In this paper, we completed a cross-sectional analysis of evidence supporting DMHTs included in NICE evaluations to examine the strength of evidence to support decision-making both at the level of individual studies and across the package of evidence from multiple studies supporting DMHTs. We also aimed to identify common evidence gaps that, if addressed, could provide more certainty on the impact of using DMHTs.

This study did not involve human participants and therefore, ethics approval and informed consent were not required.

We used a cross-sectional analysis based on NICE evaluations for DMHTs. We identified all NICE evaluations relating to DMHTs published before September 2025. These were identified based on internal knowledge and a review of published guidance on the NICE website [15]. We defined digital mental health as any topic that related to the use of software to support triage, diagnosis, treatment, or ongoing management of mental health, behavioral, and neurodevelopmental conditions. We excluded topics related to the management of physical health conditions, even if these related to mental health populations.

From each NICE evaluation, we identified included DMHTs and reviewed committee documentation to identify studies that provided supporting evidence for each of these technologies. As this review is focused on studies relating to effectiveness and cost-effectiveness, we excluded studies if they reported only patient engagement or experience and did not report outcomes relating to clinical symptoms, health-related quality of life (HRQoL), functioning, well-being, or cost and resource use. For diagnostic technologies, we also excluded studies that provided information only on diagnostic accuracy due to differences in assessing quality for these study designs, including the need to consider appropriate reference standards, measures of accuracy, and performance at threshold settings.

For each identified study, we extracted information on a series of items relating to study quality. The items were selected because they provide a broad overview of the quality of research methodology and reporting and are commonly included across risk of bias tools and preferred reporting items checklists [16,17]. These items have also been included in similar reviews of evidence supporting regulatory approval [18].

Data on these items were extracted from publicly available supporting documentation on the NICE website (eg, committee papers, assessment reports) and from published versions of supporting studies. Where information was contained in multiple sources, we relied primarily on the original publication to ensure accuracy. We also recorded whether committee documentation mentioned ongoing studies that could provide additional evidence for a technology after the conclusion of the evaluation. For evaluations for early use, we also extracted information on additional evidence that was essential or could support further review from Evidence Generation Plans.

We report the number of identified NICE evaluations for DMHTs, the number of DMHTs included within these evaluations, and the total number of included studies. For each item relating to study quality, we categorized each study according to relevant characteristics and report the number and proportion of studies with these relevant characteristics. We also reported the number of studies supporting each identified DMHT and the number and proportion of DMHTs that were supported by evidence indicating higher quality or relevance to decision-making across a subset of items. These items were based on previous research, which has examined the quality of studies supporting regulatory submission [18] and align with consensus-based reporting guidelines [16]. Finally, we report evidence gaps identified in NICE evaluations where DMHTs were for early use and had recommendations for use with further evidence generation.

We reviewed nine NICE evaluations on DMHTs (Table 1). These covered a broad range of conditions, including anxiety, depression, psychosis, insomnia, attention deficit hyperactivity disorder (ADHD), and tic disorders.. They also covered a range of uses, including triage, diagnosis, treatment, and ongoing management and support. Recommendations were available on 30 DMHTs. Four of these technologies were included in multiple evaluations for different populations. Two recommendations were made for routine use, 20 recommendations were made for use while further evidence is generated, and 20 recommendations were made for use in research only. Recommendations on two technologies have been withdrawn since initial publication, as the technologies are no longer available in England.

We identified a total of 78 studies related to 25 DMHTs. Seven of these studies provided support for technologies across two evaluations. Of the studies, 72 were publicly available. Six studies were not publicly available but were provided in confidence during NICE evaluations, and some information on these was reported in publicly available supporting documentation. Two technologies were also supported by confidential responses to requests for information during the NICE evaluation. These responses may include additional evidence that is not publicly available and was not available to us, so could not be included in this review. Summary characteristics of included studies are described in Table 2, Table 3, and Table 4.

