Opioid crisis in rural and urban counties from 2010 to 2018: in light of the index of relative rurality

doi:10.21203/rs.3.rs-2437152/v1

Download PDF

Research Article

Opioid crisis in rural and urban counties from 2010 to 2018: in light of the index of relative rurality

https://doi.org/10.21203/rs.3.rs-2437152/v1

This work is licensed under a CC BY 4.0 License

Version 1

posted

You are reading this latest preprint version

Previous studies have documented the rural-urban disparities of the opioid crisis, but rarely consider the heterogeneities within rural and urban areas, nor the changing dynamics of the rural-urban boundaries. Taking advantage of a continuous measure of rurality, this study separates counties into “becoming-more-rural” and “becoming-more-urban” groups according to the change of rurality between 2000 and 2010, and compares the determinants of the development of prescription opioid-, heroin-, and synthetic opioid-related opioid mortality in the two groups of counties. The analysis uses longitudinal data of 2010–2018 from a variety of national datasets such as the confidential Multiple-Cause of Death data, the U.S. Opioid Dispensing Rate Maps, and the American Community Survey. The results show that while increasing rurality does not predict increasing prescription opioid- and synthetic opioid-involved mortality, the growth of heroin-involved mortality could be a severe issue for becoming-more-rural counties that have already observed heroin overdose deaths. The demographic groups that have higher risk of opioid overdoses may be different in different areas, for example increasing males and people engaged in manual labor occupations are associated with increasing prescription opioid-involved mortality for becoming-more-rural counties but in becoming-more-urban areas, heroin-involved mortality correlates to a higher percentage of Hispanics. Concentrations of socioeconomically disadvantaged populations like veterans and people in poverty increase opioid overdose deaths particularly for becoming-more-urban counties, while healthcare services shown more beneficial to becoming-more-rural areas. The findings suggest that the programs and regulations to reduce opioid-involved mortality should consider the changing nature of counties in the degree of rurality.

The cliché, “addiction does not discriminate,” is no longer suitable for the opioid crisis since it has hit different populations in different ways. The unequal spread of the crisis manifested across the rural-urban spectrum. Rural areas suffer from higher opioid prescribing rates and a higher proportion of overdose deaths involving prescription opioids compared to urban areas, while urban areas observe rapid growths in heroin- and synthetic opioid-related overdose deaths in recent years [1–3]. Previous literature on the rural-urban difference in the opioid crisis rarely examines factors accounting for the with-rural and within-urban disparities in the crisis, nor considers the changing dynamics of rural-urban areas. Based on an innovative continuous measure of rurality (i.e., the index of relative rurality) [4], this study separates U.S. counties into two groups with one becoming more rural and one becoming more urban according to the temporal change of rurality between 2000 and 2010, and compares the determinants of the development of prescription opioid-, heroin-, and synthetic opioid-involved mortality from 2010 to 2018 in the two groups. The findings contribute to understanding the development of the opioid crisis in rural and urban areas, accounting for the changing rural-urban boundaries.

Within-rural and within-urban disparities in the opioid crisis

Rurality has been seen as an important dimension in shaping the geography of the opioid crisis and linking space/place with the heterogeneous experiences of different populations [3, 5–8]. Compared to urban areas, rural communities are often occupied by disadvantaged populations such as older adults and less educated people, and experience decline in both social and economic aspects [6, 9]. In addition, rural residents often lack ways to protect themselves from opioid prescribing and drug addiction compared to their urban counterparts because of the prevalence of stigma and low responsiveness of prescribing-control programs and policies [6, 10, 11]. Consequently, the mortality rate from prescription opioids is persistently higher in nonmetropolitan counties than in metropolitan counties [3]. Urban areas, by contrast, face challenges from the rapidly-growing illicit drug market that leads to a rising number of heroin-and synthetic opioid-related overdose deaths, though prescription opioid-related mortality is stabilized after control [1–3].

Given the different experiences of rural and urban areas in the opioid crisis, the determinants of the development of the opioid crisis in these two divisions may be different [5, 7, 8]. In other words, the heterogeneities in opioid outcomes within rural and within urban areas may be constructed by different factors, and even for the same factor, the magnitude of its impact could be different [5, 12, 13]. However, previous studies on the geographic disparities in the development of the opioid crisis often use a binary rural/urban division, which disguises the huge heterogeneities within the two divisions [5, 6, 14]. Though not expansive, some research has shown great within-rural variations in the opioid crisis, that economically-distressed mining- and service-dependent rural counties suffer from higher drug mortality rates [5]. Compared to rural areas that become destinations of retirees and enjoy an economic renaissance [14–16], rural counties that experience growth in the proportion of racial and ethnic minority groups who participate in low-skilled manual labor occupations and macroeconomic stressors such as higher poverty rates may suffer from higher depression and anxiety rates that lead to higher prevalence of drug use [5, 7, 9, 17]. The healthcare infrastructure also varies tremendously within rural regions [18], which may shape the development of the crisis.

Compared to that on within-rural heterogeneities, research on the variations in opioid outcomes within urban/metropolitan counties is even more limited. More educated and highly skilled population have moved to suburbs, leaving central cities with concentrations of poverty and unstable labor markets [16]. Economic distress, high control of prescription opioids, and high level of social capital may facilitate the growth of illicit drug markets that results in rising numbers of overdose deaths involved heroin and synthetic opioid in metropolitan areas [3, 19–21]. This paper focuses on both within-rural and within-urban heterogeneities in the opioid crisis and examines the differences in contributors of these heterogeneities.

The index of relative rurality

Another major limitation in previous studies on the opioid crisis across the rural-urban continuum is using categorical measures of rurality, such as the National Center for Health Statistics (NCHS) Urban-Rural classification scheme and the Rural-Urban Commuting Area (RUCA) codes [3, 7, 22]. On one hand, the categorical measures of the rural-urban continuum omit the fluidity of spatial boundaries of rural and urban areas, which results from the increasing inter-dependence and similarity between rural and urban America [15, 23, 24]. In addition, the discrete rurality measures use arbitrary thresholds and artificialize the differences among units, which makes them easily fall into the “threshold trap” that causes dangerous deceptions by separating similar cases but grouping very different cases together [25]. On the other hand, the categories of rural-urban divisions are always seen as fixed within a certain time period, thus fail to account for the ongoing dynamics of population migration and industrialization and urbanization processes.

The index of relative rurality (IRR), are designed to address the limitations of discrete rurality measures by Waldorf [25]. IRR is a continuous, multi-dimensional index to measure the degree of rurality, which is constructed based on four dimensions, including population size, population density, the extent of the urban area, and remoteness (distance to the closest metropolitan area). According to Waldorf and Kim [26], there are three advantages of the IRR over the typology-based rurality measures. First, it reflects rurality as a relative concept in that it shows how a place’s rurality is relative to that of other places. Second, as a continuous measure, the IRR is able to capture small changes over space and time than typologies. For example, empirical research shows that the IRR at the ZIP Code level is more relevant to health services research than the RUCA codes and reflects nuanced differences in rurality across observations [27]. Third, this index supports more advanced analytical models, for instance models that show linear and non-linear effect of rurality.

