A dataset of digital therapeutic device approvals to support regulatory decisions and industry research

Abstract

Digital Therapeutics (DTx), as an evidence-based, software-driven medical intervention, are rapidly developing globally with corresponding regulatory frameworks being gradually established in multiple countries. However, there is currently a lack of unified and structured data resources to support in-depth cross-national comparative and industry analysis. To address this, this study aimed to construct a standardized dataset covering approved DTx devices from four countries: US, China, Germany, and Belgium. We identified devices through systematic searches of national public databases combined with a keyword strategy. Following rigorous multi-stage screening, standardized information extraction, and expert arbitration, 507 devices meeting international DTx criteria were ultimately included, encompassing 18 product characteristics. This dataset provides a comprehensive and reliable structured data foundation for regulatory decision-making, industry research, and cross-national policy analysis pertaining to DTx.

Background & Summary

Digital Therapeutics (DTx) represents an evidence-based, software-driven intervention designed to prevent, manage, or treat medical disorders¹. Currently, DTx has been implemented across multiple critical clinical domains, including chronic disease management, mental health, and rehabilitation^2–6. Numerous studies have demonstrated that DTx offers advantages over conventional interventions, such as personalized and precise treatment, improved patient adherence, and minimal side effects^7–11. Recognizing its potential, many countries have introduced supportive policies and continuously expanded regulatory pathways to foster DTx development. For instance, the US FDA established the Digital Health Center of Excellence to refine regulatory evaluation processes¹², China’s Hainan Province introduced measures to accelerate DTx industry growth and incorporated DTx into its provincial health industry plan¹³, Germany enacted the Digital Healthcare Act¹⁴, and France is developing a fast-track system to facilitate patient access to DTx with reimbursement through public insurance¹⁵.

Against this backdrop, the number of marketed DTx products has been increasing¹⁶. Correspondingly, research on clinical trial outcomes and real-world follow-up data is also expanding^17–23. However, we observe that most current studies rely on DTx device datasets with notable limitations. On one hand, the data are often scattered across databases of various national regulatory authorities and academic publications, lacking unified metadata standards and structured coding, which hinders cross-regional and cross-product comparisons. On the other hand, the publicly available characteristics of DTx devices in existing studies and regulatory databases remain relatively limited, making it difficult to assess other critical device attributes. This restricts the ability of researchers, clinicians, and payers to adequately evaluate the long-term value of these devices.

To address this limitation, we have compiled a comprehensive and standardized dataset integrating approved DTx devices from public sources across four countries: US, China, Germany, and Belgium. The dataset encompasses 18 product characteristics, including fundamental device information (e.g., device name, approval number, approving country and authority, approval date, device classification, prescription requirements), functional attributes (e.g., intervention format and functional category, specific functions provided, target diseases, intended users, software type, digital technologies employed, mode of use, behavioral intervention technology), and manufacturer details (e.g., manufacturer name, country of origin, first-level administrative region, company size). By providing detailed device characteristics, this dataset aims to support evidence-based regulatory policymaking and facilitate comprehensive analysis of the DTx landscape for researchers, clinicians, and payers. Additionally, a portion of this dataset was analyzed in a previous study¹⁶ focusing specifically on the current approval situation of DTx. Building upon these prior data, we further expanded the DTx product characteristics by integrating the European Common Classification Grid for Digital Medical Devices (CEU Grid-DMD)²⁴, thereby providing a more comprehensive perspective on product features. This dataset not only facilitates further research into the industrial development trends of DTx but also holds significant importance for the widespread promotion and advancement of DTx globally.

Methods

Figure 1 shows the detailed workflow of the dataset construction process. This systematic approach ensures both the high quality and standardization of the DTx device dataset.

Fig. 1.

DTx device dataset construction process.

