EPA’s Center for Public Health and Environmental Assessment is comprised of multidisciplinary teams that develop assessments synthesizing the current state of the science in a variety of topic areas supporting federal and state decision-making, including under the Clean Water Act and Clean Air Act. Examples include assessments of the impacts of ozone on ecological endpoints including plant growth and foliar injury and assessments of the relationship between instream nutrient concentrations and biological endpoints. Decision-makers have a strong preference or need for assessments to consider all relevant peer-reviewed scientific literature (often hundreds to thousands of studies). The adaptation of systematic review (SR) principles for ecological risk assessment provides a framework for collecting secondary data efficiently and transparently using predefined processes to defensibly present the state of the science on a particular topic. These methods and technical advancements offer significant efficiencies in searching and screening the literature, extracting data, and developing visualizations and conducting quantitative analyses to support scientific conclusions. Here, we review some of the advancements our teams have applied to date in the assessment development process and future envisioned improvements to bring comprehensive, transparent, and timely assessments to decision-makers.
Results/Conclusions
We successfully implemented new SR tools to develop the ecological criteria of the Ozone Integrated Science Assessment (ISA). We used machine learning to map results from literature searching and citation mapping to discipline-specific topics. As a result, scientists screened and prioritized literature relevant to their expertise, and only had to manually review 1,032 out of 12,247 abstracts to identify 95% of relevant studies. Structured data entry into evidence tables then streamlined the presentation of results in the ISA. Results of the literature search and screening and the studies cited were documented in a publicly accessible database. In addition, we explored expansive literature searching and increased our evidence base for synthesizing the relationship between instream nutrient concentrations and biological endpoints by approximately 1/3 using citation mapping, targeted website searching, and expert elicitation above and beyond conventional keyword database searching. Building from advancements in literature searching and screening, future efforts will include adapting software tools for data extraction from primary literature, and employing data analysis and visualization software. This will be rolled into an evergreen workflow where our critical areas of scientific evidence are continually updated, thereby reducing barriers to delivering scientific assessments to decision-makers. Our ecological data revolution has just begun.