Effects of County-Level Opioid Dispensing Rates on Individual-Level Patterns of Prescription Opioid and Heroin Consumption: Evidence from National U.S. Data

Abstract

Objective.

We examine directly whether county-level changes in opioid dispensing rates affect individual-level prescription opioid misuse, frequency of use, and dependence, as well as the same outcomes for heroin.

Methods.

Using data from the restricted-access National Survey on Drug Use and Health, CDC retail opioid prescriptions, Prescription Drug Abuse Policy System, and U.S. Census, we applied fixed-effects models to determine whether county-level dispensing rates affected prescription opioid outcomes as intended and whether changes in rates adversely affected heroin use outcomes. We used Bayes Factors (BFs) to confirm evidence for null findings.

Results.

The sample includes 748,800 respondents aged 12 and older from 2006 to 2016. The odds of prescription opioid misuse, frequency of misuse, and dependence were 7.2% (p<.001), 3.5% (p<.01), and 10.4% (p<.001) higher, respectively, per standard deviation increase in the county-level opioid dispensing rate per 100. We found no evidence for any association between opioid dispensing rates and the three heroin outcomes. The odds ratio was non-significant according to frequentist techniques in our fixed-effects models, and Bayesian techniques confirmed very strong support for the null hypothesis.

Conclusions.

County-level opioid dispensing rates are directly associated with individual-level prescription opioid misuse, frequency of misuse, and dependence. Changes in dispensing were not associated with population shifts in heroin use. Reductions in opioid dispensing rates have contributed to stemming prior increases in prescription opioid misuse while not adversely affecting heroin use. Physicians and other healthcare providers can take action to minimize opioid dispensing for tangible benefits regarding prescription opioid misuse without adversely affecting heroin use.

Introduction

During the first decade of the 21^st century, the United States experienced a dramatic increase in the number of controlled substances prescribed and dispensed, particularly opioids. The number of opioid prescriptions dispensed from retail pharmacies increased from 174.1 million in 2000 to 256.9 million in 2009.^1–2 Over the same period, prescription opioid misuse increased dramatically,^2–7 along with a simultaneous rise in overdose rates due to these substances.⁸ As dispensing rates began to decrease in the early 2010s,⁹ fatal overdose rates due to prescription opioids plateaued. As this plateau began, heroin overdoses rose from 2010–2016 before also plateauing, followed by increases in overdoses from illicit synthetic opioids such as fentanyl.⁸ These overdose trends suggest that dispensing trends impacted prescription opioid misuse, and potentially the rise of heroin and fentanyl use. Yet, little research directly links local-level dispensing rates to individual-level use and dependence using nationally representative data, despite connections between dispensing and misuse/dependence patterns as an often hypothesized pathway to overdose trends. In this article, we examine directly whether county-level dispensing rates affected individual-level prescription opioid and heroin misuse, frequency, and dependence in a nationally representative dataset.

A body of research demonstrates a relationship between opioid dispensing patterns and adverse outcomes from opioid consumption. According to administrative and clinical records, for individuals with valid prescriptions for pain, patients with higher prescribed doses and increased days of supply are more likely to develop an opioid use disorder¹⁰ or die of an overdose.^11–14 However, studies show that a substantial proportion of prescription opioid overdose deaths are associated with nonmedical use of opioids obtained via diversion,^15–16 making evidence from patient behaviors less well-extendable to non-medical use. In the aggregate, increases and decreases in opioid dispensing are associated with concomitant deaths from prescription opioids.^17–19 Although few national studies directly consider the relationship between dispensing and prescription opioid misuse, we can infer from studies on Prescription Drug Monitoring Programs (PDMPs) as an indirect measure of opioid dispensing, as these systems of surveillance are intended to reduce unnecessary prescribing and dispensing. The evidence on whether PDMPs reduce prescriptions or reduce overdoses is mixed.^20–25 A study of use patterns using this measure found little evidence that PDMPs affect nonmedical opioid use and dependence.²⁶ Studies with broader geographic scope have tended to use state-level measures, despite evidence that the quantity of prescriptions dispensed and acute outcomes such as overdose are more localized.^22,27 For example, it is not clear that high rates of dispensing in the western counties of New York would significantly impact pharmaceutical opioid consumption in relatively low dispensing counties of Queens and Brooklyn several hundred miles away, particularly since opioid diversion commonly occurs with medication diverted from friends and family.²⁸ In sum, past studies tend to focus on the effect of dispensing on pain patient outcomes, aggregate patterns of overdoses, or indirect measures such as state-level PDMPs and also obscure trends occurring at a more local level. Despite considerable attention given to the opioid crisis, analyses that build upon this prior work to examine the nuances of the relationship between opioid dispensing and opioid misuse and dependence remain much needed. Below, we consider the alternative hypothesis that dispensing rates are associated with changes in prescription opioid misuse outcomes; as we expect an increase in misuse in harms when the dispensing rate was increasing and a decrease as it declined, this represents a two-sided hypothesis and test against a null of no association.