The innovative continuous measure of rurality enables me to assemble a longitudinal dataset of county-level opioid mortality and characteristics including IRR, and thus examine how the dynamics of changing in rurality (e.g., population change and urbanization processes) shape the development of the opioid crisis. Instead of being defined as rural/urban areas based on categorical divisions that barely change over time, counties are separated as those whose degree of rurality increases (i.e., becoming more rural) and those whose degree of rurality decreases (i.e., becoming more urban). Given the documented rural-urban disparities in the opioid crisis [3, 5, 7, 8], the developing trends of opioid outcomes in these two groups of counties as well as the drivers of these trends may be different. To my knowledge, this study is among the first analyzing the different experiences in the opioid crisis of counties based on the changes of rurality over time, which refers to, becoming more rural and becoming more urban. The research contributes to the understanding of how rural-urban differences and within-rural and within-urban heterogeneities in opioid outcomes have been constructed over time by utilizing a continuous rurality measure to capture the nuanced temporal and spatial change in the degree of rurality. Considering the three major types of opioid mortality including prescription opioid-, heroin-, and synthetic opioid-involved mortality, my analysis answers two research questions:

How do prescription opioid-, heroin-, and synthetic opioid-involved mortality rates in becoming-more-rural and becoming-more-urban counties develop over time?
How do determinants of the developments of the three types of opioid mortality differ between becoming-more-rural and becoming-more-urban counties?

The longitudinal dataset across U.S. counties for this study are assembled based on a variety of county-level national data, and stratified according to the index of relative rurality (IRR) [4]. The dependent variables, three types of opioid-related mortality rates, are constructed based on confidential Multiple-Cause of Death (MCOD) files from the National Vital Statistics System maintained by the U.S. Centers for Disease Control and Prevention (CDC) (NCHS, 2020). Prescription opioid-involved deaths are defined as fatal drug overdose deaths involving drug poisonings (X40-44, X60-64, X85, or Y10-14) and the presence of prescription opioids including methadone (T40.2 or T40.3) according to the International Classification of Diseases (ICD-10) codes. Heroin- and synthetic opioid-involved deaths are similarly defined as fatal drug overdose deaths involving drug poisonings with the presence of heroin or opium (T40.0 or T40.1) and the presence of synthetic opioids (excluding methadone) or unknown narcotics (T40.4 or T40.6), respectively. The three types of mortality rates are corresponding numbers of deaths per 100,000 persons in a county and also age-adjusted based on the 2000 U.S. standard population. To reduce annual fluctuations in counties with small populations, I calculate the average value over three-year periods between 2009–2017, and for the year 2018 the average mortality rate is computed between 2017 and 2018.

The measure of rurality, IRR, is an unweighted average of four dimensions, including logarithm of population size, logarithm of population density, distance to the closest metropolitan area, and the percentage of the urban area (as defined by the U.S. Census Bureau) of total land area, which are rescaled to range between 0 and 1. The original IRR ranges from 0 to 1, with the larger value indicating more rurality. The index is multiplied by 100 in this study. Waldorf and Kim [4] provided data on the IRR of both 2000 and 2010 for 3,134 counties. There are 1,557 counties experiencing increase in IRR between 2000 and 2010 which are referred as “becoming-more-rural” counties and 1,577 experiencing decreasing in IRR are referred as “becoming-more-urban” ones. The IRR is also included in the analysis as a covariate, and its values in 2011–2018 are obtained through extrapolation.

To analyze the temporal changes of three types of opioid-involved mortality, the sample for each type of opioid mortality excludes counties where the number of the corresponding opioid-involved deaths remains zero during 2010–2018. There are 83% (1,292 out of 1,557) of becoming-more-rural counties and 96% (1,510 out of 1,577) of becoming-more-urban counties having observed prescription opioid-involved deaths during 2010–2018, thus the sample for prescription opioid-involved mortality ends up with 2,802 counties. Only 54% (848 out of 1,557) becoming-more-rural counties have developed any heroin-involved deaths during these years and for becoming-more-urban counties the number is 82% (1,294 out of 1,577), thus the sample for heroin-involved mortality contains 2,142 counties. The synthetic opioid-involved mortality sample contains 2,657 counties, consist of 77% (1,192 out of 1,557) becoming-more-rural counties and 93% (1,465 out of 1,577) becoming-more-urban counties.

I consider a variety of sociodemographic and healthcare supply factors that may explain the development of opioid mortality. Data about demographic composition and socioeconomic characteristics are drawn from the American Community Survey (ACS) 5-year estimates [29]. I use the values of the 5-year estimates to represent the mid-year of the 5-year span (e.g., 2008–2012 5-year estimates for the year of 2010). The estimates for the year 2018 are obtained through extrapolation. Demographic composition refers to the percentage of males and the percentages of non-Hispanic blacks (blacks hereafter) and Hispanics. Socioeconomic characteristics include educational attainment, occupation, marriage status, insurance coverage, veteran status, welfare reliance and economic conditions. Educational attainment is measured by the percentage of the population aged 25 years and older with a high-school degree or above. For occupation, the percentage of people engaged in manual labor occupations is the percentage of the civilian employed population (16 years and over) who work in agriculture, forestry, fishing and hunting, mining, construction, and manufacturing. The percentage of people who are separated or divorced from their partners are also included. Insurance coverage is measured by the percentage of people aged 18 and over having no insurance coverage. The percentage of veterans and the percentage of households in which one or more current members receiving public assistance during the past 12 months are also considered in the analysis. The poverty rate and unemployment rate are used to indicate the economic conditions of a county. The poverty rate is calculated as the percentage of people whose income in the past 12 months falls below the poverty level. The unemployment rate is the percentage of people who are 16 years and older, who are in the labor force but unemployed.

A county-level social capital index introduced by the Joint Economic Committee [30] is also included. The index includes four subindices: family unity, community health, institutional health, and collective efficacy. I constructed the data of the social capital index for the year of 2010 and 2018. The 2010 index ranges from − 4.45 to 3.27 while in 2018, the range of the index narrowed to -4.17 to 3.17. The values for the years between 2010–2018 are obtained using interpolation method.

For opioid and healthcare supply, I consider opioid prescribing and the availability of primary healthcare services. The county-level opioid prescribing rates (i.e., the total number of opioid prescriptions dispensed annually per 100 annual residents) from 2010 to 2018 are based on the U.S. Opioid Dispensing Rate Maps provided by IQVIA Xponent from the CDC [31]. To ensure the values are comparable across years, the values of prescribing rates in 2017 and 2018 should be both multiplied by an inflation factor (1.019). The numbers of five types of providers with a National Provider Identifier (NPI) per 10,000 residents are used to measure the availability of primary healthcare services: active medical doctors (MDs hereafter), nurse practitioners (NPs hereafter), physician assistants (PAs hereafter), dentists, and chiropractors. All estimates from 2010 to 2018 are obtained from the Area Health Resources Files (AHRF) 2019–2020 [32].

To assemble a complete set of panel data for all variables for longitudinal analysis, the missing values of prescribing rates and social capital index in counties are imputed using Markov chain Monte Carlo model with 30 multiple imputations. The imputed values show high relative consistency (all above 0.98). The complete panel data for prescription opioids-, heroin-, and synthetic opioid-involved mortality samples have 25,218 (2,802×9), 19,278 (2,142 ×9). and 23,913 (2,657×9) county-years correspondingly. I first present a figure comparing the developing trends of the average mortality rates for three types of opioids based on separate samples of becoming-more-rural and becoming-more-urban counties throughout 2010–2018. Next, spatial distributions of counties based on the change of rurality for samples of each type of opioid mortality are shown in thematic maps. The maps also show the distribution of counties with and without any deaths involving the corresponding type of opioids. Then I compare descriptive statistics of becoming-more-rural and becoming-more-urban samples based on pooled data. T-test results are included to see if the differences between two separate samples are significant. Then three two-way fixed effects Poisson regression models are conducted for three samples of each type of opioid mortality, considering the skewed distributions of all three types of opioid mortality. Taking prescription opioid-involved mortality for example, Model 1 is based on becoming-more-rural counties, Model 2 is based on becoming-more-urban counties, and Model 3 is based on the overall sample with interaction terms (multiplying all variables and the change of rurality status). This model is conducted to show the differences between the coefficients of the two sub-samples [33], thus in Model 3 only the coefficients of the interaction terms are presented. Similarly, Models 4–6 are for the heroin-involved mortality sample and Models 7–9 are for the synthetic opioid-involved mortality sample.