The majority of studies reported conflicts of interest (n=68, 87.2%%) with funding sources from the public sector (n=38, 48.7%), industry (n=20, 25.6%), third sector (n=11, 14.1%), and insurers (n=1, 1.3%). Around half of the studies were registered or had a published protocol (n=37, 47.4%). Half of the studies recruited participants in England (n=43, 55.1%) and most used multi-center designs (n=52, 66.7%). Just over half of the studies were randomized controlled trials (n=42, 53.8%), with the remainder using non-randomized controlled (n=11, 14.1%) and non-comparative (n=20, 25.6%) designs. Half of the studies had no blinding while around a third used single blind designs with allocation status not known to outcome-assessors (n=23, 29.5%). For comparative studies, the control condition was evenly split between waiting list (n=16, 20.5%), standard care (n=21, 26.9%), and active controls (n=21, 26.9%) with one additional study reporting using a placebo control (n=1, 1.3%). Studies using active and placebo controls used a variety of approaches that provided a varying level of therapeutic content. In some cases, standard care and waiting list controls also received therapeutic content as part of their usual care. Study interventions were delivered both as part of (n=43, 55.1%) and outside of routine care (n=27, 34.6%).

For most studies, the primary endpoint was identified or could be determined due to its prominence in reporting (n=60, 76.9%). A total of 54 measures were included as primary endpoints. Half of the studies included an outcome relating to HRQoL, functioning, or well-being (n=43, 55.1%), and 34 different measures were used for these outcomes. A minority of studies included preference-based HRQoL measures that can support the generation of health utilities (n=15, 19.2%) or outcomes relating to impact on resource use (n=21, 26.9%), both of which are valuable for use in economic evaluations. Only a minority of studies reported on adverse events (n=34, 43.6%). Details on all included primary endpoints, and health-related quality of life, functioning, and well-being measures are available in Multimedia Appendix 1.

The median number of enrolled participants was 180 (IQR 300, range=10 to 67,468), with a median of 129 participants analyzed (IQR 237, range 10 to 948,294). The rate of dropout varied from 0% to over 50% of participants, although 12 studies did not report sufficient data to determine dropout from enrolment to analysis. Where the primary endpoint was identified, it was met in the majority of studies (n=60, 76.9%). The majority of studies reported on basic demographic information relating to participant age (n=63, 80.8%) and gender (n=60, 76.9%) but fewer studies reported on participant ethnicity (n=31, 39.7%).

For five technologies, no relevant supporting evidence was identified during the evaluation. Each of these technologies was recommended for use in research only. 25 technologies had relevant supporting evidence, with most of these supported by one (n=12, 40%) or two (n=4, 13.3%) studies, with a range of between 1 and 21 studies. Ongoing studies were identified for 15 technologies. These were a range of designs, including randomized controlled trials, real-world evidence, and implementation studies.

Most technologies had at least one study that used elements of a high-quality study design (Table 5). Over half of these technologies were supported by studies with randomized methods (n=15, 60%), and this was combined with other elements of high-quality study design, including notably blinding of outcome assessors (n=14, 46.7%%). Most studies also included at least one element of study design that could support NICE decision-making on the use of technologies in the NHS in England. The most common elements were inclusion of outcomes relating to HRQoL, functioning, or well-being (n=19, 76%) and recruitment in England or the rest of the UK (n=18, 60%).

Eight technologies were supported by at least one study with all elements of a high-quality study design and relevance to decision-making for use in the NHS in England. Three technologies were supported by multiple studies that each included all of these elements. Seven technologies included all of these elements but were split across 2 or more different studies.

Evidence generation plans were available for seven NICE evaluations for early use. These evidence generation plans outlined seven areas where additional evidence was essential (Table 6). The most common evidence gap that was judged as essential was on costs associated with delivering DMHTs and impact on resource use (n=6). This was followed by further evidence on effectiveness compared to relevant comparators (n=5), adverse events (n=4), appropriate outcomes (n=4), HRQoL (n=2), user experience and engagement (n=1), and quality of data from digital sources (n=1).

Four areas where additional evidence could support decision-making were also mentioned in evidence generation plans. Two of these were for evidence gaps that were essential for other evaluations, including on user experience and engagement (n=5) and HRQoL (n=4), and two related to additional evidence gaps on effectiveness for specific population subgroups (n=2) and on determining the population that would benefit most from the DMHT (n=1).