An additional concern about decreasing the number of available prescriptions is that reductions in the circulation of prescription opioids could push those with opioid dependence to illicit opioids such as heroin. While there is some evidence that many heroin users transitioned from prescription opioids,^29–33 the relationship between this transition and the availability of prescription opioids has not been thoroughly explored using national data. In fact, the CDC’s Principal Deputy Director and colleagues have argued that there is no evidence that prescribing practices have resulted in increased heroin use.¹ However, studies making this claim often rely on indirect evidence of the effect of dispensing rates on overdoses¹⁷ or a measure of dispensing such as PDMPs.²⁶ As Compton et al. describe, “Some persons certainly use heroin when they are unable to obtain their preferred prescription opioid; however, whether the increases in heroin trends in the overall population are driven by changes in policies and practices regarding prescription opioids is much less clear.”^33:155 Again, the central piece of the argument, namely whether changes in dispensing rates affected patterns of heroin use, remains uncertain. On one hand, reductions in dispensing may have had a direct effect on transitions to heroin use. In such circumstances, we would anticipate seeing growth in heroin use in locations that reduce opioid dispensing. On the other hand, the growth in heroin use may be an ancillary component of the wider opioid crisis, but not directly attributable to patterns of opioid dispensing. In examining heroin use outcomes then, our alternative hypothesis is that decreases in county-level opioid dispensing are associated with increases in heroin use outcomes, while the null is that of no relationship.

Relying upon national data from the CDC and SAMHSA, this paper builds upon prior research to examine directly the relationship between opioid dispensing rates at the county level with individual-level pharmaceutical opioid misuse and heroin use as well as dependence. The findings may have implications not only for prevention and intervention, but for clinical practice as well.

Methods

Individual-Level Data

We utilized National Survey on Drug Use and Health (NSDUH) data, a national survey of substance use conducted annually by the Substance Abuse and Mental Health Services Administration (SAMHSA) since 1971. Using a stratified, multistage area probability sampling design to produce representative national data, approximately 70,000 individuals age 12 and older are surveyed annually. Via a U.S. Census Bureau Research Data Center, we utilized restricted NSDUH data, which contains state and county FIPS codes, from 2006 to 2016 to examine individual-level use, frequency of use, and dependence for prescription opioids and heroin as dispensing rates shift. The included years represent those where all dispensing data and covariates were available.

We consider measures of use, frequency of use, and dependence for prescription opioid misuse and heroin. First, we examine a binary indicator (created by the NSDUH) of any prescription opioid misuse or heroin use in the past year. Second, for both substances, we use an ordinal measure of past year frequency of use given the sparsity of data at specific counts that preclude using count models. The categories coded from the number of days are: none, infrequent (1–11 days), intermittent (12–59 days), regular but non-daily (60–300 days), and daily (>300 days). As the measure of prescription opioid frequency changed in 2015 from a past year to past month reference period, the model specific to prescription opioid frequency does not include 2015 and 2016 data as the other models do. Finally, we examine a binary indicator of dependence for both substances, which the NSDUH defines by DSM-IV criteria.

Within the NSDUH, we included appropriate individual-level covariates within the models described below. Specifically, we accounted for birth year (including a squared term to account for nonlinearity), race/ethnicity, gender, income, educational attainment, health insurance status, employment status, marital status, and U.S. nativity. Descriptive statistics for all variables are shown in Table 1, which also provides covariate category coding.

