The CUA was based on an RCT conducted in 2023, which enrolled patients with CRD from 3 clinical centers in South Korea. The inclusion criteria were (1) age older than 19 years, (2) diagnosis of CRD (such as COPD, ILD, lung cancer, asthma, or bronchiectasis), (3) a need for rehabilitation due to respiratory symptoms or difficulties in daily life, and (4) the ability to use a mobile app. The exclusion criteria comprised (1) cognitive impairment, (2) previous participation in pulmonary rehabilitation, (3) unstable cardiovascular disease, (4) walking difficulties, and (5) pregnancy. A total of 84 participants met the inclusion criteria and were included in the full analysis set. Participants were randomly assigned to the DTx group (n=43) or the control group (n=41) using a block randomization method. Patients were followed up every 4 weeks for a total of 8 weeks from the initiation of treatment. Baseline characteristics were collected at the time of hospital visit before the intervention initiation. Demographic variables, such as sex, age, and clinical outcomes (such as 6MWD, mMRC, CAT, SGRQ, and HADS score), were assessed. In addition, patients were categorized by CRD subtype. Among CRD subtypes, ILD included idiopathic pulmonary fibrosis, diffuse ILD, nonspecific interstitial pneumonitis, and other fibrotic interstitial pulmonary diseases, based on clinical categorization.

The DTx group participated in rehabilitation sessions 3 times a week at home and visited the clinical center once every 4 weeks for examinations. Participants received an initial hands-on training session led by a research nurse to assist with app installation and usage. Each patient was equipped with a smartwatch, and the app was installed on their mobile phones.

To improve digital literacy, particularly among older adults, several strategies were incorporated into the trial design. These included the provision of user-friendly manuals with enlarged fonts and simplified visual instructions (Figure 1), as well as continuous monitoring by clinical staff through a web-based platform during the early stages of rehabilitation. Participants who failed to meet the minimum engagement criteria—defined as missing 3 consecutive exercise sessions—were contacted and supported by the study team. Participants’ adherence to the DTx program was monitored using a web-based platform. Adherence was calculated as the ratio of the actual number of completed sessions to the expected number of sessions during the interval between clinic visits. In addition, rehabilitation activities were documented after each session, which promoted self-awareness and supported improved adherence. Personalized exercise prescriptions were developed based on each participant’s baseline 6MWD test results, considering individual functional capacity. In addition, the DTx was integrated with a smartwatch to provide real-time feedback based on heart rate monitoring during exercise, supporting self-regulation and enhancing the safety of home-based rehabilitation.

The control group attended an educational session at the beginning of the 8-week trial and was advised to follow Korean pulmonary rehabilitation guidelines: aerobic exercise at 60% maximum intensity (3-5 times/wk) and anaerobic exercise at 60%-70% intensity (2-3 times/wk) [20]. Participants in the control group visited the clinical center every 4 weeks for scheduled assessments, identical to the DTx group.

Costs were calculated from a health care system perspective for the base-case analysis. Two types of costs were included: (1) medical and (2) pharmaceutical costs. Expenses were collected retrospectively after the completion of the RCT and included only those directly related to CRD, as determined by clinical judgment. Medical costs, including screening, examination, and physician consultation fees, were collected based on electronic health records from the clinical trial institution. Pharmaceutical costs were calculated by summing drug acquisition costs and pharmacy dispensing fees. Drug acquisition costs were calculated using drug quantities from the electronic health record and standard unit prices from the National Health Insurance drug list [17]. For the DTx group, the cost of using the EASYBREATH app, priced at US $93.17 for the 8-week period, was included.

As the CUA did not estimate costs beyond a 12-month horizon, no discount rate was applied. All costs were initially collected in Korean won and converted to US $ using a 2023 exchange rate of US $1=1288 Korean won.

QALY was used to assess the QoL, as it reflects both health-related QoL and life expectancy to measure overall health benefits [15]. However, the trial did not include any generic preference-based health-related QoL instruments, such as the EQ-5D, Short Form 6 Dimensions, or Health Utilities Index, which are typically required for direct utility measurement. Instead, disease-specific measures like the CAT and SGRQ were collected during the trial. These scores were mapped to utility values using validated algorithms from previous literature to estimate QALYs. In total, 3 relevant publications were identified through a literature search [21-23] (Table S1 in Multimedia Appendix 1). Among the identified algorithms, we selected algorithm by Lim et al [23] based on the following criteria: (1) lowest root mean square error and mean absolute error (Table S2 in Multimedia Appendix 1), (2) similarity in disease severity to the trial population, and (3) comparable racial characteristics (Table S3 in Multimedia Appendix 1). The study by Lim et al [23], conducted in a Korean tertiary hospital, closely aligned with the trial’s clinical context. The algorithm was internally validated using a split-sample approach and bootstrap resampling (10,000 iterations) to assess root mean square error and mean absolute error. The model with the lowest average prediction error was selected as the final algorithm [23]. Using the proposed algorithm (ordinary least squares model 3), CAT scores were mapped to EQ-5D-3L utility values. QALY was then calculated using the trapezoid rule [15].

