News & Comment

In recent years the intersection of wearable technologies and machine learning (ML) based deep learning (DL) approaches have highlighted their potential in sleep research. Yet, a recent study published in NPJ Digital Medicine highlights the generalization limitations of DL models in sleep-wake classification using actigraphy data. Here, this article discusses some of the challenges and opportunities presented by domain adaptation and self-supervised learning (SSL), innovative methodologies that use large-scale unlabeled data to bolster the generalizability of DL models in sleep assessment. These approaches not only improve sleep-wake classification but also hold promise for extending to more comprehensive sleep stage classification, potentially advancing the field of automated sleep assessment through efficient and user-friendly wearable monitoring systems.

Generative AI is designed to create new content from trained parameters. Learning from large amounts of data, many of these models aim to simulate human conversation. Generative AI is being applied to many different sectors. Within healthcare there has been innovation specifically towards generative AI models trained on electronic medical record data. A recent review characterizes these models, their strengths, and weaknesses. Inspired by that work, we present our evaluation checklist for generative AI models applied to electronic medical records.

Boussina et al. recently evaluated a deep learning sepsis prediction model (COMPOSER) in a prospective before-and-after quasi-experimental study within two emergency departments at UC San Diego Health, tracking outcomes before and after deployment. Over the five-month implementation period, they reported a 17% relative reduction in in-hospital sepsis mortality and a 10% relative increase in sepsis bundle compliance. This editorial discusses the importance of shifting the focus towards evaluating clinically relevant outcomes, such as mortality reduction or quality-of-life improvements, when adopting artificial intelligence (AI) tools. We also explore the ecosystem vital for AI algorithms to succeed in the clinical setting, from interoperability standards and infrastructure to dashboards and action plans. Finally, we suggest that algorithms may eventually fail due to the human nature of healthcare, advocating for the need for continuous monitoring systems to ensure the adaptability of these tools in the ever-evolving healthcare landscape.

The utilization of artificial intelligence (AI) in diabetes care has focused on early intervention and treatment management. Notably, usage has expanded to predict an individual’s risk for developing type 2 diabetes. A scoping review of 40 studies by Mohsen et al. shows that while most studies used unimodal AI models, multimodal approaches were superior because they integrate multiple types of data. However, creating multimodal models and determining model performance are challenging tasks given the multi-factored nature of diabetes. For both unimodal and multimodal models, there are also concerns of bias with the lack of external validations and representation of race, age, and gender in training data. The barriers in data quality and evaluation standardization are ripe areas for developing new technologies, especially for entrepreneurs and innovators. Collaboration amongst providers, entrepreneurs, and researchers must be prioritized to ensure that AI in diabetes care is providing quality and equitable patient care.

We explore the evolving landscape of diagnostic artificial intelligence (AI) in dermatology, particularly focusing on deep learning models for a wide array of skin diseases beyond skin cancer. We critically analyze the current state of AI in dermatology, its potential in enhancing diagnostic accuracy, and the challenges it faces in terms of bias, applicability, and therapeutic recommendations.

Navigating contemporary healthcare, wearable technology and smartphones are marking the dawn of a transformative era in patient observation and personalised care. Wearables, equipped with various sensing technologies (e.g., accelerometer for movement, optics for heart rate), are increasingly being recognised for their expansive potential in (remote) patient monitoring, diagnostics, and therapeutic applications which suggests a plausible move towards a more decentralised healthcare system. This shift is evident as healthcare providers and patients alike are becoming increasingly accepting of wearable-driven tools, as they enable continuous health monitoring outside of traditional clinical settings. Equally, the ubiquitous nature of smartphones, now more than mere communication tools, is being harnessed to serve as pivotal health monitoring instruments. Their added sensing capabilities with Internet of Things (IoT) driven connectivity enable a (relatively) seamless transition from conventional health practices to a more interconnected, digital age. However, this evolving landscape is not without its challenges, with concerns surrounding data privacy, security, and ensuring equitable access to digital advances. As we delve deeper into digital healthcare, we must harness the full potential of those technologies and ensure their ethical and equitable implementation, envisioning a future where healthcare is not just hospital-centric but is part of our daily lives.

Contemporary wearables like smartwatches are often equipped with advanced sensors and have associated algorithms to aid researchers monitor physiological outcomes like physical activity levels, sleep patterns or heart rate in free-living environments. But here’s the catch: all that valuable data is often collected separately because the sensors don’t always play nice with each other, and it’s a real challenge to put all the data together. To get the full picture, we may often need to combine different data streams. It’s like putting together a puzzle of our health, instead of just looking at individual pieces. This way, we can gather more useful info and better understand health (it’s called digital twinning). Yet, to do so requires robust sensor/data fusion methods at the signal, feature, and decision levels. Selecting the appropriate techniques based on the desired outcome is crucial for successful implementation. An effective data fusion framework along with the right sensor selection could contribute to a more holistic approach to health monitoring that extends beyond clinical settings.