Table 1:

Weighted Descriptive Statistics (N = 748,800)

	Mean or %	SD
Prescription Opioid Outcomes
Past year prescription opioid misuse	4.62%
Past year frequency of prescription opioid misuse
None	95.59%
Infrequent (1–11 days)	2.30%
Intermittent (12–59 days)	1.25%
Regular but non-daily (60–300 days)	0.52%
Almost daily (>300 days)	0.34%
Prescription opioid dependence	0.54%
Heroin Outcomes
Past year heroin use
Past year frequency of heroin use	99.89%
None	0.07%
Infrequent (1–11 days)	0.02%
Intermittent (12–59 days)	0.01%
Regular but non-daily (60–300 days)	0.01%
Almost daily (>300 days)	0.26%
Heroin dependence	0.15%
County-Level Opioid Dispensing Rate per 100	78.37	34.55
State-Level Policy Covariates
Prescription Drug Monitoring Program	86.88%
Good Samaritan Policy	22.67%
Any Naloxone Access Expansion	35.78%
Pain Management Clinic laws	15.41%
Medical Marijuana Laws	33.77%
County-Level Covariates
Percent Hispanic	15.37	15.30
Percent Black	12.74	12.63
Percent Unemployed	5.14	1.43
Percent Living in Poverty	10.52	4.59
Percent over 25 with a Bachelor’s	20.69	8.72
Percent Foreign Born	12.55	10.76
Percent Female Headed Households	7.24	2.04
Household Income (thousands)	54.48	14.50
Metropolitan Area Types
Large central	29.84%
Large fringe	25.63%
Medium	20.62%
Small	9.29%
Micropolitan	8.70%
Noncore	5.92%
Individual-Level Covariates
Female	51.51%
Year of Birth	1967.27	19.52
Health Insurance	86.34%
U.S. Born	84.94%
Race/Ethnicity
White	65.94%
Black	11.88%
Native American	0.52%
Hawaiian Islander/Pacific Islander	0.35%
Asian	4.80%
Multi-racial	1.44%
Hispanic	15.07%
Income Bracket
<20k	18.10%
20k-49.9k	32.01%
50k-74.9k	17.08%
75k-99.9k	12.13%
>100k	20.68%
Educational Attainment
Less than high school	13.29%
High School degree	26.37%
Some College	24.37%
BA/BS+	26.29%
12–17 year olds	9.68%
Employment Status
Full-time	46.20%
Part-time	12.36%
Unemployed	4.42%
Other/Not in labor force	27.33%
12–17 year olds	9.68%
Marital Status
Married	48.17%
Widowed	5.49%
Divorced/Separated	12.43%
Never married	33.91%

Dispensing Data

We utilized county-level opioid dispensing data from the Centers for Disease Control and Prevention (CDC), including years 2006 through 2016. Our key variable is retail opioid prescriptions dispensed for 100 people per county (CDC, 2020), which does not include administration of opioids in hospitals or treatment settings. While there are a total of 3,149 unique U.S. counties as of 2016, the number of counties in our analysis range from 2,637 in 2012 to 2,851 in 2015, with the N in other years falling in-between. Missing counties primarily result from incomplete dispensing data within the CDC database, which contains data for 87.6% to 94.0% of counties in any given year. As per the CDC,³⁴ missing data can indicate that “the county had no retail pharmacies, the county had no retail pharmacies sampled, or the prescription volume was erroneously attributed to an adjacent, more populous county according to the sampling rules used.”

Policy Data

To determine the effect of county-level dispensing rates independent of the effect of opioid-related policies placed into effect during the period of observation, we used the Prescription Drug Abuse Policy System for a comprehensive listing of policy passage in each state. These measures included state-level variables for: any prescription drug monitoring program (PDMP), any expanded naloxone access to the public, Good Samaritan laws absolving criminal or civil liability in report of an overdose, any pain clinic prescribing restrictions, and presence of medical marijuana laws. We note that medical marijuana laws - any laws that permit the use of cannabis for medical purposes - were not passed because of the opioid crisis but remain an important policy covariate pertinent to pain management and general patterns of substance use.³⁵

County- and State-level Covariates

The U.S. Census Bureau’s American Community Survey (ACS) and Decennial Censuses provided county-level time-varying covariates, specifically the unemployment rate, median household income, and percentages of households in poverty, foreign born, female-headed households, Non-Hispanic Black, Hispanic, and over 25 with a Bachelor’s. We also accounted for urbanity/rurality using the NCHS Urban-Rural Classification Scheme. We used the 5-year ACS estimates because only larger counties are available with shorter estimates. As these begin in 2009, we linearly interpolated between the 2000 census and 2009, utilizing the interpolated values for 2006 to 2008.

