Computable Phenotypes for Respiratory Viral Infections in the All of Us Research Program

Area of Science:

• Epidemiology
• Biostatistics
• Computational Biology

Background:

• Electronic health records (EHRs) offer valuable temporal data for infectious disease research.
• Optimal methods for identifying infections within EHRs are still under development.
• Respiratory viruses pose a significant public health challenge, necessitating robust identification strategies.

Purpose of the Study:

• To develop and validate computable phenotypes for identifying respiratory virus infections using EHR data.
• To assess the effectiveness of integrating billing codes, prescriptions, and laboratory results for disease phenotyping.
• To enable large-scale studies on host genetics, health disparities, and clinical outcomes related to episodic infectious diseases.

Main Methods:

• Utilized the 'All of Us' Research Program dataset comprising 265,222 participants.
• Developed computable phenotypes by integrating billing codes, prescriptions, and laboratory results within 90-day episodes.
• Analyzed cohort sizes for various respiratory viruses, including adenovirus and SARS-CoV-2.

Main Results:

• Phenotypes successfully generated cohorts for multiple viruses, ranging from 238 (adenovirus) to 28,729 (SARS-CoV-2) cases.
• Virus-specific billing codes demonstrated variable sensitivity (8-67%) but high positive predictive value (90-97%).
• Identified infections aligned with expected seasonal patterns and proportions compared to CDC data, indicating method validity.

Conclusions:

• An integrated EHR data approach is superior to individual components for identifying episodic infectious diseases.
• The developed method effectively identifies severe infections and supports large-scale epidemiological research.
• This phenotyping strategy facilitates studies on host genetics, health disparities, and clinical outcomes across various episodic diseases.