Gregoor et al. evaluated the healthcare implications and costs of an AI-enabled mobile health app for skin cancer detection, involving 18,960 beneficiaries of a Netherlands insurer. They report a 32% increase in claims for premalignant and malignant skin lesions among app users, largely attributed to benign skin lesions and leading to higher annual costs for app users (€64.97) compared to controls (€43.09). Cost-effectiveness analysis showed a comparable cost to dermatologist-based diagnosis alone. This editorial emphasizes the balance in AI-based dermatology between increased access and increased false positives resulting in overutilization. We suggest refining the diagnostic schemas with new referral pathways to capitalize on potential savings. We also discuss the importance of econometric analysis to evaluate the adoption of new technologies, as well as adapting payment models to mitigate the risk of overutilization inherent in AI-based diagnostics such as skin cancer detection.

AI-based prediction models demonstrate equal or surpassing performance compared to experienced physicians in various research settings. However, only a few have made it into clinical practice. Further, there is no standardized protocol for integrating AI-based physician support systems into the daily clinical routine to improve healthcare delivery. Generally, AI/physician collaboration strategies have not been extensively investigated. A recent study compared four potential strategies for AI model deployment and physician collaboration to investigate the performance of an AI model trained to identify signs of acute respiratory distress syndrome (ARDS) on chest X-ray images. Here we discuss strategies and challenges with AI/physician collaboration when AI-based decision support systems are implemented in the clinical routine.

Artificial intelligence systems are increasingly being applied to healthcare. In surgery, AI applications hold promise as tools to predict surgical outcomes, assess technical skills, or guide surgeons intraoperatively via computer vision. On the other hand, AI systems can also suffer from bias, compounding existing inequities in socioeconomic status, race, ethnicity, religion, gender, disability, or sexual orientation. Bias particularly impacts disadvantaged populations, which can be subject to algorithmic predictions that are less accurate or underestimate the need for care. Thus, strategies for detecting and mitigating bias are pivotal for creating AI technology that is generalizable and fair. Here, we discuss a recent study that developed a new strategy to mitigate bias in surgical AI systems.

Health and wellness/well-being are multifaceted topics further complicated when trying to understand the environmental impact. Typically, there has been a one size fits all approach when trying to understand the 3-way interaction, but that is a limited approach. Equally, measurement (of each) has often used a limited set of outcomes during short periods to provide insight. A more robust understanding of health and well-being within environments may require longitudinal/continuous assessment that holistically targets individuals. Therefore, there is a growing requirement for careful data management, individual-first methodologies, scalable research designs and new analytical approaches, e.g., artificial intelligence. That presents many challenges but interesting research opportunities for the field of digital medicine.

Digital health technologies (DHTs) have brought several significant improvements to clinical trials, enabling real-world data collection outside of the traditional clinical context and more patient-centered approaches. DHTs, such as wearables, allow the collection of unique personal data at home over a long period. But DHTs also bring challenges, such as digital endpoint harmonization and disadvantaging populations already experiencing the digital divide. A recent study explored the growth trends and implications of established and novel DHTs in neurology trials over the past decade. Here, we discuss the benefits and future challenges of DHT usage in clinical trials.

The usage of digital devices in clinical and research settings has rapidly increased. Despite their promise, optimal use of these devices is often hampered by low adherence. The relevant factors predictive of long-term adherence have yet to be fully explored. A recent study investigated device usage over 12 months in a cohort of the electronic Framingham Heart Study. It identified sociodemographic and health-related factors associated with the long-term use of three digital health components: a smartphone app, a digital blood pressure cuff, and a smartwatch. The authors found that depressive symptoms and lower self-rated health were associated with lower smartwatch usage. Female sex and higher education levels were associated with higher app-based survey completion. Here, we discuss factors predictive for adherence and personalized strategies to promote adherence to digital tools.

Artificial intelligence (AI) and natural language processing (NLP) have found a highly promising application in automated clinical coding (ACC), an innovation that will have profound impacts on the clinical coding industry, billing and revenue management, and potentially clinical care itself. Dong et al. recently analyzed the technical challenges of ACC and proposed future directions. Primary challenges for ACC exist at the technological and implementation levels; clinical documents are redundant and complex, code sets like the ICD-10 are rapidly evolving, training sets are not comprehensive of codes, and ACC models have yet to fully capture the logic and rules of coding decisions. Next steps include interdisciplinary collaboration with clinical coders, accessibility and transparency of datasets, and tailoring models to specific use cases.

Even as innovation occurs within digital medicine, challenges around equity and racial health disparities remain. Golden et al. evaluate structural racism in their recent paper focused on reproductive health. They recommend a framework to Remove, Repair, Restructure, and Remediate. We propose applying the framework to three areas within digital medicine: artificial intelligence (AI) applications, wearable devices, and telehealth. With this approach, we can continue to work towards an equitable future for digital medicine.