Analysis

We used logistic regression models with fixed-effects for state and year to determine the effect of dispensing rates.³⁶ The strength of including fixed-effects is the elimination of unobserved heterogeneity. Fixed-effects estimators are robust to any observed or unobserved time-invariant omitted variables, which removes any constant state-level effects.³⁶ As such, we included a fixed-effect for state to account for unobserved differences across locales over time. We also included a standard error cluster-correction for county to account for dependencies between individuals within counties. Finally, to provide estimates of the effect of dispensing rates independent of changes over time, all models included year fixed-effects. All analyses included statistical weights to account for the stratified, multistage probability sampling design. Since we use logistic regression, all model results are presented as odds ratios with 95% confidence intervals. Given the nature of the variable, we use ordinal logistic regression to model frequency of use. The Variance Inflation Factor indicated no issues with collinearity.

To provide additional evidence for findings, we also present Bayes Factors (BF).³⁸ In particular, BFs are useful for providing evidence for null hypotheses, given that frequentist techniques (i.e. the parameter tests in the fixed-effects models) can only conclude that the null cannot be rejected, rather than providing evidence for the null. A BF of less than 1/3 indicates evidence for the null hypothesis, with the following levels: 1/3–1/10=moderate; 1/10–1/30=strong; 1/30–1/100=very strong; <1/100=extreme.³⁷ BFs also support the alternative hypothesis when the value exceeds 3. When a theoretical prior distribution is unknown, the BF can be computed from a uniform distribution with a plausible maximum.³⁸ We take this approach here by using a Uniform[0,1] prior distribution, but also note that our conclusions were identical using Normal(0,1).

Across the years of observation, the entire NSDUH sample considered has about 748,800 respondents. All analytic sample sizes, in accordance with the restricted data agreement, are rounded to the nearest hundred.

Results

Table 1 displays descriptive statistics. For the outcomes, 4.6% of the sample in all years reported prescription opioid misuse. Regarding frequency, 95.6% reported no misuse, 2.3% infrequent misuse, 1.3% intermittent misuse, 0.5% regular but non-daily misuse, and 0.3% reported daily misuse. Across all years, 0.5% of respondents reported dependence on prescription opioids. For heroin, only 0.3% of respondents reported any use. Not surprisingly, the rates for each of the frequency categories is below 0.1%, and only 0.2% reported dependence on heroin. While several of these percentages are quite low, the substantial size of the NSDUH ensures that there are still no small cell sizes. For our main predictor of interest, the average county-level opioid dispensing rate across all respondents and years was 78.4 per 100. Temporally, average county-level opioid dispensing rates rose from 80.5 per 100 in 2006 to a high of 96.1 per 100 in 2012, at which point rates began to steadily decline.

Table 2 shows the results of our logistic regression models. Models 1 through 3 show the outcomes for prescription opioid misuse. The county-level dispensing rate is significant and positively associated with all three outcomes. From Model 1, a one-unit increase in the county-level opioid dispensing rate per 100 is associated with 0.2% increased odds of individual-level past year prescription opioid misuse (p<.001), net of opioid-related policies, county-level sociodemographic measures, individual-level correlates, and state and year fixed effects. While this magnitude appears low, it only reflects the magnitude of a one-unit increase. For example, we can consider an increase of a standard deviation in the dispensing rate (SD=34.6; see Table 1). A one SD increase in the dispensing rate is associated with 7.2% increased odds of past year prescription opioid misuse. Model 2 considers frequency of prescription opioid misuse. A one-unit increase in the county-level opioid dispensing rate is associated with 0.1% increased odds of being in a higher frequency category (p<.01). For a SD increase in the dispensing rate, the associated odds are 3.5% higher of being in a higher frequency category. Finally, according to Model 3, a one-unit increase in the county-level opioid dispensing rate is associated with 0.3% increased odds of individual-level prescription opioid dependence (p<.001). For a SD increase in the dispensing rate, the associated odds of dependence are 10.4% higher. While we interpret the increase given this expected positive relationship, we may also consider a SD decrease given that dispensing rates began to decline. In this case, a SD decrease in dispensing rates are associated with a 6.7% decreased odds of past year prescription opioid misuse, 3.4% decreased odds of being in a higher frequency category, and 9.6% decreased odds of dependence. We also note that the dispensing effects presented are nearly identical in models without the policy variables (not presented here). Further, although we have strong support using a frequentist approach, we also note that BFs for the effect of dispensing on each of these three outcomes indicate support for the alternative.

