Abstract
IMPORTANCE: Illicitly manufactured fentanyl remains a public health threat and trustworthy measurements in prevalence are crucial to public health approaches. Low prevalence behaviors, such as route of administration of illicitly manufactured fentanyl, may have shifted over time, which changes community risk profiles. OBJECTIVE: To assess the impact of bias mitigation methods in an online survey sample and quantify changes in routes of administration in illicitly manufactured fentanyl use over time. DESIGN, SETTING, AND PARTICIPANTS: This repeated cross-sectional survey included US adults 18 years and older in an online, panel-based general population sample fielded twice yearly, in spring and autumn. Corrections for demographic and nondemographic composition bias using calibration weights and removal of misclassification from careless/inattentive responses were applied. Data were collected from April 2022 to October 2024, and data were analyzed in May 2025. MAIN OUTCOMES AND MEASURES: Self-reported use of illicitly manufactured fentanyl in the past 12 months and routes of administration, which included oral, injection, smoking, or snorting. Weighted frequency and percentages were calculated. RESULTS: In the full 2022-2024 sample of 175 058 respondents where misclassification removal and calibration was applied, 50.6% (95% uncertainty interval [UI], 50.3-60.0) were female, 48.1% (95% UI, 47.8-48.4) were male, and 1.3% (95% UI, 1.2-1.3) were transgender, nonbinary, or something else, and the median (IQR) age was 47 (32-62) years. The bias-mitigated prevalence estimate of illicitly manufactured fentanyl use in the last 12 months increased from 0.7% (95% UI, 0.7-0.8) in 2022 to 1.1% (95% UI, 1.0-1.2) in 2024. Oral use of illicitly manufactured fentanyl increased from 35.9% (95% UI, 31.1-40.7) in 2022 to 44.4% (95% UI, 40.3-48.5) in 2024, which was the most common route of administration. In 2024, use by smoking was 37.9% (95% UI, 34.1-41.6), use by snorting was 27.1% (95% UI, 23.5-30.7), and use by injection was 24.5% (95% UI, 21.3-27.7). Importantly, bias mitigation cumulatively reduced the national estimate of illicitly manufactured fentanyl by 70.9% in 2024 (from 3.9% [95% UI, 3.8-4.1] when neither was applied to 1.1% [95% UI, 1.0-1.2]), an important factor when considering prevalence and change over time. CONCLUSIONS AND RELEVANCE: Results of this survey study suggest that fentanyl use has shifted toward oral use, which may contribute to observed lower mortality rates despite an increase in prevalence of use. Methods intended to reduce systematic bias have a strong influence on low prevalence behavior estimates and should be implemented for all survey-based drug use surveillance.