Identify data sources

The selection of US, China, Germany, and Belgium as key countries for data sourcing was based on the following considerations. The US FDA established a preliminary regulatory framework adapted to the technical characteristics of DTx through the Digital Health Software Precertification (Pre-Cert) Program and the De Novo pathway in 2017^25,26. In China, the provincial development plan of Hainan explicitly proposed the exploration of digital therapeutics starting in 2022, followed by the release of China’s first DTx-specific regulatory policy by the NMPA in 2025, marking the formal recognition of DTx within China’s regulatory system²⁷. Germany and Belgium, as European pioneers in DTx policy advancement, established dedicated approval processes and health insurance reimbursement mechanisms for DTx in 2019 and 2018, respectively, leading the way within the European Union²⁸. Furthermore, the US represents a technological and economic powerhouse, while China possesses a large population and a rapidly growing burden of chronic diseases²⁹. Taken together, these four countries have each undertaken significant explorations in DTx evaluation, market access, and reimbursement, collectively providing a representative overview of the current global DTx approval landscape.

The dataset was compiled from four publicly accessible sources and two industry websites used for supplementary data collection. The official sources include the US FDA³⁰, China’s NMPA³¹, the Digital Health Applications (DiGA) directory published by the German Federal Institute for Drugs and Medical Devices (BfArM)³², and the mHealthBELGIUM platform³³. To minimize omission bias, two industry websites, namely the Digital Therapeutics Alliance (DTA)³⁴ and VBData³⁵, were utilized as supplementary sources for cross-verification of the records obtained from the four primary databases.

Data collection and processing

For the US and China, a systematic keyword-based search was conducted in December 2024 (see Supplementary Note 1 and Supplementary Table 1). For Germany and Belgium, the official DiGA directory and the mHealthBelgium platform were accessed in September 2024 to retrieve all registered products. All initially identified and publicly acquired data were uniformly processed and consolidated in Excel by a trained student team following standardized protocols. This initial process yielded a total of 26,194 records (China: 20,111; US: 6,003; Germany: 55; Belgium: 25). Furthermore, a pre-screening procedure was applied to the US and Chinese data to exclude duplicate entries and devices without software components, resulting in a final set of 14,677 candidate devices (China: 11,021; US: 3,576; Germany: 55; Belgium: 25).

Rigorous data screening

Based on the DTx criteria endorsed by the DTA, we established stringent inclusion and exclusion criteria (see Table 1) for meticulous screening of candidate devices. These criteria were reviewed and approved by a domain expert (PX). Prior to formal screening, to ensure consistency in the evaluation process, two researchers independently pre-screened a random sample of 40 devices. Inter-rater reliability was quantified using Cohen’s kappa coefficient. Subsequently, both researchers independently screened all candidate devices from the US and China according to the inclusion and exclusion criteria (Table 1). Inconsistencies between raters primarily emerged in borderline cases, such as distinguishing whether a software solely provided health data tracking (which met exclusion criteria) or actively delivered an evidence-based medical intervention. Any discrepancies during the screening process were adjudicated by the domain expert (PX). This adjudication involved a consensus-based discussion where conflicting evaluations were reviewed against official device descriptions to ensure the final decision strictly aligned with the DTx definition. Ultimately, a total of 507 devices meeting the DTx standards were included (US: 192; China: 235; Germany: 55; Belgium: 25). A detailed flowchart of the screening process is presented in Fig. 2.

Table 1.

Inclusion and exclusion criteria.

Inclusion criteria

Exclusion criteria

1. Software-driven: used alone or in combination with drugs, medical devices or other therapies. 2. Supported by evidence-based evidence (randomized controlled trials, cohort studies, case-control studies, etc.). 3. Disease intervention: to achieve the prevention, management, and treatment of disease. 4. Patient-oriented: to be used by the patient and or the patient’s family, with or without medical supervision.

1. The intervening role belongs to the hardware device. 2. The software only provides data collection/data analysis/hazardous value reminder/disease diagnosis functions, which cannot prevent, manage, or treat diseases. 3. Exclusion is only for physicians and not patients and or patients’ families. 4. Not officially approved, registered or certified (not approved by FDA, NMPA or not CE marked). 5. Basic information about the product (product name, company name, intended use, etc.) is not available.

Fig. 2.

Screening process for medical devices meeting the DTx definition. NMPA: National Medical Products Administration. FDA: Food and Drug Administration. DiGA: Digital Health Applications. DTx: digital therapeutics.

