Ebook: Health. Innovation. Community: It Starts With Us

The digital health community continued to explore the transformative potential of artificial intelligence (AI) and machine learning at HIC 2024.

Health and care are increasingly delivered in a digital world, but nothing promises to change the way we work and live like AI, generative AI, and machine learning.

Other key topics in the program, including workforce, interoperability, FHIR, standards, UX and usability, and cybersecurity, continue to demonstrate how digital health leaders and innovators are steadily advancing towards connected care.

In this rapidly developing environment, the voice of the consumer and the clinician has never been so vital. It was pleasing to see the number of consumer representatives, generously supported through sponsorship, in attendance as delegates.

This year, for the first time at HIC, a symposium was held for early and mid-career researchers supported by Academic Translational Research in Digital Health. Designed for graduate and postgraduate students, PhD candidates, postdoctoral fellows, and early career faculty, it focused firmly on digital health research and supported those looking at research careers.

For the Australasian Institute of Digital Health, continually evolving and changing to serve our members, the conference has morphed into a new model: Health. Innovation. Community (HIC). It’s a new way of broadening our message, connecting the digital health ecosystem and engaging with the Australian community.

The theme of HIC 2024 is It Starts With Us. It acknowledges that, as a community, we are all responsible for initiating and creating positive change. HIC 2024 envisions a future where collective insights, shared knowledge, cross-disciplinary collaboration, and enthusiasm drive systemic improvements.

Together, the digital health community can help deliver a more responsive, inclusive, innovative, and human-centric healthcare system.

Digital health is more than the development and adoption of new technologies. It opens the door to new ways of thinking, spurring innovation in connected care that will ultimately enable accessible, equitable healthcare for all Australians.

On behalf of the Australasian Institute of Digital Health Board, I would like to thank the HIC 2024 Organising Committee, co-chaired by Dr Tanya Kelly and Professor Adam Dunn, for their leadership in delivering this important conference.

Damian Green (FAIDH)

Acting Chair

Australasian Institute of Digital Health

Deputy Director General

Corporate Services Queensland Health

Healthcare staff are typically exposed to hundreds of alarms each shift; researchers believe that the number of alarms may be more than 1000 alarms per shift. This may lead to staff becoming desensitised to the alarm, leading to a delayed or inadequate response to the alarm, known as alarm fatigue. Clinical-grade wireless smartphones are now carried by clinical staff to receive calls and notifications from a nurse call system.

Standardised nursing terminologies (SNTs) support the visibility of nursing work and documentation, enabling data sharing and comparison. An online survey assessed the knowledge and use of SNTs and revealed barriers and enablers to their use by Australian nurses. Just over half of the respondents were familiar with SNTs before the survey, a quarter reported a reasonable understanding of SNTs, just under half reported previous use of a SNT, and less than 14% indicated a current use of a SNT in their workplace. Perceived benefits to SNTs identified by respondents included a reduction in variation and the ability to evaluate the effectiveness of nursing care by measuring outcomes. Both barriers and enablers to the use of SNTs included education and training, standardisation and contextualisation across Australia, and integration into any electronic medical record system. Nurses are poorly informed on what SNTs are and how they can be leveraged to support their work and documentation. There is a need for an Australia-wide strategic approach to ensure the future of nurses’ work is visible, and SNTs are purposefully and correctly implemented across the country.

A National Nursing and Midwifery Digital Health Capability Framework was developed in 2020 to define and complement digital health knowledge and skills for professional practice. This mixed-methods study explored nurses’ and midwives’ Framework knowledge and its applicability, types and extent of nursing and midwifery informatics roles, and barriers and enablers to working in informatics. Survey respondents reported familiarity with the Framework, with content analysis identifying three themes - informatics as part of nursing or midwifery roles, descriptions and variability of informatics roles, and the need for informatics role development and recognition. Framework knowledge can be improved through standardised and defined roles and a career pathway that includes national, organisational, local, and individual support.

