Epidemiological Surveillance of Long Covid in Southern Arizona: A Comparative Study of Active vs. Passive Surveillance

Abstract

The emergence of post-COVID-19 conditions (PCC), now commonly referred to as Long COVID, presents a major public health challenge due to its heterogeneous symptomatology, uncertain clinical course, and lack of standardized diagnostic tests. Surveillance of Long COVID is essential for understanding its burden, but traditional infectious disease systems are not well suited to monitor a condition that bridges acute and chronic disease domains. To address this gap, the Centers for Disease Control and Prevention (CDC) launched Tracking the burden, distribution, and impact of Post COVID-19 conditions in diverse populations for children, adolescents, and adults (Track PCC), which integrates active surveillance through structured, participant-reported surveys with passive surveillance using electronic health record (EHR) data. This dissertation leverages Track PCC data from Arizona, a state with historically high COVID-19 cases and diverse populations, to evaluate and compare active and passive surveillance methods for Long COVID. Aim 1 developed and applied a standardized framework that harmonized survey symptoms and EHR-based diagnostic codes into 24 shared domains across cardiopulmonary, neurological, gastrointestinal, psychiatric, sensory, and dermatologic systems, while also identifying domains not amenable to cross-system comparison. Aim 2 applied this framework to quantify prevalence differences: active surveillance consistently captured a broader range and higher prevalence of symptoms, including milder and fluctuating conditions often absent from EHRs, whereas passive surveillance more frequently identified clinically coded conditions such as cardiovascular and psychiatric diagnoses. Logistic regression confirmed that surveillance system type strongly predicted symptom capture, with evidence of effect modification by age and race for selected domains. Aim 3 characterized temporal dynamics, revealing that self-reported symptom onset typically occurred within weeks of infection and persisted across surveys, whereas first provider-documented diagnoses in EHRs were often delayed by three to twelve months, varying by domain. These delays highlight the gap between lived experience and clinical recognition. Overall, findings demonstrate that while active surveillance is sensitive to early and diverse symptom capture, it is limited by follow-up and recall. Passive surveillance provides clinically verified diagnoses at scale but underrepresents less severe or inconsistently coded symptoms, leading to diagnostic delays. Together, these results underscore the complementary nature of the two approaches and suggest that hybrid models may provide the most complete and timely picture of Long COVID burden. By systematically harmonizing and evaluating surveillance approaches, this dissertation advances methodological innovation in Long COVID monitoring and informs strategies for more comprehensive, equitable, and responsive surveillance of emerging chronic conditions with infectious origins.