Introduction

Mental health and substance use disorders are common health conditions that affect the lives of countless people worldwide. In 2018, it was estimated that 20% of the Australian population had a mental health or substance use disorder (Australian Bureau of Statistics, 2018), with rates of 13.2–16.7% reported overseas (Baker, 2020; Jha et al., 2019; Stagnaro et al., 2018). Mental health and substance use disorders represent 5% of disease burden and account for around 14.3% of deaths globally (Dattani et al., 2021; Walker et al., 2015). For individuals with severe mental illness, adverse outcomes such as poorer physical health, stigma, and homelessness commonly co-occur (Dubreucq et al., 2021; Elbogen et al., 2021; Robson & Gray, 2007). This highlights the importance of accessible and multi-faceted services to ensure that people can receive the care they need. The uptake and use of mental health and substance use treatment has been found to help individuals manage symptoms of their condition(s), improving their physical and social wellbeing (Clark et al., 2018; Nakao et al., 2021). Furthermore, the sooner one access care, the less their mental health status declines (Reichert & Jacobs, 2018; Smith et al., 2018). Although effective treatments for mental health conditions exist, delays in accessing mental health care frequently occur and many affected people remain untreated (Bidargaddi et al., 2020). Thus, accessible mental health treatment and care is critical.

In recent years, virtual care delivery, which encompasses the use of e-health, m-health, and telehealth, has surged across all healthcare services due to the COVID-19 pandemic, particularly in mental health and substance use services (Ellis et al., 2021; Sorkin et al., 2021; van Kessel et al., 2022; Zhu et al., 2021). Virtual care addresses some of the barriers in accessing traditional in-person mental health and substance use services, such as poor affordability and inaccessibility (Coombs et al., 2021; Moroz et al., 2020), with telehealth of particular interest as it provides the closest approximation to in-person care. For certain mental health conditions, virtual care offers the same effectiveness of care as traditional in-person consultation (Greenwood et al., 2022; Krzyzaniak et al., 2021; Scott et al., 2022a; Scott, Clark, Scott et al., 2022a, b), but more convenience as patients can receive care from their own homes, reducing the need to travel (Berardi et al., 2024; Polinski et al., 2016). This is particularly beneficial for patients living in rural and remote areas, who might find it burdensome to travel long distances to receive treatment (Butzner & Cuffee, 2021). Studies have also shown that virtual care can reduce the wait times for outpatient services (Caffery et al., 2016; Mahmoud et al., 2021; Uscher-Pines et al., 2020; Valentine et al., 2021). Additionally, the option to seek care from home can mitigate the stigma associated with accessing mental health treatments (Kim et al., 2022; Kim & Tesmer, 2021; Philip et al., 2022).

Despite its many benefits, the implementation of virtual care does present unique challenges. Common barriers to virtual care use include a lack of access to technology, difficulties establishing strong client-patient rapport, and concerns about privacy (Hughto et al., 2021; Naal et al., 2021). To ensure that people can access the care they need, especially when virtual care is the only option, it is critical that this service delivery modality is implemented in a way that promotes effective, sustainable, and equitable uptake. One way of addressing this need is to employ a person-centred approach by identifying the features of virtual care delivery that are important to service users and service providers and most likely to influence their utilisation of a service. As a first step, it is important to identify the factors that contribute to people using (or not using) mental health and substance use services that involve an element of virtual care. As a second step, it is important to understand the trade-offs that people make between these factors (e.g., between wait time and type of provider), and the extent to which each factor leads to greater (or lesser) use of a service for different groups of people. This second step allows analyses that can predict the change in uptake for a service, with a change in a factor (e.g., how many more clients would use the service if the wait time was reduced by half) providing clear, easily implementable recommendations to services to ensure effective, sustainable, and equitable uptake.