Ordinary least squares model 3 from Lim et al [23]:

The trial protocol was approved by the Korean Ministry of Food and Drug Safety (Approval 1431). This 8-week period RCT was approved by the Institutional Review Boards at each clinical center (Catholic University Seoul St Mary’s Hospital, KC23DNDS0031; Inje University Sanggye Paik Hospital, SGPAIK 2022-11-022-001; and Inje University Haeundae Paik Hospital, HPIRB 2022-12-013-002). Written informed consent was obtained from all participants during the primary clinical trial. For the economic evaluation study, the waiver of additional informed consent was approved, considering the retrospective study design and minimal risk to participants.

The dataset used in this study was fully deidentified and did not contain any personally identifiable information. Only patient identification codes, treatment group assignment, sex, and age were collected. No additional data that could allow for the reidentification of individual patients were included. This study did not use any images or materials containing identifiable information of individual participants. In an economic evaluation study, no additional compensation was provided.

A total of 84 patients were included in the randomized trial. The baseline characteristics of the patients were comparable, except for age. Table 1 illustrates the demographic characteristics of the study population. Male participants comprised 79.1% (34/43) and 87.8% (36/41) of the DTx and control groups, respectively. The mean age in years was slightly lower in the DTx group (mean 63.4, SD 10.36 y) compared with the control group (mean 67.78, SD 6.93 y; P<.05). COPD was the most common diagnosis, with 26 (60.47%) patients in the DTx group and 23 (56.1%) patients in the control group. Baseline clinical efficacy measures, including the 6MWD, mMRC, CAT scores, SGRQ subscales, and HADS, were well matched across the groups (all P>.05).

Treatment adherence in the DTx group was assessed among participants included in the PPS at both the week 4 (n=43) and week 8 (n=40) visits. All participants demonstrated an adherence rate of ≥80% throughout the intervention period, corresponding to the “very high compliance” category [11].

This economic evaluation demonstrated that DTx improved QALYs in patients with CRD over an 8-week period. The ICUR was US $8890 per QALY, with a 95% CI ranging from –US $29,927 to US $18,602, which falls within the cost-effectiveness threshold. In addition, the VBP analysis indicated that DTx remained cost-effective up to a usage cost of US $192 based on the estimated QALY gain. Although the CI was wide, its upper bound remained below the WTP threshold (US $19,410 per QALY). This suggests that the DTx is likely to be cost-effective under most iterations. However, the wide interval and inclusion of zero in ICUR also indicate some degree of uncertainty, highlighting the need for additional sensitivity analyses.

To explore this uncertainty, PSA and DSA were conducted. In the PSA, 30.9% of ICUR were located in the northwest quadrant, indicating that DTx was more costly and less effective on the cost-effectiveness plane. Although this suggests a risk of the DTx being dominated, 58.1% of ICUR fell below the WTP line in the northeast quadrant on the plane. Furthermore, the CEAC indicated that over half of the iterations supported the cost-effectiveness of DTx at the selected WTP threshold. The DSA indicated that the ICUR was most sensitive to variation in pharmaceutical costs in the control group, likely due to underlying differences in concomitant medications across participants. These differences might be explained by variations in the initial CRD diagnoses, as patients with different conditions required different sets of medications. The tornado diagram presented that none of the results, including those using alternative mapping algorithms, exceeded the WTP threshold.

In the clinical trial, significant differences in QALY between groups were observed after week 4. This may have resulted from both the individual effects of DTx after an initial adaptation period and the cumulative benefits of exercise. This pattern aligns with major clinical guidelines, which recommend a minimum of 6 weeks of pulmonary rehabilitation to achieve clinical benefits [23,26]. The RCT was designed to capture such delayed effects. Furthermore, such time-dependent effects should be taken into account in real-world clinical settings and in future RCTs evaluating DTx.

In this study, both groups attended follow-up visits every 4 weeks, which may have introduced a Hawthorne effect. According to clinical experts, patients with CRD generally visit hospitals every 12 weeks for routine prescriptions, suggesting that both groups were more closely monitored than would typically occur in real-world settings. However, the therapeutic effect might be maintained even in real-world practice because of continuous feedback and motivation through the app. In contrast, the standard treatment group may have been more influenced by frequent follow-up visits during the trial. Therefore, the outcomes observed in this study could be interpreted as relatively conservative estimates of the DTx’s real-world effectiveness.