Evidence generation plans also provide advice on the most appropriate study designs to support further decision-making. These included robust, well-designed randomized controlled trials and approaches using real-world data with statistical methods that can adequately control for confounding variables (eg, propensity score matching, controlled interrupted time series), which were all deemed to be feasible for these technologies. For some technologies, it was suggested that some evidence gaps could be addressed through additional analysis of already available trial data.

In this study, we used a cross-sectional analysis to examine evidence supporting NICE evaluations for DMHTs. This analysis included nine NICE evaluations with recommendations for 30 DMHTs. These DMHTs were supported by a total of 78 studies.

We found that studies evaluating DMHTs often use high-quality methods that provide valuable information on effectiveness and cost-effectiveness. A number of these studies have included elements that are reportedly difficult to incorporate into evaluations of digital health technologies, including randomisation and blinding [28]. Our findings show that there is high-quality evidence available for DMHTs covering various areas of importance and that this evidence accrues across multiple studies, which build to provide a package of evidence to support decision-making. Although there remain key evidence gaps which could be addressed at earlier stages of evidence generation or during early stages of adoption in health services. In some cases, the evidence developed across these studies was sufficient to provide recommendations for routine use in the NHS in England. In other cases, evidence suggested that DMHTs could play a role in health services but there remained uncertainties about their impact and there are a series of common evidence gaps that these DMHTs needed to address prior to further review.

Despite finding supporting high-quality evidence for many DMHTs, we also found that some DMHTs that could help address health system priorities do not appear to have evidence to support claims of their effectiveness and value for money. There needs to be a coordinated focus from stakeholders across the health system, including developers, health professionals, and policymakers, to ensure that DMHTs are supported by appropriate evidence and that they are not adopted into health services until sufficient evidence is available.

Our findings on study characteristics and common evidence gaps align with previous studies on evidence for DMHTs [3,6,7]. For example, our findings suggest that despite some increases in reporting of adverse events over time, improving monitoring and reporting of adverse events remains a priority [7,10]. This review does provide a more nuanced picture of reporting of adverse events, though, and suggests that developers may focus on recording and reporting adverse events in some but not all of their supporting studies. This approach does not appear to be in line with best practice for health research [29] or with regulatory requirements [30] but shows that valuable information can be found by reviewing the full evidence package for a DMHT. Similarly, evidence on user experience and engagement has previously been cited as a challenge for DMHTs with high levels of discontinuation [11], and our findings suggest that this is a common evidence gap for DMHTs evaluated by NICE.

We also identified additional evidence gaps that have had less discussion in previous research but are important to support decision-making. In order to assess the impact of introducing DMHTs into health services, evidence needs to focus on the benefits compared with currently available care. This was identified as a key gap, as comparative evidence was either not available or studies used comparators that did not reflect standard care in England. Studies did not commonly include preference-based HRQoL measures. These measures can be used to generate health utilities, and their omission causes challenges in assessing cost-effectiveness. This could be addressed by including appropriate measures such as the EuroQol-5 Dimensions (EQ-5D), which is NICE’s preferred instrument for capturing HRQoL, in early evaluations and subsequent trials.

Evidence on the cost of delivering DMHTs and their impact on short- and long-term resource use was the most common evidence gap and was highlighted as an essential area to address. Developers should be clear about licensing and pricing models, and there needs to be an additional focus on developing evidence on the impact on contacts with health professionals, hospitalization, and other relevant outcomes. For some newer DMHTs that are able to collect and analyse large amounts of data, there was also uncertainty about the quality of this data and whether systems were being designed to capture a range of important variables for use within analysis.

Our findings also highlight that challenges that are present in other areas of mental health research are also present for DMHTs. As with previous studies, we found a lack of consistency in the outcomes used for primary endpoints in DMHT studies, as well as a wide range of possible outcomes for measuring HRQoL, functioning, and well-being [9]. Some level of variability is to be expected across different populations and different purposes. However, there was variability in studies for the same or similar DMHTs, which could have implications for assessing comparative effectiveness. Alongside this variability, we also saw a reliance on condition-specific symptom scales and more limited use of other patient-relevant outcome measures, including measures of HRQoL, functioning, and well-being, and other outcomes that matter to patients (eg, relapse, hospitalization, access to services, waiting times). Greater use of these types of outcomes could help ensure that DMHTs align with the goals of users and also address health system priorities.

