JMIR Publications

ABSTRACT

Background:

Electronic health record (EHR) data provide a unique opportunity to study COVID-19 and vaccine effectiveness but require a well-defined computable phenotype of COVID-19-like illness (CLI).

Objective:

We evaluated the performance of diagnostic codes in identifying COVID-19 cases in emergency department/urgent care (ED/UC) and inpatient settings.

Methods:

We conducted a retrospective cohort study using EHR, claims, and laboratory data from four U.S. health systems. Patients aged ≥18 years with an ED/UC or inpatient acute respiratory illness (ARI) encounter and SARS-CoV-2 PCR test during March 2020-March 2021 were included. We evaluated CLI definitions using combinations of ICD codes as follows: COVID-19-specific codes; CLI definition used in VISION network studies (VISION CLI); ARI signs, symptoms and diagnosis codes only; signs and symptoms of ARI only; random forest model definitions. We evaluated sensitivity, specificity, positive (PPV), and negative predictive value (NPV) using a positive test reference standard.

Results:

Among 90,952 hospitalizations and 137,067 ED/UC visits, 5,627 (6.2%) and 9,866 (7.2%) were positive for SARS-CoV-2, respectively. COVID-19-specific codes had high sensitivity (91.6%) and specificity (99.6%) for hospitalized patients. The VISION CLI definition maintained high sensitivity (95.8%) but lowered specificity (45.5%). All CLI definitions had lowered sensitivity for ED/UC encounters. Random forest approaches identified distinct CLI definitions with high performance for hospital encounters and moderate performance for ED/UC encounters.

Conclusions:

COVID-19-specific codes have high sensitivity and specificity for identifying SARS-CoV-2 positivity. Separate combinations of COVID-19-specific codes and ARI codes enhance the utility of CLI definitions for studies using EHR data in hospital and ED/UC settings. Clinical Trial: n/a