TRANSTHEORETICAL MODEL DEVELOPMENT FOR RETAIL HEALTH CLINIC UTILIZATION

Primary Care is considered to be in a crisis in the U.S. related to increasing rates of chronic disease, increasing numbers of patients, less physicians, and less money. This dilemma has led to the rise of what could be a disruptive innovation in the form of retail health clinics, health clinics located within retail settings like pharmacies and large retail stores. The core aim of this study was to use a sequential approach to measurement development to develop TTM measures for the Stage of Change, Decisional Balance, and Self-Efficacy for patients’ readiness to utilize retail health clinics using split half validation procedures. The sample consisted of 551 patients with a stage distribution of Precontemplation 24.4%, Contemplation 14.2%, Preparation 20.3%, Action 5.8% and Maintenance 35.3%. Table 3 reports demographics and Stage of Change. Exploratory principle components analyses produced a 2-factor (Pros α=.88; Cons α=.85) 8-item scale for the Decisional Balance measure and a 1-factor 5-item scale for the Self-Efficacy measure (α=.83). Confirmatory analyses replicated the hypothesized factor structures for both the decisional balance (CFI=.958, SRMR=.055, loadings .63-.88) and Self-Efficacy (CFI=.999, SRMR=.019, loadings .73-.84) scales. MANOVA results by stage of change were significant Wilk’s Λ= .79, F(4, 4,484)= 9.85, p<.001, multivariate η2=.076. The Self-Efficacy measure and the Pros scale of the Decisional Balance measure replicated the expected patterns across the stages. The Cons scale deviated from the expected pattern of decreasing from Precontemplation to Maintenance, actually resulting in an increase. Overall, this study supports the application of the

Care has been considered to be in a crisis in the U.S. related to increasing rates of chronic disease, increasing numbers of patients, less physicians, and less money (Lee, Bodenheimer, Goroll, Starfield, & Treadway, 2008). This dilemma has led to the rise of what could be a disruptive innovation in the form of retail health clinics across the country. Retail clinics are generally located in retail settings including pharmacies, grocery stores, and discount chains with the vast majority owned and operated by large pharmacy companies (Arthur et al., 2015). In fact, only 3 companies, CVS, Walgreens, and Target accounted for the ownership of 73% of all retail clinics in 2012. In contrast, existing hospital chains or physician groups owned just 11% (Mehrotra & Lave, 2012). Like more traditional care providers, retail clinics have a referral network for more serious or chronic illnesses and collaborate with other local providers. They generally accept most major health insurance plans and utilize electronic medical records (McKinlay & Marceau, 2012).
In many ways, the patient experience can be very similar to more traditional providers.
For example, retail clinics are generally open 7 days a week for 12 hours on weekdays and 8 hours on weekends for walk-in appointments. They provide services like vaccinations and physical exams in addition to treating a limited number of acute conditions. However, lab tests, EKGs, the diagnosis of serious medical conditions, and in many cases the management of chronic diseases are not offered. Visits are short (approx. 15 mins) and costs can be as much as 30-80% less than costs for more traditional providers of acute care. Prices are predominantly displayed and they generally accept all major insurance carries. The providers staffing retail clinics are often Nurse Practitioners (NPs) and Physician Assistants (PAs).
While data on the expansion in quantity and scope of these clinics is becoming more readily available, there is limited research into better understanding who are using these clinics and why. Rising healthcare costs have brought new found attention and interest to cost reduction strategies. Some patients are also becoming better healthcare consumers who are more likely to consider costs when selecting providers and treatment facilities. Retail health clinics not only offer an additional treatment facility option with expanded access, but have capitalized on healthcare consumerism via increased cost transparency.
Improving our understanding of what may lead patients to use retail clinics can provide valuable information for how the rise of these will impact the current healthcare structure, costs, and coordination of care. These data can also add to the understanding of what healthcare consumers value in their decisions where to obtain care and could help to predict future healthcare trends in the areas of acute and preventive care. Moreover, the possible consequences of increased retail health clinic utilization are not well understood. Expanding our knowledge about this potentially disruptive addition to the healthcare system is vital if we are to keep pace with the constantly evolving US healthcare system.

