The evidence base for this economic evaluation comprises a multicenter, double-blind RCT of MindMySkin, currently being conducted across 3 outpatient dermatological centers in Singapore, with a target recruitment of 690 participants with psoriasis, eczema, or chronic urticaria [22]. Participants are randomized to MindMySkin or the Healthy365 app, which serves as an active control comparator. Healthy365 is a general wellness app managed by the Singapore Health Promotion Board, providing health tracking, fitness challenges, and general lifestyle recommendations without specifically targeting skin disease or psychodermatological outcomes. The use of an active control preserves blinding and helps differentiate intervention-specific effects from broader nonspecific effects associated with app use and health engagement. This is particularly important for the economic evaluation, as attributing cost-effectiveness to MindMySkin’s targeted psychobehavioral content, rather than to app engagement per se, strengthens the validity of the incremental analysis and the relevance of findings for implementation decisions. For the economic evaluation, effectiveness is assessed using the Dermatology Life Quality Index (DLQI) [23], while health utility is measured using the EQ-5D-5L [24], enabling derivation of quality-adjusted life years (QALYs) across the 32-week study period.

This study conducts a cost-utility and cost-effectiveness evaluation in accordance with the Expert Delphi Consensus for Health Economics Analysis Plans for Trial-Based Economic Evaluations [25] and the Consensus on Health Economic Criteria checklist [26].

This double-blind RCT will recruit English-speaking participants aged 16 years or older with psoriasis, eczema, or chronic urticaria across 3 study sites in Singapore. The restriction to English-speaking participants reflects the current availability of MindMySkin in English only. The age cutoff of 16 years or older is intended to ensure participants have sufficient cognitive maturity to self-manage a digital psychobehavioral intervention. Participants will be randomized in a 2:1 ratio to MindMySkin or an active control (Healthy365 [27]) and followed up for 32 weeks. The full protocol has been separately published [22].

All study sites are health care institutions operating within Singapore’s publicly funded health system, in which services are financed through a combination of patient payments, patient subsidies, and institution-side government subvention. The primary base-case analysis will adopt a Singapore health care system perspective, as recommended by the Agency for Care Effectiveness Singapore [28], whereby costs to the system as a whole will be considered. A complementary analysis will be conducted from the societal perspective, including patient out-of-pocket costs and productivity costs.

The cost analysis aims to inform the cost-effectiveness associated with the deployment and maintenance of MindMySkin as a scalable service beyond the trial setting. Intervention development costs and research costs will be collected separately but excluded from the primary analysis.

MindMySkin is hypothesized to improve HRQOL through reductions in symptom burden and disability, resulting in changes in health state utility without effects on mortality or life expectancy. The economic evaluation thus assumes no life years gained, with all changes in QALYs attributable to changes in HRQOL outcomes over time. As a supportive, nonsubstitutive intervention, MindMySkin is not expected to materially alter patterns of health care use. Productivity improvements in work, school, and daily activity are anticipated in the intervention arm as downstream effects of reduced symptom burden and improved coping. These effects will be examined in secondary analyses from the societal perspective.

MindMySkin [17] is self-administered, delivered via a mobile app, and designed to complement routine dermatological care. The content is self-paced, does not require clinician involvement, and allows users to engage at their discretion, thereby reflecting real-world use beyond the trial setting.

The comparator is the current standard of care, operationalized as an active digital control using the Healthy365 app, a nationally funded general wellness app that does not include dermatology-specific or psychobehavioral therapeutic content. Healthy365 is an existing publicly funded program with fixed sunk costs independent of this study. Therefore, no app-related costs are attributed to the control arm. In a similar vein, structured psychobehavioral interventions are not part of routine dermatological care in Singapore, and the comparator is assumed to incur no additional intervention-related costs. Participants in both arms will continue to receive standard medical management, including clinic visits, investigations, and pharmacological treatments when clinically indicated.

The within-trial economic evaluation adopts a 32-week time horizon, corresponding to the final follow-up of the RCT. This horizon captures the full period over which costs and health outcomes are prospectively observed and is appropriate for assessing short- to medium-term differences in health care use, HRQOL, and symptom burden attributable to the intervention. For planning and interpretive purposes, a longer-term horizon (up to 5 years) is considered in scenario and model-based extrapolation analyses to explore the potential implications of sustained use and scale-up. This is explicitly exploratory and not part of the base-case trial-based analysis.