Data standardization and information extraction

Drawing upon existing literature on DTx^{17,20,24,36–39} to identify key dimensions and features relevant to DTx, we developed a standardized information extraction framework (Supplementary Table 2) to comprehensively analyze multi-dimensional characteristics of DTx devices. The framework encompasses 18 product features, including basic device information (e.g., device name, approval number, approving institutions, approval time, management category, medical prescription), functional attributes (e.g., functional classification, functions offered, target diseases, service users, software type, digital technology, software usage, behavioral intervention technology), and manufacturer details (e.g., manufacturer name, country of origin, first-level administrative region, company size).

Specifically, basic and approval information (such as device name, approval number, approving institutions, approval time, regulatory class, and medical prescription) was primarily sourced from publicly available records in official regulatory databases, including the FDA, NMPA, DiGA, and mHealthBELGIUM. Functional attributes and technical characteristics were preferentially extracted from official product descriptions, technical documentation, or summary information published by regulatory authorities. Functional classification was based on the DTA classification system for DTx, and was further refined with reference to prior classification approaches²⁴ for digital medical devices to identify specific function types. The functions offered by each device were extracted based on descriptions of intervention methods documented in official regulatory summaries. Target diseases were coded according to the International Classification of Diseases (ICD-10) published by the World Health Organization (WHO)⁴⁰, based on official product information. Service users and software usage patterns were categorized in accordance with official product descriptions. Software type was determined by referring to established classification frameworks from prior research³⁷, supplemented by official information. Digital technology types were primarily derived from existing literature³⁶, extracted in conjunction with official documentation, and supplemented with publicly available manufacturer information when official data were incomplete. Behavioral intervention technology was uniformly coded based on classification standards from previous studies^38,41. Manufacturer information, including manufacturer name, country of origin, first-level administrative division, and size of manufacturer, was compiled by integrating corporate registration data from official databases, supplemented by information from company websites and publicly available business registration records.

During the information extraction process, where relevant features were missing or incomplete in official databases, data were supplemented following a predefined hierarchy. Manufacturer websites, product specifications, or technical white papers were consulted first. If such information remained unavailable, industry platforms including DTA and VBData were referenced. For instances where manufacturer information was absent, publicly available business databases such as Crunchbase⁴², LinkedIn⁴³, and OpenCorporates⁴⁴ were further utilized. For variables requiring subjective judgment, such as behavioral intervention technology, data for all devices meeting DTx criteria were independently extracted by two researchers (QCF and JL) using a unified framework. To minimize bias and ensure consistent rating criteria, the raters strictly adhered to the definitions detailed in Supplementary Tables 2, 3, which presents the value ranges, classification references, and corresponding data sources for each field. Similar to the screening phase, prior to formal data extraction, the two researchers independently extracted information for a random sample of 40 devices. Inter-rater reliability was then evaluated using Cohen’s kappa coefficient based on this 40-device sample to calibrate their assessments and ensure consensus. Following this calibration step, the raters proceeded to extract data for the remaining devices. During the full extraction phase, inconsistencies most frequently occurred when differentiating between Guided and Adjunctive behavioral interventions due to overlapping functional descriptions, or when classifying multifaceted digital technologies. Any disagreements during the extraction process were adjudicated by a third expert (PX). This final decision-making process involved a joint review session where differing rationales were evaluated against predefined definitions and official product specifications, with a final classification assigned only upon reaching a unanimous consensus. The final dataset was independently reviewed by two experts (PX and JBL).

Data Records

To address the informatics challenge of real-time and the customizable comparison and visualization of the current status of DTx in major countries, we developed an open, unified online database that supports advanced queries. It has over 20 conditions, a list of results, and >100 types of visual displays. The URL of the website is https://chinadtx.org.

This dataset provides a structured and traceable informational foundation for regulatory decision-making and industry research, and is publicly available on the Open Science Framework (OSF) platform⁴⁵ for researchers to download and use for any purpose. It is provided as a single Microsoft Excel (.xlsx) file containing two sheets, each serving a distinct purpose. Sheet 1 is the primary data table, which systematically documents 507 medical devices that meet DTx criteria approved by regulatory authorities in China, US, Germany, and Belgium as of December 2024. The data are organized in a two-dimensional structure of rows (records) and columns (variables). Each record includes 18 standardized fields, such as device name, risk classification, target disease, type of digital technology, behavior change techniques, and approval date, enabling cross-national comparison, quantitative analysis, and policy simulation. For devices approved by multiple regulatory authorities, the same DTx has been merged into a single record. Sheet 2 provides standardized definitions for all fields used in Sheet 1. It corresponds to Supplementary Table 2 and offers a reference framework for standardized extraction of information from similar devices in future studies.

