598 - Correlation of International Classification of Diseases (ICD) codes to initial provider-selected antibiotic indications in hospitalized adult patients within the Duke Antimicrobial Stewardship Outreach Network (DASON)

Background: Provider-selected antibiotic indications are a measure to help track antibiotic use, and facilitate stewardship activities.  International Classification of Diseases 10^th version (ICD-10) codes have been widely used in the medical field for a variety of purposes, including billing for reimbursement, disease epidemiology, administration, and research.  The ability of these codes to accurately describe the true disease diagnosis has been questioned.  The purpose of this study is to provide insight into correlation between ICD-10 codes and provider-selected antibiotic indications recorded upon antibiotic order entry.

Methods: This multicenter, retrospective cohort study was performed using data from 17 hospitals in the DASON network. Antibiotic orders during calendar year 2019 for inpatients  ≥ 18 years of age that included an indication for use were extracted from the DASON antimicrobial stewardship assessment portal.   Orders with the antibiotic indication category of prophylaxis (medical or surgical) and other were excluded.  The primary outcome was agreement between antibiotic indication and any discharge ICD-10 code from the same hospital admission. Secondary analyses stratified results by antibiotic and indication. Descriptive statistics were used to describe outcomes.

Results: A total of 246,999 unique antibiotic orders were identified in 180,109 admissions.  After removing prophylaxis (n=75,124) and other (n=36,359), 135,516 orders were included. Most orders did not have an ICD-10 code matching the prescriber indication (92,237 [68%]). All indications except HEENT (18% mismatch) and genitourinary infections (46% mismatch) lacked a corresponding diagnosis code in more than 50% of cases (Table 1). Urinary tract infections (93%), bloodstream infections (90%), and central nervous system infections (80%) showed the highest rates of mismatch among indications (Table 1).Table 1: Correlation of ICD-10 Codes to Provider Selected Indications


Conclusion: We observed a high rate of mismatch between antibiotic indications and ICD-10 codes. Provider-selected antibiotic indications at the time of empiric treatment may be more reflective of diagnostic differential but is a poor indicator of ultimate patient diagnosis.