Food safety inspections are an essential public health function. Violations identified during inspections can lead to disease outbreaks and jeopardize the safety of employees. Agencies like the Chicago Department of Public Health provide guidance, services, and strategies to retail food establishments in order to create a healthier city. However, food safety inspectors have limited time and resources, and issues in a city are widespread. How can public health organizations make the best use of their limited resources to minimize disease outbreak?

This presentation will explore how interpretable and user friendly AI methods in text analytics and visualization can assist. Using a case study leveraging the publicly available structured data and freeform comment text from inspection reports performed in the city of Chicago, we will highlight areas of common issue with grocery and restaurant establishments. Furthermore, we will assess the most serious issues impacting sanitation including pest control and uncover the scope and breadth of sanitation problems. By identifying patterns in the physical areas of inspection that are most associated with serious sanitation issues, we will show how to provide actionable recommendations for delegating inspectors to target inspections for issues that require remedy. By the end of the session, attendees will be able to explain how analytics and AI methods can help improve the quality and effectiveness of inspections, furthering the protection of human health in the food service industry.

Highly Pathogenic Avian Influenza (HPAI) A is a global disease of One Health significance due to its multi-species infectivity, prevalence in wild bird populations, rare yet serious human infections, and threat to the domestic poultry industry. The complexity of monitoring these drastically different situations is challenging but important. The International Society for Infectious Diseases’ (ISID) Program for Monitoring Emerging Diseases (ProMED) provides free text reports with essential human case details such as influenza type, strain, exposure, date of onset, and country location. However, users of ProMED must sift through thousands of individual messages to obtain single case reports. How can analysts effectively tap this data to take relevant actions related to HPAI?

This project demonstrates the value of AI-supported text analytics techniques in extracting meaning from unstructured text. We will show how text analytics can isolate human case counts, relevant details, and provide an overarching summary of HPAI influenza A (H5Nx) cases. Additionally, structured data from US animal surveillance reports is overlayed with the text analysis from the ProMED reports to provide a cohesive view of how HPAI is impacting humans globally and animals domestically. The objective of this project is to support public models such as One Health. The objectives of this session are to explain how to leverage unstructured text reports for surveillance data, why these data provide rich sources of information to surveillance models, and how analytic systems can unify diverse views of Avian Influenza activity.

Improving the use of antibiotics in healthcare is a national priority and of One Health significance, with long-term care an important sector to its success. The Centers for Medicare & Medicaid Services (CMS) and the States conduct regular on-site inspections of nursing homes with trained surveyors to document non-compliance with Federal regulations, and as of 2017 these included specific antibiotic stewardship requirements. The outcomes of these investigations are captured in narrative statements of deficiencies (CMS form 2567). The data is too extensive for manual analysis alone, which can produce at best a qualitative assessment. How can organizations like CMS assess infection control and other issues related to nursing homes and effectively regulate?

This panel will explore how text analytics can be used to investigate stewardship issues in nursing homes, utilizing public CMS-2567s and citations for antibiotic stewardship (F881). Our work demonstrates the value of combining natural language processing, machine learning, and human expertise to derive meaningful insights from free-form text. The solution allows for identifying trends and patterns in the data, analyzed alongside structured data sources to better understand antimicrobial stewardship in nursing homes. With this approach, potential issues and emerging trends can be more readily identified and addressed.

During this session, attendees will learn how to leverage text analytics to detect problems within nursing homes and streamline the process using vast amounts of unstructured data. Additionally, they will understand how to investigate inconsistencies between severity scores assigned by surveyors and the actual issues detected within the text. By incorporating an iterative approach to continuous learning, attendees will learn to tailor solutions to specific business problems more efficiently.