Our findings have implications for future research and could help guide evidence generation for new and existing DMHTs. We identified a series of common evidence gaps that could help inform study design and could be addressed earlier during the evidence generation process. If these gaps could be addressed earlier in the process, they could increase certainty and allow a wider range of DMHTs to be recommended to support health system priorities. These findings emphasize the need for developers who are aiming for implementation in specific settings to review content from relevant HTA agencies and other decision-makers. To achieve this, there is a wide range of publicly available resources, such as the NICE Evidence Standards Framework [31], and developers may benefit from accessing expertise from advisory services, such as NICE Advice [32]. There are also similar resources that can provide information on using real-world evidence to address uncertainty [33]. We also identified that less than half of the studies supporting DMHTs were registered or had a published protocol, and this was particularly prevalent for nonrandomized studies. The NICE real-world evidence framework encourages developers to publish protocols with pre-specified analyses for studies of comparative effects [33]. This is a further practical step that could be taken to improve the transparency of evidence for DMHTs.

We also identified that there was limited or no evidence to support decision-making on the use of some DMHTs despite them aligning with areas of unmet need for the health system. For some of these DMHTs, this may be because they have been developed more recently and are actively developing evidence but were included in a NICE evaluation alongside DMHTs with more mature evidence. Developers should ensure that they generate sufficient evidence to meet regulatory requirements and to support health technology assessment [8]. This would help support consideration of the potential impact of their technologies and ensure that patients can benefit from the full range of available DMHTs, if they can be demonstrated to be safe, effective, and provide value for money.

Our study focuses on a set of NICE evaluations of DMHTs. These evaluations are prioritized by NICE because they represent a priority for the NHS in England, and it has been identified that there are at least some relevant DMHTs with sufficient evidence to make recommendations. This is a key strength of the study because it allows us to examine whether DMHTs that could support national priorities are supported by evidence. It also provides a naturalistic account of evidence available to support decision-making on use at a particular point in time. However, this approach also means there is a substantial level of selection bias in the included DMHTs and our findings may not be reflective of evidence across the broader field of digital mental health. In addition, our approach means that we have a snapshot of evidence for included DMHTs at a specific time point. DMHTs are likely to be generating evidence across their lifecycle, and this study should be revisited to explore how evidence supporting DMHTs changes and whether DMHTs are able to address evidence gaps highlighted in this study.

A further limitation is that we were reliant on evidence that was identified during NICE evaluations and is publicly available. In assessments for early use, a pragmatic approach to identification and selection of evidence is used to allow rapid assessment and to ensure applicability of evidence to the scope of an evaluation. In addition, in several cases, DMHTs were supported by unpublished data which were provided with confidentiality during NICE evaluations but could not be reflected in our findings. For both reasons, this may mean we are not able to capture all available supporting evidence. Our study also focused on evidence of effectiveness and cost-effectiveness, which play a key role in HTA. However, decision-making is also supported by information on a broader set of factors, including the potential for technologies to address unmet needs and the experiences and views of patients and health professionals. These factors could not be captured in this study.

Digital mental health technologies (DMHTs) are playing an increasing role in mental health services. To achieve their potential, DMHTs need to demonstrate that they are effective and present value for money. However, the quality of evidence for DMHTs is variable and there are concerns that evidence is not sufficient to support decision-making. Our study highlights that some DMHTs are supported by high-quality studies and that evidence to support DMHTs is likely to be developed across multiple studies. There remain key evidence gaps that could be addressed earlier in evidence generation to provide a stronger case for adoption.

Developers should ensure that their evidence generation is aligned with the expectations of decision-makers in relevant settings to reduce uncertainty in these areas. There should also be a greater focus in the field of digital mental health on generating evidence on the cost of delivering DMHTs, their impact on resource use, and on other outcomes that matter to patients, and ensuring that all of these data are compared with current and relevant care options within health services.