Rising Injection-Related Infections Amid Declining Overdose Deaths in the Southeast
Document Type
Abstract
Location
Virtual
Start Date
24-4-2026 9:00 AM
End Date
24-4-2026 3:00 PM
Description
Background: Although opioid overdose deaths have recently stabilized or declined in parts of the southeastern United States, injection-related infections (IRIs) continue to rise. This evolving syndemic of substance use disorder (SUD) and infectious complications is often under-identified due to limitations in ICD-10 coding, provider documentation variability, and underreporting of injection drug use. Consequently, health systems lack reliable tools to quantify IRI burden and evaluate trends. Objectives: To develop and validate an ICD-10–based algorithm to accurately identify four IRI phenotypes (endocarditis, vertebral osteomyelitis/epidural abscess, bacteremia & deep infections, and skin & soft tissue infections (SSTI)) and to quantify its diagnostic performance against gold-standard manual chart review. Methods: We conducted a retrospective analysis of a multi-hospital electronic medical record dataset within a single health system from January 1, 2023 to January 1, 2025. Phenotype-specific ICD-10 codes for IRI diagnoses were combined with SUD-related ICD-10 codes and temporally linked evidence of SUD using an iterative modeling approach. Validation was performed through blinded manual chart review by two independent reviewers using predefined adjudication criteria. Inter-rater reliability was assessed using Cohen’s kappa. Algorithm performance was summarized using positive predictive value (PPV). Results: The algorithm identified 710 true positive, 158 suspected positive, and 175 false positive encounters. Strict PPV was 68%, and inclusive PPV was 83%. Inter-rater reliability was strong (κ = 0.86). Over the study period, IRI encounters increased, while opioid overdose admissions remained stable. Conclusions: Injection-related infections are rising despite stabilization of overdose admissions in this southeastern health system. A validated ICD-10 algorithm enables scalable identification of IRIs without reliance on structured data fields or natural language processing. Recommendations: This validated algorithm may strengthen public health surveillance, inform resource allocation, support addiction medicine integration and infectious disease service planning, and enable health system–level monitoring of injection-related infection trends as the opioid crisis evolves beyond overdose mortality alone.
Recommended Citation
Dicken, Patrick and Bridges-Curry, Hanna, "Rising Injection-Related Infections Amid Declining Overdose Deaths in the Southeast" (2026). Medicine Research Day. 4.
https://digitalscholar.lsuhsc.edu/mrd/2026/qi/4
Rising Injection-Related Infections Amid Declining Overdose Deaths in the Southeast
Virtual
Background: Although opioid overdose deaths have recently stabilized or declined in parts of the southeastern United States, injection-related infections (IRIs) continue to rise. This evolving syndemic of substance use disorder (SUD) and infectious complications is often under-identified due to limitations in ICD-10 coding, provider documentation variability, and underreporting of injection drug use. Consequently, health systems lack reliable tools to quantify IRI burden and evaluate trends. Objectives: To develop and validate an ICD-10–based algorithm to accurately identify four IRI phenotypes (endocarditis, vertebral osteomyelitis/epidural abscess, bacteremia & deep infections, and skin & soft tissue infections (SSTI)) and to quantify its diagnostic performance against gold-standard manual chart review. Methods: We conducted a retrospective analysis of a multi-hospital electronic medical record dataset within a single health system from January 1, 2023 to January 1, 2025. Phenotype-specific ICD-10 codes for IRI diagnoses were combined with SUD-related ICD-10 codes and temporally linked evidence of SUD using an iterative modeling approach. Validation was performed through blinded manual chart review by two independent reviewers using predefined adjudication criteria. Inter-rater reliability was assessed using Cohen’s kappa. Algorithm performance was summarized using positive predictive value (PPV). Results: The algorithm identified 710 true positive, 158 suspected positive, and 175 false positive encounters. Strict PPV was 68%, and inclusive PPV was 83%. Inter-rater reliability was strong (κ = 0.86). Over the study period, IRI encounters increased, while opioid overdose admissions remained stable. Conclusions: Injection-related infections are rising despite stabilization of overdose admissions in this southeastern health system. A validated ICD-10 algorithm enables scalable identification of IRIs without reliance on structured data fields or natural language processing. Recommendations: This validated algorithm may strengthen public health surveillance, inform resource allocation, support addiction medicine integration and infectious disease service planning, and enable health system–level monitoring of injection-related infection trends as the opioid crisis evolves beyond overdose mortality alone.