Central to these approaches are so-called machine learning algorithms, methods that are ideally suited to uncover robust associations amidst complex, high-dimensional data. Machine learning algorithms have helped advances in our ability to understand the human genome, process vast unstructured data on social media sites to identify drug safety events and underpin the emergence of a range of physician decision-support tools from medical image diagnosis to prediction of unplanned hospital admissions. Predictive analytics holds the key to unlocking some of the potential of big data. Now health care needs to unlock its resistance and lack of familiarity with these techniques and embrace novel analytical methods.
This symposium will review how predictive analytics applied to big and complex data provide opportunities but also pose challenges to outcomes researchers. Case studies will illustrate how such novel analytical approaches enable new ways of doing outcomes research, e.g. comparing traditional methods to machine learning techniques in predicting non-adherence to therapy or examples of the use of predictive analytics to identify patients at risk of future events.
While the concept of value for money has been prevalent in other therapeutic areas for a number of years, assessing value and potentially managing access to innovative therapies in oncology represents a substantial shift for markets such as the Canada, Germany, and the United States.
This symposium will provide a platform to discuss the current and future trends in assessing value and managing access in oncology for these three key markets, in order to highlight key similarities as well as differences with respect to the process and evidence requirements for manufacturers bringing innovative treatments to market.
The panel will discuss approaches to determining whether to measure utility in a trial (versus an alternative study type), challenges of balancing the requirements for reimbursement and regulatory authorities in a single trial, whether EQ-5D is an appropriate measure (or an alternative measure is justified), optimal timing of assessments, and specification of analyses to utilize the power of patient-level utility data.