With regard to cost components, the SD of pharmaceutical costs exceeded that of medical costs. Variability in pharmaceutical costs arose from differences by patient in concomitant medication and its unit price, while other medical costs are more homogeneous because of the regular schedules in the trial. Given the skewed nature of cost data in economic evaluations [27], the observed variation in pharmaceutical costs was considered acceptable. In the DTx group, although smartwatch integration was possible, it was not mandatory. Therefore, the cost of the smartwatch was not included in the DTx-related cost components.

To examine the cost-effectiveness across population heterogeneity, the subgroup analysis was conducted. DTx was cost-effective across all age groups, suggesting that the user-friendly interface and in-person training sessions effectively addressed differences in digital literacy. Additionally, DTx demonstrated particularly high cost-effectiveness among patients with ILD. The ICUR was substantially lower in the ILD group at US $559 per QALY compared with US $19,134 per QALY in the COPD group. Patients with ILD had higher baseline disease severity and incurred greater medical costs, which may have allowed more room for clinical and economic improvement. Although the cost of DTx was the same across groups, the ILD subgroup achieved larger QALY gains and greater reductions in health care costs, resulting in a more favorable ICUR. These findings suggest that disease characteristics may influence the cost-effectiveness of DTx. However, given the small sample size in the subgroup, these results should be interpreted with caution.

Additionally, when analyzed from a limited societal perspective, the incremental cost difference was minimal (US $1.46) because the unit costs for transportation and informal caregiving were identical between groups. The number of hospital visits influenced the cost difference because of its alignment with the protocol-specified schedule.

A Markov model was developed to estimate long-term outcomes. Over a 1-year time horizon, the model projected higher cumulative costs (US $1443) and QALY gains (0.3354) compared with the base-case analysis. Since the model applied a fixed transition probability observed during the trial, it implicitly assumed that the treatment effect remained constant over time. This assumption may have led to an overestimation of the incremental QALY gain and a corresponding underestimation of the ICUR (US $4302 per QALY). Therefore, these results should be interpreted with caution, given the overprojection of long-term treatment effects.

Previous studies have shown that telemedicine-based pulmonary rehabilitation programs are more cost-effective than traditional programs [28-31]. Although telerehabilitation is a part of digital health technology, it differs from software medical device (DTx), which has clinically proven efficacy and safety [32]. There are economic evaluations of telerehabilitation [33], but no study has specifically evaluated the cost-effectiveness of DTx in pulmonary rehabilitation.

Economic evaluation studies of DTx for managing other chronic diseases such as hypertension, chronic heart failure, and type 2 diabetes have demonstrated its superior cost-effectiveness compared with standard care or telehealth interventions [34-36]. Although the characteristics of each condition and DTx differ, mobile software apps have demonstrated their potential as important tools for managing chronic diseases. These findings are consistent with our study on DTx in patients with CRD, further supporting the cost-utility of DTx in managing chronic conditions.

In South Korea, several studies have used cost-benefit analysis to evaluate telemedicine, artificial intelligence–based health care services, and mobile health care programs [37,38]. However, cost-benefit analysis is less preferred in health care technology assessments because of financial equity concerns arising from variations in WTP [39]. In contrast, CUA uses the more widely accepted QALY metric, providing acceptable evidence for health care decision-making.

South Korea operates a single-payer health care system, but reimbursement for any DTx has not yet been implemented. Nonetheless, active policy discussions are ongoing regarding DTx coverage and pricing. In this context, the US $192 VBP identified in this study for an 8-week session is considered relatively moderate, compared with in-person pulmonary rehabilitation sessions available in Korea. For comparison, the reimbursement rate for conventional pulmonary rehabilitation for CRD is approximately US $1377 for an 8-week course, which is higher than the proposed VBP for DTx. Internationally, Germany provides an example of negotiated pricing for digital health applications. For example, the Somnio app (mementor DE GmbH) was approved at 52% of the manufacturer’s proposed price (€224.92 in 90 days, as of 2022) [40]. Although DTx is not yet publicly reimbursed in Korea, the findings from this CUA and VBP analysis can serve as evidence for assessing the reimbursement feasibility of DTx.

To the best of our knowledge, this is the first study to evaluate the cost-effectiveness of DTx for pulmonary rehabilitation and its pricing, with a focus on the value. The analysis was conducted from a health care system perspective, focusing on direct medical costs relevant to decision-makers. These results suggest that a DTx-based rehabilitation program is a cost-effective option for both patients and insurers. This study provides evidence supporting the feasibility of implementing DTx in the Korean health care environment. To support sustainable adoption within the health care system, appropriate pricing strategies are essential.

In Korea, discussions are ongoing regarding how to incorporate DTx pricing into the existing market framework. Given that the Korean health care system operates under a fee-for-service structure and a single-payer insurance system, there is limited flexibility for introducing new types of payment schemes. In this context, we adopted a VBP approach to propose a price that reflects the health outcomes gained.

