Opening Remarks
As part of the Pacific Symposium on Biocomputing 2021
January 7, 2021
Artificial intelligence (AI) and other computational and bioinformatic approaches have become a critical component of biomedical research. The wealth of available medical data and pertinent research questions have driven experts across many scientific fields to begin developing computational methods to drive innovation in medical research. However, AI in healthcare is often labelled as “disruptive” a word simultaneously embracing its innovative nature, while warning against its turbulent impact on a broad range of health-care related disciplines. As a result, many healthcare stakeholders continue to be reserved, and even outright resistant, to AI advances for clinical outcomes.
Healthcare stakeholders include researchers across a variety of disciplines: clinicians, patients, insurers, legislators, lawyers, economists, UN agencies, governments, private and non-profit organizations, to name a few. There is no straightforward strategy for creating meaningful involvement mechanisms across many healthcare stakeholders. In this workshop, we aim to invite talks focusing on AI approaches in biomedical research from diverse and inclusive research teams, with expertise that spans different academic and professional disciplines, or who have collaborated with or studied the perspective of various stakeholders of computational healthcare research. Specifically, talks will emphasize both lessons learned from collaborative research as well as how the collaboration influenced the design, interpretation and over-all positioning of the results, and provide advice for how other researchers can engage in their stakeholder community.
10:00 AM PST
10:05 AM PST
The COVID-19 pandemic has highlighted the importance of telemedicine. The field of field of biomedical data science now has more motivation than ever to innovate mobile medical tools that fill gaps in the healthcare ecosystem. In this talk, I will describe the opportunity space, barriers to translation and ways to overcome them, and describe several solutions from my lab and others that have, or nearly have, crossed the divide to reimbursed clinical use.
10:30 AM PST
Tracking health indicators is crucial for successfully managing chronic illness, yet many health informatics tools do not adequately support the tasks and communication required. We used qualitative methods to construct a conceptual model of shared health informatics from interviews with and presentations given by people managing a range of chronic illnesses. From our data we identified the need for a health informatics model that (1) incorporates the ongoing nature of tracking work and (2) represents the social dimension of tracking for illness management. Our model depicts communication, information, collection, integration, reflection, and action work in the social context of the person with chronic illness, informal carers, health care providers, and community members. We will also share insights from this study and other human-computer interaction work into incorporating the user’s perspective into research and design of tools to support health.
10:55 AM PST
Lack of access to online resources and digital infrastructures means that many people in remote and rural communities are prevented from equitably participating in online commerce, education, data sharing, knowledge mobilization, democratic processes, and more. This is particularly true in northern communities in Canada where cellular infrastructure and internet are non-existent or substantially lacking. This presentation will describe work that has been led by the Inuit community of Rigolet, Nunatsiavut to bridge the digital divide through the use of wireless mobile ad-hoc networks.
11:20 AM PST
Understanding human biology is important to address disease states, develop remedies, and improve survival from illness. The approach for research to solve or address challenges in biology involves people to observe, collate, investigate, interpret, and report critical findings, using tools to aid in the collection of data. Diverse person groups of problem solvers outperform the groups of best individuals at solving these complex problems, increasing innovation and resolving challenges at faster paces. Additionally, diversity of teams, particularly ethnic diversity, consistently outperforms the non-diverse to enhance scientific impact. Tools such as computers that utilize artificial intelligence algorithms learn based on inputs into the learning processes; similarly to the lack of diverse person teams, the predictive output of these algorithms may not have pertinent findings that apply to all populations if the inputs were not wholly inclusive. There is no doubt that AI and machine learning improves prognosis and diagnostic accuracy and improve preciseness of findings on radiological and pathological imaging. As with most genetic datasets published, the input data of which the machine learns from often come from single homogenous and not diverse populations or datasets, and the output might not apply to those populations for disease accuracy and prognosis.
11:45 AM PST
Increased scrutiny of the risks and benefits of artificial intelligence (AI) applications in healthcare highlights the need for real-world evaluations of patient-centered outcomes. The complexity of healthcare, compounded by the user- and context-dependent nature of AI applications, calls for a multifaceted approach beyond traditional in silico evaluation of AI. I will discuss the need for designing and evaluating patient-centered AI through phased research, drawing analogies to clinical research frameworks for drugs and medical devices.
12:10 PM PST
Research into artificial intelligence tools aimed at improving clinical outcomes need to evaluate not only technical performance, but socio-technical performance outcomes. In this talk, I will be presenting how the development of Sepsis Watch, a tool which assists clinicians in the early diagnosis and treatment of sepsis, challenged us to develop a new paradigm of evaluating medical AI in a social context. One of the critical factors influences the potential of Sepsis Watch to improve septic patient outcomes was the integration of the tool into existing social and professional dynamics. I will emphasize the importance stakeholders who are required to act as ‘AI Diplomats’ and are responsible for much of the repair work needed to successful launch AI tools in a way that minimizes disruptions in hierarchies, expertise, and work cultures. As well, I’ll present the four key values necessary for the translation of biocomputing research into medical practice: rigorously defining the problem, building relationships with key stakeholders, respecting professional discretion, and creating an ongoing feedback loop with stakeholders.
12:35 PM PST