Data are from treatment facilities in the 50 US states and the District of Columbia. OTP indicates opioid treatment program.
A, Population-weighted mean driving time in each county from census tract mean centers of populations (MCPs) to the closest OTP. B, Population-weighted mean driving time in each county from census tract MCPs to the closest pharmacy. C, Population-weighted mean difference in 1-way driving times to the closest OTP and pharmacy from census tract MCPs in each county. Population-weighted mean driving times were longer to OTPs than to pharmacy in all counties. Gray indicates that a county had no driving routes from census tract MCPs included in the analysis.
The cumulative proportion of US residents living in each county type with a driving time to the closest opioid treatment program (OTP) and closest pharmacy equal to or less than a given time. County types are based on the 2013 National Center for Health Statistics Urban-Rural Classification Scheme.
eMethods. Supplementary Methods
eTable 1. Characteristics of United States Census Tracts
eTable 2. Driving Times in Minutes to the Closest Opioid Treatment Program and Pharmacy
eTable 3. State-Level Estimates of Population One-Way Mean Driving Time, Driving Distance, and Driving Costs to OTPs and Pharmacies
eTable 4. Driving Times in Minutes to the Closest Opioid Treatment Program and Pharmacy
eTable 5. Driving Distances in Miles (Mean and Median) to the Closest Opioid Treatment Program and Pharmacy
eTable 6. Sensitivity Analysis Without Restriction That Driving Times to Both OTPs and Pharmacies Be 12 Hours or Less
eFigure 1. Flowchart of United States Census Tracts Included in Analysis
eFigure 2. Population-Weighted Mean Driving Times by State
eFigure 3. Population-Weighted Mean Driving Distances by State and County
eFigure 4. Population-Weighted Mean One-Way Driving Costs Using Internal Revenue Service Estimate of Variable Costs of Driving in 2019
eFigure 5. Population-Weighted Mean Fuel Costs of a One-Way Using Average State Gas Price From 2017 and an Estimated Fuel Efficiency of 22.3 Miles per Gallon
County-level Estimates of Population-Weighted One-Way Mean Driving Time, Driving Distance, and Driving Costs to OTPs and Pharmacies
Customize your JAMA Network experience by selecting one or more topics from the list below.
Kleinman RA. Comparison of Driving Times to Opioid Treatment Programs and Pharmacies in the US. JAMA Psychiatry. 2020;77(11):1163–1171. doi:10.1001/jamapsychiatry.2020.1624
What are the driving times to opioid treatment programs vs pharmacies in the US?
In this cross-sectional study of driving times, the population-weighted mean 1-way driving time from census tract mean centers of population was 20.4 minutes to opioid treatment programs and 4.5 minutes to pharmacies, a statistically significant difference. Differences in driving time, distance, and cost between 1-way trips ending at opioid treatment programs and pharmacies were largest in micropolitan and noncore counties.
In this study, population-weighted mean driving times from US census tract MCPs were longer to OTPs than to pharmacies, suggesting that driving times to methadone maintenance treatment in the US may be reduced with pharmacy-based dispensing of methadone maintenance.
Methadone maintenance is an effective treatment of opioid use disorder, but federal regulations in the US restrict methadone dispensing to opioid treatment programs (OTPs). In Australia, Canada, and the UK, patients can obtain methadone maintenance from community pharmacies.
To compare driving access to methadone maintenance treatment between OTP and pharmacy dispensing models.
Design, Setting, and Participants
This descriptive cross-sectional study assessed driving times from census tract mean centers of population to OTPs and pharmacies. Census tracts from the 50 US states and the District of Columbia (based on the 2010 US Census) were included if their population was greater than 0, if their mean center of population (MCP) was within 3 miles of the road network, and if the 1-way driving times from the census tract MCP to both an OTP and a pharmacy were 12 hours or less. Data analyses were performed from November 15, 2019, to April 18, 2020.
Main Outcomes and Measures
The primary outcome was the population-weighted mean driving time from census tract MCPs to OTPs and pharmacies in the US. Census tract MCPs are population-weighted geographic centroids of residents living in each census tract. Driving times were estimated using historical average driving speeds.