These steps are best achieved through preference elicitation methodologies, such as discrete choice experiments (DCEs), which use a systematic process to directly ask individuals for their preferences on a product or service. Compared to qualitative methodologies, preference elicitation methods can produce more precise preference data as individuals are explicitly communicating their likes, dislikes, and priorities, within realistic scenarios, comparable across all respondents (Soekhai et al., 2019). As such, they are ideal methods for empirically determining which factors are most important to service users and providers, and therefore likely to influence their utilisation of the services (Clark et al., 2014; Ryan et al., 2008). In a DCE, participants are presented with a series of realistic hypothetical products or services (e.g., a health service), which are comprised of key attributes (e.g., waiting time) that are believed to influence utilisation of the product/service. The expression of these attributes differs between the hypothetical products/services through pre-defined levels, such as one, two or three weeks (for waiting time). Participants indicate which product/service they prefer and/or would use. Their choices, paired with the attribute levels presented to them, provide information on the ‘ideal’ product/service. An evidence-based approach, often using systematic reviews and qualitative methods, is utilised to select the attributes and levels included in a DCE or other preference elicitation study (Ryan et al., 2008). A synthesis of the features (i.e., attributes and levels) included in preference elicitation studies, is thus an efficient method of exploring what important factors have been found, and what gaps in knowledge remain. This systematic review aimed to determine what factors have been found to influence service user and service provider preferences for virtual service delivery in mental health and substance use care.

Methods

The research team used Covidence (Veritas Health Innovation, 2021) to conduct screening, full-text review, quality assessment and data extraction. EndNote (The EndNote Team, 2013) was used to import exported references from the databases and the manual search into Covidence. A review protocol was created and made available to the public using the Open Science Framework on September 29th, 2021 (DOI: https://doi.org/10.17605/OSF.IO/ZBG3J).

Search Strategy

A systematic search was conducted across seven health, economics, and social science databases: PubMed, PsycINFO, EconLit, MEDLINE, CINAHL, Academic Search Ultimate, and ProQuest Central (Consumer Health, Health & Medical Collection, Nursing & Allied Health, Psychology, Public Health & Social Science databases). The search was conducted by one independent reviewer (CM) for all articles published from database inception until May 23rd 2023, using search terms related to preference elicitation of service modality in mental health and substance use populations (Table 1). Additionally, a manual search was conducted by four independent reviewers (CM, JW, KD, LW) of the reference lists of relevant excluded review articles and included articles, as well as relevant excluded protocol articles and dissertations, to identify relevant peer-reviewed original research articles.

Table 1 Database search terms
Full size table

Eligibility Criteria

Titles and abstracts of articles were screened for relevance by two independent reviewers (CM, FKL, KD or LW), with conflicts resolved either via consensus or a third reviewer (FKL). Articles that were screened as being relevant moved to the full-text review stage where they were reviewed against eligibility criteria by CM and one other independent reviewer (CT, GLHF or KD), with conflicts resolved by a third independent reviewer (LW). Articles were included if they were (i) written in English, (ii) reported on peer-reviewed original research, and (iii) used a preference elicitation measure (e.g., discrete choice experiment) for virtual service delivery in either (a) mental health or (b) substance use service settings. Articles were excluded if they (i) were written in a language other than English, (ii) were a review article, conference paper, book, dissertation, grey literature, or not peer-reviewed, (iii) used non-human participants, (iv) were not focused on either (a) mental health or (b) substance use services, v) did not use a preference elicitation measure, vi) did not include a virtual care modality, or vii) did not report on service delivery features.

Quality Assessment and Data Extraction

The Mixed Methods Appraisal Tool (version 2018; Hong, 2018; Pluye et al., 2009) was used by two reviewers (CM, LW) to appraise the quality of the studies included in the systematic review. We completed the two screening questions applicable to all study designs and the five questions specific to the ‘quantitative descriptive’ study design, using the tool’s yes/no/can’t tell response options. Discrepancies in the quality appraisal of the two reviewers were discussed until consensus was achieved. No studies were excluded based on quality; however the quality of the studies was considered when interpreting the results.