Table 2:

Fixed-Effects Results for Effect of County-Level Opioid Dispensing Rates on Individual-Level Prescription Opioid and Heroin Outcomes in the NSDUH from 2006–2016, Odds Ratios and 95% Confidence Intervals

	Prescription opioids			Heroin
	Model 1: Past year misuse	Model 2: Frequency of past year misuse	Model 3: Dependence	Model 4: Past year use	Model 5: Frequency of past year use	Model 6: Dependence
County-level opioid dispensing rate	1.002*** (1.001,1.002)	1.001** (1.001,1.002)	1.003*** (1.001,1.004)	0.999 (0.996,1.001)	0.998 (0.994,1.002)	0.999 (0.996,1.003)

Note:

^**p < .01;

^***p<.001.

Exponentiated coefficients from logistic regression displayed (binary for past year misuse/use and dependence; ordinal for frequency); 95% confidence intervals in parentheses. All models include state and year fixed-effects and cluster-corrected standard errors for county, and utilize sampling weights. Models include all covariates shown in Table 1 (see Appendix Table A1 for full models). Regression coefficients represent the effect of a one-unit increase in county-level opioid dispensing rate; see text for magnitude of a standard deviation increase. Bayes Factors indicate very strong support for the null hypothesis for the heroin coefficients.

Models 4 through 6 in Table 2 show the same three outcomes for heroin. There are no significant associations between the county-level opioid dispensing rate and these individual-level heroin outcomes. Using the alternative hypothesis that decreases in opioid dispensing increases the heroin outcomes, BFs from the three models are 0.0050, 0.0084 and 0.0052, respectively, demonstrating very strong support for the null hypothesis.

Discussion

The substantial rise and subsequent decrease in prescription opioid misuse and dependence is often attributed to changes in opioid dispensing practices.^1–9 This paper builds upon past studies to move beyond indirect outcomes such as overdose^17–19 and predictors such as PDMPs,^20–26 geographically broad measures of dispensing, and administrative and clinical data on patient populations.^10–16 In using a direct measure of county-level opioid dispensing linked to a nationally representative individual-level dataset in the NSDUH, we find evidence of an association between higher levels of opioid dispensing and increased odds of past year prescription opioid misuse, past year frequency of use, and dependence. Thus, increases in dispensing at the local level enabled growth in misuse early on, while efforts to curb opioid prescriptions during more recent years appear had an effect on reducing prescription opioid misuse and dependence. Thus, results support the conclusion that efforts to reduce opioid prescriptions have directly impacted non-medical use of pharmaceutical opioids.

Beyond the linkages of dispensing patterns to prescription opioid misuse, we find no evidence that shifts in local-level opioid dispensing affected odds of heroin use, frequency of heroin use, or heroin dependence, with the frequentist fixed-effects models confirmed by Bayesian techniques. This suggests that trends in heroin use may be an ancillary component of the opioid crisis rather than directly attributable to patterns of opioid dispensing. Given that many heroin users have transitioned from prescription opioid misuse,^29–33 there has been a reasonable fear that well-meaning attempts to curb opioid prescriptions could result in additional opioid misusers making this transition. However, we found no evidence that this occurred at the population level, such that efforts should continue to reduce opioid prescriptions to levels required for patient care without fears that reductions may drive up heroin use. These results speak directly to Compton and colleagues’ contention that increases in heroin trends in the overall population may not have been driven by changes in policies and practices regarding prescription opioids.³³ These findings also cohere with research on discontinuation of opioids, as recent work shows that discontinuation without tapering was the norm for long-term opioid therapies and that changes in prescribing due to prescription drug monitoring programs do not increase abrupt discontinuation.³⁸ Nonetheless, as reductions in prescribing and dispensing continue, expansion of substance abuse treatment programs should occur and remain available for those who misuse prescription opioids to prevent transitions to heroin (or synthetic opioids) that have been noted in the literature.^29–33

Notably, since an effect of dispensing on prescription opioid misuse remains after accounting for opioid-related policies, there are likely other mechanisms operating in the links between dispensing and opioid misuse beyond state policy implementation. We contend that these effects are attributable to more general shifts within the institution of medicine beyond those defined by public policy. General attention to the opioid and overdose crisis may have made providers more cognizant of their own prescribing behaviors.¹ For example, physicians may alter their own prescribing behavior after becoming aware of overdoses within the community.³⁹ Moreover, as the crisis became apparent, emphasis for physicians to consider nonopioid pain management alternatives increased.⁴⁰ Thus, although other research has shown that policies such as prescription drug monitoring programs and pain clinic regulations reduce opioid dispensing,^17,26 our results suggest that the effect of changes in dispensing on patterns of opioid misuse are not solely attributable to such policies and suggest broader changes in prescriber behavior.

