Abstract
Purpose: As a rapidly growing new health care delivery model in the United States, retail clinics have been the subject of much debate and controversy. Located physically within a retail store, retail clinics provide simple acute and preventive services for a fixed price and without an appointment. Some hope that retail clinics can be a new safety-net provider for the poor and those without a primary care physician. To better understand the potential for retail clinics to achieve this goal, we describe the sociodemographic characteristics of the communities in which they operate.
Methods: We created an inventory of all retail clinics in the United States and determined the proportion that are in a health professional shortage area (HPSA). We defined each retail clinic's catchment area as all census blocks that were less than a 5-minute driving distance from the clinic. We compared the sociodemographic characteristics of the population within and outside of these retail clinic catchment areas.
Results: Of the 982 clinics in 32 states, 88.4% were in an urban area and 12.5% were in an HPSA (20.9% of the US population lives within an HPSA). Compared with the rest of the urban population, the population living within a retail clinic catchment area has a higher median household income ($52,849 vs $46,080), is better educated (32.6% vs 24.9% with a college degree), and is as likely to be uninsured (17.7% vs 17.0%). In a multivariate model, the census block's median household income had the strongest association with whether the census block was in a retail clinic catchment area (odds ratio, 3.63; 95% CI, 3.26–4.05; median income, ≥$54,779 vs <$30,781, respectively).
Conclusions: We found that relatively few retail clinics are located in HPSAs and that, compared with the rest of the urban population, the population living in close proximity to a retail clinic has a higher income.
Offering a novel method of health care delivery, retail clinics have garnered significant interest from patients, politicians, physicians, and health plans. They are called retail clinics because they are physically located in retail stores like grocery stores, drugstores, or “big box” stores such as Wal-Mart. Retail clinics provide walk-in care for a limited number of acute illnesses and preventative care services for a fixed price. Visits are usually covered by most health insurance plans.1 Generally staffed by nurse practitioners, retail clinics focus on patient convenience by requiring no appointment and offering night and weekend hours. The number of retail clinics has increased rapidly, and it is estimated that there were 3 million patient visits to retail clinics in 2008.2
As the number of retail clinics has increased, several physician associations have raised concerns about the retail clinic model. They worry that retail clinics can possibly increase fragmentation of care, provide inferior care, and adversely impact the delivery of preventative care.3–9 Most retail clinics are owned by for-profit drugstore chains. Some have worried that, in an effort to increase pharmacy sales, retail clinic providers will overprescribe medications.4,5 However, policymakers, including California Governor Arnold Schwarzenegger,10 Pennsylvania Governor Ed Rendell,11 and Senator John McCain,12 have supported the growth of retail clinics and included them in their health reform proposals. Potential cited benefits of clinics include improving access to care and providing affordable care for the poor and underserved.13,14
Despite the interest and controversy generated by retail clinics, there has been little empirical evaluation of their impact.1,15–21 The goals of this study were to describe the location of retail clinics, determine whether they are disproportionately located in areas with a scarcity of health providers, and describe the sociodemographic characteristics of the population that lives close to a retail clinic.
Methods
Inventory of Retail Clinics
We compiled a list of retail clinic companies based on several sources: recent industry reports, analysis from a clinic consulting company (merchantmedicine.com), and general web searches.22,23 To our knowledge there have been no hard definitions of retail clinics in previous literature. We therefore chose to define retail clinics based on the following 3 criteria: (1) location within a retail store, (2) staffing by nurse practitioners or physician assistants, and (3) a limited menu of services with specified fixed prices. We did include some companies whose clinics did not satisfy all these criteria but embodied the character of the model. For example, we included QuickHealth clinics, which are located in retail stores such as Wal-Mart and use a menu of services but are staffed by physicians. We chose to include 2 clinic operators that target transient populations: AeroClinic, which operates in airports, and Roadside Med, which operates in Pilot Travel centers that are located next to highways. These clinics serve employees of their respective retail hubs in addition to travelers. We verified the list of companies with 2 experts on the retail clinic industry: MaryKate Scott, an independent consultant, and Caroline Ridgeway from the Convenient Care Association, which is a lobbying organization for the retail clinic industry. We obtained the addresses of all clinics from the retail clinic companies’ web sites between June and August 2008.
Mapping of Retail Clinics
Using geospatial imaging software (ArcInfo version 9.3, ESRI, Inc., Redlands, CA) we mapped the location of each clinic. Many retail clinics have opened in newly built neighborhoods that do not map directly to an identified street in the ESRI street database, which is several years old. For 28% of clinics, we used Google maps to correct misspelled addresses, identify alternate city names or zip codes, or isolate the closest address of a location that was able to mapped. If we used the closest address that was able to be mapped, it was within 1 to 2 blocks of the actual retail clinic location.