Figure 1 shows significantly divergent developing patterns for three types of opioid mortality among counties with the corresponding type of opioid-involved deaths. Overall, the trends for two groups of counties are similar. The average county-level prescription opioid-involved mortality rates are persistently higher for becoming-more-urban counties than becoming-more-rural counties from 2010 to 2018, and in both groups of counties, the rates have been declining especially since 2015. In 2018, the average prescription opioid-involved mortality rate is 5.4 per 100,000 for becoming-more-urban counties and 4.8 for becoming-more-rural counties. The average heroin-involved mortality rates have increased since 2010 and become stable since 2016 for both becoming-more-rural and becoming-more-urban counties. The heroin-involved mortality rates in becoming-more-rural counties are higher in recent years, with 4.6 per 100,000 in 2018, compared to 4.3 in becoming-more-urban counties. The average synthetic opioid-involved mortality rates have increased rapidly for both groups of counties in these years, from around 3 per 100,000 in 2014 to over 8 per 100,000 in 2018.

The spatial distributions of becoming-more-rural and becoming-more-urban counties for three samples of different types of opioid mortality are shown in Fig. 2. Panel 1 is for the prescription opioid-involved mortality sample, and Panel 2 and 3 are for the heroin- and synthetic opioid-involved mortality samples. Overall, most counties in the West, along the East coast and Florida are becoming more urban, while many counties in the Northeast, the Great Plains and the South are becoming more rural. The maps also show that some counties in the Great Plains, which are also mostly rural, have no opioids-involved mortality throughout the time period (2010–2018) in this study. For the heroin-involved mortality sample, most counties in the Great Plains and some counties in the West and the South are excluded from our analysis, which also indicates that the heroin-related deaths have emerged in urban areas first.

A general overview of differences between counties that become more rural and become more urban is presented in Table 1. Based on the pooled data for 9 years (i.e., 2010–2018), there are large and statistically significant differences between these two groups. The average county-level prescription opioid-involved mortality rate is 5.37 per 100,000 for becoming-more-rural counties and 5.80 for becoming-more-urban counties, and the average county-level heroin- and synthetic opioid-involved mortality rates for these two groups are 3.00 versus 2.81 and 4.35 versus 4.20. The differences for prescription opioid- and heroin-involved mortality are statistically significant. For all three samples, becoming-more-rural counties usually have lower percentages of racial and ethnic minorities (i.e. blacks and Hispanics), people with high school education and above, uninsured people and veterans, but have higher percentages of males, people engaged in manual labor occupations, people who are separated or divorced, and households receiving public assistance compared to becoming-more-urban counties. Though becoming-more-rural counties have lower prescribing rates and more social capital, poverty rates are higher and the availability of health service providers (including MDs, NPs, PAs, and dentists) are significantly more limited in these counties. This is consistent with previous literature that more disadvantaged populations are concentrated and left behind in remote and less-developed rural areas (Keyes et al., 2014; Monnat & Pickett, 2011). In addition, unemployment rates and the numbers of chiropractors are similar for both groups of counties. It is important to note here that these crude comparisons are only based on the pooled data thus may be concealing. In later analysis, the longitudinal structures of these data are considered.

Table 2 presents the results of three fixed effects models based on 30 imputation datasets of the sample of prescription opioid-involved mortality. Model 3 shows that a number of determinants, including the IRR, demographic composition, occupation composition, social capital, and opioid and healthcare supply factors, have significantly different impacts on prescription opioid-involved mortality for the two groups of counties. For becoming-more-rural counties, a one unit increase in the IRR is associated with approximately 20% (1-exp(-0.230) = 0.206) decrease in the prescription opioid-involved mortality rate, while for becoming-more-urban counties, the association between changes in the IRR and the prescription opioid-involved mortality rate is not significant, controlling for all other covariates. Increasing percentages of Hispanics are associated with decreasing prescription opioid-involved mortality rates for becoming-more-rural counties but with increasing prescription opioid-involved mortality rates for becoming-more-urban counties. For becoming-more-rural counties, every one percent increase in the male population and the population engaging in manual labor occupations correlate to 3.9% (exp(0.038)-1 = 0.039) and 1.7% (exp(0.017)-1 = 0.017) increase in prescription opioid-involved mortality rate respectively, but these associations do not exist for becoming-more-urban counties. Increasing social capital is associated with decreasing prescription opioid-involved mortality for becoming-more-urban counties while this relationship is not found among becoming-more-rural counties. The opioid supply factors also have different effects for both groups. While the increasing prescribing rate will increase the prescription opioid-involved mortality rate, the magnitude of this effect is significantly larger for becoming-more-urban counties, which is that an increase in the prescribing rate per 100 persons correlates to 0.4% (exp(0.004)-1 = 0.004) percent increase in the prescription opioid-involved mortality rate, compared to 0.2% (exp(0.002)-1 = 0.002) for becoming-more-rural counties. Increasing availability of healthcare providers including MDs, NPs, and PAs is associated with increasing prescription opioid-involved mortality for becoming-more-urban counties, but these associations are not found for becoming-more-rural counties. Lastly, some socioeconomic factors play similar roles in shaping the prescription opioid-involved mortality for both groups of counties. Increasing percentages of veterans and uninsured people and decreasing unemployment rates are associated with increasing prescription opioid-involved mortality rates for both groups. In addition, increasing availability of chiropractors are associated with increasing prescription opioid-involved mortality rates for both groups.

Table 3 shows results of fixed effects Poisson regression models for heroin-involved mortality rates. The IRR, demographic composition, socioeconomic factors, and healthcare supply factors all have different impacts on the heroin-involved mortality for both groups of counties. Increasing IRR is significantly associated with increasing heroin-involved mortality for becoming-more-rural counties, while for becoming-more-urban counties there is no such association. Increasing percentages of males are associated with increasing heroin-involved mortality for both groups of counties but for becoming-more-urban counties the magnitude of this impact is significantly larger. One percent increase in people who are separated or divorced from their partners will increase heroin-involved mortality rates in becoming-more-rural counties by 4.3% (exp(0.042)-1 = 0.043), but decrease heroin-involved mortality rates in becoming-more-urban counties by 1.7% (1-exp(-0.017) = 0.017). Increasing percentages of uninsured people, veterans, and people in poverty correlate to increasing heroin-involved morality only for becoming-more-urban counties but not for becoming-more-rural counties. Increasing unemployment rates are associated with decreasing heroin-related mortality but this association is significantly stronger for becoming-more-urban counties. Prescribing rates are not associated with heroin-involved mortality for both groups. Increasing numbers of NPs are associated with decreasing heroin-involved mortality for becoming-more-rural counties, while increasing numbers of dentists correlate to increasing heroin-involved mortality for becoming-more-urban counties. A one unit increase in the number of chiropractors (per 10,000) will increase the heroin-involved mortality rate by 7.9% (exp(0.076)-1 = 0.079) for becoming-more-rural counties but decrease the heroin-involved mortality rate by around 17% (1-exp(-0.189) = 0.172) for becoming-more-urban counties. Some determinants have similar impacts on heroin-involved mortality rates for both groups. Increasing percentages of the population with high school education and above and increasing social capital are associated with increasing heroin-involved mortality for both groups of counties, while increasing percentages of the black population and increasing numbers of MDs decrease heroin-involved mortality for both.

