Currently Viewing:
The American Journal of Managed Care April 2013
More Comprehensive Discussion of CRC Screening Associated With Higher Screening
David M. Mosen, PhD, MPH; Adrianne C. Feldstein, MD, MS; Nancy A. Perrin, PhD; A. Gabriela Rosales, MS; David H. Smith, RPh, PhD; Elizabeth G. Liles, MD; Jennifer L. Schneider, MPH; Ronald E. Myers, PhD; and Jennifer Elston-Lafata, PhD
Comparative Effectiveness of Early Natalizumab Treatment in JC Virus-Negative Relapsing-Remitting Multiple Sclerosis
Jonathan D. Campbell, PhD; R. Brett McQueen, PhD; Augusto Miravalle, MD; John R. Corboy, MD; Timothy L. Vollmer, MD; and Kavita Nair, PhD
Health Plan Innovations in Delivery System Reforms
Karen Ignagni
Prescription Opioid Abuse: Challenges and Opportunities for Payers
Nathaniel P. Katz, MD, MS; Howard Birnbaum, PhD; Michael J. Brennan, MD; John D. Freedman, MD; Gary P. Gilmore, RPh; Dennis Jay, BS; George A. Kenna, PhD, RPh; Bertha K. Madras, PhD; Lisa McElhaney, BS; Roger D. Weiss, MD; and Alan G. White, PhD
Effects of Coverage Gap Reform on Adherence to Diabetes Medications
Feng Zeng, PhD; Bimal V. Patel, PharmD, MS; and Louis Brunetti, MD
Primary Nonadherence to Statin Therapy: Patients' Perceptions
Teresa N. Harrison, SM; Stephen F. Derose, MD, MS; T. Craig Cheetham, PharmD, MS; Vicki Chiu, MS; Southida S. Vansomphone, PharmD; Kelley Green, RN, PhD; Kaan Tunceli, PhD; Ronald D. Scott, MD; Elizabeth Marrett, MPH; and Kristi Reynolds, PhD, MPH
Trends in Hospital-Based Childbirth Care: The Role of Health Insurance
Katy B. Kozhimannil, PhD, MPA; Tetyana P. Shippee, PhD; Olusola Adegoke, MPH, MBBS; and Beth A. Virnig, PhD, MPH
Disease Management Programs: Barriers and Benefits
Racheli Magnezi, MBA, PhD; Galit Kaufman, RN, MA; Arnona Ziv, MBA; Ofra Kalter-Leibovici, MD; and Haim Reuveni, MD
Use of Web-Based Shared Medical Records Among Patients With HIV
James D. Ralston, MD, MPH; Michael J. Silverberg, PhD, MPH; Louis Grothaus, MA; Wendy A. Leyden, MPH; Tyler Ross, MA; Christine Stewart, PhD; Steven Carzasty, MSW; Michael Horberg, MD, MAS; and Sheryl L. Catz, PhD
Currently Reading
Retail Clinic Utilization Associated With Lower Total Cost of Care
Andrew Sussman, MD; Lisette Dunham, MSPH; Kristen Snower, MBA; Min Hu, MPH; Olga S. Matlin, PhD; William H. Shrank, MD; Niteesh K. Choudhry, MD, PhD; and Troyen Brennan, MD

Retail Clinic Utilization Associated With Lower Total Cost of Care

Andrew Sussman, MD; Lisette Dunham, MSPH; Kristen Snower, MBA; Min Hu, MPH; Olga S. Matlin, PhD; William H. Shrank, MD; Niteesh K. Choudhry, MD, PhD; and Troyen Brennan, MD
Retail clinic use is associated with lower overall total cost of care based on a matched-pair analysis.
Objectives: To better understand the impact of retail clinic use on a patient’s annual total cost of care.

Study Design: A propensity score matched-pair, cohort design was used to analyze healthcare spending patterns among CVS Caremark employees in the year following a visit to a MinuteClinic, the retail clinics inside CVS pharmacies.

Methods: De-identified medical and pharmacy claims for CVS Caremark employees and their dependents who received care at a retail clinic between June 1, 2009, and May 31, 2010, were matched to those of subjects who received care elsewhere. High-dimensional propensity score and greedy matching techniques were used to create a 1-to-1 matched cohort that was analyzed using generalized linear regression models.

Results: Individuals using a retail clinic had a lower total cost of care (–$262; 95% confidence interval, –$510 to –$31; P = .025) in the year following their clinic visit than individuals who received care in other settings. This savings was primarily due to lower medical expenses at physicians’ offices ($77 savings, P = .008) and hospital inpatient care ($121 savings, P = .049). The 6022 retail clinic users also had 142 (12%) fewer emergency department visits (P = .01), though this was not related to significant cost savings.