All 1682 unique locations of OTPs were included, and 69 475 unique pharmacy locations were included after geocoding. A total of 72 443 census tracts were included in the analysis. The mean population-weighted driving time from census tract MCPs was 20.4 minutes (95% CI, 20.3-20.6 minutes) to OTPs and 4.5 minutes (95% CI, 4.4-4.5 minutes) to pharmacies (P < .001). Differences in driving time, distance, and cost between 1-way trips ending at OTPs and pharmacies were largest in micropolitan and noncore counties.
Conclusions and Relevance
In this study, population-weighted mean driving times from US census tract MCPs were longer to OTPs than to pharmacies.
Methadone is a US Food and Drug Administration (FDA)–approved treatment for opioid use disorder (OUD) that is associated with reduced overdose and all-cause mortality.1 In the US, methadone maintenance treatment of OUD is primarily dispensed through opioid treatment programs (OTPs).2 There are fewer than 1700 OTP locations across the US where patients can access methadone maintenance treatment,3 and many rural counties have shortages of OTPs.4 Although office-based prescribing of buprenorphine for OUD has increased in recent years, methadone remains an important component of OUD treatment, with guidelines recommending methadone maintenance for patients with unsuccessful buprenorphine treatment trials.5 More than 380 000 people in the US received methadone maintenance treatment in 2017 despite regulations restricting its use and shortages of OTPs.6
Methadone is dosed daily, and federal regulations require patients to travel to dispensing facilities up to daily for medication dispensing.2 Regional samples suggest that driving times to OTPs are longer in rural areas than urban areas,7-9 and patients with longer driving distances have shorter duration in methadone maintenance treatment.10 In Australia, Canada, and the UK, patients can obtain daily dispensed methadone at community pharmacies.11 In the US, pharmacies can dispense methadone prescribed for analgesia, but only pharmacies registered as medication units under the auspices of an OTP can dispense methadone for OUD.12,13 Given that widespread pharmacy-based dispensing of methadone maintenance is an alternative model of care that has the potential to enhance access to methadone maintenance treatment, this study compared driving access to OTPs and pharmacies in the US. The a priori hypothesis was that the population-weighted mean driving time from census tract mean centers of population (MCPs) in the US was lower to the closest pharmacy than to the closest OTP.
The primary and secondary outcomes of this cross-sectional study, along with the statistical analysis plan for these outcomes, were prespecified in a protocol that was uploaded to GitHub on November 15, 2019.14 Data analyses were performed from November 15, 2019, to April 18, 2020. The pharmacy database used for the analysis was changed on 2 occasions after posting the protocol because of database unavailability (first change) and to provide more comprehensive coverage of pharmacies (second change). The Stanford University institutional review board determined that the study was not human subjects research; therefore, informed consent was not required. This study followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guideline.
Census tract MCPs are population-weighted geographic centroids for all residents in the census tract.15 Census tracts are “relatively permanent small-area geographic divisions of a county or statistically equivalent entity.”16 Census tract populations and MCP geographic coordinates for the 50 US states and the District of Columbia were obtained from the 2010 US Census.15
The OTP addresses were obtained from Substance Abuse and Mental Health Services Administration (SAMHSA) Opioid Treatment Locator on February 3, 2020.3 The OTPs in US overseas territories were removed.
Community pharmacy addresses were obtained from a database of all active National Provider Identifiers using organizations identified as a community or retail pharmacy (January 2020 update).17 All health care practitioners and organizations in the US that are covered entities under the Health Insurance Portability and Accountability Act are required to have a National Provider Identifiers.18 Pharmacies outside the 50 states and the District of Columbia were removed.