Table 4 shows the comparison of determinants for synthetic opioid-involved mortality rates in becoming-more-rural and becoming-more-urban counties. The IRR, percentage of Hispanics, socioeconomic factors and the availability of healthcare providers have different impacts on synthetic opioid-involved mortality rates in the two groups of counties. The change in the IRR is negatively associated with the change in synthetic opioid-involved mortality for both counties, while the magnitude of this association is much larger for becoming-more-rural counties. Decreasing percentages of the black population, increasing percentages of veterans and increasing percentages of households receiving public assistance are associated increasing synthetic opioid-involved mortality only for becoming-more-urban counties. A one unit increase in poverty rates increases synthetic opioid-involved mortality rates by 2.7% (exp(0.027)-1 = 0.027) for becoming-more-urban counties but only 0.9% (exp(0.009)-1 = 0.009) for becoming-more-rural counties. Increasing percentages of the uninsured population and decreasing unemployment rates are associated with increasing synthetic opioid-involved mortality rates for both groups of counties but these associations are significantly stronger for becoming-more-rural counties. A one unit increase in the number of MDs (per 10,000) increases synthetic opioid-involved mortality rates by 2.6% (exp(0.026)-1 = 0.026) only for becoming-more-urban counties. Increasing numbers of NPs are associated with increasing synthetic opioid-involved mortality for both groups but the association is significantly stronger for becoming-more-urban counties. A one unit increase in the number of chiropractors (per 10,000) decreases synthetic opioid-involved mortality by 6.3% (1-exp(-0.065) = 0.063) for becoming-more-rural counties but oppositely, increases synthetic opioid-involved mortality by 6.3% (exp(0.061)-1 = 0.063) for becoming-more-urban counties. Lastly, both samples observe that increasing percentages of males and people engaged in manual labor occupations, increasing social capital, and increasing availability of PAs are associated increasing synthetic opioid-involved mortality.

Taking advantage of a continuous measure of rurality, this study separates the U.S. counties into two groups as becoming-more-rural and becoming-more-urban according to whether the degree of rurality increases or decreases between 2000 and 2010, and examines the developing trends and determinants of three different types of opioid mortality, including prescription opioid-, heroin-, and synthetic opioid-involved mortality, during the years of 2010–2018. The results show significant differences in the developing trends of three types of opioid-involved mortality rates as well as in the determinants of the development of opioid-involved mortality for becoming-more-rural and becoming-more-urban counties. The study helps understanding the relationship between rurality and opioid-involved mortality, which further reflects the rural and urban difference in facing the hit of the opioid crisis.

It should be noted first that the analysis is only based on counties that have observed any prescription opioid-, heroin-, or synthetic opioid-involved deaths during 2010–2018, with a larger number of becoming-more-rural counties, mostly in the Great Plains area, being excluded since they have observed no such deaths compared to becoming-more-urban counties. This is consistent with evidence that some rural counties in the Great Plains have the lowest prescribing rates [34] and illicit drug markets are firstly developed in urban and metropolitan areas [3, 8, 21]. The overall developing trends of three opioid-involved mortality rates are similar for both groups of counties, especially that the average synthetic opioid-involved mortality has very similar values between two groups during its rapid growing period (i.e., 2014–2018). The average prescription opioid- and heroin-involved mortality rate has been stabilized and even declined after 2016 for both groups, but compared to becoming-more-urban counties, the average prescription opioid-involved mortality in becoming-more-rural counties is consistently lower while the average heroin-involved mortality rate is higher since 2014.

Though sharing some similarities, the determinants accounting for the development of opioid-involved mortality are largely different for becoming-more-rural and becoming-more-urban counties. First, the change in the degree of rurality may lead to different challenges in both groups of counties. Consistent with previous research showing that urban and metropolitan areas suffer from more rapid growth of drug overdose deaths involving synthetic opioids than rural areas [8, 35], the degree of rurality negatively correlates to synthetic opioid-involved mortality, thus becoming-more-urban counties are at a high risk of more deaths involving synthetic opioids. For becoming-more-rural counties that have heroin-involved deaths, increasing rurality correlates to increasing heroin-involved mortality, which is unexpected given that heroin use in rural areas is not as prevalent as in urban areas [6]. Thus, heroin-involved overdoses in rural areas are growing and warrant attention. Another unexpected finding is that increasing rurality does not necessarily predict the increases in the prescription opioid-involved mortality, though prescribing rates in rural areas are found higher than that in urban areas [3, 34]. This indicates that overdoses involving prescription opioids continue to be a problem in areas that becoming more urbanized.

Second, for the two types of counties, the change in opioid-involved mortality is shaped by different demographic and socioeconomic characteristics, showing that the groups with greater vulnerability in rural and urban areas may be different. For becoming-more-rural counties, the percentages of males and people engaged in manual labor occupations positively correlate with prescription opioid-involved mortality, indicating a particular disadvantage of rural counties that are dependent on labor-intensive industries. In general, counties with higher percentages of racial and ethnic minority groups have lower opioid-involved mortality, which is aligned with studies showing that these people have been hit less by the crisis compared to the white population [9, 36]. However, I also find that those becoming-more-urban counties with increasing percentages of Hispanics suffer from increasing prescription opioid-involved mortality, which may indicate a higher risk of Hispanic residents in urban areas compared to other groups. Regarding family distress, increasing percentage of people who are separated or divorced from their partners correlates to increasing prescription opioid- and heroin-involved mortality only for becoming-more-rural counties, but for becoming-more-urban counties separation or divorce decreases heroin-involved deaths. There are several explanations for this divergence. On one hand, family distress may bother rural residents more and increase their drug use due to the higher prevalence of traditional and conservative values in rural areas compared to urban areas. On the other hand, social networks may play more important roles in facilitating the circulation of illicit drugs in urban counties than in rural counties. It is found that urban drug users are more likely to obtain opioids from their family and friends than their rural counterparts are [19, 20]. Thus, separation or divorce may cut off this channel of heroin distribution for urban residents.

In addition to the supportive evidence shown in this study for the “deaths of despair” arguments that people who are socioeconomically marginalized suffer from a greater risk of drug addiction and overdose [8, 9, 37, 38], the disadvantages of the concentration of groups at risk are found particularly stronger for becoming-more-urban counties in the development of heroin- and synthetic opioid-involved mortality, compared to becoming-more-rural counties. The possible explanation is that the “drug environments” (i.e., illicit drug supplies), are more developed in urban areas than in rural areas [1–3], thus the illicit opioid crisis in urban areas is more driven by the demand-side factors and hit the groups at higher risk of socioeconomic distress and drug abuse and addiction more compared to that in rural areas. Nonetheless, due to the growing interdependency between rural and urban areas and the growing illicit drug markets (especially for illicitly manufactured fentanyl) in rural areas [3, 5], rural residents that have been disadvantaged and marginalized would increasingly suffer from the crisis. For example, the percentage of the uninsured population is found to be associated with synthetic opioid-involved mortality in becoming-more-rural counties and this association is significantly stronger than that in becoming-more-urban counties.

Third, the contrary impacts of social capital on the development of the opioid crisis, which include protecting people from drug addiction and overdoses through family and community supports [39–41] and facilitating the circulation of drugs [34, 42, 43], are shown in becoming-more-urban counties, while becoming-more-rural counties are found to suffer more from the downside of social capital. For both groups of counties, increasing social capital is related to increasing heroin- and synthetic opioid-involved mortality, which may be explained by the positive impact of social capital on the growth of illicit drug markets [44, 45]. Regarding prescription opioid-involved mortality, however, social capital show protective impacts only for becoming-more-urban counties but not for becoming-more-rural counties. This may be explained by the conservative cultures and great social and kinship networks in rural communities that restrict personal privacy and autonomy and thus cultivate the drug use stigma and prevent people from seeking treatment of addiction [6, 45–51]. These characteristics of rural communities can be strengthened by more social capital, therefore residents in rural counties with more social capital may be exposed to higher risk of drug misuse and overdose, leading to less protection that social capital has on prescription opioid-involved deaths in rural areas.

