Are Real-World Data and Evidence Good Enough to Inform Health Care and Health Policy Decision-Making?

Abstract

We are entering the era of digitized real-world data to inform health-care and health policy decisions---including medical claims, electronic health records, sensor/real-time monitoring data, etc. This will increasingly be merged with novel other data sources---including purchasing preferences self-reporting of experience via social networks, weather, etc. Together this big data has the potential to revolutionize decision-making and help to realize the goal of a ``learning health-care system.'' However, there are concerns that the data are not of sufficient quality---both in terms of accuracy and completeness. Are data good enough today? If not, when will we know when they are good enough? There are multiple ongoing and potential applications of real-world data/big data including understanding the epidemiology of disease and unmet medical need, informing the development of precision medicines, informing health-care benefit design, informing quality improvement efforts, informing health technology assessments regarding access to and pricing of new therapies, assessing the incidence/prevalence of adverse events associated with marketed medications to inform regulatory labelling, informing bedside shared decision-making between patient and provider, and informing regulatory labelling decisions regarding indications, dosing, benefits in subpopulations, etc. The criteria for what are good enough are not the same across these applications. For some uses, the data we have today are already good enough. For other uses, it remains controversial whether the data are good enough. A framework is discussed that allows end users to decide whether data are of sufficient quality to inform decision-making.