Limitations

Although this study has numerous strengths, we must consider some limitations. First, counties are imperfect measures of geographic space, and measuring dispensing rates at this level can obfuscate important within-county level differentiation. The size and number of counties also vary considerably across states. However, they are a substantial improvement over measuring dispensing at the state level given the local nature of opioid prescription dispensing. Second, while we included a large battery of individual-, county-, and state-level controls as well as state and year fixed-effects, we recognize other factors may affect patterns of use and dependence. The use of fixed-effects provides results robust to observable and unobservable static state-level factors, increasing confidence in our results. However, unmeasured time-varying factors, including those related to survey methodology, remain a possible source of confounding given the observational nature of the data. Third, with repeated cross-sectional data, we cannot examine within-person change. National population geocoded individual longitudinal data on prescription opioid misuse and heroin, particularly of a sample size that allows the examination of these uncommon outcomes, remains unavailable. Lastly, we recognize that changes in opioid dispensing may also affect illicit fentanyl use, however, we cannot make any conjectures about fentanyl or fentanyl-adulterated heroin. NSDUH only recently began collecting this information, which limits examinations of fentanyl’s relationship with opioid dispensing. The results for heroin provide some promise that such a shift for fentanyl might not accompany dispensing changes, but this remains an open question.

Conclusions

Our findings indicate that county-level rates of opioid dispensing had a direct effect on individual-level opioid misuse and dependence, but reductions in dispensing did not have any adverse effects on heroin use. Institutionally driven changes among prescribers, potentially shaped by both professional recognitions of the problem as well as policy implementation, may have helped to curb the prescription opioid crisis; however, these changes do not appear to have altered heroin use (in either direction) following shifts in dispensing at the county-level. We recommend that medical providers continue to monitor patterns of prescribing and dispensing, as well as states continue to pursue policies that temper unnecessary opioid prescriptions.

Appendix

Appendix Table A1:

Fixed-Effects Models for Individual-Level Prescription Opioid and Heroin Outcomes in the NSDUH from 2006–2016, Odds Ratios and 95% Confidence Intervals