Lastly, increasing opioid supply and availability of healthcare providers generally increases opioid-involved mortality for becoming-more-urban counties while for becoming-more-rural counties the impacts of these factors are mixed. On one hand, increasing NPs are found to decrease prescription opioid- and heroin-involved mortality for becoming-more-rural counties, and the positive impact of opioid prescribing on prescription opioid-involved mortality is less strong for becoming-more-rural counties than for becoming-more-urban counties. A possible explanation is that since rural communities are underserved by healthcare providers and prescribers [3, 12], the healthcare shortage may be severer in counties that continue to lose population and become more rural. Thus in these areas, opioid prescribing may not be as harmful as that in urban areas, and residents in counties with more NPs may enjoy better healthcare services and are less likely to die from overdoses. Differently, for urban areas increasing numbers of primary care providers may reflect more demand for health problems and opioid prescriptions, thus are associated with more deaths involving both prescription opioids and illicit opioids use. On the other hand, for becoming-more-rural counties, increasing availability of NPs and PAs predicts increasing synthetic opioid-involved mortality. Moreover, opioid prescribing is negatively associated with synthetic opioid-involved mortality. These findings indicate that the development of synthetic opioid-involved mortality in rural areas may be very different from the other two types of opioid mortality. Previous literature has shown that in the current wave of the opioid crisis, synthetic opioids have become popular alternatives to prescription opioids and gradually occupied traditional illicit drug markets because of the greater potency, lower costs, and easier distribution [52, 53], and supply-side factors are major drivers of the growth of deaths involving synthetic opioids [54–56]. In counties that are becoming more rural and socioeconomic marginalized, higher percentages of primary care services may also indicate more developed synthetic opioids markets, thus correlate to higher synthetic opioid-involved mortality. In addition, the detection of synthetic opioids especially the novel ones is particularly harder than other types of opioids [57, 58], which also makes it difficult for rural healthcare providers to discover and treat synthetic opioid abuse, thus undermines the protective impacts of healthcare services on synthetic opioid overdoses.

The study is subject to several limitations. First, to compare the coefficients for two samples, the study includes the interaction of the “becoming-more-urban or becoming-more-rural” variable with all other factors into the models. The assumption of this approach is that all other covariates are the same between these two groups, which may overestimate the differences in the coefficients. For further research, the comparison could be done with more advanced methods. Second, as discussed previously, one of the most important factors is the cultural differences in rural and urban areas, but it is not easy to measure and include in the analysis. The differences in the impacts of measured variables, like social capital and the availability of healthcare services, may be explained by different cultural attributes (e.g., patient-provider relationships, privacy, drug use stigma) in these two areas. Third, the ecological fallacy and modifiable areal unit problem (MAUP) are inevitable for ecological studies. The findings need to be further evaluated in future studies based on other different levels.

Nevertheless, the study contributes to the extant literature by identifying different drivers of the development of the opioid crisis in counties that are becoming more rural and becoming more urban. The findings show that though the developing trends of opioid-involved mortality rates in the two groups of counties are generally similar, the drivers of the developments are different. Demand-side factors play more significant roles in facilitating the growth of opioid overdose deaths, especially those involving illicit drugs, in becoming-more-urban counties than in becoming-more-rural counties, possibly due to the higher availability of illicit drugs in urban areas. Becoming-more-urban counties suffer more from the concentration of socioeconomically disadvantaged groups and increasing numbers of healthcare providers than becoming-more-rural counties. However, despite the beneficial impacts of increasing availability of healthcare services on decreasing the prescription opioid- and heroin-involved deaths in becoming-more-rural areas, synthetic opioid markets may first develop in relatively better-off rural counties, leading to an inevitable growth of synthetic opioid overdoses. The findings suggest that policies and regulations for controlling the development of the opioid crisis should be localized and adjusted according to the change of the rurality of counties. For areas that are being urbanized and gaining population, communities and populations that are socioeconomically deprived and demand for more opioid and healthcare supplies deserve particular attention in the crisis, while for areas that are experiencing population loss and growing marginalized, promoting healthcare equity as well as enforcing illicit drug market disruption strategies are both important.

Funding

The study is supported by the Fall 2020 Benevolent Association Research and Creative Activity Grant at the University at Albany and the 2021 Dissertation Award from the Rural Sociological Society.

Conflicts of interest

The author has no conflicts of interest to declare that are relevant to the content of this article.

Data availability

The data that support the findings of this study are from a variety of national datasets, including restricted datasets from confidential Multiple-Cause of Death (MCOD) files from the National Vital Statistics System maintained by the U.S. Centers for Disease Control and Prevention, which are not publicly available, and publicly available datasets, which are the U.S. Opioid Dispensing Rate Maps (https://www.cdc.gov/drugoverdose/rxrate-maps/index.html), the Index of Relative Rurality (https://purr.purdue.edu/publications/2960/1), the Joint Economic Committee Social Capital Project (https://www.jec.senate.gov/public/index.cfm/republicans/2018/4/the-geography-of-social-capital-in-america), the American Community Survey (https://usa.ipums.org/usa/), and the Area Health Resources Files (https://data.hrsa.gov/data/download).

Code availability

The codes used for this analysis are for STATA and R, and are available upon request.

Authors' contributions

Feinuo Sun is the sole author of this manuscript.

Acknowledgement

The study is supported by the Fall 2020 Benevolent Association Research and Creative Activity Grant at the University at Albany and the 2021 Dissertation Award from the Rural Sociological Society. The author also would like to thank Dr. Tse-Chuan Yang, Dr. Kate Strully, Dr. Glenn Deane and Seulki Kim for their help to improve the paper.