Counties were classified based on the 2013 National Center for Health Statistics (NCHS) Urban Rural Classification Scheme, the most recent urban-rural classification scheme released by the US Centers for Disease Control and Prevention.19 Counties are classified as large central metro if they have a population of at least 250 000 and are located within metropolitan statistical areas (MSAs) with at least 1 million people. Large fringe metro counties are in MSAs with at least 1 million people but do not meet other large central metro criteria. Medium metro counties are within MSAs with 250 000 to 999 999 people. Small metro counties are in MSAs with fewer than 250 000 people. Micropolitan counties lie within micropolitan urban clusters of 10 000 to 49 999 people, and noncore counties do not meet criteria for metropolitan or micropolitan classification. Census tracts were additionally classified using the 2010 Rural-Urban Commuting Area (RUCA) codes.20 The RUCA codes classify census tracts based on population density, urbanization, and daily commuting patterns to nearby population centers.21
Methadone-dispensing locations for all OTP addresses that contained a post office box were verified (and corrected if necessary) by checking facility websites or calling facilities. Batch geocoding was performed using ArcMap software, version 10.7.1 (ESRI) and a US street address locator from StreetMap Premium 2019, North America, release 3.22,23 The following default locator settings were applied: spelling sensitivity of 80% and match threshold of 85%. The OTP addresses that were unmatched or tied after batch geocoding were verified using information from facility websites or by calling the facility if a website was unavailable or unclear. Batch geocoding was performed again using newly confirmed addresses. The remaining tied and unmatched addresses were matched during individual review (eMethods in Supplement 1). Pharmacy addresses were batch geocoded using the same parameters and software, with unmatched or tied pharmacies excluded from further analyses. All matched addresses were matched at the “StreetAddress” level of specificity, which reflects an interpolated location along a street within an address range.24
The great-circle distance from each MCP to all OTPs and pharmacies was calculated using the Haversine formula.25 From each MCP, driving time and driving distances were calculated to the 10 OTPs and pharmacies with the smallest great-circle distances, with the closest OTP and pharmacy defined as the OTP and pharmacy with the shortest driving times among these. Driving times and distances were determined using a Python software, version 2.7 script (Python Software Foundation) and ArcGIS Network Analyst (origin destination cost matrix) (ESRI) with Street Maps Premium 2019, North America, release 3.22,23 Vehicle speed was modeled based on average driving speeds from historical traffic data.26 The maximum allowed (1-way) driving time was 12 hours (to allow a return trip within 24 hours), and the search tolerance for mapping locations to the road network was 3 miles.
Driving routes from a census tract MCP were included in the analyses only if there was a driving route from both the MCP to an OTP and from the MCP to a pharmacy that was 12 hours or less. The primary outcome was the population-weighted mean driving time from census tract MCPs to OTPs and pharmacies in the US. The a priori hypothesis was that the population-weighted mean of driving times from census tract MCPs to OTPs was longer than to pharmacies. The secondary outcomes were the population-weighted mean driving times to OTPs and pharmacies for each county type. For each county type, the hypothesis was that the population-weighted mean of driving times to OTPs was longer than that to pharmacies. A sensitivity analysis without the 12-hour limit was also performed.
A weighted 1-sample, 2-tailed t test on the differences between OTP and pharmacy driving times was conducted to test the hypotheses associated with the primary and secondary outcomes. P < .05 was considered to be statistically significant for the primary outcome. A Bonferroni correction for 7 statistical tests (6 secondary outcomes plus the primary outcome) was applied to all secondary outcome assessments. The 95% CIs of the means were calculated using weighted t statistics.
A Welch analysis of variance (ANOVA) was used to determine whether driving time to treatment facilities vary by NCHS county classification, calculated separately for OTPs and pharmacies using unweighted driving times. Weighted cumulative incidence plots for the NCHS county classification were generated by plotting the cumulative proportion of individuals with a specific drive time or shorter. Driving costs were calculated by multiplying the 1-way distance of a driving route by (1) the 2019 Internal Revenue Service standard mileage rate for deducting medical expenses ($0.20 per mile) based on the variable cost of operating a vehicle27 or (2) the mean cost of motor gasoline per state in 2017 with a fuel efficiency of 22.3 miles per gallon, reflecting the mean fuel efficiency of light-duty vehicles in the US in 2017.28,29
County and state population-weighted mean driving times, distances, or costs were calculated by weighting the value from each census tract by its population for all census tracts in the region. Details about county changes since the 2010 US Census are included in the eMethods in Supplement 1. The 95% CIs were calculated using weighted t statistics.