	Prescription opioids			Heroin
	Model 1: Past year misuse	Model 2: Frequency of past year misuse	Model 3: Dependence	Model 4: Past year use	Model 5: Frequency of past year use	Model 6: Dependence
County-level Opioid Dispensing Rate per 100	1.002*** (1.001,1.002)	1.001** (1.001,1.002)	1.003*** (1.001,1.004)	0.999 (0.996,1.001)	0.998 (0.994,1.002)	0.999 (0.996,1.003)
State-Level Policy Covariates
Prescription Drug Monitoring Program	1.037 (0.962,1.118)	1.058 (0.974,1.150)	0.949 (0.764,1.178)	1.019 (0.746,1.393)	1.257 (0.784,2.015)	1.055 (0.678,1.641)
Good Samaritan Policy	1.016 (0.956,1.081)	1.005 (0.928,1.089)	1.009 (0.843,1.207)	0.746* (0.596,0.935)	0.925 (0.657,1.303)	0.671* (0.492,0.916)
Any Naloxone Access Expansion	0.961 (0.898,1.029)	0.952 (0.876,1.035)	0.843 (0.704,1.010)	1.158 (0.901,1.488)	0.955 (0.625,1.461)	1.224 (0.849,1.766)
Pain Management Clinic laws	0.998 (0.931,1.071)	0.990 (0.909,1.079)	0.938 (0.778,1.131)	1.176 (0.878,1.576)	1.324 (0.839,2.087)	1.244 (0.806,1.920)
Medical Marijuana Laws	1.020 (0.949,1.097)	1.011 (0.924,1.107)	0.998 (0.804,1.238)	1.102 (0.839,1.448)	1.014 (0.681,1.511)	1.015 (0.745,1.384)
County-Level Covariates
Percent Hispanic	0.997 (0.994,1.000)	0.997 (0.993,1.000)	1.000 (0.992,1.008)	1.014 (0.995,1.034)	1.016 (0.994,1.039)	1.009 (0.986,1.032)
Percent Black	1.002 (0.998,1.005)	1.001 (0.996,1.005)	1.006 (0.996,1.016)	1.006 (0.990,1.023)	0.991 (0.972,1.011)	1.002 (0.985,1.019)
Percent Unemployed	1.045*** (1.021,1.070)	1.060*** (1.030,1.091)	1.029 (0.969,1.092)	1.086 (0.990,1.191)	1.216** (1.079,1.370)	1.079 (0.966,1.206)
Percent Living in Poverty	0.993 (0.980,1.006)	0.997 (0.983,1.011)	1.003 (0.979,1.027)	0.947* (0.906,0.990)	0.932* (0.876,0.991)	0.945* (0.900,0.992)
Percent over 25 with a Bachelor’s	1.009*** (1.004,1.014)	1.009*** (1.004,1.015)	0.991 (0.981,1.002)	1.013 (0.997,1.030)	1.025* (1.003,1.047)	1.018 (0.998,1.039)
Percent Foreign Born	0.999 (0.994,1.003)	0.997 (0.992,1.002)	1.003 (0.993,1.014)	0.986 (0.963,1.009)	0.983 (0.954,1.013)	0.998 (0.974,1.022)
Percent Female Headed Households	1.001 (0.977,1.025)	1.000 (0.974,1.028)	0.989 (0.933,1.049)	1.028 (0.945,1.119)	1.087 (0.961,1.229)	1.079 (0.980,1.188)
Household Income (thousands)	0.999 (0.995,1.003)	1.000 (0.996,1.005)	1.008* (1.000,1.017)	0.995 (0.980,1.009)	0.997 (0.975,1.018)	0.990 (0.974,1.005)
Metropolitan Area (vs. Large central)
Large central	0.953 (0.882,1.031)	0.944 (0.862,1.034)	0.961 (0.806,1.147)	0.780 (0.580,1.048)	0.927 (0.605,1.419)	0.899 (0.645,1.253)
Medium	0.955 (0.889,1.025)	0.959 (0.883,1.043)	0.960 (0.806,1.142)	0.821 (0.631,1.067)	0.974 (0.690,1.374)	0.684* (0.497,0.941)
Small	0.927 (0.851,1.010)	0.925 (0.837,1.021)	0.945 (0.776,1.152)	0.754 (0.519,1.095)	0.900 (0.565,1.433)	0.818 (0.498,1.343)
Micropolitan	0.869** (0.792,0.955)	0.862** (0.773,0.961)	0.849 (0.686,1.052)	0.636** (0.451,0.897)	0.670 (0.438,1.026)	0.557** (0.358,0.864)
Noncore	0.846** (0.755,0.949)	0.815** (0.713,0.931)	0.821 (0.604,1.117)	0.686 (0.441,1.067)	1.233 (0.705,2.156)	0.572* (0.330,0.990)
Individual-Level Covariates
Female (vs. Male)	0.815*** (0.788,0.843)	0.814*** (0.782,0.847)	0.749*** (0.680,0.825)	0.509*** (0.447,0.579)	0.478*** (0.390,0.586)	0.530*** (0.442,0.636)
Health Insurance	0.774*** (0.737,0.813)	0.749*** (0.711,0.789)	0.696*** (0.625,0.776)	0.578*** (0.499,0.669)	0.532*** (0.418,0.677)	0.543*** (0.448,0.658)
U.S. Born	1.717*** (1.597,1.846)	1.889*** (1.736,2.055)	3.434*** (2.659,4.435)	3.337*** (2.045,5.443)	2.514* (1.118,5.651)	3.860*** (1.971,7.562)