Conclusions: This study found that retail clinic use was associated with lower overall total cost of care compared with that at alternative sites. Savings may extend beyond the retail clinic visit itself to other types of medical utilization.

Am J Manag Care. 2013;19(4):e148-e157
If access to retail clinics can help to lower total medical expenditures while providing high quality of care, retail clinics could be an important component of the systems of care that provide comprehensive services at lower cost.

  • Total cost of care savings associated with retail clinic usage.

  • Savings were derived from decreased physician office and hospital visit expenses.

  • Use of retail clinics associated with reduced number of emergency department visits.
Medical walk-in clinics located in retail stores have become increasingly prevalent since first appearing in 2000.1 There are now more than 1400 retail clinics and the volume of patients treated increased more than 10-fold between 2007 and 2009 (personal e-mail communication with Tine Hansen-Tuton, MPA, JD, Convenient Care Association, June 6, 2012).1-5 These clinics generally employ a nurse practitioner or physician assistant to care for patients with a defined set of conditions that include acute medical problems (eg, sore throat, upper respiratory infection) and vaccinations on a walk-in basis, without appointments.6 Recently, monitoring services for chronic diseases and general physical examinations have been added.

Some have argued that the retail clinic could represent a positive development in American healthcare, suggesting that the care is both high quality and low cost.7 Quality of care at retail clinics has been assessed by several investigators. Mehrotra and colleagues found that quality of care for patients treated at retail clinics with sore throat, ear infection, and urinary tract infection was at least as good as that at alternative sites of care on 14 objective quality measures.8 Another study demonstrated 99% provider adherence to clinical practice guidelines for sore throat treatment.9 Studies of pediatric patients have revealed similarly high levels of adherence to clinical practice guidelines at retail clinics.10 Meanwhile, the costs per episode of illness have been observed to be 40% to 80% lower than those for services provided in physician offices, urgent care sites, and emergency departments (EDs).8

The impact of retail clinic medicine on the total cost of care has not been rigorously studied. After a literature review, we found only 1 study on total medical costs in the 6 months following a retail clinic visit, suggesting they are decreased, but this study did not examine the cost of pharmacy expenses or the total cost of care for patients seen at retail clinics.11 One hypothesis would suggest that retail clinics serve as a lower-cost substitute site of care for needed services. Alternatively, total costs of care might be higher for retail clinic users if the clinics stimulate increased utilization, especially in light of the access that they provide, including weekend and evening walk-in care without appointments.12 If patients additionally seek care from their regular care providers for the condition for which they were seen in the retail clinic, total cost and utilization might be higher. Similarly, if they seek care at retail clinics for conditions that would otherwise not be treated by professionals (self-care), overall utilization might be higher. Understanding the effect of retail clinic care on the total cost of delivered healthcare is essential as we seek new models to deliver high-quality care at lower overall costs.


A propensity score matched-pair, cohort design was used to analyze the healthcare spending pattern of CVS Caremark employees and dependents who were continuously eligible for medical and pharmacy benefits between June 1, 2008, and May 31, 2011. MinuteClinic is the largest retail clinic chain in the United States, with approximately 600 clinics in CVS pharmacies in 25 states.

Data Sources

All de-identified medical and pharmacy claims were obtained for all members with services between June 1, 2008 and May 31, 2011 following internal and external guidelines designed to guarantee confidentiality and integrity of personal health information. The medical claims identified by a computer generated unique subject identification number were:date of service, diagnosis codes as defined by International Classification of Disease, Ninth Revision (ICD-9), procedures identified by the Current Procedure Terminology, Version 4 code (CPT-4), gross cost of service, hospital revenue code, the provider tax ID number, and the member’s status as insured employee or dependent. The pharmacy claims identified by a computer generated unique subject identification number were: fill date of the prescription, generic product index (GPI) code, National Drug Codes (NDC), and gross cost of the drug. Internal de-identified employment data was aggregated and used to identify the environment of the subjects’ workplace for the CVS Caremark employees.