One reviewer (CM) independently extracted the following data from the included articles: country, setting (e.g., hospital), service type (i.e., mental health or substance use), sample size, participant type (e.g., service user or provider), participant demographics (e.g., age, gender), preference elicitation tool (e.g., DCE or rating scale), and study outcomes. For quality assurance purposes, 20% of the extracted data was checked by another reviewer (LW). A meta-analysis was not feasible due to heterogeneity in the attributes and levels of the studies included. Instead, results are presented as a qualitative synthesis.

Results

While we sought to synthesise preferences for features of virtual mental health and substance use service delivery in this review, we did not identify any studies that directly investigated this. However, 19 studies were identified by our search strategy that met our eligibility criteria, which are subsequently included in this review. We included these studies because they examined preferences related to some component of virtual care. Additionally, these studies provide evidence on what features are generally important to service users and providers when making decisions about their care and the services they access. As such, the following sections mainly report on preferences for aspects of mental health and substance use treatment services other than virtual care. Figure 1 depicts the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) flowchart detailing the number of articles at each stage of the review.

Fig. 1
figure 1

PRISMA flowchart of the review process

Full size image

Study Characteristics

Table 2 presents the study characteristics and participant characteristics. The 19 studies were conducted by 12 research teams. Most studies were published in 2010 or after (95%), were conducted in North America (63%), and elicited preferences for solely mental health service features (74%) using DCEs/conjoint analysis (90%). Sample sizes ranged from 42 to 1984 participants (mean = 374, median = 218). Seven of the 19 studies used more than one recruitment method, with face-to-face (32%) and online (26%) being the most frequently employed. Participants were primarily women in most studies (n = 14), with only one study focusing solely on men (Dwight-Johnson et al., 2013). Mean age ranged from 35 to 59 years with a large spread as most participants were in their early 20’s to mid-30’s. Most participants had undertaken or attained more than a high school degree (61-100%) in 12 of the 19 studies. Of the studies that reported on language (n = 8), English was most frequently reported as preferred/first language (71-91%), but two studies had Spanish-speaking participants as the majority (Dwight-Johnson et al., 2004, 2010). Three studies (Hawke et al., 2021a, b; Klein & Cook, 2010) reported on residential location, with most participants residing in a metropolitan area (68-79%) compared to regional and rural.

Table 2 Study characteristics of included articles
Full size table

Most of the studies (n = 17) included participants that were service users; people with a mental health or substance use condition (n = 11), carers/family members (n = 3), university students (n = 2), people from the general population (n = 3), and health authority employees (n = 1). Service users who were people with a mental health or substance use condition, and their carers/family members, were drawn from various settings including outpatient mental health clinics, primary care, and mental health organisations’ social media channels. Only four studies included participants that were service providers (clinicians and/or service administrators). Service provider background was not provided in three of the four studies. However, the one that did report on professional background had a sample comprised mainly of psychotherapists/psychiatrists. Regarding the mental health/substance use focus, depression was the most common across the studies (n = 7). Other focuses included anxiety (n = 2), tobacco smoking (n = 1), and opioid use disorder (n = 1). Ten studies did not focus on a particular mental health and/or substance use disorder. Of the studies reporting on treatment history (n = 9), most service user participants (62-100%) were either currently being treated or had previously been treated for their mental health or substance use conditions.

Quality Assessment of Included Studies

Table 3 reports the percentage of MMAT (Hong, 2018; Pluye et al., 2009) items that met methodological requirements within each article after consensus. Agreement between the two reviewers (CM, LW) was very good, with agreement percentages resulting as either 71% (9 articles), 86% (7 articles), or 100% (3 articles) for the 19 included articles. All 19 articles included clear research questions and reported on data that addressed those questions. Of the 19 articles, two (Lau et al., 2012; Phillips et al., 2021) had excellent methodological quality with 100% of items meeting requirements. Some methodological constraints, however, were identified in 17 of the articles. The most common limitations of articles were not demonstrating a sample representative of the target population (80% failed to meet item) and not having low risk of non-response bias (80% failed to meet item). This was often due to the samples being over representative of females and highly educated people, low response rates, and not reporting analyses comparing the demographics of people who did and did not participate. Despite these concerns, the quality of the studies overall was very good, with 14 of the 19 articles having relevant sampling strategies, appropriate measurements, and appropriate statistical analyses.

