Strategies for population-level identification of Post-Acute Sequelae of COVID-19 through health administrative data

Cristina Mazzali¹Pietro Magnoni^1*Alberto Zucchi²Giovanni Maifredi³Luca Cavalieri D'Oro⁴Maria Letizia Gambino⁵Anna Clara Fanetti⁶Pietro Giovanni Perotti⁷Marco Villa⁸Maria Grazia Valsecchi⁹Daria Vigani¹⁰Claudio Lucifora¹⁰Antonio Giampiero Russo¹

• ¹Epidemiology Unit, ATS Città Metropolitana di Milano, Milan, Italy
• ²Epidemiology Unit, Agenzia di Tutela della Salute di Bergamo, Bergamo, Italy
• ³Epidemiology Unit, Agenzia di Tutela della Salute di Brescia, Brescia, Italy
• ⁴Epidemiology Unit, Agenzia per la Tutela della Salute Brianza, Monza, Italy
• ⁵Epidemiology Unit, Agenzie di Tutela della Salute dell'Insubria, Varese, Italy
• ⁶Epidemiology Unit, Agenzia di Tutela della Salute della Montagna, Sondrio, Italy
• ⁷Epidemiology Unit, Agenzia di Tutela della Salute di Pavia, Pavia, Italy
• ⁸Epidemiology Unit, Agenzia di Tutela della Salute della Val Padana, Mantua, Italy
• ⁹Bicocca Bioinformatics Biostatistics and Bioimaging Centre (B4), Universita degli Studi di Milano-Bicocca Dipartimento di Medicina e Chirurgia, Monza, Italy
• ¹⁰Department of Economics and Finance, Universita Cattolica del Sacro Cuore, Milan, Italy

Post-acute sequelae of COVID-19 (PASC) encompass several clinical outcomes, from new-onset symptoms to both acute and chronic diagnoses, including pulmonary and extrapulmonary manifestations. Health administrative data (HAD) from health information systems allow population-level analyses of such outcomes. Our primary aim was to identify clinical conditions potentially attributable to SARS-CoV-2 infection, and the types of HAD and 'diagnostic criteria' used for their detection.Methods. We performed a literature review to identify HAD-based cohort studies assessing the association between SARS-CoV-2 infection and medium-/long-term outcomes in the general population. From each included study, we extracted data on design, algorithms used for outcome identification (sources, coding systems, codes, time criteria/thresholds), and whether significant associations with SARS-CoV-2 infection were reported.We identified six studies investigating acute and chronic conditions grouped by clinical domain (cardiovascular, respiratory, neurologic, mental health, endocrine/metabolic, pediatric, miscellaneous). Two studies also addressed the onset of specific symptoms.Cardio/cerebrovascular conditions were most studied, with significant associations reported for deep vein thrombosis, heart failure, atrial fibrillation, and coronary artery disease. Conditions in other domains were less investigated, with inconsistent findings. Only three studies were designed as test-positive vs test-negative comparisons.Discussion. Heterogeneity in data sources, study design, and outcome definitions hinder the comparability of studies and explain the inconsistencies in findings about associations with SARS-CoV-2 infection. Rigorously designed studies on large populations with wide availability of data from health information systems are needed for population-level analyses on PASC, and especially on its impact on chronic diseases and their future burden on healthcare systems.