Coding Long COVID: Characterizing a new disease through an ICD-10 lens

BackgroundNaming a newly discovered disease is a difficult process; in the context of the COVID-19 pandemic and the existence of post-acute sequelae of SARS-CoV-2 infection (PASC), which includes Long COVID, it has proven especially challenging. Disease definitions and assignment of a diagnosis code are often asynchronous and iterative. The clinical definition and our understanding of the underlying mechanisms of Long COVID are still in flux, and the deployment of an ICD-10-CM code for Long COVID in the US took nearly two years after patients had begun to describe their condition. Here we leverage the largest publicly available HIPAA-limited dataset about patients with COVID-19 in the US to examine the heterogeneity of adoption and use of U09.9, the ICD-10-CM code for "Post COVID-19 condition, unspecified." MethodsWe undertook a number of analyses to characterize the N3C population with a U09.9 diagnosis code (n = 21,072), including assessing person-level demographics and a number of area-level social determinants of health; diagnoses commonly co-occurring with U09.9, clustered using the Louvain algorithm; and quantifying medications and procedures recorded within 60 days of U09.9 diagnosis. We stratified all analyses by age group in order to discern differing patterns of care across the lifespan. ResultsWe established the diagnoses most commonly co-occurring with U09.9, and algorithmically clustered them into four major categories: cardiopulmonary, neurological, gastrointestinal, and comorbid conditions. Importantly, we discovered that the population of patients diagnosed with U09.9 is demographically skewed toward female, White, non-Hispanic individuals, as well as individuals living in areas with low poverty, high education, and high access to medical care. Our results also include a characterization of common procedures and medications associated with U09.9-coded patients. ConclusionsThis work offers insight into potential subtypes and current practice patterns around Long COVID, and speaks to the existence of disparities in the diagnosis of patients with Long COVID. This latter finding in particular requires further research and urgent remediation.

Who has long-COVID? A big data approach

BackgroundPost-acute sequelae of SARS-CoV-2 infection (PASC), otherwise known as long-COVID, have severely impacted recovery from the pandemic for patients and society alike. This new disease is characterized by evolving, heterogeneous symptoms, making it challenging to derive an unambiguous long-COVID definition. Electronic health record (EHR) studies are a critical element of the NIH Researching COVID to Enhance Recovery (RECOVER) Initiative, which is addressing the urgent need to understand PASC, accurately identify who has PASC, and identify treatments. MethodsUsing the National COVID Cohort Collaboratives (N3C) EHR repository, we developed XGBoost machine learning (ML) models to identify potential long-COVID patients. We examined demographics, healthcare utilization, diagnoses, and medications for 97,995 adult COVID-19 patients. We used these features and 597 long-COVID clinic patients to train three ML models to identify potential long-COVID patients among (1) all COVID-19 patients, (2) patients hospitalized with COVID-19, and (3) patients who had COVID-19 but were not hospitalized. FindingsOur models identified potential long-COVID patients with high accuracy, achieving areas under the receiver operator characteristic curve of 0.91 (all patients), 0.90 (hospitalized); and 0.85 (non-hospitalized). Important features include rate of healthcare utilization, patient age, dyspnea, and other diagnosis and medication information available within the EHR. Applying the "all patients" model to the larger N3C cohort identified 100,263 potential long-COVID patients. InterpretationPatients flagged by our models can be interpreted as "patients likely to be referred to or seek care at a long-COVID specialty clinic," an essential proxy for long-COVID diagnosis in the current absence of a definition. We also achieve the urgent goal of identifying potential long-COVID patients for clinical trials. As more data sources are identified, the models can be retrained and tuned based on study needs. FundingThis study was funded by NCATS and NIH through the RECOVER Initiative.

GPS-estimated foot traffic data and venue selection for COVID-19 serosurveillance studies

The initial phase of the COVID-19 pandemic in the US was marked by limited diagnostic testing, resulting in the need for seroprevalence studies to estimate cumulative incidence and define epidemic dynamics. In lieu of systematic representational surveillance, venue-based sampling was often used to rapidly estimate a communitys seroprevalence. However, biases and uncertainty due to site selection and use of convenience samples are poorly understood. Using data from a SARS-CoV-2 serosurveillance study we performed in Somerville, Massachusetts, we found that the uncertainty in seroprevalence estimates depends on how well sampling intensity matches the known or expected geographic distribution of seropositive individuals in the study area. We use GPS-estimated foot traffic to measure and account for these sources of bias. Our results demonstrated that study-site selection informed by mobility patterns can markedly improve seroprevalence estimates. Such data should be used in the design and interpretation of venue-based serosurveillance studies.