198. Chart Validation of an Algorithm for Identifying Patients with Intravenous Drug Use-Associated Endocarditis Using Administrative Code Data

BACKGROUND: Studies using administrative data have described increasing rates of intravenous drug use (IVDU)-associated infective endocarditis (IE) in the United States. These studies used International Classification of Disease (ICD) diagnosis codes to identify hospitalized patients with IE and any illicit drug use (i.e., opioid, amphetamine, cocaine or sedative), but were hindered by absence of specific ICD codes for IVDU. We reviewed charts to determine the positive predictive value (PPV) of ICD codes for identifying patients with IE and IVDU. METHODS: We examined national Veterans Affairs (VA) administrative data from January 2010 to December 2017 to identify patients hospitalized for a first episode of potential IVDU-associated IE based on inpatient ICD 9 and 10 codes for both IE and any illicit drug use, the algorithm used to identify IVDU-IE in most prior studies. We randomly selected 100 of these patients nationally and reviewed hospital charts to confirm clinical documentation of: (1) IE, (2) any illicit drug use, and (3) current or past IVDU. RESULTS: We identified 340 patients with concurrent ICD codes for IE and drug use, increasing from 28 in 2010 to 51 in 2017 (82% increase). In chart review of 100 randomly selected patients, the PPV of ICD codes was 93% (95% CI 88–98%) for a documented clinical diagnosis of IE; 96% (95% CI 92–100%) for documented drug use by any route; and 63% (95% CI 53–73%) for documented IVDU. Among the 37% of patients without clinically documented IVDU, 30% (i.e.,11% of total patients) had clinical documentation stating that drug use was only by non-IV routes, 59% (22% of total) had documented drug use without mention of route of use, and 11% (4% of total) had clinical documentation that patients denied any drug use. CONCLUSION: The incidence of first hospitalization for IE among patients with ICD codes for drug use increased by 82% from 2010 to 2017 in VA care. Concurrent ICD codes for illicit drug use had moderate PPV for identifying IVDU in setting of IE, largely due to identification of patients using drugs without documented intravenous use. There is a need to develop more accurate case-finding algorithms for identifying patients with IVDU-associated endocarditis, for both epidemiologic surveillance and quality improvement applications. DISCLOSURES: All authors: No reported disclosures.