While Fast Healthcare Interoperability Resources (FHIR) clinical terminology server enables quick and easy search and retrieval of coded medical data, it still has some drawbacks. When searching, any typographical errors, variations in word forms, or deviations in word sequence might lead to incorrect search outcomes. For retrieval, queries to the server must strictly follow the FHIR application programming interface format, which requires users to know the syntax and remember the attribute codes they wish to retrieve. To improve its functionalities, a natural language interface was built, that harnesses the capabilities of two preeminent large language models, along with other cutting-edge technologies such as speech-to-text conversion, vector semantic searching, and conversational artificial intelligence. Preliminary evaluation shows promising results in building a natural language interface for the FHIR clinical terminology system.

Encapsulating a patient’s clinical narrative into a condensed, informative summary is indispensable to clinical coding. The intricate nature of the clinical text makes the summarisation process challenging for clinical coders. Recent developments in large language models (LLMs) have shown promising performance in clinical text summarisation, particularly in radiology and echocardiographic reports, after adaptation to the clinical domain. To explore the summarisation potential of clinical domain adaptation of LLMs, a clinical text dataset, consisting of electronic medical records paired with “Brief Hospital Course” from the MIMIC-III database, was curated. Two open-source LLMs were then fine-tuned, one pre-trained on biomedical datasets and another on a general-content domain on the curated clinical dataset. The performance of the fine-tuned models against their base models were evaluated. The model pre-trained on biomedical data demonstrated superior performance after clinical domain adaptation. This finding highlights the potential benefits of adapting LLMs pre-trained on a related domain over a more generalised domain and suggests the possible role of clinically adapted LLMs as an assistive tool for clinical coders. Future work should explore adapting more advanced models to enhance model performance in higher-quality clinical datasets.

Social media offers a rich source of real-time health data, including potential vaccine reactions. However, extracting meaningful insights is challenging due to the noisy nature of social media content. This paper explores using large language models (LLMs) and prompt engineering to detect personal mentions of vaccine reactions. Different prompting strategies were evaluated on two LLM models (GPT-3.5 and GPT-4) using Reddit data focused on shingles (zoster) vaccines. Zero-shot and few-shot learning approaches with both standard and chain-of-thought prompts were compared. The findings demonstrate that GPT-based models with carefully crafted chain-of-thought prompts could identify the relevant social media posts. Few-shot learning helped GPT4 models to identify more of the marginal cases, although less precisely. The use of LLMs for classification with lightweight supervised pretrained language models (PLMs) found that PLMs outperform LLMs. However, a potential benefit in using LLMs to help identify records for training PLMs was revealed, especially to eliminate false negatives, and LLMs could be used as classifiers when insufficient data exists to train a PLM.

Hospital bed occupancy serves as an important indicator of healthcare system efficiency, directly impacting patient care quality and staff workload. This study delves into the efficacy of midnight census, a conventional method for assessing bed occupancy, in supporting hospital operational planning. Historically, the midnight census has been utilised to gauge bed occupancy; however, its reliability is debated due to fluctuations throughout the day. This paper presents an analysis of 5.5 years of patient flow data from one of the hospitals in Queensland, Australia, scrutinising the statistical associations between different occupancy levels, e.g., midnight, peak, average, and minimum. The findings shed light on the efficacy of the midnight census and suggest the adoption of an hourly-based occupancy rate for more accurate capacity planning and management.

The HOTspots digital surveillance platform (HOTspots) is a critical technology of the HOTspots Surveillance and Response Program. It provides timely point-of-care access to pathology and demographic data from previously underserved regions. Co-designed with clinicians, epidemiologists, and health policy makers, the platform provides the evidence-base to empower efficient clinical management of patients with antimicrobial resistant (AMR) infections and supports national disease surveillance efforts in Australia. The pathway from conceptualisation to deployment for the HOTspots digital surveillance platform is described.

