Estimating the Prevalence of Opioid use Disorder in the Cincinnati Region using Probabilistic Multiplier Methods and Model Averaging

• Main Authors: Peter J. Mallow, Nila Sathe, Michael Topmiller, Jennifer Chubinski, Dillon Carr, Roni Christopher
• Format: Article
• Language: English
• Published: Columbia Data Analytics, LLC
• Series: Journal of Health Economics and Outcomes Research
• Online Access: http://jheor.scholasticahq.com/article/9729-estimating-the-prevalence-of-opioid-use-disorder-in-the-cincinnati-region-using-probabilistic-multiplier-methods-and-model-averaging.pdf

**Background:** Opioid use disorder (OUD) and its consequences have strained the resources of health, social, and criminal justice services in the Cincinnati region. However, understanding of the potential number of people suffering from OUD is limited. Little robust and reliable information quantifies the prevalence and there is often great variation between individual estimates of prevalence. In other fields such as meteorology, finance, sports, and politics, model averaging is commonly employed to improve estimates and forecasts. The objective of this study was to apply a model averaging approach to estimate the number of individuals with OUD in the Cincinnati region. **Methods:** Three individual probabilistic simulation models were developed to estimate the number of OUD individuals in the Cincinnati Core Based Statistical Area (CBSA). The models used counts of overdose deaths, non-fatal overdoses, and treatment admissions as benchmark data. A systematic literature review was performed to obtain the multiplier data for each model. The three models were averaged to generate single estimate and confidence band of the prevalence of OUD. **Results:** This study estimated 15 067 (SE 1556) individuals with OUD in the Cincinnati CBSA (2 165 139 total population). Based on these results, we estimate the prevalence of OUD to be between 13 507 (0.62% of population) and 16 620 (0.77% of population). **Conclusions:** The method proposed herein has been shown in diverse fields to mitigate some of the uncertainty associated with reliance on a single model. Further, the simplicity of the method described is easily replicable by community health centers, first-responders, and social services to estimate capacity needs supported by OUD estimates for the region they serve.