Additionally, we calculated basic descriptive statistics for the 507 included DTx devices. The detailed results, stratified by approval country and dimensional characteristics, are presented in Supplementary Tables 4–7 for China, the US, Germany, and Belgium, respectively. Specifically, the approved devices in China are primarily used to treat diseases of the nervous system. Those in the US are mainly used for the management or monitoring of endocrine, nutritional and metabolic diseases. The approved devices in Germany are primarily used to treat mental and behavioural disorders. Those in Belgium are mainly used for the management or monitoring of endocrine, nutritional and metabolic diseases, as well as diseases of the circulatory system.

Technical Validation

Data integrity and consistency

In the dataset collected in this study, approval date information is unavailable for the 25 DTx devices approved in Belgium, due to the fact that their official registration platform (mHealthBelgium) does not publicly disclose detailed records of approval dates. The temporal trends in data density relative to approval time for the remaining three countries (US, China, and Germany) are shown in Fig. 3.

Fig. 3.

Annual approval data density for DTx devices.

Inter-rater agreement for the identification of DTx devices was assessed using Cohen’s kappa coefficient, which yielded a value of 0.812 (p < 0.01), indicating a high level of agreement between the two reviewers. Additionally, the consistency of information extraction for key categorical variables was also evaluated using Cohen’s kappa, which demonstrated generally high agreement across features. The specific kappa values for each variable are presented in Fig. 4.

Fig. 4.

Inter-rater agreement for key DTx feature coding assessed by cohen’s kappa coefficient. BIT: Behavioral intervention technology.

It should be noted that the distribution of included devices across countries is inherently unbalanced, with substantially fewer records from Germany and Belgium compared to the US and China. This discrepancy primarily arises from differences in data source characteristics and regulatory pathways. Specifically, data from Germany and Belgium were obtained from the DiGA directory and the mHealthBelgium platform, respectively. These databases do not assign specific DTx numbers, and both function as curated registries that include only products that have successfully undergone formal evaluation and reimbursement approval processes. Therefore, these sources represent pre-selected subsets of digital therapeutics rather than the full spectrum of available or emerging products. In contrast, data collection for the US and China was based on broader regulatory databases using keyword-based search strategies, enabling more comprehensive identification of candidate devices. Additionally, compared with the US and China, European regulatory and reimbursement frameworks for digital therapeutics impose relatively stringent evidence and access requirements, which may further limit the number of products entering these official registries⁴⁶.

Data accuracy verification

Following data extraction, a data auditing process was implemented to ensure accuracy. First, terminological consistency across all field features was checked and standardized. Second, cross-validation of device approval information against original sources revealed 11 devices with multinational approvals (Table 2). Comparative analysis confirmed consistency in extracted information across different regulatory jurisdictions with no outliers identified. To avoid duplicate counting and ensure dataset independence and accuracy, these multinational devices were deduplicated, retaining only one record per device. The approving authorities field was annotated with combined labels (e.g., “FDA-NMPA”) to indicate approvals from both US and Chinese regulatory agencies. Finally, the consolidated dataset underwent expert review to guarantee data reliability.

Table 2.

DTx devices approved by multiple countries.