There is growing evidence for the benefits of eHealth interventions with Aboriginal and Torres Strait Islander people. Yet, there is a lack of guidance for culturally safe, relevant, and sustainable initiatives with Aboriginal and Torres Strait Islander peoples and organisations. To this end a research program was established to develop a roadmap for eHealth with Aboriginal and Torres Strait Islander peoples. The current phase of the research program is a review of the literature aimed at identifying the important characteristics of eHealth interventions with Aboriginal and Torres Strait Islander people. Thirty-nine publications reporting on a variety of eHealth modalities with Aboriginal and Torres Strait Islander people were identified. To assess the cultural quality of the final papers, the authorship applied the Aboriginal and Torres Strait Islander Quality Appraisal Tool (QAT). Results from the appraisal demonstrated significantly higher QAT scores between studies, including more Indigenous authors. This further substantiates the importance Aboriginal and Torres Strait Islander ways of knowing, doing, and being incorporating Indigenous worldviews and leadership have on the cultural quality of eHealth research studies.

This manuscript describes the conception and development of a novel, innovative digital health and informatics learning module designed specifically for entry-to-practice physiotherapy university programs. The design process involved consultation with stakeholders, alignment with contemporary digital health competency guidelines for health professional education, and educational design workshopping with faculty to ensure relevance and success. Key curriculum components include modules on health system transformation, design-thinking approaches, solution refinement and innovation pitching in the context of digital health. The subject intended learning outcomes (SILOs) were focused on digital health transformation, addressing the need for a curriculum on digital health transformation. This tertiary module aims to equip university graduates with essential knowledge and skills to thrive in a digitally enabled healthcare system by offering this framework for future health professional education in the digital age.

Advances in cancer treatment have improved patient outcomes and survival in recent decades. Increased complexity, duration, and individualisation of treatment protocols present an important challenge for care teams monitoring adherence to best-practice care. A rigid rules-based system for flagging outliers is not fit for purpose, as there are sound reasons for deviating from baseline protocols, such as the management of treatment side effects to a tolerable degree, however the methods for determining the bounds of appropriateness for variation are not well studied or understood. The development of digital representations to inform cancer care delivery in a timely and continuing manner is crucial. This scoping review seeks to identify gaps in current methods and propose a novel approach to digitally represent patient journeys in clinically meaningful visual and computational forms. These methods can be combined to produce real-time, clinically applicable tools such as group-level business-intelligence dashboards (are processes and resources adequate to ensure that patients are being treated according to best practice?) as well as individual-level decision support (what is the likely outcome for this patient if treatment is stopped early based on prior data?) and day to day clinical workflows (what has happened to this patient so far?)

Healthcare in the 21st century is experiencing tumultuous times of turbulent and tortuous change, characterised by an aging population, an increasing chronic disease burden, and inadequate workforce capacity to meet this burgeoning demand. The development of digital capabilities for health service managers of today and tomorrow requires an approach that transcends traditional adult learning education and training trajectories. A five-step process for developing an andragogical approach to health service management competency development in the digital context is proposed. This process includes qualifying the capabilities required of health service managers in the digital age and expediating digital transformation within the Australian healthcare environment, informed by empirical research; linking teaching approaches for digital health using the five tenets of adult learning; and coalescing the competencies needed to contextualise knowledge and skills development requirements for the 5th Industrial Revolution. Importantly, formal and informal education and training for health service managers should focus on competency transferability, which requires trainers and educators to understand the transforming context and challenges facing health service managers in the healthcare industry. It is incumbent on the digital health community to build capacity and enable workforce development to inculcate sustainable influence for lasting change in the healthcare system.