Table 3 Results on the mixed methods appraisal tool assessing the quality of the included articles
Full size table

Study Outcomes: Preferences for Characteristics of Virtual Service Delivery for Mental Health and Drug Care

Seventeen studies employed DCEs, and two studies used rating surveys. The attributes from the DCEs and questions from the rating surveys were grouped thematically to synthesise the mental health and substance use care features included in the studies. After thematic synthesis, 41 unique features were identified for mental health and substance use care across all the studies (Fig. 2). Features were themed into one of four overarching aspects of mental health and substance use care to aid in synthesis: ‘service features’, ‘treatment features’, ‘clinician features’ and ‘additional supports’. The following sections describe the statistically significant preferences of features that were included in four or more studies, as features identified in fewer studies did not provide sufficient evidence to synthesise. Preferences of features were not presented separately for service users and service providers due to the small number of studies in the latter group. However, where relevant, the participant group is noted in the presentation of findings. All features identified in each theme and their associated preferred option can be found in Table 4.

Fig. 2
figure 2

A stacked column chart showing the number of studies that included each of the 41 unique attributes

Full size image
Table 4 Preferences for features included in the studies
Full size table

Service Feature Preferences

Eighteen features were identified that aligned with the theme of service features. All studies included at least one of the features in this theme, with 10 features included in four or more studies.

Service Location: Preferences varied regarding the location of service delivery, with some studies reporting a preference for clinical settings (Bastien et al., 2021; Becker et al., 2016; Cunningham et al., 2017; Dwight-Johnson et al., 2010; Hawke et al., 2021b; Tauscher et al., 2023) and others reporting a preference for the home environment (Batterham & Calear, 2017; Becker et al., 2016; Cunningham et al., 2017; Dwight-Johnson et al., 2013). Becker et al. (2019) found that participants preferred community-based offices, over options such as their home, family doctor’s office or hospital. Demographic factors influenced location preferences in some studies. In a United States (US) study, Dwight-Johnson et al. (2013) found that men of Mexican origin preferred their home whereas White non-Hispanic men had no significant preference. Becker et al. (2016) reported that participants who were predicted to access conventional services (i.e., mostly current patients/family members, male, ≤ high school education) preferred clinics or hospitals while those who were predicted to access convenient services (i.e., most clinicians, female, education > high school) preferred their home. Lastly, Cunningham et al. (2017) found that university students with lower intent to use face-to-face mental health services (more likely to be men) preferred their home, compared to university students who wanted services with alternative (e.g., exercise, diet) treatments offered (more likely to be women) where a campus student health centre was preference.

Delivery Mode and Format: Eleven studies identified preferences for modes of delivery (Bastien et al., 2021; Batterham & Calear, 2017; Becker et al., 2016; Cunningham et al., 2017; Klein & Cook, 2010; Lau et al., 2012; Lokkerbol et al., 2019a,b; Phillips et al., 2022; Phillips et al., 2021; Tauscher et al., 2023), with most reporting a preference for in-person delivery. Two studies (Lokkerbol et al., 2019a,b) identified a preference for a combination of in-person and tech-enabled delivery. Only one study (Lau et al., 2012) found a preference for tech-enabled delivery (i.e., via telephone) over in-person, in a sub-group of participants who favoured individual therapy. Four studies reported on the format of service delivery, which focused on e-mental health programs (i.e., web-based psychoeducation materials and/or videos) (Batterham & Calear, 2017; Klein & Cook, 2010; Muntingh et al., 2019; Phillips et al., 2021). Two Australian studies found that participants from the general population preferred an informative text format for e-mental health programs (Batterham & Calear, 2017; Klein & Cook, 2010) while one study found that participants recruited from the general population in Germany preferred a video format (Phillips et al., 2021). The fourth study did not find a significant preference for delivery format among Dutch patients with a remitted depressive or anxiety disorder (Muntingh et al., 2019).

