(P-449) Characterizing “long-covid” Using Real World Data: Post-discharge Clinical Course Among Patients Initially Hospitalized for COVID-19

Background: Emerging evidence suggests that a growing number of patients experience long-term adverse effects from COVID-19. However, clinical characteristics of these late sequelae remain ill-defined. The increasing aggregation of real-world data (RWD) including electronic medical records (EMR) from multiple health systems represents an opportunity to characterize at scale the range of medically attended disease processes, symptoms, and predisposing factors that may benefit from treatment.

Objectives: Among patients initially hospitalized for COVID-19, describe post-discharge clinical course including: (1) incidence and reason for readmission, (2) death, and (3) other health care utilization (COVID-19 testing and results, ambulatory encounters, pharmacotherapy course).

Methods: This retrospective cohort analysis uses data from Health Catalyst’s COVID-19 EMR Database, a longitudinal patient-level database containing data from 18 U.S. provider systems, containing >400,000 patients (as of 01/10/2021) with COVID-19 confirmed by lab test or ICD-10 diagnosis. We identified hospitalized patients with a primary discharge diagnosis of U07.1 who were discharged alive between 01/01/2020 – 09/10/2020. Eligible patients were followed for up to 90 days post-discharge. Associations between patient characteristics at index hospitalization and selected outcomes (readmission, death) were evaluated using proportional hazards regression.

Results: Among 10,315 patients hospitalized with COVID-19, 39.7% had a subsequent hospital-based interaction (12.4% readmitted; 27.3% with other hospital-based care) and 4.3% expired within 90 days post-discharge. Of patients readmitted, 41.7% were readmitted within 7 days, 32.4% within 8-30 days, and 26.0% within 31-90 days. Overall, compared with patients with no readmission, patients readmitted were older (mean: 65 vs 59 years) and more likely to be Black or African American (16.7% vs. 13.5%). Common ICD-10 diagnoses (at 3-digit level) at readmit included: Z79 Long-term drug therapy (70.8%), U07 COVID-19 (55.3%), E78 Disorders of lipoprotein metabolism (46.4%), D6* Anemias and Coagulation defects (47.3%), J96 Respiratory failure (44.9%), and J12 Viral pneumonia (35.8%).

Conclusions: Our findings illustrate that EMR-sourced RWD can be leveraged to characterize longitudinal impact of rapidly evolving emergent infectious diseases like COVID-19. Future work will expand to report on comparator series and identify subgroups at highest risk for long-term sequelae of COVID-19.