Graduation Year

2022

Document Type

Dissertation

Degree

Ph.D.

Degree Name

Doctor of Philosophy (Ph.D.)

Degree Granting Department

Epidemiology and Biostatistics

Major Professor

Jason L. Salemi, Ph.D.

Co-Major Professor

Etienne Pracht, Ph.D.

Committee Member

Chighaf Bakour, Ph.D.

Committee Member

Amber Gum, Ph.D.

Committee Member

Benjamin Jacob, Ph.D.

Committee Member

Kevin J. Kip, Ph.D.

Keywords

GIS, ICD coding, marketing, MESF, opioid use disorder, prescribing

Abstract

Many factors contribute to the opioid epidemic in Florida, including, but not limited to, physician over prescribing, opioid manufacturer malfeasance (direct marketing to physicians, hospitals, and universities, as well indirect political contributions, and deliberately obfuscating the impact of the addictive properties of their various drugs), and as the epidemic approaches its 25th year, sociodemographic characteristics. This body of work examines three of these. First, because an opioid naïve persons’ typical initial encounter with the drug is through a physician, we sought to shed light on the impact of direct pharmaceutical marketing to physicians. We identified opioid providing physicians from the Medicare Part D and matched them to those in the Sunshine Act (record of payments to individual physicians) using Poisson regression with propensity score matching to help control for confounding. While we failed to find a dose-response relationship between opioid prescribers and marketing efforts, we still noted at the highest frequencies of interactions, a threshold effect, where physicians were much more likely to have higher prescription rates than their opioid prescribing peers. We feel that this sets the stage for more detailed research and that these findings add to the body of work on the unhealthy influence of pharmaceutical marketing.

Second, we set out to look at the area effect of pharmaceutical marketing (defined as the rate of physician encounters per year in each county) on clusters of opioid overdose rates of emergency department (ED) encounters and inpatient (IP) admissions. We used data from Medicare Part D, the Sunshine Act, and Florida Department of Health (FDOH), and, as FDOH reports the data at the county level, aggregated all as such. Then, using techniques from geographic information systems (GIS) and linear mixed modeling (LMM) we attempted to identify factors that might be associated with clustering of overdose rates between 2016 and 2019. We failed to note any significant relationship between either IP or ED overdose rate clusters and our main outcome of interest, the quantity of opioid marketing, however, we did note that counties with greater Hispanic and non-Hispanic Black individuals were more likely to be found in a cluster, and that treatment resources saw an inverse association between the IP and ED clustering. We feel this study opens the doors to using more robust GIS informed modeling techniques.

Last, we sought to examine the effect of the opioid epidemic on IP admissions and ED encounters over time, relative to the changeover of the International Classification of Diseases versions nine (ICD-9) to ten (ICD-10) in the last quarter of 2015. We obtained quarterly Florida hospital discharge data from the Agency for Healthcare Administration (AHCA), parsed according to IP versus ED and whether opioids were related to the presenting problem as well as co-morbid conditions. We used interrupted time series modeling to examine both the immediate effect, and any change in trends occurring post ICD changeover. We noted large immediate level changes in all groups and increase in ED encounters overall with a decrease in overall IP admissions, as well as increases in non-White ED encounters, and increases in the number of co-morbid conditions. We feel this shows that opioid trend data should be analyzed with care and that ICD-9 and ICD-10 data related to opioids should be examined separately. Additionally, our findings on the co-morbid conditions can be used to inform researchers interested in designing interventions. Overall, this dissertation adds to the body of work and sets the stage for future efforts aimed at understanding the opioid epidemic.

Download

Share

COinS