Source of approval	Device name	Manufacturer	Source country
FDA - NMPA	ANDON HEALTH CARE MANAGEMENT SYSTEM SOFTWARE, AG-608 SINGLE & AG-608 MULTI BLOOD GLUCOSE MONITORING SYSTEMS	ANDON MEDICAL CO., LTD.	China
FDA - NMPA	Sparse Sample Pk Profile And Dosing Software	Baxalta US Inc.	US
FDA - NMPA	ECG App	Apple Inc.	US
FDA - NMPA	CureSight	NovaSight Ltd.	Israel
NMPA - FDA	Portable (Ultrasonic) Nebulizer	Dongguan SIMZO Electronic Technology Co.Ltd.	China
NMPA - FDA	Irregular Rhythm Notification Feature (IRNF)	Apple Inc.	US
FDA - mhealthbelgium	FibriCheck	Qompium Nv	Belgium
FDA - mhealthbelgium	Sunrise Sleep Disorder Diagnostic Aid	Sunrise	Belgium
FDA - mhealthbelgium	Cochlear Baha 6 Max Sound Processor Cochlear Baha Fitting Software 6 Cochlear Baha Smart App	Cochlear Americas	US
FDA - mhealthbelgium	ACCU-CHEK CONNECT DIABETES MANAGEMENT APP	ROCHE DIAGNOSTICS CORPORATION	Switzerland
FDA - NMPA - mhealthbelgium	MiniMed Connect kit, MiniMed Connect uploader, MiniMed Connect app, CareLink Connect (CareLink Personal)	MEDTRONIC MINIMED, INC.	US

NOTE: The symbol “-” here refers to joint approval devices, like “FDA-NMPA” refers to devices were approved by the US and China. In addition, “FDA-NMPA” indicates the order of approval of the device, i.e., the device was approved in the US before it was approved in China. The device approved in Belgium then lacked approval time, so it was placed last by us.

Additionally, to minimize omission bias and validate the completeness of our dataset, we cross-verified our initial search results against two supplementary industry databases: the DTA and VBData. This validation process successfully identified 12 eligible DTx products that were not captured through the initial official database queries. These omissions stemmed from the inherent boundaries of our primary retrieval strategies. Specifically, for the US database, our search relied on a predefined set of FDA product classification codes (as detailed in Supplementary Note 1), meaning products registered under unlisted codes were bypassed. For the Chinese database, our strategy relied on specific keywords (‘software’ and ‘system’), which inevitably missed a small number of eligible devices that did not use these exact terms in their names. As detailed in our screening flowchart (Fig. 2), these 12 supplementary devices (8 approved by the FDA and 4 by the NMPA) were subsequently incorporated into our dataset. Furthermore, beyond identifying omitted devices, these industry platforms were instrumental in validating the extracted content. Specifically, we conducted a comparative analysis on 18 devices listed in the DTA product library and 12 devices featured in VBData (as explicitly indicated in the ‘Data sources’ column of the dataset). This analysis demonstrated that the features extracted from official regulatory databases were highly consistent with the information reported on the industry websites. In cases where official regulatory databases lacked detailed product descriptions, the missing fields were successfully cross-referenced and supplemented using the comprehensive product monographs available on DTA and VBData.

Limitations

Although the constructed dataset possesses considerable scale and representativeness, it is subject to several limitations. First, significant heterogeneity in data formats, documentation standards, and degree of accessibility across national data sources complicated the process of data collection and integration. Although multiple measures were implemented to maximize data completeness and accuracy, potential biases could not be fully eliminated. For instance, the absence of detailed approval timing information from the Belgian regulatory platform resulted in missing data for this information. Second, device identification in this study relied primarily on a keyword-based search strategy. Despite its systematic design, it remains possible that some eligible products may be overlooked, thereby introducing a certain degree of selection bias. Nevertheless, it should be noted that, among currently available public resources, this dataset remains one of the larger and more comprehensive collections in the DTx field. Furthermore, as the dataset includes only four countries (US, China, Germany, and Belgium), selected for their representative roles in DTx regulation and market development, the findings may not fully capture the global diversity and overall trends in DTx adoption.

Author contributions

J.B.L., J.L. and Q.C.F. contributed to the study design, execution, quality control, and manuscript revision. S.P.L. and Y.C.L. contributed to the quality supervision and manuscript revision. P.X. contributed to the study design, data collection, quality control, and manuscript drafting. Q.C.F., J.K.Z., L.F.Z., L.Y., H.Q.Z., Y.F.H. and L.D. contributed to the data search strategy and data collection. All authors were involved at each stage of manuscript preparation and approved the final version. P.X. and Y.C.L. did critical revision of the manuscript for important intellectual content. All authors have access to all the data in the study. J.B.L., J.L., P.X. have accessed and verified all the data in the study.

Data availability

The dataset is available for download from the OSF project page⁴⁵ and can also be accessed via the website (https://chinadtx.org).

Code availability

No custom code was used in this study.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s41597-026-07221-3.