Director, OHDSI Center Northeastern University & University of Oxford University of Oxford, NDORMS Portland, United States
Background: In the 1990s, healthcare facilities had begun the transition from paper records to electronic health record (EHR)s. The increasing accessibility to digital health data and the rising costs and recognized limitations of randomized clinical trial (RCT)s made room for the emergence of real-world data (RWD). While life science organizations from biopharma companies to contract research organizations (CROs) have embraced RWD, this is the first study to our knowledge to assess the extent to which RWD is utilized in public health research.
Objectives: This study examines trends in the use of RWD for public health research over the past 5 years (2017-2021) and to summarize the most used data sources among those research articles that use RWD.
Methods: Applying a search algorithm to PubMed we identified original research articles published in the American Journal of Public Health (AJPH) from 2017 to 2021. Using the U.S. Food and Drug Administration’s definition of RWD, abstracts were reviewed to determine whether RWD was the primary source for analytic data. To build reliability of assessing whether the primary source for analytic data satisfied the RWD definition, for the first 500 manuscripts any articles in question were brought to a small committee for consensus. Articles which used RWD were then categorized based on the primary source of that data.
Results: Of the 1,041 articles reviewed, the percentage of articles which utilized RWD remained consistent from 2017 to 2021, ranging from 22% in 2019 to 29% in 2021 (p-value for trend=0.5). The most common source was vital statistics, fatality, and mortality, comprising 21% of RWD articles, closely followed by surveillance data, encompassing 20% of RWD articles. There were more articles with social media as the primary source of RWD in 2020 (4% of RWD articles) than in all other years combined. Key sources of RWD for regulatory decision making such as EHR, claims data, prescription data, and patient registries comprised 29% of RWD sources in AJPH public health research.
Conclusions: The use of RWD in AJPH for public health research has remained consistent over the past five years. Within specific areas of public health, the use of RWD and specifically EHR, claims and registry data remains limited. Further assessment of the utility of RWD in public health is needed to determine whether this limited use of RWD result from RWD not having elements needed for this research or if RWD is underutilized in public health.
