and March 31, 2021 (period 2), represents the beginning of the COVID-19 pandemic–related public health restrictions and the rapid implementation of virtual mental health care. Operating mental health services through a virtual medium was relatively new during this time [5,8,34], including in the study clinics in which virtual care was very limited in use prior to the pandemic [7]. Therefore, this required adjustments and adapting to this new service delivery method for both clinicians and clients. Clinicians had to learn how to use virtual care and work through this new virtual medium; it took time to develop these skills [34]. Furthermore, clinicians had to review safety considerations pertaining to using virtual care with the client, which took up time within an appointment, and thus, it took longer to get to the work of identifying treatment goals and providing treatment [34]. Moreover, clients had to adjust to this new way of receiving care [35]. Furthermore, within the first year of the pandemic, the proportion of clients referred to the urgent triage stream increased, as did the average HEADS-ED score [36], which may reflect greater treatment needs and, thus, may be why we observed longer duration of treatment during this time. Perez et al [37] found that during the COVID-19 pandemic, youth attended more virtual care appointments, likely due to increased accessibility and increased mental health needs. However, these virtual sessions were often shorter than in-person sessions, possibly due to the more frequent visits [37]. This may help explain why we observed longer episodes of care during this time.

While our measures of clinical presentation and acuity did not affect our findings, there are likely important client factors that affect service need and attendance but are not adequately captured in administrative data. For example, at the onset of the pandemic, with public health restrictions in place, the availability of in-person treatment was limited; therefore, in-person appointments were held only in cases in which there were safety concerns. As such, those receiving in-person care during this time may have been systematically different from those receiving virtual care [34,38]. Together, these considerations could contribute to virtual episodes of care initially being longer or consisting of more visits while adjusting to this new way of care.

When restrictions were lifted and choice was available, clients, caregivers, and clinicians may have continued using the virtual modality in part due to concerns regarding the risk of contracting COVID-19 disease [7]. However, in period 4, our most recent period, from April 1, 2022, to March 31, 2023, we observed fewer virtual episodes of care (n=413) during this time compared with the start of the COVID-19 pandemic (n=1130). Again, those choosing virtual care during this time may be systematically different in terms of their presenting concerns and treatment needs, which may for the majority be less acute and less complex than those choosing in-person care, and thus, virtual episodes of care during this time may be shorter [38]. Furthermore, due to virtual care having been established for at least 2 years by this time, the efficiency and proficiency in using technology, as well as more efficient and accessible processes and policies, had likely improved for both clients and clinicians over time [34].

Administrative data are an important resource for both research and routine quality improvement activities, and this study benefits from the use of a rich administrative dataset that captures detailed individual-level clinical and service use information from across a child and youth ambulatory mental health service. These data provide greater detail on care episodes and patients than the more commonly studied administrative health databases (eg, physician billing and hospital discharge abstract databases). However, the use of administrative data within this study presents some limitations regarding the information that is available. For example, the complexity of a client’s case, their goals of treatment, or their reasons for seeking mental health care (which may differ from the preset categories representing clinician-identified clinical presentations) are not captured but may also impact treatment duration. These factors likely impact treatment duration regardless of modality; however, they could interact with modality in several ways. For instance, virtual care may be preferred or avoided, based on client needs or preferences. Other factors, such as the social determinants of health, including socioeconomic status, health literacy, and family status, which would impact engagement in care and barriers to care, would also be useful to help predict treatment trajectories [39,40]. These factors may also vary across treatment modalities, as certain modalities may present or mitigate barriers for certain clients.

Additionally, the number of clinicians involved in an episode of care could also impact treatment duration as the client may have multiple encounters with multiple providers within an episode and contribute to increased treatment duration. While we accounted for the addition of Specific Partnership work, which would likely involve additional clinicians in a client’s care, we did not explicitly capture the number of clinicians involved in treatment. The use of in-person or virtual care may also vary across providers, based on preferences or clinical requirements. Finally, we do not have a way to confirm that the end of a given episode of care represents the clinical closure in which the treatment goals were met, mutual agreement to end care with limited benefit, or whether clients were lost to follow-up. Future studies would benefit from the distinction of clinical or administrative closures of care episodes as well as the inclusion of routine outcome measures to support conclusions regarding the relative effectiveness of virtual care as reflected by treatment length.

With the advancement of virtual care as a treatment modality following the onset of the COVID-19 pandemic and the evolving client treatment needs, evidenced by changes in patterns of mental health service use and acuity, this area of research has become increasingly important, especially with the risk of virtual care being abandoned as we transition back to prepandemic (or pre public health restrictions) operations and processes, thus reducing equitable access to mental health services.

This research demonstrates that the duration of treatment for virtual care is comparable with in-person care. While virtual episodes of care were longer immediately following the onset of the pandemic, these differences were small, and by the most recent period of observation, virtual episodes of care were shorter. These findings challenge the perception that virtual care routinely takes longer than in-person treatment. Consequently, as virtual and in-person episodes of care showed broadly similar treatment durations, it supports the continuation of virtual mental health services as a treatment option for children and adolescents, ensuring that those who benefit from virtual care can maintain necessary access to services without undue impact on service capacity.

Further research is necessary to better understand how in-person versus virtual treatment modalities contribute to the effectiveness of care in terms of client mental health outcomes and goals of care. Future studies would also benefit from more comprehensive information on individual client characteristics, such as sociodemographic information and measures of complexity, to help explain any observed differences in treatment duration. Furthermore, information on predictors of outcomes may help explain variations in patterns of service use. Additionally, while our study provided information regarding how treatment modality impacts the duration of episodes of care, the efficacy of virtual care requires further examination. Measures of efficacy in the form of routine outcome measures, alongside information regarding expected treatment timelines, would be important in program-level decisions surrounding the continuation of virtual care and, at the clinical level, supporting shared decision-making.