In the provision of patient-centered care, healthcare professionals face a challenging triad that is analogous to the classical “Three-Body Problem” in physics, whereby any shifts in the dynamics of medical knowledge, narrative competence and digital technologies can tip the balance of healthcare education and practice. Drugs/Medications are often caught in the dichotomy of being both “heroes” and “villains”, which underscores the necessity of a balanced, informed, and empathetic approach to patient care. Drawing inspiration from the popular novel and Netflix series, the dynamic interplay between the science of medical knowledge, the art of empathy and narrative competence, and the advancement of healthcare technologies to harmonise the diverse, yet interconnected domains, of medication management, medical humanities and digital health is demonstrated – in the emerging field of Digital Health Humanities. The novel project called M.A.G.I.C. is revealed, that blends the rigour of medical science with the nuanced understanding of patient narratives, and the cutting-edge potential of digital innovations like Generative Artificial Intelligence and the Metaverse, to create a holistic, learner-centered approach for the education of the healthcare professions.

Machine Translation (MT) has emerged as a crucial tool in bridging language barriers. In health settings, MT is increasingly relevant due to the diversity of patient populations, the dominance of English in medical research, and the limited availability of human translation services. Improvements in MT accuracy have prompted a re-evaluation of its suitability in contexts where it was once deemed impractical. This scoping review with meta-analysis delved into the appropriateness and limitations of MT in health, including in medical education, literature translation, and patient-provider communication. A keyword search in PubMed, PubMed Central, and IEEE Xplore produced peer-reviewed literature that focused on MT in a health context, published from 2018 to 2023. Analysis and mapping of full-text articles revealed 33 studies among 2,589 returned abstracts, indicating that MT is still unsuitable for direct use in patient interactions, due to clinical risks linked to insufficient accuracy. However, MT was showing promise further away from patients, for translation of medical articles, terminology, and educational content. Further research in improving MT performance in these contexts, coverage of under-studied languages, and study of the existing usages of MT are recommended.

The COVID-19 pandemic has negatively affected individuals with chronic conditions, such as type 2 diabetes mellitus (T2DM). The full extent of the impact however remains unknown, mainly due to the limited research availability. This study examines the pandemic’s impact on T2DM diagnosis and management in the United States. The methods include a literature review and an online survey of American healthcare professionals regarding their experiences of T2DM during the pandemic. Findings indicated significant reductions in healthcare utilisation among T2DM patients and a decline in the quality of care for this population. These reductions may have been attributed to fewer HbA1C tests being performed and emergency department visits, with a high proportion of individuals experiencing uncontrolled diabetes and receiving treatment intensification, especially among racial/ethnic minority groups, rural populations, and those with comorbidities. Effective strategies are needed to support T2DM regular follow-up and self-management, tailored to patient needs and culturally appropriated. Technologies like telemedicine can help address these needs, potentially reducing healthcare costs and improving clinical outcomes and quality of life for people with T2DM.

This paper describes clinicians’ views on the structure and content of an electronic discharge summary (EDS). A sample EDS template was developed by building on existing Australian guidelines to illustrate some of the proposed elements required for a high-quality clinical document. Surveys were widely disseminated to gather feedback and perspectives of hospital and primary care clinicians. A pragmatic approach to this study was underpinned by a strong evidence base and informed by implementation science methods. Key themes were identified, including variability in workflow and clinical needs, digital maturity, and digital health literacy of the clinical workforce. Understanding different workflows and priorities between hospital and primary care clinicians was a significant barrier to implementing a high-quality EDS. The strong consensus for change from both hospital and primary care clinicians, however, signaled the workforce’s readiness as a potential enabler of high-quality EDS documentation.

The “Emergency Department Pathology Order Support Tool” (ED-POST) is an electronic laboratory test ordering decision support tool that aims to decrease variation in test ordering practices. As part of a larger project on the co-design, development, and evaluation of ED-POST, this study aimed to explore the workflow nuances that might affect the intended use of the digital decision support tool. Semi-structured, in-depth interviews were conducted with 15 ED clinicians involved in the laboratory test ordering process across the development and evaluation phases of ED-POST. Participants identified the expanded role of registered nurses in test ordering and the practice of ordering tests that are outside the ED’s scope as contextual characteristics that can affect the use and perceived utility of the proposed ED-POST tool. Reconciling “work-as-imagined” with “work-as-done” in the design and development of electronic interventions is important in achieving interventions to improve the safe and effective use of pathology tests.

