Technology and Data in Healthcare

A tide of health data and digital technologies led by artificial intelligence (AI) is sweeping away long-held preconceptions about global health and healthcare access and provision. Healthcare systems are complex and are in a constant state of flux, characterized by incremental cycles of learning, adaptation, and change. Making healthcare changes endure is difficult. A 2017 global review across 60 health systems found that four factors were common in successful efforts to achieve effective, lasting change: (a) the acorn-to-oak tree principle (that is, a small initiative that tackles a specific problem may lead to systemwide impacts and change, instead of large, systemwide, long-duration reforms); (b) the data-to-information-to-intelligence principle (the role of information technology and data is becoming more critical for delivering efficient and appropriate care, but must be converted into useful intelligence); (c) the many hands principle (concerted action among stakeholders is key); and (d) the patient- as-the-preeminent-player principle (individuals are at the center of change).

In all four of these areas, technology and data are relevant. Technology and data—designed and implemented iteratively with and by patients and stakeholders—are clearly catalytic components of the current wave of healthcare system changes. Digital technology and data can add immense value to health systems and to the health of the population in several ways. Digital technology strengthens health systems, improves health financing, makes public health more effective, and reaches underserved populations. Used effectively, digital technology makes health services more personal, prevents health care costs from increasing, reduces differences in care, and makes the provision of health services easier. Digital technology supports progress toward universal health coverage. Digital technology and data by themselves are not sufficient to realize universal health coverage, but they are necessary. Governments and other actors may not be able to address health challenges fully without them.

Why the time is ripe for Digital Health?

In responding to COVID-19, new digital technologies were rapidly deployed, creating momentum. The COVID-19 pandemic brought about substantial changes in the way health care services are delivered, leading to an unprecedented surge in the use of digital tools for health service provision, the promotion of public health, and the administration of COVID-19. Technology was useful and important not only in emergency type services that COVID-19 required, but also for the durability of routine health services. The pandemic offered an opportunity to accelerate the implementation of digital health solutions that may have been recognized as options prior to COVID-19 and to understand the preconditions that favored the implementation of such solutions.

The amount of data on health and health care is growing at staggering rates, but these data are not yet being used to maximum potential. Countries have multiple sources of data about health. Some are generated in the health sector, while other data on health are located outside the health sector, such as data on social grants or the educational status of an individual. Up to 2025, health data are anticipated to exhibit the highest compound annual growth rate of data in any sector. The academic literature—approximately 30 percent of which is relevant for the health sector—has exponentially increased in the last several decades. More than half of it is available in open-access journals. These data are not being used. It takes, on average, 15+ years for new medical evidence to be translated into routine clinical practice. That duration represents almost almost half the career of a practicing healthcare professional.

A large number of organizations have or are developing digital health strategies, but these have not yet been implemented. Many digital health strategies focus on cybersecurity, telehealth, digital patient administration, electronic health records, data analytics, performance management, and workflow simplification.

Inhibitors of broader digital health adoption

There is increasing awareness of what digital health is and what it offers, but digital health is not uniformly understood. Local market conditions directly impact the level of digital health maturity and utilization. There is a lack of appreciation of the extent of foundational investments needed (such as connectivity) to make digital health interventions work. Awareness about the need for health standards for data exchange, security, data protection, and hardware has increased, but more should be done. Selecting the digital interventions to implement is difficult, as is figuring out how to integrate these interventions into the existing architecture and systems. This is a key inhibitor of broader digital health adoption. Data access is a constant challenge, as is the lack of the ability to analyze the available data given the siloed nature of locations. Skills in analyzing data are a significant issue because of the skill gap.

Furthermore, changes facing the healthcare are beyond the notion of a new technology or digital solution to digitize an existing process: more profound than that, the changes brewing in the healthcare are existential in nature -- person-centered health care, embracing new medical and health discoveries, the integration of previous separate disciplines – an expanded understanding of what is necessary to live life well. Healthcare is set to transform into a system that's centered around the patient, focusing on virtual and at-home treatments as well as linked outpatient care. Driven by data and analytics, it will prioritize value and bear risks while embracing transparency and interoperability. Enabled by cutting-edge medical technologies, it will become an integrated yet fragmented system.

Opportunity (and Risk) associated with Generative AI

The new generation of conversational software (initially, mainly chatbots), such as Open AI’s ChatGPT is a computer program that responds to a question (or prompt). What is special about it is that it uses neural networks and a vast amount of publicly available data to formulate its responses. It relies on algorithms that were already trained in language structure on large datasets, and, so, it does not need significant new data for further prediction. Accurate and conversational medical chatbots as entry points to healthcare. COVID-19 unleashed a rush in many countries to set up telemedicine services. If medically accurate chatbots were available for initial triaging and trusted by the public and providers, they could make telemedicine more effective. A medical chatbot could perform the first level of triage, prior to referral to a health facility (if needed).

It is the Generative AI Analytics that has potential for healthcare especially if is has been trained on medical texts and images and has a foundational medical language in which it can communicate. AI image generators, including Dall-E and Stable Diffusion, represent another family of GenAI software that can generate original images from scratch once the image has been described in words or it can extract text from images. Thus, it can interpret images and explain them in words or create images based on word descriptions. These technologies have application potential in the healthcare, particularly because the analytics that underpin Generative AI tools are pointed inward using health records and other medical data. Applications require careful thought. These tools can augment the efforts of healthcare professionals and streamline how services are provided. So, they are augmented intelligence tools rather than artificial intelligence tools.

Because these generative AI solutions are pre-trained, they do not require large datasets, complicated machine learning, or high-level skills to operate. However, with promise comes potential peril. There are real challenges that need urgent answers:

Regulation - Medical devices that use AI to derive (potential) diagnoses or recommend treatments need to be subject to rigorous regulatory processes akin to pharmaceutical approval processes (regulation either as medical devices or as medical products). Regulatory mechanisms should also include human-in-the-loop principles to implement practices that allow for humans to have oversight for validating models.
Trust - GenAI tools will only enjoy wide scale use if they are trusted and valued as an integral part of the health care delivery network, augmenting the work of health workers, making their jobs easier and allowing them more time with patients.
Data protection - Patient privacy will be more at risk than ever in what could be a rush to bring all the disparate medical data together. It’s important to ensure that personal data is used only for limited and identifiable legitimate purposes, that only that data which are necessary for the purpose are collected/processed, that data subjects have certain rights over their data, and that data collectors/processors have certain obligations in how they handle personal data.
Bias - Bias in AI systems and the ethics surrounding the development and use of these systems has been a long-standing debate and GenAI solutions are not exceptions to these issues. Addressing bias in AI systems will require continuous research in understanding and removing bias, regulatory mechanisms to establish responsible processes for mitigating risks, medical education and improving access to representative and high-quality health data.

As healthcare undergoes this next wave of transformation to reduce inefficient processes incrementally, and deliver new, better and seamless services, technology and data will be an integral part of it. But, technology will not drive the change: it will support, augment and accelerate the changes that health systems will continue to undergo. Looking forward, big data models, telemedicine, predictive medication, wearable sensors and a wealth of new platforms and apps could help us rethink how the world provides, accesses and manages health and healthcare.

Data and Digital Technologies for better Healthcare Access