Open challenges

Unlike previous therapeutics, DTx support rapid updates to their software to align with changes to best-practice clinical care guidelines or use adaptive algorithms that can improve over time as they encounter new data. This makes their use and predictability towards outcomes more challenging, and there is a pressing need for new regulatory frameworks in this field. To improve their ability to evaluate these increasingly complex DTx, the FDA launched a Software Precertification Pilot Program in 2017 [8]. Nine DTx companies selected for the program will help inform the development of a Total Product Lifecycle (TPLC) regulatory framework that would allow companies meeting precertification (“Excellence Appraisal”) criteria to market lower-risk devices with a Streamlined Review process followed by Real-World Performance monitoring. While the TPLC helps clarify the regulatory landscape for iterative DTx development, much work remains to ensure that this model is effective at producing safe and reliable DTx. As of the most recent update in September 2020, the FDA has drafted their metrics for Excellence Appraisal and an overview of the Streamlined Review process, and have begun exploring methods for collecting Real-World Performance data. This presents a significant challenge as collecting high quality data and modeling patient outcomes are non-trivial tasks, and are active areas of research in medical informatics [9]. For example, it is not clear whether medical orders for DTx are captured consistently in electronic medical record systems.

Reimbursement policies for DTx also remain under development as SaMDs are still being evaluated for treatment efficacy and cost-effectiveness. US-based pharmacy benefit managers Express Scripts and CVS Health both released digital health formularies in 2019 to identify recommended digital health solutions for insurers and physicians [10,11], but these only covered a limited number of medical specialties and many of the solutions are geared towards non-DTx mobile health apps that do not require FDA approval based on the agency’s “enforcement discretion” policies [8]. The Centers for Medicare and Medicaid Services (CMS) has also provided limited guidance, publishing some reimbursement codes to cover collaborative care models involving both clinical care and app usage, but not yet issuing codes for the use of DTx as a standalone treatment. SaMDs also do not fall under a particular Medicare benefit category, so breakthrough DTx are currently not eligible for coverage under new Medicare Coverage of Innovative Technology policies, further restricting their access to patient populations [12]. Finally, another major challenge with DTx reimbursement is capturing the value of the device itself, as well as the time that physicians spend monitoring patients and reviewing data from the apps.

As with any device storing and transmitting patient data, privacy and security are critical for DTx, especially those with frequent software updates or connections to other devices or electronic medical record systems. The FDA requires all medical devices companies to identify and mitigate cybersecurity risks as part of premarket submissions, but best-practice standards for data management and privacy are constantly evolving [13]. Another technical challenge, particularly for artificial intelligence and machine learning (AI/ML)-based DTx, is evaluating the safety and efficacy of algorithms in diverse patient populations. For example, training datasets may not always capture social determinants of health or other biases that affect the performance of healthcare algorithms [14]. Methods to recognize and remove such biases are paramount to ensuring the delivery of equitable healthcare, and device labeling requirements that clarify how these models were trained, or how they derived their outputs, may help promote transparency in AI/ML-based DTx [15].