The primary estimate of the potential user base for the intervention, over which fixed intervention costs are allocated, is derived using a site-based approach reflecting uptake within participating clinical settings. Each participating site will provide standardized information for a reference year (eg, 2025), including the annual number of unique dermatology outpatients and the disease distribution for atopic dermatitis/eczema, psoriasis, and chronic urticaria. Where diagnosis-specific counts are unavailable, percentage estimates will be derived from a small sample of representative clinic sessions.

Each site will estimate the proportion of patients with sufficient English proficiency, educational level, and digital literacy to engage with the intervention. To ensure consistency across sites, sampling of a fixed number of clinic sessions (eg, 5) will be used where applicable. In addition, each site will estimate the proportion of suitable patients who would realistically use the app if offered as part of routine care, informed by direct patient queries where feasible or, alternatively, by clinician judgment. This approach is anchored to patients attending the recruiting sites only and does not assume broader population-level uptake.

A second estimate of the potential user base will be derived using a platform-anchored approach to define a lower bound of potential uptake. This approach will be based on Intellect’s (the health technology platform hosting MindMySkin’s content) existing active user base and estimated prevalence of atopic dermatitis or eczema, psoriasis, and chronic urticaria. This approach assumes that only individuals who are already active users of the Intellect platform and have one of these dermatological conditions would adopt the skin-specific modules. The conservative estimation of true uptake ensures that economic benefits are more likely to be underestimated than overestimated.

Broader health care resource use will be captured using a resource use questionnaire adapted from the Core Items for a Standardized Resource Use Measure, which was developed through an expert Delphi consensus process [29]. Resource categories for the primary analysis will include domains most plausibly influenced by MindMySkin, including outpatient dermatology clinic visits, other relevant dermatological visits, emergency department attendances, inpatient admissions related to skin disease, and dermatology-related treatments and medications. Certain items and domains that were judged to be less relevant to the target ambulatory outpatient dermatology population (eg, hospice and rehabilitation care arrangements) were removed to reduce response burden, misattribution, and noise introduced by random variation in all-cause health care use over the study period, improving internal validity and avoiding inappropriate cost inflation. The adapted questionnaire covering resource categories in Table 1 is being administered at weeks 16 and 32 using a 4-month recall period.

Resource use will be self-reported in natural units, reflecting the actual volume of services consumed by each participant during the study period, and valued using institution-specific billing data and unit costs. In the absence of unit cost data, patient charges before subsidies will be used as a proxy for the cost of service provision. This is consistent with the practice in Singapore’s public health care institutions, which typically charge for services on a cost-recovery basis [30,31], and is an approach used in prior Singapore studies [32,33]. Additional studies further support the use of charges where the primary interest is the estimation of incremental cost differences between groups [34]. Medication costs will be valued using retail and institutional drug formulary prices obtained from institutional pharmacy schedules and the national drug information database.

A broader societal perspective will be considered by including patients’ out-of-pocket expenditures, costs associated with productivity losses, and monetized time costs (Table 1). Out-of-pocket expenditures and nonmedical costs (including complementary or alternative treatments and transportation) will be captured using the same self-administered resource use questionnaire, administered at weeks 16 and 32 alongside the primary health care resource use data collection.

Productivity losses will be assessed at baseline and at weeks 8 and 16 using the Work Productivity and Activity Impairment (WPAI) questionnaire [35], including the school version where applicable. Absenteeism, presenteeism, overall work or school impairment, and activity impairment will be derived using standard WPAI scoring algorithms. All participants will also report weekly hours spent managing their skin condition (eg, skin care routines, obtaining medications, and attending medical visits).

Weekly time losses, captured using a 7-day recall period, will be extrapolated over the relevant observation intervals and monetized using the national average hourly wage [36] to estimate time costs of disease management. Productivity losses associated with school absenteeism and impairment will be interpreted as near-term foregone or delayed paid-work productivity because all participants are aged 16 years or older, consistent with economic evaluation practice [37,38]. For the youth group, the national average wage for the 15- to 29-year age group will be applied [36].

Time spent using the MindMySkin app will be recorded but not costed, as engagement is expected to be brief, flexible, and unlikely to displace paid work or other productive activities, consistent with recommended practice for valuing time in digital intervention costing.

In Singapore, published cost-to-health care provider data are not available for most of the health care resources used. Hence, hospital charges are commonly used as proxies for economic costs in trial-based evaluations. As public hospitals operate on a cost-recovery basis with regulated pricing and minimal profit margins, patient charges (before subsidies) are considered reasonable approximations of the underlying cost of service from the health care system perspective.

