Clinical Scorecard: Assessment of Opioid Misuse Rates Across New York State Counties from 2007 to 2018 Using a Bayesian Spatio-Temporal Model

At a Glance

Key Highlights

• Incorrect definition and coding of opioid overdose emergency department visits can bias model inputs.
• National Survey on Drug Use and Health (NSDUH) data may underestimate opioid misuse due to underrepresentation and self-report bias.
• Collaboration between public health agencies and academic partners is essential for accurate modeling and data interpretation.

Guideline-Based Recommendations

Diagnosis

• Use ICD-10-CM T codes specific to opioid substances (T40.0, T40.1, T40.2, T40.3, T40.4, T40.6) for identifying opioid overdose events.
• Exclude X and Y ICD-10 codes related to poisoning deaths and visits coded as 'adverse effect' or 'sequela' from overdose counts.

Management

• Incorporate multiple data sources including administrative health data and surveys to improve opioid misuse prevalence estimates.
• Engage in public health-academic partnerships to access granular data and improve model accuracy.

Monitoring & Follow-up

• Provide convergence diagnostics and goodness-of-fit statistics when modeling opioid misuse prevalence.
• Consider age range specifications in model inputs to ensure consistency with survey data.

Risks

• Bias in prevalence estimates due to underrepresentation of homeless, incarcerated, and non-responding individuals in surveys.
• Potential inaccuracies from using aggregated data without person-level linkage across systems.

Patient & Prescribing Data

New York State residents with potential opioid misuse, including underrepresented groups such as homeless and incarcerated individuals

Current prevalence estimates may underestimate true opioid misuse burden; improved data linkage and modeling approaches are needed to inform resource allocation and interventions.

Clinical Best Practices

• Use validated ICD-10-CM codes specific to opioid overdose for surveillance and research.
• Collaborate with public health agencies to obtain detailed administrative data for modeling.
• Interpret modeled opioid misuse prevalence with caution, acknowledging limitations of survey data and model inputs.
• Invest in public health-academic partnerships to enhance data quality and modeling methodologies.

References

• Santaella-Tenorio et al. article on opioid misuse prevalence modeling
• Statewide Planning and Research Cooperative System (SPARCS) data and ICD-10-CM coding guidelines
• Wang et al. study on opioid use disorder prevalence using capture-recapture analysis in Massachusetts