A spatiotemporal analysis of opioid prescriptions in Indiana from 2015 to 2019

Abstract

People living in rural communities are more likely to receive opioid prescriptions, partly due to job-related injuries. State-level interventions have reduced opioid prescribing; however, rural/urban disparities persist due to differences in demographics and prescribing practices, particularly in states with large rural populations like Indiana. While spatiotemporal analyses have explored aspects of the opioid crisis, spatiotemporal patterns of opioid prescribing have not been sufficiently studied. This study utilizes a sample of Medicaid claims data from the Indiana Family and Social Services Administration from 2015 to 2019 to analyze spatiotemporal patterns of patients receiving at least one opioid prescription across Indiana. The goal was to analyze patient demographics and track prescription hotspot movement over time in rural and urban areas. We analyzed data for 107,574 Medicaid enrollees who received opioid prescriptions during the study period. We found that most patients in the cohort resided in urban areas, with the number of patients who were prescribed opioids and resided in rural areas decreasing at a faster rate. We conducted a negative binomial regression analysis to examine the relationship between various demographics (sex, age, race/ethnicity, and urban/rural classification) and the number of patients receiving at least one opioid prescription over time. Our findings indicate that older patients, patients identifying as females, patients who identify as White, and patients living in urban areas, are more likely to receive at least one opioid prescription. Additionally, the interaction effects revealed that patients from all demographic groups were more likely to receive at least one opioid prescription if they lived in urban areas, compared to those living in rural areas. Using Local Moran's I as a local spatial autocorrelation measure, we identified clusters highlighting transitions from rural to urban areas over time. In 2015-2016, three significant clusters emerged within rural-surrounded 3-digit ZIP codes (472, 474, 476), based on the Rural-Urban Commuting Area Codes. Over time, significant clusters shifted towards urban or mixed areas, possibly influenced by state guidelines and legislation. These findings enhance the understanding of opioid prescription dynamics and identify patterns in opioid prescribing rates in terms of the proportion of patients receiving opioid prescriptions among urban vs. rural communities in Indiana.

Keywords: Geospatial analysis; Indiana; Moran's I; Opioid prescribing rates; Opioid prescriptions; Urbanicity.

Fig. 1

Workflow for extraction of the cohort

Fig. 2

Classification of patients receiving at least one opioid prescription by rural and urban areas in Indiana. A). Proportion of patients receiving an opioid prescription by rural and urban classification from the FSSA database and total Indiana Medicaid population from ACS. B). Localization of the 3-digit ZIP codes and the classification in rural and urban areas. Green 3-digit ZIP codes are classified as “Urban” and blue 3-digit ZIP codes are classified as “Rural”. The urban/rural classification was extracted from RUCA Codes and transferred to a ZIP code approximation. To obtain the classification by 3-digit ZIP codes, a weighted average was calculated using the average of the RUCA codes by 5-digit ZIP codes and the total estimate of the population by each 5-digit ZIP code aggregated by the first 3-digit ZIP codes. The RUCA Codes encompass a number from 1 to 10 for the primary classification and a decimal part to indicate if there are other connections between rural and urban areas. Primary RUCA Codes are classified as isolated (RUCA 10), small town (RUCA 7–9), micropolitan (RUCA 4–6), or metropolitan (RUCA 1–3). RUCA 4 is a micropolitan area core with a primary flow within an urban cluster of 10,000 to 49,999 (large UC), so we considered RUCA 4–10 as “rural” and RUCA 1–3 as “urban”. Mann-Whitney U test between rural and urban population from our cohort and Indiana Medicaid from the ACS showed no significant differences (p-value > 0.05)

Fig. 3

Global Moran’s I statistics across 2015 to 2019. Global Moran’s I determine the overall spatial concentration of the proportion of patients receiving at least one opioid prescription in Indiana. A higher value for Global Moran’s I indicate highly concentrated areas. Global Moran’s I indicate the presence of spatial concentration but not the locations for the spots

Fig. 4

Spatial clusters with normalized patients receiving an opioid prescription by 3-digit ZIP codes, Indiana, 2015–2019. Local Moran’s I identify the specific location of the core clusters or centers of the agglomerated zones of patients receiving an opioid prescription. There are four statistics of LISA, where the first field of each of those refers to the level of proportion of patients in the core or center of the cluster and the second field indicates the level of proportion of patients in surrounding zones. High-High label means that the core and the neighbors have a high number of patients receiving opioid prescriptions. High-low and Low-high are called outliers and describe the behavior of having a core with a different number of patients in comparison with their surrounding neighbors. The significance level is calculated under the null hypothesis of random spatial distribution with a p-value < 0.05. The names of the cities presented in the figure are the 10 most populated cities in Indiana