Jalal H, Buchanich JM, Roberts MS, Balmert LC, Zhang K, Burke DS. Changing dynamics of the drug overdose epidemic in the United States from 1979 through 2016. Science (80- ). 2018;361.
Mack KA, Jones CM, Ballesteros MF. Illicit Drug Use, Illicit Drug Use Disorders, and Drug Overdose Deaths in Metropolitan and Nonmetropolitan Areas—United States. Am J Transplant. 2017;17:3241–52.
Monnat SM. Opioid Crisis in the Rural U.S. 2020;30:117–43.
Waldorf B, Kim A. The Index of Relative Rurality (IRR): U.S. County Data for 2000 and 2010. Purdue Univ. Res. Repos. 2018.
Monnat SM. The contributions of socioeconomic and opioid supply factors to U.S. drug mortality rates: Urban-rural and within-rural differences. J Rural Stud [Internet]. Elsevier; 2019;68:319–35. Available from: https://doi.org/10.1016/j.jrurstud.2018.12.004
Rigg KK, Monnat SM, Chavez MN. Opioid-related mortality in rural America: Geographic heterogeneity and intervention strategies. Int J Drug Policy [Internet]. Elsevier; 2018;57:119–29. Available from: https://doi.org/10.1016/j.drugpo.2018.04.011
Pear VA, Ponicki WR, Gaidus A, Keyes KM, Martins SS, Fink DS, et al. Urban-rural variation in the socioeconomic determinants of opioid overdose. Drug Alcohol Depend. Elsevier; 2019;195:66–73.
Peters DJ, Monnat SM, Hochstetler AL, Berg MT. The Opioid Hydra: Understanding Overdose Mortality Epidemics and Syndemics Across the Rural-Urban Continuum. Rural Sociol. 2020;85:589–622.
Dasgupta N, Beletsky L, Ciccarone D. Opioid Crisis: No Easy Fix to Its Social and Economic Determinants. Am J Public Health. 2018;108:182–6.
Ayres I, Jalal A. The impact of prescription drug monitoring programs on U.S. Opioid prescriptions. J Law, Med Ethics. 2018;46:387–403.
Sears JM, Edmonds AT, Fulton-Kehoe D. Tracking Opioid Prescribing Metrics in Washington State (2012-2017): Differences by County-Level Urban-Rural and Economic Distress Classifications. J Rural Heal. 2019;0:1–15.
Shoff C, Yang TC, Kim S. Rural/Urban Differences in the Predictors of Opioid Prescribing Rates Among Medicare Part D Beneficiaries 65 Years of Age and Older. J Rural Heal. 2020;00:1–11.
Yang TC, Kim S, Shoff C. Income Inequality and Opioid Prescribing Rates: Exploring Rural/Urban Differences in Pathways via Residential Stability and Social Isolation. Rural Sociol. 2020;0:1–24.
Jensen L, Monnat SM, Green JJ, Hunter LM, Sliwinski MJ. Rural Population Health and Aging: Toward a Multilevel and Multidimensional Research Agenda for the 2020s. Am J Public Health. 2020;
Lichter DT, Brown DL. Rural America in an urban society: Changing spatial and social boundaries. Annu Rev Sociol. 2011;37:565–92.
Thiede BC, Butler JLW, Brown DL, Jensen L. Income Inequality across the Rural-Urban Continuum in the United States, 1970–2016. Rural Sociol. 2020;
Lichter DT, Jensen L. Rural America in Transition: Poverty and Welfare at the Turn of the Twenty-First Century. Rural Dimens. Welf. Reform. 2002.
James WL. All rural places are not created equal: Revisiting the rural mortality penalty in the United States. Am J Public Health. 2014;104:2122–9.
Monnat SM, Rigg KK. Examining Rural/Urban Differences in Prescription Opioid Misuse Among US Adolescents. J Rural Heal. 2016;32:204–18.
Young AM, Havens JR, Leukefeld CG. A comparison of rural and urban nonmedical prescription opioid users’ lifetime and recent drug use. Am J Drug Alcohol Abuse. 2012;38:220–7.
Zoorob M. Fentanyl shock: The changing geography of overdose in the United States. Int J Drug Policy. Elsevier; 2019;70:40–6.
Monnat SM, Peters DJ, Berg MT, Hochstetler A. Using census data to understand county-level differences in overall drug mortality and opioid-related mortality by opioid type. Am J Public Health. 2019;109:1084–91.
Johnson KM, Lichter DT. Metropolitan Reclassification and the Urbanization of Rural America. Demography. Demography; 2020;
Lichter DT, Ziliak JP. The Rural-Urban Interface: New Patterns of Spatial Interdependence and Inequality in America. Ann Am Acad Pol Soc Sci [Internet]. 2017;672:6–25. Available from: https://doi.org/10.1177/0002716217714180
Waldorf B. A Continuous Multi-dimensional Measure of Rurality: Moving beyong the threshold measures. Annu Meet Long Island, California, July 24-27. 2006;1–29.
Waldorf B, Kim A. Defining and Measuring Rurality in the US: From Typologies to Continuous Indices. Work Ration Rural Classif April Natl Acad Sci Eng Med Washington, DC Available http//sites Natl org/DBASSE/CNSTAT/DBASSE_160632 [November 2015]. 2015;
Inagami S, Gao S, Karimi H, Shendge MM, Probst JC, Stone RA. Adapting the Index of Relative Rurality (IRR) to Estimate Rurality at the ZIP Code Level: A Rural Classification System in Health Services Research. J Rural Heal. 2016;32:219–27.
National Center for Health Statistics (NCHS). Compressed Mortality File, 1999-2018 as compiled from data provided by the 57 vital statistics jurisdictions through the Vital Statistics Cooperative Program [Internet]. Hyattsville, MD; 2020. Available from: http://wonder.cdc.gov/cmf-icd10.html
U.S. Census Bureau. American Community Survey [Internet]. IPUMS. 2021 [cited 2020 Jan 31]. Available from: https://usa.ipums.org/usa/
Joint Economic Committee (JEC). The Geography of Social Capital in America [Internet]. 2018 [cited 2020 Apr 1]. Available from: https://www.jec.senate.gov/public/index.cfm/republicans/2018/4/the-geography-of-social-capital-in-america
Centers for Disease Control and Prevention (CDC). U.S. Opioid Dispensing Rate Maps [Internet]. Opioid Overdose. 2020 [cited 2020 May 1]. Available from: https://www.cdc.gov/drugoverdose/maps/rxrate-maps.html
HRSA. Area Health Resources Files. HRSA data Wareh. 2019.
Clogg CC, Petkova E, Haritou A. Statistical Methods for Comparing Regression Coefficients Between Models. Am J Sociol. 1995;100.
Sun F. Rurality and opioid prescribing rates in U.S. counties from 2006 to 2018: A spatiotemporal investigation. Soc Sci Med. 2022;296.
Hedegaard H, Spencer MR. Urban–Rural Differences in Drug Overdose Death Rates, 1999–2019. NCHS data brief, no 403. 2021;1–8.
Grigoras CA, Karanika S, Velmahos E, Alevizakos M, Flokas ME, Kaspiris-Rousellis C, et al. Correlation of Opioid Mortality with Prescriptions and Social Determinants: A Cross-sectional Study of Medicare Enrollees. Drugs. 2018;
Case A, Deaton A. Deaths of Despair and the Future of Capitalism. Deaths Despair Futur. Capital. 2020.
Hollingsworth A, Ruhm CJ, Simon K. Macroeconomic conditions and opioid abuse. J Health Econ. 2017;56:222–33.
Stumbo SP, Yarborough BJH, Janoff SL, Yarborough MT, McCarty D, Green CA. A qualitative analysis of family involvement in prescribed opioid medication monitoring among individuals who have experienced opioid overdoses. Subst Abus. 2016;
Monnat SM. Factors Associated With County-Level Differences in U.S. Drug-Related Mortality Rates. Am J Prev Med [Internet]. Elsevier Inc.; 2018;54:611–9. Available from: http://dx.doi.org/10.1016/j.amepre.2018.01.040
Zoorob M, Salemi JL. Bowling alone, dying together: The role of social capital in mitigating the drug overdose epidemic in the United States. Drug Alcohol Depend [Internet]. Elsevier Ireland Ltd; 2017;173:1–9. Available from: http://dx.doi.org/10.1016/j.drugalcdep.2016.12.011
Eitan S, Emery MA, Bates MLS, Horrax C. Opioid addiction: Who are your real friends? Neurosci Biobehav Rev. 2017;83:697–712.
Jonas AB, Young AM, Oser CB, Leukefeld CG, Havens JR. OxyContin ® as currency: OxyContin ® use and increased social capital among rural Appalachian drug users. Soc Sci Med [Internet]. Elsevier Ltd; 2012;74:1602–9. Available from: http://dx.doi.org/10.1016/j.socscimed.2011.12.053
Bichler G, Malm A, Cooper T. Drug supply networks: A systematic review of the organizational structure of illicit drug trade. Crime Sci. 2017;6.
Portes A, Landolt P. The Downside of Social Capital. Am Prospect. 1996;26:18–22.
Hewell VM, Vasquez AR, Rivkin ID. Systemic and individual factors in the buprenorphine treatment-seeking process: A qualitative study. Subst Abus Treat Prev Policy. 2017;
Jalali M, Botticelli M, Hwang R, Koh HK, McHugh RK. The Opioid Crisis: A Contextual Framework and Call for Systems Science Research. SSRN Electron J. 2019;1–19.
Olsen Y, Sharfstein JM. Confronting the stigma of opioid use disorder - And its treatment. JAMA - J. Am. Med. Assoc. 2014.
Weisheit RA, Donnermeyer JF. Change and continuity in crime in rural America. Crim justice. 2000;
Deller S, Deller M. Spatial Heterogeneity, Social Capital, and Rural Larceny and Burglary. Rural Sociol. 2012;77:225–53.
Keyes KM, Cerdá M, Brady JE, Havens JR, Galea S. Understanding the rural-urban differences in nonmedical prescription opioid use and abuse in the United States. Am J Public Health. 2014;104:52–9.
Pardo B, Caulkins J, Kilmer B, Pacula R, Reuter P, Stein B. The Synthetic Opioid Surge in the United States: Insights from Mortality and Seizure Data. Synth. Opioid Surge United States Insights from Mortal. Seizure Data. 2019.
Baldwin GT, Seth P, Noonan RK. Continued Increases in Overdose Deaths Related to Synthetic Opioids. JAMA. 2021;325.
Mars SG, Rosenblum D, Ciccarone D. Illicit fentanyls in the opioid street market: desired or imposed? Addiction. 2019;114.
Ciccarone D. The triple wave epidemic: Supply and demand drivers of the US opioid overdose crisis. Int. J. Drug Policy. 2019.
Ciccarone D. The rise of illicit fentanyls, stimulants and the fourth wave of the opioid overdose crisis. Curr. Opin. Psychiatry. 2021.
Karila L, Marillier M, Chaumette B, Nicolas F, Amine B. New Synthetic Opioids: Part of a new addiction landscape. Neurosci. Biobehav. Rev. 2019.
Armenian P, Vo KT, Barr-Walker J, Lynch KL. Fentanyl, fentanyl analogs and novel synthetic opioids: A comprehensive review. Neuropharmacology. 2018.