Frequency and Timing: Seven studies investigated session frequency (Bastien et al., 2021; Dwight-Johnson et al., 2013; Katz et al., 2020; Lokkerbol et al., 2019a,b; Muntingh et al., 2019; Tauscher et al., 2023), with most finding either no significant preference, or a preference for weekly sessions. Muntingh et al. (2019) found that participants preferred quarterly sessions with a professional, but this was supplemented by a weekly virtual self-guided mental health program. The only study to focus on smoking cessation found veteran outpatients preferred five or more follow-up sessions after an initial session, compared to four or fewer sessions (Katz et al., 2020). Timing of appointments was reported in five studies (Becker et al., 2016, 2019; Hawke et al., 2021a, b; Lau et al., 2012), with mixed preferences (e.g., no appointments needed, weekday afternoons). However, services offering a wider range of appointment times were preferred by participants in two studies (Hawke et al., 2021a, b). Lastly, wait time was included in eight studies (Becker et al., 2016; Cunningham et al., 2017; Hawke et al., 2021a, b; Lau et al., 2012; Lokkerbol et al., 2019a,b; Tauscher et al., 2023), with all except one reporting a preference for no wait time.

Convenience: Six studies reported on convenience of appointments (Dwight-Johnson et al., 2004, 2010, 2013; Hawke et al., 2021a, b; Katz et al., 2020), with most finding a preference for including optional telephone appointments, bus passes, e-health service options and online booking options to make attending appointments easier. Alternatively, Katz et al. (2020) reported a negative preference for certain convenience strategies, such as telephone counselling and unscheduled counselling on request, compared to scheduled, in-person counselling in a Veterans Affairs clinic.

Other: Six studies included features related to extra treatment supports (Batterham & Calear, 2017; Becker et al., 2016, 2019; Dwight-Johnson et al., 2013; Hawke et al., 2021a, b), with all reporting a preference for services that offered treatment for physical or substance use problems in addition to mental health challenges. Service engagement features were included in five studies (Becker et al., 2016, 2019; Cunningham et al., 2017; Hawke et al., 2021a, b), with all reporting a preference for services that educate the community about mental health or involve people with lived experience in the design and feedback process. Cost was included as a feature in four studies (Dwight-Johnson et al., 2010, 2013; Phillips et al., 2021; Tauscher et al., 2023). All studies reported a preference for lower cost services, with the highest preference for free services.

Treatment Feature Preferences

Nine features were identified that aligned with the theme of treatment features. All studies, except one (Bastien et al., 2021), included at least one of the features in this theme, with four features included in four or more studies.

Treatment Type: Seven studies reported on preferences for treatment type (Becker et al., 2019; Cunningham et al., 2017; Dwight-Johnson et al., 2004, 2010, 2013; Klein & Cook, 2010; Muntingh et al., 2019). Most studies found that participants preferred a combination of counselling and medication or their choice of treatment (i.e., medication, counselling, or both) (Becker et al., 2019; Cunningham et al., 2017; Dwight-Johnson et al., 2004, 2010). Dwight-Johnson et al. (2013) reported that men of Mexican origin preferred counselling whereas White non-Hispanic men preferred medication. In another study (Klein & Cook, 2010), participants who favoured eHealth preferred an internet-based program with therapist-assistance, while those that did not favour eHealth preferred medication. It should be noted that this study did not present traditional in-person counselling as a specific option. Muntingh et al. (2019) was the only study not to find a significant preference for treatment type. However, this study compared four different psychotherapies rather than comparing counselling and medication.

Treatment Format: Six studies reported on treatment format (Becker et al., 2019; Dwight-Johnson et al., 2004, 2010; Lau et al., 2012; Lokkerbol et al., 2019a,b), with the majority finding a preference for individual sessions over group sessions. Becker et al. (2019) reported that one group of participants (mostly clinicians) preferred small group sessions alongside individual sessions, while the other group of participants (mostly patients) preferred individual sessions only. Dwight-Johnson et al. (2010) reported no significant preference for treatment format.