Analyses were conducted using R software, version 3.6.1 (R Foundation for Statistical Computing).30 County and state choropleths are based on an equal area reprojection of 2018 boundary shapefiles from the US Census Bureau, with adjustments to Hawaii and Alaska for display.31,32
All 1724 OTPs in the 50 states and the District of Columbia were batch geocoded, resulting in 1682 unique OTP locations after address verification, hand review, and removal of duplicate locations (Figure 1). A total of 77 636 of 81 073 community and retail pharmacies in the US were successfully batch geocoded, resulting in 69 475 unique pharmacy locations after removal of duplicate locations (Figure 1).
There were 73 057 census tracts in the 50 states and the District of Columbia as of the 2010 US Census, of which 72 502 had at least 1 resident and an MCP within 3 miles of the road network (eFigure 1 in Supplement 1). A driving route of 12 hours or shorter was determined from 72 443 census tract MCPs to both an OTP and a pharmacy (eTable 1 in Supplement 1). A total of 36 census tracts in Hawaii and Alaska, including 0.05% of US residents, had driving routes of 12 hours or less to pharmacies but not to OTPs. No census tract MCP had a driving route to an OTP without a corresponding driving route to a pharmacy.
The primary outcome of mean population-weighted drive time was 20.4 minutes (95% CI, 20.3-20.6 minutes) to the closest OTP and 4.5 minutes (95% CI, 4.4-4.5 minutes) to the closest pharmacy (t = 189.4, P < .001) (Table 1). Driving times to both OTPs and pharmacies were positively skewed. The median drive time across the US to OTPs was 12.9 minutes (interquartile range, 7.8-23.1 minutes) and 3.2 minutes to pharmacies (interquartile range, 2.0-5.1 minutes) (eTable 2 in Supplement 1). A total of 98.0% of US residents lived in census tracts with a shorter drive time to the closest pharmacy than the closest OTP. A total of 17.8% of the population lived in census tracts with a 1-way drive time of 30 minutes or greater to OTPs, and 0.4% of the population had a drive time of 30 minutes or greater to pharmacies. A total of 5.3% of the population had a drive time of 60 minutes or greater to OTPs, and 0.07% of the population had a drive time of 60 minutes or greater to pharmacies.
Population-weighted mean driving times to OTPs and pharmacies were calculated for each county, revealing regional variations in access (Figure 2 and Supplement 2). Several clusters of counties had notably high drive times to OTPs, with a large cluster in the Mountain census division (Montana, Idaho, Wyoming, Nevada, Utah, Colorado, Arizona, and New Mexico), North Dakota, South Dakota, Nebraska, and Kansas (Wyoming did not have any OTPs in the state) (Figure 2A). State-level variation had similar patterns (eFigure 2 and eTable 3 in Supplement 1). These regions had considerably lower driving times to pharmacies than to OTPs (Figure 2C and eFigure 2 in Supplement 1).
Testing the prespecified secondary hypotheses, population-weighted driving time was longer to OTPs than to pharmacies within each NCHS urban-rural county classification (Table 1). Classifying census tracts by 2010 RUCA code similarly revealed higher driving times to OTPs than to pharmacies within each primary RUCA classification (eTable 4 in Supplement 1).
Significant variability was found by county classification in the mean driving time to OTPs (Welch ANOVA on unweighted driving times, F = 4430.1; P < .001) and in mean driving time to pharmacies (Welch ANOVA on unweighted driving times, F = 1087.0; P < .001). Population-weighted driving times to both OTPs and pharmacies were longer in nonmetropolitan county types than in metropolitan county types (Table 1).
In large central metro counties, 0.6% of individuals had a drive time longer than 30 minutes to OTPs, and 0.07% of individuals had a more than 30-minute drive to pharmacies (Figure 3). In noncore counties, 85.4% of individuals had a greater than 30-minute drive to an OTP, and 36.8% of individuals had a greater than 60-minute drive. A total of 2.5% of individuals in noncore counties lived more than a 30-minute drive from a pharmacy, whereas 0.6% of individuals lived more than a 60-minute drive away.
