Unraveling COVID-19: a large-scale characterization of 4.5 million COVID-19 cases using CHARYBDIS

Daniel Prieto‐Alhambra, Kristin Kostka, Talita Duarte‐Salles, Albert Prats‐Uribe, Anthony G. Sena, Andrea Pistillo, Sara Khalid, Lana Yin Hui Lai, Asieh Golozar, Thamir M. Alshammari, Dalia Dawoud, Fredrik Nyberg, Adam Wilcox, Alan Andryc, Andrew E. Williams, Anna Ostropolets, Carlos Areia, Chi Young Jung, Christopher A. Harle, Christian Reich, Clair Blacketer, Daniel R. Morales, David A. Dorr, Edward Burn, Elena Roel, Eng Hooi Tan, Evan Minty, Frank DeFalco, Gabriel de Maeztu, Gigi Lipori, Heba Alghoul, Hong Zhu, Jason Thomas, Jiang Bian, Jimyung Park, Jordi Martínez Roldán, Jose Posada, Juan M. Banda, Juan Pablo Horcajada, Julianna Kohler, Karishma Shah, Karthik Natarajan, Kristine E. Lynch, Li Liu, Lisa M. Schilling, Martina Recalde, Henry M. Spotnitz, Mengchun Gong, Michael E. Matheny, Neus Valveny, Nicole G. Weiskopf, Nigam H. Shah, Osaid Alser, Paula Casajust, Rae Woong Park, Robert Schuff, Sarah Seager, Scott L. DuVall, Seng Chan You, Seok Young Song, Sergio Fernández‐Bertolín, Stephen Fortin, Tanja Magoč, Thomas Falconer, Vignesh Subbian, Vojtech Huser, Waheed‐Ul‐Rahman Ahmed, W. Carter, Yin Guan, Yankuic Galvan, Xing He, Peter R. Rijnbeek, George Hripcsak, Patrick Ryan, Marc A. SuchardResearch Square 2021 · DOI

Abstract

Background: Routinely collected real-world data (RWD) have great utility in aiding the novel coronavirus disease (COVID-19) pandemic response. Here we present the international Observational Health Data Sciences and Informatics (OHDSI) Characterizing Health Associated Risks, and Your Baseline Disease In SARS-COV-2 (CHARYBDIS) framework for standardization and analysis of COVID-19 RWD.

Methods: We conducted a descriptive cohort study using a federated network of data partners in the United States, Europe (the Netherlands, Spain, the UK, Germany, France, and Italy), and Asia (South Korea and China). The study protocol and analytical package were released on June 11, 2020, and are iteratively updated via GitHub.

Findings: We identified three non-mutually exclusive cohorts of 4,537,153 individuals with a clinical COVID-19 diagnosis or positive test, 886,193 hospitalized with COVID-19, and 113,627 hospitalized with COVID-19 requiring intensive services. All comorbidities, symptoms, medications, and outcomes are described by cohort in aggregate counts and are available in an interactive website.

Interpretation: CHARYBDIS findings provide benchmarks that contribute to our understanding of COVID-19 progression, management, and evolution over time. This can enable timely assessment of real-world outcomes of preventative and therapeutic options as they are introduced in clinical practice.

Introduction

Early in the COVID-19 pandemic, the global medical community faced a "tragic data gap." While millions were being infected, clinical data remained siloed within individual hospitals or national health systems, often stored in incompatible formats. This fragmentation made it nearly impossible to compare disease progression or treatment outcomes across different regions or populations in real-time.

The CHARYBDIS (Characterizing Health Associated Risks, and Your Baseline Disease In SARS-COV-2) project was established to bridge this gap. By leveraging a global network of researchers and data partners, the project created a framework to ingest diverse Real-World Data (RWD)—including electronic health records (EHRs) and insurance claims—and transform them into a standardized format for large-scale, international analysis.

Methods: The Federated Network Approach

The researchers employed a federated network approach through the OHDSI (Observational Health Data Sciences and Informatics) community. Instead of pooling sensitive patient-level data into a central repository—a process fraught with legal and privacy hurdles—the analysis code was sent to the data. This allowed each site to maintain control over its patient records while contributing to a global aggregate.

Standardizing Data with OMOP CDM

A critical technical component was the use of the OMOP Common Data Model (CDM). Participating sites mapped their local data (such as various versions of ICD codes) into this standardized schema. This ensured that clinical concepts like "hospitalization," "diabetes," or "intensive care" were defined consistently across a New York hospital, a South Korean clinic, and a European health system.

Study Design

The descriptive cohort study tracked three non-mutually exclusive groups:

  1. Diagnosed: Individuals with a clinical COVID-19 diagnosis or positive test.
  2. Hospitalized: Patients admitted to the hospital with COVID-19.
  3. Intensive Services: Patients requiring ICU care or mechanical ventilation.

The entire analytical package was released as open-source on GitHub, ensuring transparency and allowing for iterative updates as the pandemic evolved.

Global Scale and Datasets

The study analyzed 26 databases across nine countries, representing one of the largest COVID-19 characterization efforts to date:

  • North America: The USA contributed 96% of the diagnosed cohort, primarily through large insurance claims databases like IQVIA and Optum.
  • Europe: Data was sourced from the UK, Netherlands, Spain, Germany, France, and Italy. Spain provided the largest European cohort (SIDIAP-Spain, n=124,305).
  • Asia: Contributions came from South Korea (HIRA database) and China.

In total, the study characterized over 4.5 million diagnosed individuals, nearly 900,000 hospitalized patients, and over 113,000 patients requiring intensive services.