Other: Treatment effectiveness was reported by eight studies, with all finding a preference for more effective treatment (Becker et al., 2016; Cunningham et al., 2017; Katz et al., 2020; Lau et al., 2012; Muntingh et al., 2019; Phillips et al., 2021, 2022; Tauscher et al., 2023). However, two studies reported a preference for services that other people with mental health problems found helpful, over those supported solely by research evidence or clinical opinion (Becker et al., 2016; Cunningham et al., 2017). Four studies (Becker et al., 2019; Hawke et al., 2021a, b; Katz et al., 2020) reported on collaborative decision-making, with all reporting a preference for service providers to work collaboratively with the patient in shaping the treatment plan.

Clinician Feature Preferences

Five features were identified that aligned with the theme of clinician features. Nine of the 19 studies included at least one of the features in this theme, with two features included in four or more studies.

Professional Background: Five studies reported a preference for a psychologist or psychiatrist, compared to a mental health nurse, social worker, general practitioner, or (peer) counsellor (Becker et al., 2016, 2019; Cunningham et al., 2017; Dwight-Johnson et al., 2013; Klein & Cook, 2010). One study (Katz et al., 2020) reported a preference for a clinical counsellor (e.g., nurse), rather than a non-clinical counsellor (e.g., health coach) or peer counsellor (coaching by another veteran). Participant demographics introduced some variability in preferences across studies. For example, Dwight-Johnson et al. (2013) found that men of Mexican origin preferred a psychiatrist yet identified no significant preference in White non-Hispanic men. Becker et al. (2016) noted that participants who favoured traditional mental health care preferred a psychologist or psychiatrist while participants who favoured convenient mental health care preferred a mental health nurse. A subsequent study by the same team reaffirmed this, finding that one group of participants (mostly mental health professionals) preferred that mental health treatment for service users was provided by a mental health nurse, while another group (mostly patients) preferred that mental health treatment for service users was provided by a psychologist or psychiatrist (Becker et al., 2019). Lastly, Cunningham et al. (2017) found two distinct groups of preferences for provider amongst university students; those that preferred a psychologist or psychiatrist, and those that preferred a peer counsellor. The former group was characterised by hesitation to access in person services, whilst the latter group were characterised by favouring services with alternative treatments (e.g., exercise, diet).

Cultural Background: Six studies identified preferences for the cultural background of clinicians (Becker et al., 2019; Becker et al., 2016; Cunningham et al., 2017; Dwight-Johnson et al., 2013; Hawke et al., 2021a, b), with all but one (Dwight-Johnson et al., 2013) reporting a statistically significant preference. Five studies found a preference for culturally sensitive services or the option of talking to a service provider from their own cultural background. However, two studies identified specific exceptions based on participant sub-groups. Cunningham et al. (2017) found that university students hesitant to access in-person services had no significant preference related to the cultural background of the service provider. Hawke, Thabane, Wilkins, Hawke et al. (2021b) found that caregivers who prioritised a high level of involvement in a young person’s care preferred that the cultural background of service providers not be a determining factor.

Additional Supports Preferences

Nine features were identified that aligned with the theme of additional supports. Ten of the 19 studies included at least one of the features in this theme, with only one feature included in four or more studies.

Family/Caregiver Involvement: Six studies (Becker et al., 2016, 2019; Dwight-Johnson et al., 2010, 2013; Hawke et al., 2021a, b) reported a preference for involving families/caregivers in the treatment process if the patient chooses.

Discussion

Virtual care delivery offers many benefits to users of mental health and substance use services; however, it also has many barriers to its use. A user-centred approach which identifies the preferred features of telehealth services can help to ensure that telehealth is implemented in a way that will be used effectively and by all. Although there are specific advantages to telehealth, knowledge about user preferences for virtual care more broadly can help to inform this user-centred approach.

