Leveraging Natural Language Processing to Identify Veterans Who Inject Drugs to Assess Preexposure Prophylaxis and Sexually Transmitted Infection Testing Services at the Veterans Health Administration - Report - MDSpire
Advertisement
Leveraging Natural Language Processing to Identify Veterans Who Inject Drugs to Assess Preexposure Prophylaxis and Sexually Transmitted Infection Testing Services at the Veterans Health Administration
NLP Identifies Veterans Who Inject Drugs to Evaluate PrEP and STI Services
Overview
An NLP dashboard identified 507 potential veterans who inject drugs (PWID) within the Veterans Health Administration, confirming 78 cases through chart review. Despite high engagement with mental health and social work services, significant gaps were found in preventive care delivery, including low rates of HIV preexposure prophylaxis and STI screening.
Background
People who inject drugs (PWID) face elevated risks for infectious diseases but are difficult to identify in healthcare systems due to lack of specific diagnostic codes for injection drug use. The Veterans Health Administration (VA) has comprehensive electronic health records but manual identification of PWID is resource intensive. Natural language processing (NLP) offers a promising method to analyze unstructured clinical notes to detect PWID and improve access to harm reduction services such as syringe programs, HIV preexposure prophylaxis (PrEP), and STI testing. The VA is committed to expanding these services to reduce infections and improve outcomes among veterans with substance use disorders.
Data Highlights
Metric
Value
Total veterans reviewed
502,075
Potential PWID identified by NLP
507
Confirmed PWID by chart review
78 (15%)
Injection substances (among confirmed PWID)
Opiates 49%, Cocaine 41%, Methamphetamines 37%
HIV prevalence
6%
Hepatitis C antibody positivity
45% (28% viremic)
Hepatitis B exposure
13%
Engagement with mental health services
94%
Engagement with social work
82%
Seen infectious disease specialists
29%
Received syringes
12%
Recent gonorrhea/chlamydia screening
26%
Received HIV PrEP
1 individual
Key Findings
The NLP dashboard efficiently identified PWID within a large veteran population, confirming 78 cases from 507 flagged individuals.
Nearly half of confirmed PWID injected opiates, with substantial use of cocaine and methamphetamines.
High prevalence of HIV (6%), hepatitis C (45%), and hepatitis B exposure (13%) was observed among PWID.
Most PWID were engaged with mental health (94%) and social work services (82%), but only 29% saw infectious disease specialists.
Preventive care gaps were significant: 88% had not received syringes, 74% lacked recent STI screening, and only one individual received HIV PrEP.
Chart reviews were completed rapidly (1-2 minutes), demonstrating the dashboard's efficiency in clinical workflow.
Clinical Implications
NLP tools can enhance identification of PWID in healthcare systems, enabling targeted delivery of harm reduction and preventive services. Despite frequent contact with mental health and social work providers, PWID often miss critical infectious disease care, highlighting the need for integrated multidisciplinary collaboration. Expanding infectious disease specialist involvement and improving access to syringe services, STI screening, and HIV PrEP could reduce infection risks in this vulnerable population.
Conclusion
The NLP dashboard is a valuable tool for identifying veterans who inject drugs and reveals substantial gaps in preventive care despite high engagement with supportive services. Leveraging existing therapeutic relationships while enhancing infectious disease collaboration can improve comprehensive care for PWID within the VA system.
References
Veterans Health Administration 2024 -- Utilizing NLP to Detect Veterans Engaged in Injection Drug Use