Table 1

Descriptive statistics for the samples of becoming-more-rural and becoming-more-urban counties^b
	Prescription opioid-involved mortality		Heroin-involved mortality		Synthetic opioid-involved mortality
	Becoming-more-rural	Becoming-more-urban	Becoming-more-rural	Becoming-more-urban	Becoming-more-rural	Becoming-more-urban
	Mean (St.D)	Mean (St.D)	Mean (St.D)	Mean (St.D)	Mean (St.D)	Mean (St.D)
Prescription opioid-involved mortality rate	5.37 (7.03)	5.80 (5.79) ^***
Heroin-involved mortality rate			3.00 (4.43)	2.81 (3.98) ^**
Synthetic opioid-involved mortality rate					4.35 (6.76)	4.20 (6.28)
IRR	52.53 (8.79)	45.95 (9.96) ^***	50.35 (9.66)	44.47 (9.85) ^***	52.06 (8.88)	45.61 (9.82) ^***
Percent males	50.08 (2.37)	49.88 (2.16) ^***	50.08 (2.43)	49.70 (1.90) ^***	50.05 (2.30)	49.86 (2.10) ^***
Percent non-Hispanic blacks	8.97 (15.56)	9.24 (12.38)	8.35 (14.70)	9.39 (12.26) ^***	9.18 (15.78)	9.30 (12.39)
Percent Hispanics	7.54 (12.64)	9.48 (12.61) ^***	8.04 (13.88)	9.88 (13.06) ^***	7.42 (12.54)	9.54 (12.76) ^***
Percent high school and above	85.26 (6.40)	86.22 (6.32) ^***	85.54 (6.25)	86.53 (6.17) ^***	85.26 (6.33)	86.23 (6.34) ^***
Percent manual labor occupations	27.73 (7.61)	23.56 (7.66) ^***	26.52 (7.79)	22.82 (7.57) ^***	27.41 (7.63)	23.46 (7.59) ^***
Percent separated or divorced	13.86 (2.66)	13.66 (2.52) ^***	14.03 (2.54)	13.62 (2.46) ^***	13.94 (2.60)	13.67 (2.51) ^***
Percent uninsured	17.35 (7.92)	17.79 (7.99) ^***	16.45 (7.53)	17.35 (7.81) ^***	17.19 (7.90)	17.61 (7.81) ^***
Percent veterans	9.49 (2.56)	9.70 (2.97) ^***	9.56 (2.58)	9.66 (2.98) ^*	9.50 (2.56)	9.68 (2.96) ^***
Percent on public assistance	2.62 (1.83)	2.39 (1.56) ^***	2.70 (1.59)	2.40 (1.39) ^***	2.60 (1.69)	2.37 (1.36) ^***
Poverty rate	16.91 (6.51)	15.48 (6.05) ^***	16.56 (5.97)	15.19 (5.89) ^***	16.95 (6.41)	15.41 (5.98) ^***
Unemployment rate	7.19 (3.69)	7.30 (3.23) ^*	7.41 (3.30)	7.33 (3.09)	7.30 (3.63)	7.29 (3.13)
Social capital index	0.04 (0.98)	-0.16 (0.88) ^***	-0.04 (0.90)	-0.20 (0.85) ^***	0.00 (0.97)	-0.17 (0.88) ^***
Prescribing rate	83.14 (49.98)	86.56 (43.25) ^***	85.02 (50.43)	87.29 (42.54) ^**	83.23 (48.75)	86.83 (42.60) ^***
Medical doctors per 10,000	9.93 (14.34)	15.45 (18.14) ^***	11.99 (17.03)	16.87 (19.02) ^***	10.34 (14.81)	15.76 (18.31) ^***
Nurse practitioners per 10,000	4.41 (3.66)	4.48 (3.60)	4.56 (3.86)	4.68 (3.69) ^*	4.41 (3.61)	4.52 (3.61) ^*
Physician assistants per 10,000	2.20 (2.71)	2.63 (3.51) ^***	2.32 (2.73)	2.71 (3.63) ^***	2.17 (2.58)	2.61 (3.50) ^***
Dentists per 10,000	3.99 (2.91)	4.58 (2.61) ^***	4.29 (3.14)	4.79 (2.61) ^***	4.02 (2.58)	4.64 (2.60) ^***
Chiropractors per 10,000	2.16 (1.91)	2.13 (1.55)	2.18 (1.68)	2.16 (1.49)	2.13 (1.78)	2.16 (1.56)
N (county-years)	11,628	13,590	7,632	11,646	10,728	13,185
^b Significance indicates the t-test results of the comparison between two samples.
Level of significance: ^* p < 0.05, ^ p < 0.01, ^* p < 0.001

Table 2

Two-way fixed effects Poisson regression models for prescription opioid-involved mortality rates based on samples of becoming-more-rural (N = 1,292) and becoming-more-urban (N = 1,510) counties
	Model 1 (becoming-more-rural)		Model 2 (becoming-more-urban)		Model 3 (comparison)
	Coef.	S.E.	Coef.	S.E.	Coef.	S.E.
IRR	-0.230^***	0.046	-0.003	0.014	-0.227^***	0.048
Percent males	0.038^***	0.007	-0.001	0.008	0.039^***	0.010
Percent non-Hispanic blacks	-0.021^**	0.008	-0.008	0.007	-0.014	0.011
Percent Hispanics	-0.022^**	0.007	0.017^*	0.008	-0.039^***	0.010
Percent high school and above	0.000	0.003	0.001	0.003	-0.001	0.005
Percent manual labor occupations	0.017^***	0.002	-0.005^*	0.002	0.022^***	0.003
Percent separated or divorced	0.007^*	0.003	-0.003	0.004	0.010	0.005
Percent uninsured	0.015^***	0.002	0.013^***	0.002	0.001	0.002
Percent veterans	0.021^***	0.005	0.026^***	0.005	-0.004	0.007
Percent on public assistance	0.011^*	0.005	0.009	0.005	0.002	0.007
Poverty rate	0.004	0.002	0.006^*	0.003	-0.002	0.003
Unemployment rate	-0.015^***	0.003	-0.022^***	0.003	0.007	0.004
Social capital index	-0.024	0.037	-0.208^***	0.034	0.183^***	0.051
Prescribing rate	0.002^***	0.000	0.004^***	0.000	-0.002^***	0.000
Medical doctors per 10,000	-0.005	0.003	0.007^**	0.003	-0.012^**	0.004
Nurse practitioners per 10,000	-0.012^**	0.003	0.025^***	0.004	-0.036^***	0.005
Physician assistants per 10,000	0.003	0.006	0.032^***	0.007	-0.029^**	0.009
Dentists per 10,000	-0.005	0.007	-0.010	0.009	0.006	0.011
Chiropractors per 10,000	0.092^***	0.013	0.086^***	0.016	0.006	0.020
Year
2011	0.060^**	0.017	0.018	0.016	0.042	0.023
2012	0.107^***	0.019	-0.023	0.017	0.131^***	0.025
2013	0.142^***	0.021	-0.030	0.019	0.172^***	0.028
2014	0.183^***	0.024	-0.009	0.022	0.193^***	0.032
2015	0.226^***	0.027	0.054^*	0.026	0.172^***	0.037
2016	0.265^***	0.030	0.075^*	0.029	0.190^***	0.042
2017	0.238^***	0.033	0.033	0.033	0.205^***	0.047
2018	0.228^***	0.037	-0.006	0.037	0.234^***	0.052
Level of significance: ⁺ p < 0.1, ^* p < 0.05, ^ p < 0.01, ^* p < 0.001