Creating Retail Clinic Catchment Areas
We defined the catchment area around each retail clinic as a 5-minute driving distance in each direction. The catchment algorithm relies on road-based travel to define the point boundaries in every direction and then makes linear connections between these point boundaries to create a boundary loop. The travel time calculated incorporates average driving speed on the type of road (eg, faster speeds on highways), but does not incorporate typical levels of traffic. We chose a 5-minute driving distance to define the minimum size of catchment areas because previous research estimates 5 minutes as the time persons are willing to travel to some retail stores.24 We looked at a 10-minute driving distance catchment area, as well, to test the sensitivity of our analyses.
Defining Population Within and Outside Retail Clinic Catchment Areas
We compared the sociodemographic characteristics of the populations within and outside the retail clinic catchment areas. Because the vast majority of retail clinics were in urban areas, our analyses focus on only the US urban population. Urban areas have been defined by the Census Bureau25 and include regions with a population greater than 500 to 1000 people per square mile or adjoining areas. Using ArcGIS (ESRI, Inc.) we selected all urban census blocks whose centroids lay within a retail clinic catchment area. Census blocks are the smallest increment of census data; there were approximately 8.5 million census blocks in the 2000 census.25 They contain limited aggregated information from a 100% survey of the population and average 33 people per block. Blocks are aggregated into block groups, which are aggregated into census tracts, which are aggregated into counties, which are aggregated into states. After evaluating several possible aerial interpolation methods, we felt the census block centroid analysis to be the most logical for our purposes given the granularity of the data.
We acquired US Census data at the block level for race, ethnicity, age, and sex from 2000 (the most recent detailed data available). Because income, education, and urban population data are not available only at the block level, we imputed these data from the census tract. For example, if 22% of the tract had a college education, we assumed that 22% of the population of each block in that tract had a college education. On average, 120 blocks make up census tracts, which contain an average of 4000 people. To our knowledge, the most granular health insurance data are available at the county level. We obtained data about health insurance rates at the county level in 2005 from the Census Bureau's Small Area Health Insurance Estimates division and applied this data to the block level.
We created a multivariate model to determine characteristics of a census block that were associated with the census block being within a 5-minute retail clinic catchment area. The unit of analysis was the census block (n = 2,379,547). We set the outcome variable as a binary “within” or “outside of” a retail clinic catchment area and used the census block characteristics listed above (eg, percentage of census block that was nonwhite race) as predictor variables. We limited our analysis to urban census blocks with ≥10 residents to ensure a stable estimate of census characteristics. Because of collinearity with median income, we did not include 2 variables in the model: percentage of residents with a college education and percentage of residents below the poverty line. Insurance rates were not included because they were pulled from the county level. Because of a nonlinear relationship between predictor and outcome variables, we chose to divide the census block characteristics into quartiles. Many of the predictor variables were collected at the level of the census tract, so we accounted for clustering using proc surveylogistic software (version 9.1, SAS Institute, Cary, NC). We conducted a sensitivity analysis using a binary outcome variable of “within” or “outside of” a 10-minute catchment area.
We used geospatial imaging software to determine the fraction of retail clinics located in primary care health professional shortage areas (HPSAs). HPSA geographic boundaries are defined by the Health Resources and Services Administration.26
Results
As of August 2008, 42 different clinic companies operated 982 retail clinics in the United States (Table 1). Although more than 40 operators run retail clinics, the 5 largest (MinuteClinic, Take-Care, Little Clinic, TargetClinic, and Redi-Clinic) run 82% of the clinics. Clinics were located most frequently in urban areas (88.2% of clinics) and 99.1% of the population that lives within 5 minutes of a retail clinic lives in an urban area. Across the nation, 12.5% of the clinics are located in an HPSA whereas 20.9% of the general US population lives in an HPSA.27
Retail Clinic Companies Operating Clinics as of August 2008
Compared with other urban residents, the population that lives within a 5-minute retail clinic catchment area has a higher median household income ($52,943 vs $46,080), is better educated (32.6% vs 24.9% with a college education), is less likely to live below the poverty line (10.2% vs 12.6%), and has a higher proportion of the population that is Hispanic or Latino (17.6% vs 15.4%) (Table 2). They have similar rates of being uninsured (17.7% vs 17.0%) and there were no notable differences in age or sex. The demographic comparison was substantively the same when we used a 10-minute catchment area.
