We searched the following electronic bibliographic databases of published studies: Embase (Ovid), MEDLINE (Ovid), PsycINFO (Ovid), Health Management Information Consortium, HMIC (Ovid), and Epistemonikos between August 29 and September 2, 2023. We also checked the reference lists of retrieved papers for other relevant materials and searched Google to identify gray literature such as reports and policy documents. An information specialist conducted the search with terms based on MeSH and text words from key papers that were identified during scoping. The review followed an unpublished protocol and was not registered.

The search was divided into three sets of terms targeting the expected types of studies that (1) aimed to develop a taxonomy (eg, terms including taxonomy, typology, categorization, and group), (2) provided a narrative review of the field of digital mental health (eg, terms including field, future, advance, emerging, and state), and (3) completed a systematic review grouped by type of technology (eg, terms including review, synthesis, and analysis). More details on the search strategy are available in Multimedia Appendix 1.

Studies were considered eligible for inclusion if they provided an approach to categorizing or grouping digital technologies according to their characteristics and had a specific focus on mental health. The digital technologies of interest in this review are software that aims to have a direct or indirect impact on an individual’s health outcomes through use by the individual or a professional. We did not include digital technologies that aim to improve population-level care or improvements in efficiency. This is in line with the definition used in other frameworks [10,11]. Both qualitative and quantitative studies were also included if they met these inclusion criteria. Studies were not included if they extracted characteristics of technologies but did not provide clear groupings. We only included English language studies.

One team member (LA) completed the initial screening of studies by title. A second team member then screened potentially relevant studies by title and abstract (GH). For eligible studies, data were extracted using a standardized data extraction form by 1 team member (GH) and reviewed by a second team member (FE).

Information extracted included the aim and scope of the study, type of study, and details of domains used to group technologies. Initially, domains used to group technologies were extracted according to the name provided by the authors. Domains were then reviewed for similarity and pooled if appropriate. The assessment of whether a domain was present within a study was made based on the prominence given by the authors. Domains were included where a list of types of technology was provided within a paper, where a characteristic was provided as a subheading, or where a characteristic was used to define a subgroup within an analysis. For example, one paper describing a framework for mental health in primary care provides a list of domains that can be used to categorize DMHTs [14]. Similarly, in a systematic review on interventions for university students, DMHTs are divided into domains in the narrative synthesis of results [20].

The findings of the rapid review and challenges associated with categorizing DMHTs were considered by an expert working group during a web-based meeting in November 2023. This is a group that has been convened to meet and provide guidance for the duration of a Wellcome-funded research project [21].

The group was formulated to provide expertise on digital mental health from across the field. Experts were selected based on their experience of designing and researching DMHTs, using DMHTs in clinical practice, and managing the deployment of DMHTs in health services. The group comprises representatives from the Medicines and Healthcare products Regulatory Agency and NICE (including members of the project team and other experts from these organizations), representatives from NHS England, a variety of health professionals with experience of mental health conditions (eg, clinical psychology, nursing, psychiatry, and general practice), academics, representatives from health innovation networks, and lived experience advisers.

During the meeting, the group was asked to consider a series of questions through facilitated discussions. These questions related to (1) the value of developing a framework to categorize DMHTs, (2) the practical purposes of a framework, and (3) the drawbacks or challenges of developing a single framework for these purposes. Emerging themes from these discussions were identified and considered by the research team.

Within smaller breakout groups, members were then asked to consider (1) which of the domains identified within the review would be most important to include in a taxonomy, (2) whether additional domains that captured other characteristics of DMHTs were needed, and (3) whether descriptions of the domains reflected anticipated content and whether adaptations were needed to these descriptions. Based on this, groups were asked to define which domains should be included within a new framework for categorizing DMHTs and how they should be conceptualized. The domains proposed for inclusion and other considerations were collated to support the development of the new framework.

To support these discussions, the group were provided a summary of the rapid review methods, and results described within this paper, along with a sample of the publications included within the review.

The search returned a total of 3603 unique results from databases and other sources (Figure 1). Of these, 3544 were excluded by title and abstract not adequately reflecting the subject matter, and 59 of these full-text papers were subsequently reviewed. A further 3 studies that were not identified by the search were included after hand searching references and included in the full-text review. After review, 23 papers were considered eligible for inclusion.

Most studies were systematic reviews of the effectiveness or quality of technologies and grouped technologies to support analysis (8/23, 35%). The remaining studies either provided groupings of types of technologies as part of commentaries or narrative reviews of the field of digital mental health (7/23, 30%), aimed to develop their own taxonomy (5/23, 22%), or explored the implementation of DMHTs in mental health services with qualitative methods (1/23, 4%). In total, 1 (4%) study was a consensus statement on the use of DMHTs in the United States, and one further was public information for Canadian people on how to select appropriate technologies.

Most studies included DMHTs across a range of conditions and stages of the treatment pathway (15/23, 65%), but some had a more specific focus. This included anxiety and depression (2/23, 9%), comorbid mental and physical conditions (1/23, 4%), mental health during COVID-19 (1/23, 4%), youth mental health or university students (1/23, 4%), management in primary care (1/23, 4%), and ante- and postnatal mental health (1/23, 4%).

The majority of studies did not define how domains used to categorize technologies were generated (11/23, 48%), although some outlined underlying methods based on previous taxonomies (4/23, 18%) or coding characteristics of relevant technologies in systematic reviews (6/23, 26%). One study reported using workshops to develop consensus on appropriate groupings, and one was based on themes arising from qualitative interviews.

Ten domains were identified across the included studies in line with the method outlined earlier. These are condition, setting, platform or system, function, subfunction, professional input, type of communication, type of intervention, sophistication, and consumable resource use. Definitions for these domains are provided in Textbox 1.

Studies included domains on professional input (11/23, 48%), type of platform or system used (10/23, 43%), target condition (6/23, 26%), function (5/23, 22%) and subfunction (2/23, 9%) of the technology, type of intervention (4/23, 17%) with each of these was focused on the type of psychological therapy, the setting (4/23, 17%), timing of communication with professionals (2/23, 9%), whether a consumable resource was used (2/23, 9%), and technical sophistication (1/23, 4%). Details on the domains included within each study are shown in Table 1.

Based on the findings of the literature review and reflections from the working group, we developed a conceptual framework for DMHTs. The conceptual framework includes 8 domains that represent the characteristics of DMHTs (Figure 2). They each represent important characteristics of DMHTs that are likely to influence the benefits they can deliver for individuals and the wider system and risks that will need to be mitigated during use.

The proposed conceptual framework also uses a “tagging” approach, where multiple attributes within a single domain could be attributed to a technology (eg, the population targeted by a DMHT), rather than seeking to develop an approach using “buckets” to assign a particular technology to a discrete group based on a characteristic.