<![CDATA[Background: Currently, there is a lack of precision count variable models for mapping fentanyl fatalities. The primary objective of this article is to develop a predictive count variable model for mapping county level fentanyl related deaths using scalable zip code capture point census data.Subjects and Method: This ecological study focused on all zip codes within Hillsborough County, Florida. The target population included residents across these zip codes, with fentanyl related mortality data aggregated per area. Total population sampling was applied using secondary data from census and mortality records. The dependent variable was the count of fentanyl related deaths, while independent variables included sociodemographic indicators obtained from the U.S. Census Bureau. Variable measurements were based on standardized public data sources. Data were analyzed using a multicount Poisson regression model. As no overdispersion was detected (variance inflation factor <10), neither negative binomial regression nor stepwise regression was required. Spatial analysis and autocorrelation were conducted using ArcGIS, with the primary predictor further interpolated to identify geographic patterns.Results: Variable selection for the primary predictor was performed by observing the relationship between the standard error of each tested independent variable and its associated Z score. Given the identified relationship between fentanyl mortality and white populations, from the selection process, a spatial autocorrelation hot and cold spot analysis was conducted. This analysis identified zip codes with the highest and lowest predicted likelihood of fentanyl caused deaths (as opposed to deaths where fentanyl was merely present). The identified zip code locations were 33647 and 33810 for the hot spots.Conclusion: Count variable models and autocorrelation hot/cold spot mapping offer a methodological framework for future modeling efforts to predict locations of fentanyl mortality for preven-tative means.]]>
