Syndromic Surveillance for COVID-19, Massachusetts, February 2020–November 2022: The Impact of Fever and Severity on Algorithm Performance

• Published in: Public health reports (1974) Vol. 138; no. 5; pp. 756 - 762
• Main Authors: Cocoros, Noelle M., Willis, Sarah J., Eberhardt, Karen, Morrison, Monica, Randall, Liisa M., DeMaria, Alfred, Brown, Catherine M., Madoff, Lawrence C., Zambarano, Bob, Sljivo, Selsebil, Nagavedu, Kshema, Klompas, Michael
• Format: Journal Article
• Language: English
• Published: Los Angeles, CA SAGE Publications 01.09.2023; SAGE PUBLICATIONS, INC
• Subjects: Algorithms; Alignment; Classification; COVID-19; COVID-19 diagnostic tests; Diagnosis; Electronic health records; Electronic medical records; Fever; Health surveillance; Illnesses; Infectious diseases; Influenza; Medical records; Public Health Methodology; Respiratory tract; Respiratory tract infection; Revisions; Signs and symptoms; Surveillance systems; Trends; Viral diseases; United States > US; Massachusetts; COVID-19; syndromic surveillance; influenza-like illness
• ISSN: 0033-3549; 1468-2877; 1468-2877
• DOI: 10.1177/00333549231186574

Objectives: Syndromic surveillance can help identify the onset, location, affected populations, and trends in infectious diseases quickly and efficiently. We developed an electronic medical record–based surveillance algorithm for COVID-19–like illness (CLI) and assessed its performance in 5 Massachusetts medical practice groups compared with statewide counts of confirmed cases. Materials and Methods: Using data from February 2020 through November 2022, the CLI algorithm was implemented in sites that provide ambulatory and inpatient care for about 25% of the state. The initial algorithm for CLI was modeled on influenza-like illness: an International Classification of Diseases, Tenth Revision, Clinical Modification (ICD-10-CM) diagnosis code for COVID-19 and an ICD-10-CM diagnosis code suggesting severe lower respiratory tract infection or ≥1 ICD-10-CM diagnosis code for upper or lower respiratory tract infection plus fever. We generated weekly counts of CLI cases and patients with ≥1 clinical encounter and visually compared trends with those of statewide laboratory-confirmed cases. Results: The initial algorithm tracked well with the spring 2020 wave of COVID-19, but the components that required fever did not clearly detect the November 2020–January 2021 surge and identified <1% of weekly encounters as CLI. We revised the algorithm by adding more mild symptoms and removing the fever requirement; this revision improved alignment with statewide confirmed cases through spring 2022 and increased the proportion of encounters identified as CLI to about 2% to 6% weekly. Alignment between CLI trends and confirmed COVID-19 case counts diverged again in fall 2022, likely because of decreased COVID-19 testing and increases in other respiratory viruses. Practice Implications: Our work highlights the importance of using a broad definition for COVID-19 syndromic surveillance and the need for surveillance systems that are flexible and adaptable to changing trends and patterns in disease or care.