Study Cohort

Subjects were considered retail clinic users (exposed) if they were seen at CVS MinuteClinic locations. Retail clinic users (exposed) and nonusers (unexposed) were classified based on the presence or absence of medical claims billed by a MinuteClinic provider for services in the study entry period between June 1, 2009, and May 31, 2010. Index dates were defined for each member to create a subject-specific baseline covariate assessment period and follow-up period as the 365 days before and after the subject’s index date, respectively. The index date was defined for the exposed subjects based on their first retail clinic visit between June 1, 2009, and May 31, 2010. Subjects were excluded from analysis if they visited the retail clinic in the year prior to June  1, 2009, but not during the study entry period. This exclusion criterion allowed for a 12-month washout period for unexposed subjects. Subjects who were  classified as unexposed could have visited the retail clinic at some point after May 31, 2010, but not at any time prior to that date. The index dates for unexposed subjects were frequency matched to their exposed counterparts by calendar time. To account for outliers, all subjects who had total cost of care  more than $100,000 per year in either the baseline or follow-up period were removed from the analysis. All subjects were required to have at least 1 medical  claim and at least 1 ICD-9 code for a condition that could be treated at the retail clinic in the follow-up period. However, neither restriction was  applied to the baseline covariate assessment period. Retail clinic visits were defined by claims billed from MinuteClinic provider tax ID numbers only; use of  other retail clinics is unknown. Physician office visits (including urgent care, mental health visits, and rehabilitation visits) were defined by the place of  service code for office and urgent care centers that were not billed by MinuteClinic provider tax ID numbers. ED visits were defined by place of service code  for ED care. Inpatient hospital-based care was identified by place of service codes for inpatient hospital and included claims for inpatient consulting,  radiologic exam interpretation, and general inpatient care. Outpatient hospital care was defined by place of service codes for both outpatient hospital and  ambulatory surgical center, including procedures such as surgical pathology, screening mammography, endoscopy, and eye surgery. All medical claims that  did not fit these service categories were classified as “other” locations, and included laboratory and pathology expenses, radiology expenses, and other  provider expenses such as physical therapy.


The medical claims in the subject’s baseline period were used to create several behavior and utilization variables to describe and balance patient characteristics. Healthcareseeking behavior was defined by the presence of at least 1 ICD-9 or CPT-4 code for a preventive healthcare service (see Appendix). Chronic illness status (present/absent) was defined as presence of asthma, depression, chronic obstructive pulmonary disease, diabetes mellitus, coronary artery disease, or hypertension as indicated by ICD-9 codes for these diseases. Medical visit day count was defined as the sum of distinct days the subject had medical services claims, including hospital inpatient visit days. Distance was calculated as the Euclidean distance from the subject’s residence to the closest  MinuteClinic in miles at the time of the initial MinuteClinic visit. Age was calculated as of the index date; sex and state of residence were identified. Claims  in the follow-up period were categorized into 6 different locations of service based on the place of service code and provider tax ID number on the medical claims. From the internal employment data, subjects were defined as working in a retail pharmacy with a retail clinic on-site, as working in a retail store with no retail clinic on-site, as dependent subjects, or as subjects who worked at “other” sites of employment, which included office, distribution center, and mail center pharmacy employees.

Propensity Score Matching

A propensity score was calculated for each subject using the high-dimensional propensity score SAS macro.13 The propensity score included 500 empirically identified covariates (190 CPT-4 codes, 159 National Drug Codes, 151 ICD-9 codes), in addition to age, sex, state of residence, distance from residence to the nearest MinuteClinic, healthcare-seeking behavior, chronic illness status, and number of medical visits. The greedy matching algorithm was used to create 6022 matched pairs out of the 7545 exposed and 44,368 unexposed subjects who met study eligibility criteria.14 A maximum distance of 0.01 in propensity scores and no more than 7 days of difference in index dates between the exposed and unexposed members of the pair were allowed.


Total costs of care per subject in both the baseline assessment and follow-up periods were calculated as the sum of gross (ie, combined patient and insurer) service costs on the paid medical claims and gross prescription costs on paid pharmacy claims. An analysis of the location of care (eg, ED, hospital, physician office, retail clinic, other) based on claims expense was undertaken, comparing exposed and unexposed subjects.

Statistical Analyses

McNemar and paired t tests were performed to test the differences between groups. To control for remaining significant differences after matching, we fit 3 generalized linear regression models, using generalized estimating equations with spending between exposed and unexposed subjects after adjusting for the matched design and other covariates (age, sex, state, distance to MinuteClinic location, healthcare-seeking behavior, chronic illness status, work location). Since this generalized estimating equation model does not allow for values of zero, $0.01 was applied for individuals who spent nothing at the pharmacy in the study period. All analyses were performed with SAS software, version 9.1 (SAS Institute Inc, Cary, North Carolina).


Among the 51,913 potentially eligible individuals, 7545 used a MinuteClinic retail clinic and 44,368 did not. A total of 12,044 subjects were selected, creating 6022 matched pairs for analysis. Characteristics of both the unmatched and matched populations are displayed in Table 1.

Unmatched Population

The unmatched population of 7545 retail clinic users had few meaningful differences in measurable demographic characteristics from the 44,368 unexposed subjects. The only difference of note is average distance to the nearest MinuteClinic; exposed subjects lived an average of 6.2 miles from the nearest retail clinic and unexposed subjects lived an average of 32.9 miles from the nearest retail clinic.

Copyright AJMC 2006-2019 Clinical Care Targeted Communications Group, LLC. All Rights Reserved.
Welcome the the new and improved, the premier managed market network. Tell us about yourself so that we can serve you better.
Sign Up