Mean driving distances were longer to OTPs than to pharmacies, with increasing distances in less urban counties (eTable 5 in Supplement 1). The population-weighted mean 1-way driving distance in the US was 15.6 miles (95% CI, 15.4-15.8 miles) to the closest OTP and 2.3 miles (95% CI, 2.2-2.3 miles) to the closest pharmacy. The median driving distance to OTPs was 7.5 miles (interquartile range, 3.5-17.2 miles), and the median driving distance to pharmacies was 1.1 miles (interquartile range, 0.6-2.2 miles).
Driving costs were higher to OTPs than to pharmacies in all regions using the variable cost and fuel-cost models (Table 2). State-level and county-level regional variations in cost were similar to driving distance variation (eFigures 3-5 in Supplement 1). In all counties, driving time, distance, and both cost estimates were lower in drives to pharmacies than to OTPs (Supplement 2).
A sensitivity analysis was conducted, removing the requirement that both the closest OTP and pharmacy be within a 12-hour drive from the census tract MCP. An additional 29 census tracts in Alaska had driving times from the MCP to the closest OTP or pharmacy that were greater than 12 hours and were included in the sensitivity analysis. Results were similar to those of the main analysis (eTable 6 in Supplement 1).
This study found that a pharmacy-based model of dispensing methadone maintenance treatment has the potential to decrease driving times, distances, and costs for US residents compared with the current OTP-based dispensing model. In all counties, the population-weighted mean driving time to a pharmacy was lower than the driving time to an OTP. Micropolitan and noncore counties had the largest differences between driving times to OTPs and driving times to pharmacies.
The analysis builds on previous studies of driving access to OTPs in several ways. First, the analysis comprehensively covered the US, showing county, state, and regional variations in access and highlighting regions with limited access. Supplement 2 provides a searchable table with results for each county in the US. Second, this analysis provided a nationwide comparison of drive times to OTPs with a potential alternative system of methadone maintenance dispensing. Third, this study incorporated estimates of driving distances and costs, which are important considerations for many patients. Fourth, this analysis included a methodologic improvement that results in a more accurate estimate of driving times. By calculating driving times to the nearest 10 OTPs and pharmacies from each MCP and then identifying the facility by lowest driving time as closest, this analysis accounted for situations in which the nearest treatment facility with the shortest great-circle distance was different from the facility with the shortest driving time. This study also provided, to my knowledge, the first nationwide analysis of driving times to pharmacies in the US, with prior pharmacy access studies being conducted on a regional or state level.9,33,34
Decreasing driving time, distance, and costs for patients obtaining methadone maintenance may help achieve several public health goals. Patients dissuaded from obtaining treatment because of extended travel, particularly patients with disabilities, unreliable access to transportation, or from rural regions, would have reduced barriers to care. Quality of life may be increased for the more than 380 000 individuals currently receiving methadone treatment if less time is spent commuting.6 Direct financial costs of commuting can be considerable for patients in rural regions when accounting for multiple round-way trips each week. Decreasing distances driven may be associated with reduced motor vehicle collisions and injuries. Reducing commuting time may help facilitate other aspects of recovery from OUD and may be associated with enhanced occupational opportunities for patients.
Several barriers constrain expansion of the existing OTP network. OTPs are governed by the Narcotic Addict Treatment Act of 1974 and subject to extensive regulations promulgated by the US Drug Enforcement Agency (DEA) and the FDA. OTPs must be certified by SAMHSA; be accredited by a SAMHSA-approved accreditation body; submit to inspections and surveys by SAMHSA, the DEA, and accreditation bodies; maintain records on receipt, storage, and dispensing of medications; follow federal rules about patient care; and provide psychosocial services.13 In addition, OTPs must comply with state and local laws, including zoning bylaws that limit clinic placement.35
Several strategies to increase access to methadone maintenance have been proposed, including prescribing methadone through primary care clinics and federally qualified health centers,7,36 increased use of mobile methadone vans,37 and increased use of medication units.38 Medication units are satellite units of OTPs permitted to dispense methadone and obtain urine samples.13 The large-scale registering of pharmacies as medication units under the auspices of willing OTPs would be a potential mechanism for increasing pharmacy-based dispensing of methadone, although considerable regulatory burdens would continue (eg, medication units must obtain all methadone doses from the overseeing OTP). Alternatively, pharmacy-based dispensing under the direction of a prescribing practitioner could be permitted through a congressional amendment to the Narcotic Addict Treatment Act and FDA and DEA regulatory changes.