Sociodemographic Characteristics of Urban Population Within and Outside of a Retail Clinic Catchment Area
We created a multivariate model to determine characteristics of a census block that were associated with being within a 5-minute retail clinic catchment area. The characteristic with the strongest association was median household income. Compared to census blocks with a median household income <$31,781, census blocks where the median income was ≥$54,779 were more likely (odds ratio, 3.63; 95% CI, 3.26–4.05) to be within a 5-minute catchment area (Table 3). Because of the large number of census blocks (>2 million) all other variables had statistically significant associations, but the magnitude of the associations was relatively smaller (odds ratio, <2).
Census Tract Characteristics Associated With Being in a Retail Clinic Catchment Area (Multivariate Model)
Discussion
Supporters of the retail clinic model have argued that its growth could improve access to care in general and in particular for the underserved, including the poor and those with little access to primary care physicians. We find that retail clinics are not preferentially located in communities with these demographics. A relatively small fraction of clinics are located in an HPSA, and the income of census blocks within 5 minutes of a retail clinic was higher than that of the rest of the urban population.
How can these findings influence the continuing debate about retail clinics? First, it should dispel the notion that retail clinic companies are specifically targeting the underserved. Rather, except for the differences in income and education, the urban residents that live within a retail clinic catchment area are similar to the urban population as a whole. This may not be surprising. Retail clinics are most commonly run by for-profit companies who want to reach as broad a segment of the whole population as possible. It is also important to note that primary care physicians’ offices are often preferentially located in higher-income areas.28 Second, the vast majority of retail clinics are in urban areas. This might make sense from a business perspective as companies try to reach as many clients as possible, but rural areas are most in need of new care options.
There are several key limitations to our analyses. The decision about where retail clinic companies place a retail clinic is likely influenced by variables outside our analysis, such as available partner retail stores (eg, availability of a Walgreens in the area), suitable store layout for a retail clinic (eg, a corner of the store with available plumbing), and store foot traffic. We used 5-minute and 10-minute driving distances to define retail clinic catchment areas, but this was not informed by previous access to care literature. This literature has generally focused on other types of situations, such as how distance to a radiation oncologist might impact a patient's decision to undergo breast-conserving surgery, and therefore was not applicable to retail clinics.29 Therefore, we used data about typical travel times to a grocery store as a proxy.24 We recognize that patients vary in the distance they are willing to travel; other factors such as physical geography and personal driving patterns will influence who is willing to drive to a clinic. For example, patients in rural areas are probably willing to drive a longer distance. As noted above, we looked at the entire population within a catchment area, recognizing that this population is the “possible clientele” rather than the “probable clientele” of the retail clinic. For example, an elderly patient who has a strong relationship with her primary care physician may live in the catchment area but will not likely go to the retail clinic.
There were several analytic issues related to using census data that introduced some error in our results. Several sociodemographic characteristics in our analysis were not available at the census-block level. For example, health insurance data are available only at the county level. To include these data, we chose to impute available county-level data about insurance rates to the census block level. Because counties can encompass areas much larger than a typical retail clinic catchment area, we have a limited ability to detect differences in insurance rates of those within and outside a retail clinic catchment area. Therefore, insurance-related results should be interpreted with caution. Previous research has found that retail clinics attract patients who are less likely to use insurance to pay for a visit than patients who visit a physician.1 It might be that those without insurance might be preferentially seeking retail clinics for care. Lastly, by necessity we used census data from 2000 whereas clinic addresses and HPSA boundaries were from 2008. All of these census-related methodological issues introduce some error in our results, but whether they bias our findings is unclear.
Conclusions
Retail clinics are a rapidly growing segment of the US health care system that has garnered significant interest and controversy. We believe our study represents one of the first in the peer-reviewed literature to describe on a national scale who can visit a retail clinic. The communities surrounding retail clinics are, on average, wealthier and better educated than the rest of the nation. These results can help inform the ongoing policy debate about the potential harms and benefits of the retail clinic model.
Notes
This article was externally peer reviewed.
Funding: This work was supported by a grant (08-1136) from the California Health Care Foundation and a career development award from the National Center for Research Resources, a component of the National Institutes of Health. The funders had no role in the design and conduct of the study; the collection, management, analysis, and interpretation of the data; or the preparation, review, or approval of the manuscript.
Conflict of interest: none declared.
- Received for publication February 20, 2009.
- Revision received May 12, 2009.
- Accepted for publication May 20, 2009.
