Health practice is increasingly influenced by digital systems. The reliance on information and communication technologies to enable how we understand human health conditions has become a fundamental requirement for practitioners and patients alike. It was a privilege to join more than 2000 attendees at the Australasian Institute of Digital Health’s[1] Healthcare Innovation Community (HIC) Conference in Brisbane, Qld, Australia (Aug 5–7, 2024) and experience the depth of the discourse and innovations happening in this space.
The Australasian Institute of Digital Health is a professional organization launched in 2020 as Australia and New Zealand’s leading forum in health informatics and digital health. Their priorities are to promote leadership and advocacy, workforce development, and community engagement within the field of digital health. The Healthcare Innovations Community event was a major draw for the health informatics community in the region, attracting delegates from government agencies, industry, start-up companies, and public health providers. Conspicuous, however, was the lack of a strong presence from research centers and academic universities, i.e., researchers engaged in data analytics and translational research.
The program for this conference consisted of Ted-talk style symposia and several focused sessions ranging from artificial intelligence (AI)-enabled care (generative AI, digital transformation, workforce improvements, workflow pathways, patient security, digital risk management) to compliance with FAIR Data Principles (including Fast Healthcare Interoperability Resources).[2] The focus on digital health encompassed how computer technology is integrated into clinical systems and patient care pathways. This technology was promoted simply as “tools” and “aids” for the health practitioners and service providers to make their lives easier, ensure fairness and equity in patient care, and optimize efficiency in how doctors use their time. A common point of discussion was how digital health systems will allow doctors “more time with their patients to give them the human interaction they need,” which was quickly countered with the observation that digital health systems allow doctors “to see more patients in a day!” There was an emphasis on how digital technology will create fairness within the health system through improved regulation of services, and the government-aligned speakers focused on how the public will have more agency over who accesses their own health records in the future.
With an incredibly full and diverse program, amazing presentations, a strong vendor display, and interactive sessions, the conference lacked one significant point of discussion—considerations about the nature of data itself. It was recognized that digitally enabled patient care could capture, generate, use, and understand data. How we achieve each of these elements has its own caveats and concerns, the solution of which is found in how we do the other. Data generation involving patients requires secure storage that must be regulated, which, as we create digital systems that generate more and more data, creates a need for exponential growth in enterprise architecture within our health infrastructure.[3]
“Data” in their most basic form has always underpinned healthcare. Patient records reflect how sick people are diagnosed and treated and how they respond. These data can provide a latent understanding of disease biology, symptomatic interpretation, treatment selection, clinical observations, patient management pathways, workforce distribution, interdisciplinary communication between practitioners, and the overall health benefits of our clinical care services. Digital health has emerged as and is indeed driven by the “data deluge”[4] of its own making. None of the digital systems discussed or on display within HIC2024 would be viable if not for the need to understand and benefit from ever-complex and voluminous data, which have, circuitously, been generated by the implementation of the digital systems on display and discussed. For example, generation of electronic patient records has shifted our medical records office from necessary hospital administrative departments to centers of health informatic inquiry and research opportunities.[5] Discussions with one vendor during HIC24 highlight this trend. This vendor provides software that allows clinicians to provide and receive patient referrals through a secure online portal, allowing hospitals to shift away from using fax machines to send referral letters. This simple, practical service may now incorporate AI-based technology to predict expected referral rates, thereby allowing hospitals to better manage patient transitions to specialists.[6]
Data analysis is now poised to become a complementary discipline within hospitals to support physicians, radiologists, and pathologists, as well as nursing and allied health domains, but not all data are usable or meaningful. Machine learning and AI-based models are data-intensive methods, requiring major standardized data collection efforts to provide the necessary training test and validation datasets to build robust and generalizable models. Such datasets often have excessively imbalanced numbers of patients represented within the different classes being compared, leading to model overfitting. Genomic data can yield highly dimensional data due to the large number of genes, but these data are often from a limited number of biospecimens studies, presenting analysts with a “curse of dimensionality” problem.[7] Advances in digital microscopy technology allow pathologists to benefit from deep-learning models that decipher libraries of high-resolution digital images and potentially predict underlying molecular and genetic events to be used in diagnosis.[8] Yet, the complexity of the tissue sections and image capture variability creates noisy data that prevent simple models from generalizing, thus preventing standardization of use across a health system. Similarly, generative AI allows better segmentation of radiographic, mammographic, and magnetic resonance imaging, which has shown promise in improving lesion identity, but digital analysis has yet to show a reliable benefit over a well-trained human examination.[9]
As we move into a Healthcare 4.0[10] digital world, the primary issues that need to be dealt with do not necessarily pertain to the technology. Rather, the quality of data and its management are essential to ensuring that digital health systems accurately reflect the human condition. Issues such as where the data comes from and how it should be managed to be truly meaningful need to co-align with how it will be understood, trusted, and applied to advance patient care. For machine-learning algorithms to have an impact, they need to accurately represent the data used to build the models. To impact patient care and clinical outcomes, the application of data analysis will need to occur at the clinical bedside, requiring real-time collection and immediate assessment of complex data. Physicians will be confronted with new “black boxes” containing data that will explain their patient’s diseases. This will require new levels of human–computer interactions, as healthcare providers are asked to engage with software that is more than a “tool” but rather an instructive guide for their decision-making, involving behavioral challenges as they work out whether to believe the computer over their own intuition. Healthcare providers will likely benefit from advanced training and education in math and computer science as much as from biology and chemistry as we shift from biospecimens to digital data as a new determinant for patient care decisions. Finally, this begs the question: are we entering into a self-perpetuating cycle of digital data generation, requiring ever more powerful digital systems to manage, store, and understand it so that it can be rolled out across our health services? And as this happens, how do we prevent losing sight of the humans (clinicians and patients alike) that digital health systems are designed to benefit? Such questions will likely emerge for future HIC conferences.
References
Competing Interests
Sources of support: None. Conflicts of interest: None.
Author notes
Vu H, Catchpoole DR. Data data everywhere: harnassing digital health.