This analysis has several limitations. First, OTPs that do not offer methadone maintenance treatment were included in the analysis, leading to an underestimation of driving times to OTPs where individuals could obtain methadone maintenance treatment. In the 2018 National Survey of Substance Abuse Treatment Centers, only 1360 of 1519 responding OTPs (89.5%) provided outpatient methadone maintenance.39 Although affiliated medication units were not included in this analysis, only 2 of 34 states responding to a SAMHSA survey reported having medication units.40 This analysis did not incorporate information about insurance coverage, which may vary by OTP and affect patients’ ability to access care.
Second, driving routes estimated in this study originated from census tract MCPs from the 2010 US Census instead of the individual residences of individuals eligible for methadone maintenance treatment. Population distributions may have changed since 2010, and the geographic distribution of residences of individuals eligible for methadone maintenance may differ from the overall population. By modeling driving times from MCPs to treatment facilities, the modeled driving times were likely to be lower than the mean of driving times for each person in a census tract, especially if treatment facilities were located near MCPs. This study used average driving speeds over the road network, which may not reflect driving speeds when individuals are trying to commute to OTPs or pharmacies (eg, during rush hour). Although this study estimated driving times, many individuals with OUD may not have access to personal motor vehicles and rely on public transportation or other forms of commuting.
Third, experience in the UK and Australia suggests that not all pharmacies would dispense methadone. In a survey of community pharmacies in England in 2005, 79% dispensed or were willing to dispense opioid substitution therapy (methadone or buprenorphine), and of community pharmacies already dispensing opioid substitution therapy, 92% were willing to supervise consumption.41 In Australia, 2627 of 5762 community pharmacies (46%) served as dosing sites for opioid substitution therapy.42,43 Actual driving times to pharmacy-based methadone dispensing sites would likely be longer than the driving times to pharmacies estimated in this analysis.
Fourth, this analysis did not assess driving access to other FDA-approved treatments for OUD, such as buprenorphine. Another study,44 however, found that buprenorphine access has remained restricted in many parts of the US; as of the end of 2017, 42.3% of US counties did not have any Drug Addiction Treatment Act–waivered practitioners who could prescribe buprenorphine for OUD.
Fifth, this study did not address whether specific treatment outcomes (eg, treatment retention and overdose mortality) would be better with either dispensing model. A retrospective study45 of patients prescribed methadone maintenance in Ontario found that patients who chose to receive methadone through pharmacies instead of clinic dispensing had lower treatment retention, although confounders associated with patient choice may have affected these results. In the US, patients living closer to OTPs are less likely to leave treatment,10 and treatment retention is associated with reduced mortality. Providing US veterans in rural areas the option of buprenorphine teleprescriptions in a pilot study was associated with comparable treatment retention to in-office buprenorphine prescribing.46 Interim methadone maintenance dispensed from an OTP with only emergency counseling had superior treatment retention to wait list controls, despite limited use of counseling services.47 Telehealth availability of emergency counseling and remote physician support may help address some of the limitations of pharmacy-based dispensing.
In this study, population-weighted mean driving times from US census tract MCPs were longer to OTPs than to pharmacies. Benefits of in-person administration of methadone at OTPs should be balanced with the low feasibility of commutes in many rural regions of the US. Pilot studies may help determine whether pharmacy-based dispensing of methadone maintenance treatment for OUD is effective, safe, and feasible in the US.
Accepted for Publication: April 22, 2020.
Corresponding Author: Robert A. Kleinman, MD, Department of Psychiatry and Behavioral Sciences, Stanford University School of Medicine, 401 Quarry Rd, Stanford, CA 94305 (firstname.lastname@example.org).
Published Online: July 15, 2020. doi:10.1001/jamapsychiatry.2020.1624
Author Contributions: Dr Kleinman had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Conflict of Interest Disclosures: None reported.
Additional Contributions: David Medeiros, MA, Stanford University, Stanford, California, provided instruction about ArcGIS software. He was not compensated for his work.