This systematic review sought to identify what factors influenced preferences for virtual service delivery in mental health and substance use settings. We found no studies in which the primary research question was to investigate how different service features influenced preferences for different models of virtual care. However, nineteen studies met our eligibility criteria as they examined some component of virtual care (e.g., included a modality attribute) as part of investigating preferences for other care models (e.g., different models of treatment). Features identified in the studies were grouped into four thematic categories: service features, treatment features, clinician features, and additional supports. Although the aim was to understand preferences for virtual care delivery, to be included in this review studies only had to have at least one feature related to virtual care – they were not required to include, or report on, a direct comparison of in-person care with virtual care. As such, only nine of the 41 features identified were related to virtual care delivery. Most studies focused on treatment features and only included one feature related to virtual delivery, for example, as an alternative to in-person care. These studies typically reported the influence the virtual modality had on overall preferences or likelihood of uptake (e.g., offering telephone appointments increased treatment acceptance), but not how the other features influenced preferences or uptake of virtual care. These studies can thus provide information on the overall preferences for (or against) virtually provided care, but they do not provide information for service providers seeking to design a virtual care service that minimises the specific barriers associated with virtual care delivery and predict its uptake based on different service features. Four studies examined preferences for online mental health programs (Batterham & Calear, 2017; Klein & Cook, 2010; Muntingh et al., 2019; Phillips et al., 2021), but focused on self-help tools without clinician contact. Thus, whilst the preferences for additional e-health supports are covered, the specific features of virtual care services in mental health and substance use settings remain largely unexplored in preference elicitation studies.

Within the theme of service features, most studies reported a preference for in-person services, delivered at home or in a clinical setting once per week with no wait time. While some studies identified a preference for receiving treatment at home, compared to clinical locations, this preference was dependent on demographics factors, such as gender and cultural background. Although one of the benefits of virtual care is its ability to be ‘provided at home’, it is important to note that the finding from this review of a preference for home-based mental health treatment does not necessarily equate to a preference of virtual care because this finding was independent from modality. Lower education levels were associated with less preference for virtual care. Potential reasons for this may be that people with lower levels of education are less aware of digital technology, more hesitant to use health technology, and less likely to have access to digital technology (Lee et al., 2022).

Within the theme of treatment features, most participants preferred a combination of individual counselling and medication (or a choice of the two) with an emphasis on treatment effectiveness. Preference for treatment type was dependant on cultural background, with men of Mexican origin preferring counselling and White men preferring medication in a US-based sample (Dwight-Johnson et al., 2013). Treatment preference and culture have been linked in previous literature, with some studies identifying a lower preference for treating mental illness with medication among certain cultural groups, such as Latin Americans and African Americans (Lee et al., 2021; Vázquez et al., 2021). This suggests that perhaps the dominant medical model does not always fit with different cultures, and therefore, a model of virtual care which strives to replicate traditional face-to-face care might fail to meet specific cultural needs. As such, future implementations and user-centred designs of telehealth should ensure that different cultural groups and their preferences are incorporated into the model of care.

Within the theme of clinician features there was tendency for demographic factors of the service user to influence cultural background and profession preferences about the service provider. There was a preference for treatment to be provided by a psychologist or psychiatrist with the same cultural background as the service user, and different groups of people (e.g., Mexican vs. White men) reported differences in preferences for provider profession. Further, preferences regarding profession were linked to preferences for other features, such that participants who favoured convenient (i.e., telehealth) mental health care preferred a mental health nurse whereas those who liked traditional (e.g., in-person) mental health care preferred a psychologist or psychiatrist (Becker et al., 2016). These findings highlight that choice of provider could be used to ensure equal uptake of telehealth across different groups, however this is limited by the availability and diversity of the providers. Future research could investigate if increased implementation of virtual care across more services can help to address these availability issues, and whether the choice of service provider is sufficiently important to service users to overcome other barriers to virtual care. At the same time, there may be a need to educate providers to be more culturally responsive and educate users of the advantages of accessing more readily available providers (e.g., mental health nurses) whilst waiting for more traditional but in demand providers (e.g., psychiatrists, psychologists).

