A multicenter evaluation of computable phenotyping approaches for SARS-CoV-2 infection and COVID-19 hospitalizations

Rohan Khera,Joseph S Ross,Benjamin D Pollock,Camille Knepper,Bobak J Mortazavi,Harlan M Krumholz,Nilay D Shah,H Patrick Young,Zhenqiu Lin,William G Jenkinson,Karen Wang,Albert I Ko,David Peaper,Cynthia A Brandt,Richard A Martinello,Elitza S Theel,Wade L Schulz,Veer Sangha,Frederick Warner

doi:10.1038/s41746-022-00570-4

Abstract

Diagnosis codes are used to study SARS-CoV2 infections and COVID-19 hospitalizations in administrative and electronic health record (EHR) data. Using EHR data (April 2020–March 2021) at the Yale-New Haven Health System and the three hospital systems of the Mayo Clinic, computable phenotype definitions based on ICD-10 diagnosis of COVID-19 (U07.1) were evaluated against positive SARS-CoV-2 PCR or antigen tests. We included 69,423 patients at Yale and 75,748 at Mayo Clinic with either a diagnosis code or a positive SARS-CoV-2 test. The precision and recall of a COVID-19 diagnosis for a positive test were 68.8% and 83.3%, respectively, at Yale, with higher precision (95%) and lower recall (63.5%) at Mayo Clinic, varying between 59.2% in Rochester to 97.3% in Arizona. For hospitalizations with a principal COVID-19 diagnosis, 94.8% at Yale and 80.5% at Mayo Clinic had an associated positive laboratory test, with secondary diagnosis of COVID-19 identifying additional patients. These patients had a twofold higher inhospital mortality than based on principal diagnosis. Standardization of coding practices is needed before the use of diagnosis codes in clinical research and epidemiological surveillance of COVID-19.

Highlights

The COVID-19 pandemic has led to the rapid adoption of realworld evidence to guide the treatment of and the public health response to a novel pathogen[1–5]
Clinical predictive models that rely on appropriate case classification and studies that track the long-term effects of severe acute respiratory syndrome (SARS)-CoV-2 infection may be biased if case definitions are inaccurate or capture only subsets of individuals infected with SARS-CoV-2
In this study from two large health systems with academic and community-based practices, we evaluate the accuracy of various approaches to identify people with SARS-CoV-2 infection and COVID-19 hospitalizations based on diagnostic codes and laboratory testing results from the electronic health record (EHR)

Summary

INTRODUCTION

The COVID-19 pandemic has led to the rapid adoption of realworld evidence to guide the treatment of and the public health response to a novel pathogen[1–5]. The identification of both SARSCoV-2 infection and COVID-19 hospitalization is of current clinical and regulatory importance given the need for case identification for epidemiologic surveillance to track the infections, mortality, and vaccine effectiveness. In this study from two large health systems with academic and community-based practices, we evaluate the accuracy of various approaches to identify people with SARS-CoV-2 infection and COVID-19 hospitalizations based on diagnostic codes and laboratory testing results from the EHR. We assess how cohort definitions affect the evaluation of outcomes through an assessment of inhospital mortality across these cohorts

RESULTS

Computable phenotype accuracy for COVID-19 hospitalization at Yale

DISCUSSION

METHODS

CODE AVAILABILITY