Race/Ethnicity (vs. White)
Black	0.542*** (0.503,0.583)	0.495*** (0.454,0.540)	0.367*** (0.291,0.462)	0.418*** (0.305,0.572)	0.285*** (0.172,0.472)	0.334*** (0.220,0.506)
Native American	0.996 (0.845,1.174)	1.037 (0.854,1.258)	1.123 (0.711,1.773)	0.756 (0.419,1.366)	0.667 (0.338,1.315)	0.611 (0.280,1.337)
Hawaiian Islander/Pacific Islander	0.777 (0.557,1.083)	0.767 (0.503,1.171)	0.750 (0.335,1.679)	0.757 (0.346,1.656)	0.690 (0.143,3.329)	0.609 (0.142,2.616)
Asian	0.505*** (0.449,0.567)	0.496*** (0.434,0.567)	0.422*** (0.283,0.627)	0.136*** (0.063,0.290)	0.206*** (0.087,0.489)	0.033*** (0.007,0.154)
Multi-racial	1.000 (0.898,1.113)	0.923 (0.804,1.058)	0.877 (0.698,1.101)	0.725 (0.465,1.129)	0.843 (0.484,1.467)	0.706 (0.379,1.312)
Hispanic	0.693*** (0.656,0.733)	0.654*** (0.611,0.700)	0.495*** (0.401,0.610)	0.615** (0.450,0.840)	0.443** (0.261,0.750)	0.654* (0.467,0.916)
Income bracket (vs. <20k)
20k-49.9k	0.909*** (0.874,0.946)	0.917*** (0.874,0.963)	0.904 (0.808,1.011)	0.732*** (0.621,0.863)	0.820 (0.613,1.097)	0.647*** (0.509,0.822)
50k-74.9k	0.823*** (0.784,0.864)	0.842*** (0.795,0.891)	0.768*** (0.668,0.882)	0.605*** (0.470,0.779)	0.541*** (0.383,0.762)	0.594*** (0.436,0.808)
75k-99.9k	0.759*** (0.712,0.808)	0.797*** (0.742,0.856)	0.704*** (0.591,0.840)	0.574*** (0.438,0.751)	0.540** (0.368,0.792)	0.562** (0.398,0.793)
>100k	0.757*** (0.716,0.800)	0.764*** (0.715,0.817)	0.795** (0.680,0.930)	0.729* (0.565,0.942)	0.712 (0.495,1.025)	0.653** (0.479,0.890)
Educational Attainment (vs. < High School)
High School degree	0.823*** (0.781,0.868)	0.823*** (0.776,0.873)	0.778*** (0.682,0.888)	0.693*** (0.584,0.824)	0.702* (0.534,0.924)	0.739** (0.595,0.917)
Some College	0.856*** (0.810,0.905)	0.866*** (0.813,0.921)	0.720*** (0.630,0.822)	0.676*** (0.556,0.823)	0.757* (0.575,0.997)	0.763* (0.595,0.978)
BA/BS+	0.609*** (0.569,0.651)	0.609*** (0.564,0.658)	0.322*** (0.266,0.390)	0.193*** (0.126,0.294)	0.274*** (0.157,0.479)	0.182*** (0.112,0.296)
12–17 year-olds	0.522*** (0.485,0.560)	0.464*** (0.428,0.504)	0.506*** (0.418,0.612)	0.298*** (0.219,0.408)	0.258*** (0.172,0.387)	0.154*** (0.091,0.258)
Employment Status (vs. Full-time)
Part-time	1.021 (0.971,1.074)	0.996 (0.935,1.061)	1.136 (0.977,1.321)	1.380** (1.111,1.714)	1.354 (0.999,1.835)	1.180 (0.886,1.573)
Unemployed	1.297*** (1.215,1.384)	1.259*** (1.161,1.365)	1.882*** (1.634,2.168)	2.891*** (2.330,3.589)	2.046*** (1.527,2.741)	2.948*** (2.322,3.743)
Other/Not in labor force	1.044 (0.987,1.104)	1.007 (0.942,1.076)	1.735*** (1.510,1.993)	2117*** (1.703,2.631)	1.763*** (1.281,2.426)	2.321*** (1.764,3.054)
Marital status (vs. Married)
Widowed	1.630*** (1.396,1.903)	1.469*** (1.173,1.839)	2.554*** (1.760,3.706)	4.833*** (2.856,8.177)	3.040* (1.216,7.600)	3.583*** (1.689,7.603)
Divorced/Separated	1.523*** (1.426,1.626)	1.529*** (1.416,1.650)	1.986*** (1.666,2.367)	3.132*** (2.282,4.297)	3.664*** (2.282,5.882)	2.850*** (1.916,4.238)
Never married	1.706*** (1.619,1.797)	1.728*** (1.626,1.836)	2.031*** (1.745,2.364)	4.211*** (3.170,5.593)	4.390*** (2.960,6.511)	4.544*** (3.067,6.731)

Note:

^*p < .05;

^**p < .01;

^***p<.001.

Exponentiated coefficients from logistic regression displayed (binary for past year misuse/use and dependence; ordinal for frequency); 95% confidence intervals in parentheses. All models include state and year fixed-effects, cluster-corrected standard errors for county, cohort effects measured by birth year (squared), and utilize sampling weights.