Within the theme of additional supports, most features were only included in one or two studies each and were therefore excluded from the synthesised results. There is thus scope for future research to explore whether any of these miscellaneous additional supports consistently found to be important. One feature included in several studies was a preference for family/caregiver involvement in treatment planning. A recent systematic review shed light on the benefits and drawbacks of family involvement in mental health treatment (Cameron et al., 2022). Family involvement was found to provide support, give comfort, and help monitor symptoms for people receiving mental health treatment. However, at times patients were excluded from discussions about their mental health, or family members lacked mental health literacy, hindering patient treatment and recovery.

Overall, our findings show that there is no ‘one size fits all’ approach for mental health service design and delivery. We found a preference for in-person care over virtual care delivery, whether it be face-to-face or with a virtual component. However, service user and provider preferences for specific, different virtual models of care remain unknown. None of the included studies examined preferences for virtual care in populations where it may be most useful, such as people living in rural areas, due to the burden associated with traveling for face-to-face services. The issues regarding virtual care access in these populations has been explored in other research outside the scope of this review (Mseke et al., 2023). Given the advantages of preference elicitation methods in user-centred design, it is important to implement a preference elicitation study that identifies target population preferences before implementing a virtual care service.

Limitations and Directions for Future Research

This review, and the studies included in this review, are not without limitations. Most studies conceptualised virtual care as an adjunct (e.g., additional app-based support work), rather than an optional substitute to in-person care due to infection, difficulty travelling etc. Additionally, while the findings help to determine people’s preferences for one modality over another, they provided little information on which features of mental health and substance use services would increase uptake of telehealth care, or to what extent. While many qualitative studies have examined the facilitators and barriers to digital health care, preference elicitation methods such as DCEs allow researchers to determine the trade-offs that people make between different sets of preferred features. The experimental design and statistical analysis can enable conclusions, such as `people will switch from preferring in-person to virtual care when it decreases their wait time for an appointment by one week’. These types of conclusions can be easily utilised in implementation frameworks and feasibility assessments of new services. As such, it is important that future preference elicitation studies are designed and analysed such that they explore the preferences for, and predicted utilisation of, different models of telehealth as a supplement (and where appropriate, substitute) for traditional in-person care.

Almost 80% of the included studies either had a sample that that did not compare to their target population, or the authors did not assess whether they had recruited a representative sample. Most samples were skewed towards women, people with higher education levels, and people from White, English-speaking backgrounds. Consequently, the study findings are not representative. Future preference elicitation studies should aim for more representative, un-biased samples by using targeted recruitment methods, rather than recruiting from the general population online. We were not able to compare preferences between service users and service providers within this review, as we only found four studies on providers. Service users and service providers likely have different preferences, which is important to understand when designing and implementing a service so that both groups accommodated. Our review is limited by not being able to draw any conclusions about these groups’ preferences, so another avenue of future research is to compare these groups more directly.

Lastly, since most of the included studies were DCEs, the results presented in this systematic review may not be indicative of real-world utilisation of these services. Most of the DCEs did not include an ‘opt out’ option (i.e., the ability to choose neither option). Consequently, we do not know whether people would choose the said service option if presented with it in real-life circumstances; a dilemma discussed at length in the DCE literature (Determann et al., 2019; Lancsar & Donaldson, 2005; Quaife et al., 2018). Future DCEs in this area should include an ‘opt out’ or ‘status quo’ option, so that the hypothetical preferences (i.e., the DCE results) are more reflective of real-world preferences, and crucially, utilisation.

Conclusion

To our knowledge, this is the first systematic review to synthesise preference elicitation studies of mental health and substance use services with a virtual care element. Our results suggest that people are more likely to access a mental health or substance use service when it is convenient, affordable, effective, personalised, and face-to-face. Most studies focused on preferences for different mental health treatment programs, where the programs included some form of virtual element (in person care vs. phone call). No studies directly investigated preferences for, or predicted utilisation of, different modalities of services (telehealth vs. traditional in-person care). Thus, our review highlights the need for future preference elicitation studies to specifically investigate which factors influence people’s decision to access telehealth services in both mental health and substance use settings.