Assessment of Opioid Misuse Rates Across New York State Counties from 2007 to 2018

Overview

This letter critiques the methodology used to estimate opioid misuse (OM) prevalence in New York State counties, highlighting concerns about data definitions, survey biases, and model outputs. It emphasizes the need for collaboration between public health agencies and academic partners to improve the accuracy of opioid misuse estimates.

Background

Estimating the burden of opioid misuse is critical for monitoring trends, identifying disparities, and guiding resource allocation and interventions. The article by Santaella-Tenorio et al. attempted to model OM prevalence using publicly available surveillance data and survey estimates. However, limitations in data definitions, survey representativeness, and modeling approaches may affect the accuracy of these estimates. Robust empirical data and collaborative approaches are needed to enhance the precision of opioid misuse surveillance.

Data Highlights

The critique identifies several key data issues: (1) Misclassification of emergency department visits for opioid overdose due to incorrect ICD-10-CM code usage; (2) Underestimation of OM prevalence from the National Survey on Drug Use and Health (NSDUH) due to under-representation of vulnerable populations and self-report bias; (3) Inconsistencies where statewide OM prevalence estimates are lower than the sum of county-level estimates; (4) Lack of clarity on age ranges used in county-level inputs, potentially affecting prevalence calculations.

Key Findings

• The definition of opioid overdose emergency department visits used in the model was incorrect, including inappropriate ICD-10-CM codes and failing to exclude subsequent visits.
• NSDUH-based OM prevalence estimates likely underestimate true prevalence due to sampling and reporting biases.
• Statewide OM prevalence estimates from the model were paradoxically lower than the sum of county-level estimates, suggesting model convergence or fit issues.
• Age range specifications for county-level data inputs were not provided, possibly contributing to discrepancies in prevalence estimates.
• Successful opioid use disorder prevalence estimation requires person-level data linkage across multiple administrative systems, which is challenging but feasible as demonstrated in Massachusetts and Kentucky.
• Public health-academic partnerships are essential to improve data quality, model inputs, and ultimately the accuracy of substance use prevalence estimates.

Clinical Implications

Clinicians and public health professionals should interpret modeled opioid misuse prevalence estimates with caution, recognizing potential underestimation and data limitations. Collaborative efforts between academic researchers and public health agencies can enhance data accuracy, enabling better-informed clinical and policy decisions to address the opioid crisis. Improved surveillance methods may lead to more targeted interventions and resource allocation.

Conclusion

Accurate estimation of opioid misuse prevalence is challenged by data limitations and modeling complexities. Strengthening partnerships between public health entities and academic researchers is vital to develop more reliable surveillance tools that can better inform responses to the opioid epidemic.

References

1. Santaella-Tenorio et al. -- Estimating opioid misuse prevalence in New York State
2. Wang et al. -- Opioid use disorder prevalence estimation using capture-recapture analysis in Massachusetts