Federated Automated Survey Tool (FAST)

Summary/Abstract Public health officials within both acute and chronic disease realms have relied predominantly on survey data to gather information on disease prevalence, behavioral models, risk populations, risk probability, and disease progression. Conventional surveys are subject to a number of known limitations, such as respondentsandapos; reluctance to participate, social desirability biases, lag time between questionnaire design, data collection, and availability of results, and intermittent coverage of important topics due to associated implementation costs. Further, disease control experts and policy makers lack access to real-time data and efficient tools to provide contextual awareness vis-à-vis surveys that are implemented for disease surveillance and program management. The implications of not having a timely and broader understanding of the environment and community affects the representativeness and demographic specificity of assessments and of the data used to drive policies and interventions.The proposed Federated Automated Survey Tool (FAST) will be developed as a collaboration among Barron Associates, Inc., George Mason University, and University of Virginia researchers. FAST will be an analytics platform that can be used by public health officials, clinical care investigators, institutional administrators, and others to more easily survey targeted cohorts regarding acute and chronic diseases (e.g., influenza, coronavirus, high blood pressure, etc.) and other indicators (e.g., depression prevalence, tobacco use, substance abuse, etc.) by harnessing social media (e.g., Twitter) or other web/electronic data. Based on both automated and tailorable investigator inputs, the proposed FAST platform will facilitate the construction of appropriate interrogations of social media and web data to yield prospective and longitudinal insights to answer user-initiated questions.The FAST analytics platform will enable local, national, and worldwide surveys on geographically- and demographically-targeted social media and web users based on their Tweets, posts, emails, search, and other web data and metadata. The FAST platform will utilize sophisticated text analytics and novel survey construction and analysis techniques. The survey results will then be analyzed automatically to gain insights and answer a diverse set of questions regarding targeted geographic- and demographic-specific prevalence and severity estimates. These can be one-off surveys, pre- and post-intervention surveys, or online, real-time, longitudinal surveys. As an example of the latter, school administrators could track national or more localized (i.e., geo-tagged) student social media posts in real time regarding issues such as drinking, drug use, stress, depression, or suicide, enabling administrators to better tailor services offered to students and/or detect the need for interventions.The FAST platform will employ a consolidated approach that makes it relatively easy for non-experts to create, administer, and survey social network and electronic data of nearly any cohort. With FAST, the full range of probability sampling techniques (e.g., simple random samples, stratified random samples, etc.) will be available to end-users, along with the corresponding estimated variance and bound on the error of the estimate.Project Narrative The FAST analytics platform will efficiently provide for deeper insights into health behaviors as they are occur- ring, leading to improved policy development and the delivery of interventions. Such insights can be used to monitor changes associated with program interventions as they are implemented or between surveys, identify new hypotheses to investigate, and support decision making. The FAST analytics platform will operate on a subscription-based freemium model to support markets that include public health systems, health care providers, researchers, companies, journalists, governments, institutions, and others who wish to efficiently conduct past, present, or future surveys without the requirement for substantial text mining expertise and at relatively low implementation cost.