Westat Experts Present at AAPOR 2025

Westat’s public opinion and survey methodology experts are heading to the 80th Annual American Association for Public Opinion Research (AAPOR) Conference, in St. Louis, MO, May 14-16. This year’s conference theme is “Reshaping Democracy’s Oracle: Transforming Polls, Surveys, and the Measurement of Public Opinion in the Age of AI,” and large language models (LLMs), natural language processing (NLP), and machine learning (ML) will be key topics of discussion. Westat is a sponsor of the conference.

We sat down to talk with Westat staff to learn what attendees can expect from their presentations.

Raising Response Rates Through Real-Time Adaptation

Westat works on a national longitudinal panel survey on healthcare use and expenditures that was historically reliant on paper-based supplemental self-administered questionnaires (SAQs). This survey had a strong decline in response rates for both in-person interviews.

Westat’s Lena Centeno, MA, will provide an in-depth analysis of how embracing new technologies allowed us to quickly pivot to a multimode approach to the survey, pushing the envelope on innovation and significantly raising response rates. She will highlight the procedural changes implemented to maximize response rates and enhance data quality.

Join Centeno on Friday, May 16, for Optimizing Multimode Contact Procedures to Maximize Response.

Getting Quality Data Through Targeted Respondent Identification

Establishment surveys are important tools for informing policy and economic decisions within organizations, but their quality depends on the knowledge of the respondents. Kelly Daley, PhD, will discuss strategies for locating target respondents, focusing on challenges such as gatekeeper barriers, organizational complexity, and respondent burden. We examine approaches, including pre-contact screening, hierarchical referrals, and the use of multimode data collection (e.g., web, phone, mail, and in-person) to improve respondent identification and data accuracy, and, ultimately, improve data quality and response rates.

Join Daley on Thursday, May 15, for Efforts to Identify the Most Knowledgeable Person and the Challenges Associated.

Collecting Data on the Health of America’s Veterans

Recent research suggests that offering both online and paper response options simultaneously, with a bonus for online participation, may increase survey response rates. Preliminary results from the U.S. Department of Veterans Affairs Gulf War Era Cohort Study support this.

April Fales, MS, will share the innovative approaches we used in this study and their implications for future surveys. Discover how these findings can enhance data collection methods and improve engagement with veterans, ultimately meaning policymakers have better quality data on veterans’ health.

Join Fales on Thursday, May 15, for Impacts of a Sequential or Concurrent Choice+ Approach in a Longitudinal Study of Veterans.

Improving Cognitive Interviewing With Large Language Models

Cognitive interviewing has long been used to test and develop survey questions. Although it is an effective method, analyzing data and summarizing findings are both labor-intensive and time-consuming. LLMs have the potential to help reduce the burden.

Hanyu Sun, PhD, will discuss using a fine-tuned LLM to analyze cognitive interviews and identify errors in survey questions, using annotated transcripts by human experts for training and testing, focusing on ways to minimize errors during this process in order to save time and resources while not sacrificing data quality.

Join Sun on Thursday, May 15 for Leveraging Large Language Models for Cognitive Interviewing: A Proof of Concept.

She will also be presenting on ML methods: Use Machine Learning Methods to Evaluate Survey Questions By Mode.