The primary health outcome for the cost-effectiveness analysis will be the DLQI, a disease-specific HRQOL scale ranging from 0 (no impairment) to 30 (maximum impairment). This is collected at baseline and at weeks 4, 8, 16, 24, and 32.

The EQ-5D-5L will be used for the cost-utility analysis, with utilities obtained from the Singapore value set for the measure [24,39]. Four skin-related bolt-ons will also be collected [40,41] and valued using available tariffs. The EQ-5D-5L data will be collected at baseline and at weeks 8, 16, and 32. The area under the curve for the 32-week study follow-up will be used to estimate the DLQI and QALYs gained over the study period. Differences between the comparator groups in QALY estimates will be assumed to result only from changes in HRQOL outcomes, with no change in life years.

Parameter, methodological, and structural uncertainty will be addressed through a series of prespecified sensitivity analyses conducted from the perspective of the primary health care system in various ways.

Sampling uncertainty in costs and outcomes will be characterized using nonparametric bootstrapping with 10,000 replications, resampling values with replacement. Bootstrap outputs will be used to derive 95% CIs for incremental costs, incremental effects, and ICERs. Uncertainty surrounding the ICERs will be illustrated using cost-effectiveness planes, while cost-effectiveness acceptability curves will be generated to depict the probability of the intervention being cost-effective across a range of WTP thresholds.

Parameter uncertainty will be explored through deterministic sensitivity analyses in which key cost inputs are varied within empirically derived bounds. To assess the influence of extreme cost observations, winsorization at the 95th percentile will be applied, whereby individual-level costs exceeding this threshold will be replaced with the 95th percentile value. Sensitivity analyses will also examine alternative assumptions regarding intervention costs and the projected user base to reflect uncertainty in implementation scale and market pricing.

Methodological uncertainty will be examined by estimating ICERs at an intermediate time point (week 16) in addition to the final follow-up at week 32, and by applying an alternative missing data strategy using complete-case analysis instead of multiple imputation.

Structural uncertainty will be assessed by testing alternative analytic assumptions, including the inclusion of intervention development costs under in-house development scenarios and the application of lower real-world effectiveness assumptions relative to trial conditions.

The impact of alternative assumptions regarding the potential user base and intervention uptake on fixed-cost allocation and per-user intervention costs will be explored.

Heterogeneity will be assessed using a NMB regression framework, with interaction terms between treatment assignment and subgroup indicators to identify prognostic groups in which the intervention is cost-effective.

Adoption and engagement subgroups will be defined using app-derived use metrics, including overall module completion and the frequency and duration of app use. Demographic and socioeconomic subgroups will be examined by age group, education level, and employment status to assess distributional equity. Disease-related heterogeneity will be explored through primary dermatological diagnosis and baseline disease severity, including both clinician-assessed and patient-reported measures. Psychographic subgroups will examine baseline use of digital and mental health apps, resilience and self-efficacy, personality type, and the presence of comorbid anxiety or depression.

If within-trial results suggest a high likelihood of cost-effectiveness over the trial follow-up, a model-based extrapolation will be undertaken to examine longer-term costs and outcomes over a 5-year horizon using a cohort Markov model with 6-month cycles. Health states will reflect clinically relevant dermatological trajectories, including controlled disease, disease flare, uncontrolled disease, and death. Transition probabilities, cost, and utility parameters will be informed by trial data where available and supplemented by local data, published literature, and expert input. Key assumptions will be explicitly stated and explored in sensitivity analyses, while model validity will be assessed through internal consistency checks and face validity. Where trial data are insufficient to reliably parameterize a Markov model, a simplified decision tree–based scenario analysis will be used.

Ethics approval for this study has been obtained from the National Healthcare Group Domain Specific Review Board (2022/00751). As the economic evaluation is conducted as a secondary analysis embedded within the parent RCT, no separate ethics application was required. Written informed consent is obtained from all participants prior to enrollment, including consent for collection of health care resource use and quality of life data for economic evaluation purposes. For minors, parental or guardian consent is obtained alongside participant assent, in accordance with institutional requirements. All data are deidentified prior to analysis, with participant information stored on password-protected institutional servers accessible only to authorized study personnel. Participants receive reimbursement of up to SG $80 (SG $1=US $ 0.77 as of June 29, 2026) upon completion of study assessments, disbursed incrementally across the different stages of the trial.

The study incorporates the insights and perspectives of patient and public representatives, from the development and curation of the psychotherapeutic modules to providing feedback on the study design. These members will also support the analysis and dissemination of the study results. There are currently 3 patient representatives involved in this study.