Table 3

Two-way fixed effects Poisson regression models for heroin-involved mortality rates based on samples of becoming-more-rural (N = 848) and becoming-more-urban (N = 1,294) counties
	Model 4 (becoming-more-rural)		Model 5 (becoming-more-urban)		Model 6 (comparison)
	Coef.	S.E.	Coef.	S.E.	Coef.	S.E.
IRR	0.408^***	0.089	-0.042	0.024	0.449^***	0.092
Percent males	0.027^*	0.012	0.069^***	0.014	-0.042^*	0.018
Percent non-Hispanic blacks	-0.061^***	0.014	-0.038^**	0.012	-0.023	0.018
Percent Hispanics	-0.013	0.010	-0.041^**	0.014	0.028	0.017
Percent high school and above	0.024^***	0.006	0.019^**	0.006	0.005	0.009
Percent manual labor occupations	-0.006	0.004	0.007	0.004	-0.013^*	0.006
Percent separated or divorced	0.042^***	0.006	-0.017^*	0.007	0.059^***	0.010
Percent uninsured	-0.003	0.004	0.010^**	0.004	-0.013^*	0.005
Percent veterans	0.001	0.010	0.052^***	0.010	-0.051^***	0.014
Percent on public assistance	-0.015	0.010	0.011	0.010	-0.026	0.014
Poverty rate	0.009	0.005	0.038^***	0.005	-0.029^***	0.007
Unemployment rate	-0.012^*	0.006	-0.042^***	0.006	0.030^***	0.008
Social capital index	0.484^***	0.078	0.356^***	0.081	0.128	0.115
Prescribing rate	0.000	0.001	-0.001	0.001	0.000	0.001
Medical doctors per 10,000	-0.021^***	0.005	-0.020^***	0.004	-0.001	0.007
Nurse practitioners per 10,000	-0.021^**	0.007	0.001	0.007	-0.022^*	0.010
Physician assistants per 10,000	0.010	0.011	-0.001	0.011	0.011	0.016
Dentists per 10,000	-0.020	0.014	0.068^***	0.016	-0.088^***	0.021
Chiropractors per 10,000	0.076^**	0.027	-0.189^***	0.028	0.266^***	0.039
Year
2011	0.155^**	0.050	0.312^***	0.040	-0.156^*	0.064
2012	0.538^***	0.048	0.677^***	0.040	-0.139^*	0.062
2013	0.845^***	0.049	0.980^***	0.041	-0.135^*	0.064
2014	1.176^***	0.052	1.259^***	0.046	-0.082	0.069
2015	1.420^***	0.057	1.492^***	0.052	-0.072	0.077
2016	1.527^***	0.063	1.634^***	0.058	-0.107	0.086
2017	1.531^***	0.068	1.681^***	0.065	-0.151	0.094
2018	1.469^***	0.075	1.673^***	0.071	-0.204^*	0.103
Level of significance: ⁺ p < 0.1, ^* p < 0.05, ^ p < 0.01, ^* p < 0.001

Table 4

Two-way fixed effects Poisson regression models for synthetic opioid-involved mortality rates based on samples of becoming-more-rural (N = 1,192) and becoming-more-urban (N = 1,465) counties
	Model 7 (becoming-more-rural)		Model 8 (becoming-more-urban)		Model 9 (comparison)
	Coef.	S.E.	Coef.	S.E.	Coef.	S.E.
IRR	-0.461^***	0.050	-0.155^***	0.019	-0.306^***	0.053
Percent males	-0.040^***	0.009	-0.033^**	0.011	-0.007	0.014
Percent non-Hispanic blacks	-0.009	0.009	-0.040^***	0.009	0.031^*	0.013
Percent Hispanics	-0.026^***	0.007	0.000	0.010	-0.026^*	0.013
Percent high school and above	0.002	0.004	-0.001	0.005	0.003	0.006
Percent manual labor occupations	0.009^***	0.003	0.010^**	0.003	-0.001	0.004
Percent separated or divorced	-0.016^***	0.004	-0.006	0.005	-0.010	0.006
Percent uninsured	0.020^***	0.002	0.013^***	0.003	0.007^*	0.003
Percent veterans	0.009	0.006	0.046^***	0.007	-0.036^***	0.009
Percent on public assistance	0.011	0.006	0.054^***	0.007	-0.043^***	0.010
Poverty rate	0.009^**	0.003	0.027^***	0.004	-0.018^***	0.005
Unemployment rate	-0.059^***	0.003	-0.037^***	0.004	-0.022^***	0.005
Social capital index	0.225^***	0.050	0.196^***	0.043	0.029	0.067
Prescribing rate	-0.001^*	0.000	0.000	0.000	-0.001	0.001
Medical doctors per 10,000	-0.001	0.003	0.026^***	0.004	-0.028^***	0.005
Nurse practitioners per 10,000	0.009^*	0.004	0.042^***	0.005	-0.034^***	0.006
Physician assistants per 10,000	0.068^***	0.007	0.071^***	0.009	-0.003	0.011
Dentists per 10,000	-0.011	0.010	-0.011	0.012	0.000	0.016
Chiropractors per 10,000	-0.065^***	0.016	0.061^**	0.021	-0.126^***	0.026
Year
2011	0.053	0.028	-0.036	0.026	0.089^*	0.038
2012	0.118^***	0.029	-0.083^**	0.027	0.201^***	0.040
2013	0.245^***	0.030	-0.014	0.029	0.259^***	0.042
2014	0.460^***	0.032	0.193^***	0.032	0.267^***	0.045
2015	0.778^***	0.035	0.528^***	0.035	0.249^***	0.049
2016	1.123^***	0.038	0.873^***	0.039	0.250^***	0.055
2017	1.318^***	0.042	1.065^***	0.043	0.252^***	0.060
2018	1.393^***	0.046	1.125^***	0.048	0.268^***	0.066
Level of significance: ⁺ p < 0.1, ^* p < 0.05, ^ p < 0.01, ^* p < 0.001

No competing interests reported.

Download PDF

Version 1

posted

You are reading this latest preprint version

Opioid crisis in rural and urban counties from 2010 to 2018: in light of the index of relative rurality

Status:

Version 1

Abstract

Figures

1. Introduction

2. Data And Methods

3. Results

4. Conclusion And Discussion

Declarations

References

Tables

Additional Declarations

Status:

Version 1