HEART OF THE MATTER: UNRAVELLING AND OVERCOMING "BIAS" IN CARDIOVASCULAR RESEARCH (CCC-102)

***Kahoot! will be used for this workshop. Participants can download the application on their iOS or Android device prior to the session or join through their device browser during the session***

Session Description: "The findings of this study may be subject to potential biases” - a routine statement in almost all scientific papers. Although sources of bias are frequently recognized in health research, the implications that bias may have on study results are often ignored, poorly explained, and not quantified, thus leading to false or misleading conclusions. Bias is defined as “an error in the conception and design of a study – or in the collection, analysis, interpretation, reporting, publication, or review of data – leading to results or conclusions that are systematically (as opposed to randomly) different from truth” (Porta, 2016). Bias can be introduced into various study designs, occur at any stage of the research process, and diminish the validity of results. These errors are often complex, subtle, and difficult to detect, which complicates interpretation, distorts scientific reasoning, and muddles knowledge dissemination.

Bias can have serious consequences for cardiovascular care and public health initiatives. Cardiovascular research is the foundation for identifying risk factors, developing preventative measures, and improving treatments for cardiovascular disease. Inaccurate or misleading research, as a result of bias, may compromise the effectiveness of health interventions, reduce the generalizability of research findings to diverse patient populations, or offset clinical decision-making. To evaluate bias in cardiovascular research, novel analytical strategies, such as quantitative bias analysis, and assessment tools have emerged. As artificial intelligence (AI) approaches provide new opportunities to transform cardiovascular medicine, surveillance of intrinsic and extrinsic biases that affect quality control and algorithmic fairness have become increasingly important. For example, the collection of biased data used to train AI algorithms can alter accuracy and reduce the generalizability of models to marginalized populations. Careful attention to data collection, model development, and ongoing evaluation of AI algorithms are crucial to recognizing and reducing bias.

The focus of this workshop will be to introduce and summarize key bias-related topics in cardiovascular research. This workshop will begin with three short presentations to: (1) provide an overview of the most common types of biases observed in cardiovascular epidemiologic studies; (2) highlight strategies to evaluate and minimize the impact of bias using case-based examples and assessment tools, (3) describe how bias may influence the development and interpretation of cardiovascular AI algorithms. Attendees will then participate in an interactive, game-based session to apply their knowledge gained throughout this workshop. This workshop is supported by the Canadian Cardiovascular Research Collaboratory and Institute for Circulatory and Respiratory Health.