Discussion

In this retrospective cohort study, we inferred medical- and illicit-source fentanyl exposure by linking UDT results with medication records from one of the largest U.S. EHR databases using a sequential time-window screening method. We found that illicit-source fentanyl exposure was associated with a sustained increase in nonfatal opioid overdoses over the study period, together with substantially greater clinical burden and risks of opioid-related harmful outcomes than medical-source fentanyl exposure. We further observed distinct demographic and clinical characteristics between exposure groups and elevated risks associated with illicit-source fentanyl initiation across diverse clinical subgroups. These findings are consistent with the evolving landscape of the opioid overdose crisis, in which illicitly manufactured fentanyl has emerged as a major contributor.11 Together, these findings provide one of the first national-scale assessments of the clinical impact of illicit-source fentanyl exposure in the U.S. using individual-level EHR data and highlight the need to distinguish these two populations for surveillance, risk stratification, and targeted interventions.

Fentanyl remains an important synthetic opioid for managing severe pain.4,5 Although highly effective, its high potency also increases the risk of misuse and overdose.7,26,27 In recent years, fentanyl has become a major driver of the opioid overdose crisis in the U.S.,13 largely due to its increasing presence in the illicit drug market.10,11 This shift has prompted significant public health responses, including expanded distribution of naloxone,28 enhanced drug monitoring, and stricter regulatory measures to prevent the spread of fentanyl-related substances. A major challenge in addressing the overdose epidemic is the lack of comprehensive data-driven studies on illicit fentanyl use, as a significant proportion of overdose cases involve fentanyl obtained outside of formal healthcare or prescription systems.11,12 Previous studies have used UDT data to track nonprescribed fentanyl co-detection with other substances,18,19 yet they have provided limited insight into how the source of fentanyl exposure influences opioid-related harmful outcomes. This study addresses this gap by differentiating medical- and potential illicit-source fentanyl exposure at the individual level, enabling direct comparison of longitudinal patterns and associated clinical risks by source of fentanyl exposure.

Marked differences were observed between the medical- and illicit-source fentanyl exposure cohorts. Individuals in the medical-source cohort were older and had a higher burden of chronic conditions, such as cancer, cardiovascular disease, diabetes, and hypertension,29 diseases consistent with legitimate therapeutic indications for fentanyl use. This clinical profile was further reflected by a substantially higher number of fentanyl medication records. In contrast, individuals in the illicit-source cohort were younger,30 less likely to be married, and had a markedly higher prevalence of mental health disorders, substance use disorders, and opioid-related harmful outcomes. This greater clinical burden may reflect differences in exposure context, including uncontrolled dosing, increased likelihood of polysubstance use, and lack of medical supervision. These findings highlight the distinct demographic and clinical differences between the medical-source and illicit-source fentanyl exposure populations.

A particularly noteworthy finding was the dynamic temporal pattern in nonfatal opioid overdoses driven by medical-versus potential illicit-source fentanyl exposures. Overdose cases driven by medical-source fentanyl increased modestly over time, whereas the illicit-source cohort showed a sustained increase through 2023. These patterns indicate that illicit-source fentanyl remains a significant and growing threat to public health nationwide. This shift in overdose driving force underscores the critical impact of policy changes, such as opioid prescription restrictions, on reducing medical-driven fentanyl misuse. However, the continued rise in overdoses linked to illicit fentanyl highlights a persistent and escalating crisis that requires urgent attention.

Furthermore, illicit-source fentanyl initiation was associated with increased short-term risk of opioid-related harmful outcomes compared with medical-source initiation. Individuals initiating fentanyl from illicit sources experienced approximately a two-fold higher risk of opioid abuse, and three-fold higher risks of opioid overdose and dependence within 30 days, with elevated risks also observed over extended follow-up periods (∼180 days). These associations were also observed in analyses restricted to the illicit-source cohort, suggesting that the increased risks are not solely explained by differences between exposure groups, but also reflect the importance of the source of initiation. Across all clinical subgroups, illicit-source initiation was associated with increased risk of outcomes, regardless of the underlying clinical conditions. These findings suggest that the adverse impact of illicit-source fentanyl initiation extends across a broad range of clinical populations and is not limited to specific high-risk groups.

This study has several limitations. First, illicit fentanyl exposure was inferred from fentanyl UDT results. Patients without UDT records were excluded from the current analysis, which may have led to an incomplete assessment of the impact of illicit fentanyl exposure. In addition, UDT-based exposure classification may be subject to misclassification, which could lead to underestimation of the true association. However, UDT provides an objective measure of drug exposure, offering greater reliability than self-reported or survey-based data. Importantly, compared with previous studies that had smaller sample sizes and cross-sectional designs, this study is one of the first to estimate the impact of illicit fentanyl use across all U.S. states over an extended period. The patterns observed in this study are also supported by previous studies.11,12,30 Second, although substance use disorders were included in the analyses to account for potential polysubstance use, they may not fully capture all concurrent substance use. Third, patients classified as the medical-source may have been exposed to illicit fentanyl between the time of medical administration and the UDT. However, given that these patients were likely from inpatient settings and the time window between medical fentanyl use and testing was short, the likelihood of illicit exposure in this group is low. Fourth, we were unable to perform a comparative analysis for fentanyl-associated events at the state level due to small sample sizes in several states. Instead, we conducted analyses based on major U.S. regions, which still provided valuable insights into regional variations in opioid-related harmful outcomes. Fifth, fatal opioid overdose could not be examined because cause-of-death information was not available in Epic Cosmos. Sixth, healthcare encounters occurring outside Epic Cosmos may not be captured, which may limit the generalizability of our findings. However, because Epic Cosmos is broadly representative of the U.S. population seeking healthcare, the potential impact on our findings is likely to be limited. Seventh, missingness in UDT data was primarily due to incomplete laboratory information in UDT records, such as missing specimen collection dates or laboratory values. This suggests that missingness was likely related to data completeness; however, we were unable to assess whether it could be considered missing at random. The potential impact of missing data on the study findings therefore cannot be completely excluded. Finally, the COVID-19 pandemic may have influenced healthcare utilization, access to care, substance use patterns, and clinical outcomes. However, the potential impact of the pandemic was not explicitly evaluated in the present study.

In summary, this cohort study represents one of the first and largest observational studies to estimate the scale and impact of illicit-source fentanyl exposure on the national level and among major geographic regions in the U.S. The evolving dynamics of fentanyl-driven harmful outcomes across U.S. healthcare systems suggest that illicit-source fentanyl use continues to play a major role in the ongoing overdose epidemic, even as harms associated with medical-source fentanyl use remain comparatively modest. Our findings underscore the complexity of the underlying mechanisms contributing to the opioid-associated public health crisis and highlight the need for a deeper understanding of the interactions between medical and illicit fentanyl use. Future research should focus on characterizing high-risk populations who transition from medical-to illicit-source fentanyl use to better inform targeted prevention and intervention strategies. In addition, analyses of healthcare utilization, care quality, and drug market dynamics were not addressed in this study and represent important directions for future research.