However, diverse health care systems may require tailored pricing strategies in economic evaluations to reflect the economic value of DTx. For instance, subscription-based pricing models should consider both the number of eligible patients and real-world usage rates to avoid price overestimation. Outcome-based pricing requires clearly defined clinical end points and infrastructure for performance monitoring, which could be supported by integration with wearable technologies with DTx. Tiered pricing reflects differences in disease severity, adherence, or expected benefit, and it allows high-value patient groups to be prioritized through differentiated prices. Once integrated into the health care system, pricing strategies should align with both the structure of the health care system and technological characteristics of DTx—such as outcome measurability and variability in patient usage—to ensure its sustainable implementation.

This study has several strengths. To the best of our knowledge, this is the first study in South Korea to evaluate the cost-effectiveness of DTx. The analysis was based on data from a confirmatory clinical trial, which enhances the internal validity of the findings. Additionally, the analysis used outcome data from 3 clinical centers covering a range of patients with CRDs, including those with COPD and ILD. High treatment adherence in the DTx group further supports the reliability of the observed effects. Furthermore, when the QALY gain of the DTx group was reduced in the DSA, the intervention remained cost-effective. The robustness of the results was confirmed by additional sensitivity analyses.

However, this study also has several limitations, including study design, data and statistical analysis, and real-world applicability. First, regarding study design and sample size, the study period was limited to 8 weeks, which did not capture long-term cost-effectiveness. Although a 1-year Markov model analysis was conducted, it was based on assumptions rather than observed long-term data. Further research with an extended follow-up period is needed. In addition, the small sample size of the trial may limit the generalizability of the results. However, the sample size was calculated with a 15% dropout rate in mind, and the robustness of the findings was supported through various sensitivity analyses.

Second, regarding data collection and analytical aspects, nonmedical expenses and productivity losses were excluded from the base case analysis due to limited data. As a result, the broader societal burdens were not fully captured, which may have led to an underestimation of the therapeutic value of DTx. However, given the mild disease severity of participants, the impact of productivity losses is likely to be minimal. To address this limitation, a scenario analysis from a limited societal perspective, including non–health care costs, was conducted based on consultations with clinical experts and previous literature.

For utility estimation, utility values were derived using a mapping algorithm rather than directly obtained from the EQ-5D-3L. As the RCT included only CRD-specific health-related QoL instruments (CAT, SGRQ, and HADS), we reviewed available mapping algorithms. Disease-specific instruments may be more responsive and sensitive to particular conditions [41]. However, condition-specific measures may have limitations in generalizability, as they could not fully reflect the impact of side effects and comorbidities on health valuation. In addition, condition labels and focusing effects, which result from narrower descriptive systems, may further limit their broader applicability [42]. Although we selected a mapping algorithm developed in a population similar to ours, it was based on cross-sectional data. This may limit to reflect longitudinal changes in individual health status. To address uncertainty in utility estimates, we applied various mapping algorithms in sensitivity analyses and selected those that best reflected disease severity, minimized prediction error, and aligned with the study design.

Additionally, there is a statistically significant difference in baseline age observed between the 2 groups, with the DTx group being younger on average. This difference may have contributed to the favorable outcomes observed in the DTx group. Although subgroup analyses stratified by an age threshold of 65 years confirmed that DTx remained cost-effective, baseline age was not adjusted as a covariate, which may have introduced residual confounding.

Third, concerning external validity and real-world applicability, the study design required both groups to attend routine follow-ups every 4 weeks, which may have inflated medical costs for the DTx group. However, in real-world settings, patients using DTx typically require routine follow-ups only every 6-12 weeks. Finally, the high adherence observed in the DTx group may have led to an overestimation of its efficacy. This adherence was likely influenced by continuous monitoring and feedback from study administrators and the lack of a double-blind design inherent to digital therapeutics. Furthermore, because adherence data for the control group were unavailable—as adherence in the DTx group was derived from the app’s built-in recording function—the relationship between treatment effects and adherence levels could not be assessed. Therefore, while the DTx group achieved superior outcomes under controlled conditions, such results may not be fully generalizable to real-world settings where adherence is typically lower. Nevertheless, adherence may be supported to some extent in real-world settings through clinician support on web-based platforms and the DTx’s self-recording feature, which encourages long-term engagement in rehabilitation.

This study showed that the EASYBREATH app, a DTx-assisted pulmonary rehabilitation program, is more cost-effective than standard treatment. Cost-effectiveness was maintained up to a usage cost of US $192 per 8-week period. These findings support DTx-based pulmonary rehabilitation as an effective treatment option for managing patients with CRD and provide valuable insights for health care decision-makers. Further research is necessary to evaluate long-term effectiveness and sustainability.