The opioid crisis has had a profound impact on U.S. veterans, prompting the Veterans Health Administration (VHA) to implement stringent guidelines for opioid prescriptions. A recent study explored how dual-system healthcare usage—where veterans receive care from both VHA and non-VHA providers—affects the risk of Opioid Use Disorder (OUD). Using advanced deep neural networks (DNN) and explainable artificial intelligence, researchers analyzed 856,299 patient records from VA Medical Centers in Washington DC and Baltimore (2012-2019). The findings revealed that dual-system use significantly increases OUD risk, especially when interacting with demographic and clinical factors. Interestingly, older patients who are dual-system users face heightened OUD risks, while a history of other drug use mitigates this risk. This research underscores the need for targeted interventions to protect vulnerable veteran populations.
Understanding the Elevated Risk Among Dual-System Users
Dual-system healthcare utilization among veterans presents unique challenges in managing opioid prescriptions. The study found that nearly 17% of the analyzed cohort had OUD, identified through natural language processing of clinical notes and ICD diagnoses. Dual-system users were more likely to have OUD, driven by fragmented care and lack of coordinated information sharing between healthcare systems. The DNN model confirmed that dual-system use is a significant risk factor, alongside prior opioid or substance use. Importantly, the interaction between dual-system use and certain demographic and clinical factors further complicates the risk profile.
Older age was generally associated with a lower OUD risk but interacted positively with dual-system use, indicating that older patients enrolled in dual systems may be particularly vulnerable. Conversely, a history of other drug use interacted negatively with dual-system use, suggesting a protective effect against OUD. These nuanced interactions highlight the importance of personalized care strategies for different patient profiles. For instance, younger veterans with multiple comorbidities and those receiving care from multiple systems require special attention. The study also revealed that baseline conditions like PTSD, depression, chronic pain, and TBI played a role in increasing OUD risk among dual-system users.
Implications for Policy and Future Research
The implications of this study extend beyond understanding risk factors; they provide valuable insights for policymakers and healthcare providers. The research underscores the necessity of improving care coordination between VA and non-VA systems to mitigate the elevated OUD risk among dual-system users. Advanced AI models can play a crucial role in identifying high-risk subgroups, enabling targeted interventions. For example, older veterans without a history of substance use might benefit from enhanced monitoring and support when enrolled in dual systems.
Moreover, the study highlights the underdiagnosis of OUD in electronic health records (EHRs), as NLP methods identified over eight times more cases compared to ICD-based diagnoses. This discrepancy suggests that traditional diagnostic methods may overlook many cases, necessitating the integration of NLP tools into routine clinical practice. Future research should explore the underlying causes of these interactions and incorporate social and community factors to develop comprehensive prevention strategies. Additionally, investigating the impact of treatment dose and duration on OUD outcomes could provide further insights into effective management practices.