There are numerous behavioural, social and environmental factors that influence the symptomatology of a chronic health condition. These factors and how they manifest are often very specific to the individual, which creates challenges for applying macro population health approaches and insights to guide treatment. An artificial intelligence system, referred to as a non-axiomatic reasoning system (NARS), is presented. Learning in NARS is incremental and ongoing. A practical application of NARS in chronic pain management is demonstrated, as NARS can establish associations with behavioural activities that might exacerbate pain levels and revise the strengths of these associations over time. The system has potential application in any condition requiring patient-centric adaption.

An innovative chatbot incorporates a drawing tool allowing users to draw pictures that symbolise the nature of their chronic pain. Rather than simply ask ‘What did you draw?” what if the chatbot could engage the user to provide insight into symbols and images related to pain and suffering? The system was tested on images that included features of clinical interest in a first step toward a chatbot able to ‘see’ and discuss visual input. Such a chatbot has the potential for retrospective and prospective applications. *Contains childhood sexual abuse references*

An urgent global imperative is to ensure health workforces have a solid digital health capability to deliver consumer-centred services now and into the future. To address a lack of formal recognition of health informatics knowledge and skills in Australia, the Certified Health Informatician Australasian (CHIA) was launched ten years ago. In that time, 3125 CHIA candidates registered to have their capabilities assessed using a core set of health informatics competencies. This paper presents the demographic profile of people undertaking the examination and trend data of the pass rate over the decade, and identifies possible factors that may lead to higher achievement. Certifications such as CHIA are critical in the professionalisation of digital health specialists and health informatics.

Patient discharges from hospital to residential aged care facilities (RACFs) are often delayed due to inefficiencies in matching them to suitable available beds. To investigate the improvements that digital solutions can offer here, case-mix adjusted Length of Stay (LOS) of patients discharged to RACFs from a major metropolitan Victorian hospital were compared before and after the introduction of a digital solution for matching patient needs and preferences to available RACFs places. The study found that after the digital solution was implemented, the period where a patient would wait in the hospital for a RACF placement post the Aged Care Assessment Services (ACAS) assessment, the LOS reduced by 26.0% [95% confidence interval (95%CI):8.5%-40.5%]. The corresponding decrease in complete LOS for the After period was 16.5% (95%CI=1.5%-29.4%). This equates to a significant reduction in time spent by vulnerable patients in hospital while potentially freeing up 88,805 (95%CI:28,934- 137,864) patient bed days and delivering a cost saving of $105,993,947 (95%CI:$34,651,867-$165,105,956) if applied to the reported number of hospital patient days used by those eligible and waiting for residential age care nationally across the country in 2021-22.

Patient portals can support care delivery and have demonstrated mixed impacts on patients, clinicians and health services. These variable impacts are primarily due to barriers to uptake and challenges associated with successfully implementing patient portals. As little is currently known about how implementation barriers can be addressed in practice, this project used an action research approach to evaluate and optimise the implementation of a patient portal in a real-world setting. Whilst the study is ongoing, the preliminary findings are reported in this paper, along with plans to use the findings to improve the portal in the next round of the action research cycle.

Considering consumer preferences, this paper proposes computer-readable policy expressions to govern data sharing and use in consent-based health information exchange. Consent automation is an essential component of digital health transformation efforts, enabling better consumer choices in using their sensitive data while unlocking new values through policy-aware workflow expressions. The approach is based on the HL7 FHIR standard while using the obligation and responsibility concepts formally defined in the ISO ODP Enterprise Language (ODP-EL) standard. These concepts were expressed using a domain specific language (DSL) to support code generation for integration with digital health applications. The approach enables improved interoperability, considering consent, privacy, security requirements, and the ability to adopt new business, regulative or jurisdictional policies. Additionally, it accommodates generative AI solutions for faster alignment with existing document-based policies, ensuring clear provenance. Finally, consent automation and a broader policy framework in the context of the latest Australian Government initiatives in digital health are discussed.