Key Clinical Insights

Demographic Patterns

The data revealed distinct demographic patterns across regions. In the US and Europe, while diagnoses were common among younger adults (ages 25–60), hospitalizations and intensive care requirements were overwhelmingly concentrated in the 60–80 age bracket.

A consistent sex gap was also observed: across most regions, women were slightly more likely to be diagnosed, but men were significantly more likely to be hospitalized or require intensive care. South Korea was a notable exception, showing a higher proportion of women hospitalized.

Comorbidity Benchmarking

The study quantified the baseline health of COVID-19 patients, reinforcing known risk factors. Hospitalized patients exhibited dramatically higher rates of hypertension, type 2 diabetes, and chronic kidney disease compared to the general diagnosed population, establishing these as primary indicators for severe disease.

Symptom Reporting and Coding Lag

Interestingly, symptoms such as anosmia (loss of smell) were recorded in less than 1% of cases in most databases. This likely reflects "coding lag," where clinicians lacked specific codes for new symptoms during the early stages of the pandemic, rather than a lack of symptom prevalence. This highlights a limitation of RWD: it is only as good as the clinical documentation at the point of care.

Discussion and Limitations

The CHARYBDIS project demonstrates that a federated model can produce high-quality, large-scale research without moving patient data, offering a blueprint for future pandemic responses. However, the study has limitations:

  • Descriptive Nature: The findings describe clinical journeys but do not establish causal relationships between treatments and outcomes.
  • Temporal Bias: The data primarily reflects the "initial wave" (early to mid-2020) and does not account for later variants (Delta, Omicron) or the impact of vaccination.
  • Geographic Gaps: There is a lack of data from low-to-middle-income countries, which the authors are actively seeking to address.

Conclusion

The CHARYBDIS framework has produced over 22,200 unique aggregate characteristics, all hosted on a publicly available interactive website. These benchmarks are vital for understanding the progression and management of COVID-19 and provide a foundation for evaluating new therapeutic and preventative options as they are introduced in clinical practice.

Summary updated

Abstract

<b>Background:</b> Routinely collected real world data (RWD) have great utility in aiding the novel coronavirus disease (COVID-19) pandemic response [1,2]. Here we present the international Observational Health Data Sciences and Informatics (OHDSI) [3] Characterizing Health Associated Risks, and Your Baseline Disease In SARS-COV-2 (CHARYBDIS) framework for standardisation and analysis of COVID-19 RWD. <b>Methods:</b> We conducted a descriptive cohort study using a federated network of data partners in the United States, Europe (the Netherlands, Spain, the UK, Germany, France and Italy) and Asia (South Korea and China). The study protocol and analytical package were released on 11 <sup>th</sup> June 2020 and are iteratively updated via GitHub [4]. <b>Findings:</b> We identified three non-mutually exclusive cohorts of 4,537,153 individuals with a clinical <i>COVID-19 diagnosis or positive test,</i> 886,193 <i>hospitalized with COVID-19</i> , and 113,627 <i>hospitalized with COVID-19 requiring intensive services</i> . All comorbidities, symptoms, medications, and outcomes are described by cohort in aggregate counts, and are available in an interactive website: https://data.ohdsi.org/Covid19CharacterizationCharybdis/. <b>Interpretation:</b> CHARYBDIS findings provide benchmarks that contribute to our understanding of COVID-19 progression, management and evolution over time. This can enable timely assessment of real-world outcomes of preventative and therapeutic options as they are introduced in clinical practice.
BibTeX
@inproceedings{prietoalhambra2021unraveling,
  author = {Daniel Prieto‐Alhambra and Kristin Kostka and Talita Duarte‐Salles and Albert Prats‐Uribe and Anthony G. Sena and Andrea Pistillo and Sara Khalid and Lana Yin Hui Lai and Asieh Golozar and Thamir M. Alshammari and Dalia Dawoud and Fredrik Nyberg and Adam Wilcox and Alan Andryc and Andrew E. Williams and Anna Ostropolets and Carlos Areia and Chi Young Jung and Christopher A. Harle and Christian Reich and Clair Blacketer and Daniel R. Morales and David A. Dorr and Edward Burn and Elena Roel and Eng Hooi Tan and Evan Minty and Frank DeFalco and Gabriel de Maeztu and Gigi Lipori and Heba Alghoul and Hong Zhu and Jason Thomas and Jiang Bian and Jimyung Park and Jordi Martínez Roldán and Jose Posada and Juan M. Banda and Juan Pablo Horcajada and Julianna Kohler and Karishma Shah and Karthik Natarajan and Kristine E. Lynch and Li Liu and Lisa M. Schilling and Martina Recalde and Henry M. Spotnitz and Mengchun Gong and Michael E. Matheny and Neus Valveny and Nicole G. Weiskopf and Nigam H. Shah and Osaid Alser and Paula Casajust and Rae Woong Park and Robert Schuff and Sarah Seager and Scott L. DuVall and Seng Chan You and Seok Young Song and Sergio Fernández‐Bertolín and Stephen Fortin and Tanja Magoč and Thomas Falconer and Vignesh Subbian and Vojtech Huser and Waheed‐Ul‐Rahman Ahmed and W. Carter and Yin Guan and Yankuic Galvan and Xing He and Peter R. Rijnbeek and George Hripcsak and Patrick Ryan and Marc A. Suchard},
  title = {{Unraveling COVID-19: a large-scale characterization of 4.5 million COVID-19 cases using CHARYBDIS}},
  booktitle = {Research Square},
  year = {2021},
  doi = {10.21203/rs.3.rs-279400/v1},
}
Figure