REVIEW OF LITERATURE
The retail health clinic industry began in 2000 with the opening of QuickMedX clinics in Minnesota and the industry has seen substantial growth since (Leppel, 2010).
There were questions in 2007 about the continued growth and sustainability of clinics with as few as 60 clinics at the beginning of 2006 (Tu & Cohen, 2008). However, those concerns seemed to diminish quickly with the number of clinics rising dramatically over the next few years. According to Professional Pulse (Professional Pulse, 2016), there were approximately 1,900 clinics in existence by 2014. The number of clinics is expected to exceed 2,800 by the end of 2017 supporting more than 11 million annual appointments according to a report by Accenture (Accenture, 2015). Retail clinics may be here to stay.

Services Provided
Retail clinics focus the care they provide on a limited number of common acute conditions. These conditions generally have widely accepted treatment guidelines and generally do not require follow-up appointments making them ideal for treatment in the retail settings (Dalen, 2016). Approximately 5% of cases that present at retail clinics fall outside the scope of their practice and in these cases, retail clinics refer patients to other available providers like urgent care or emergency departments in hospitals (Mehrotra, Wang, Lave, Adams, & McGlynn, 2008). By far, the most common presenting illness is upper respiratory infections accounting for approximately 61% of all visits. Preventive exams and vaccinations also account for a substantial portion of visits, 22% of all visits (Weinick, Burns, & Mehrotra, 2010).
While the scope of retail clinics has been limited to date, there are efforts currently underway to expand into the areas of chronic care management, public health related interventions, and supplementing the care they provide via telemedicine.
These shifts have large implications for the role of retail clinics and have led to the formation of partnerships between retail clinics and larger healthcare systems. For example, CVS, operator of approximately 1,000 retail clinics, currently has affiliations with more than 50 healthcare organizations including the Cleveland Clinic, Henry Ford Health System, and Kaiser Permanente (Dalen, 2016). These partnerships, along with efforts made by independent retail clinics, are creating a shift away from fragmented care and may actually facilitate connected health care system growth and access.
As the reach of retail clinics continues to expand, their ability to treat chronic illness continues as well. Indeed, most of the major players in the retail health business have expanded into some areas of chronic care. For example, Walgreens is now offering management services for asthma, diabetes, and high cholesterol (Appleby, 2013). Clinics operated by WalMart now have the capability to diagnose, treat, and manage a wide range of chronic illness including hypertension, dyslipidemia, and COPD in addition to diabetes and asthma (Chang, Brundage, & Chokshi, 2015). CVS offers many of these same services, and is expanding into weight management.
QCare Clinics, partnered with ShopRite grocery stores, developed behavioral health screen kiosks placed in the waiting rooms of retail clinics to screen for common mental health conditions (Bacharach, Frohlich, Garcimonde, & Nevitt, 2015). This highlights the potential for areas of further expansion into public health domains such as mental health screening, smoking cessation, alcohol reduction, and HIV screening.
There is a precedent in other countries for community pharmacies to be points of care for such interventions. For example, pharmacies in the United Kingdom are using the Transtheoretical Model of Behavior Change, which has proven to be effective and cost efficient when delivered in community pharmacies. Weight management interventions have also proven to be feasible in these settings and early research has shown positive short-term results (Brown et al., 2016). Pharmacies are also practical and appealing for HIV screening because at-risk populations often lack PCPs or medical homes, cannot afford the costs of traditional settings, and may require repeat testing (Dugdale, Zaller, Bratberg, Berk, & Flanigan, 2014).
A relatively new expansion for retail clinics has been to leverage the use of telemedicine technologies. CVS announced in 2015 that they were partnering with three leading direct-to-patient telemedicine services to bring these services to their instore clinics (CVS Health, 2015). In such a system, patients are offered the opportunity to be treated remotely by a physician with the assistance of an on-site nurse. Early data on these services has been positive with 32% actually preferring a telehealth visit over a traditional in-person visit and 70% reporting that they were highly satisfied with the experience, would use it again, and would recommend it to others. Of those that utilized the service, 80% were insured, 70% were female, and 59% had a primary care provider (Polinski et al., 2016).

Benefits of Retail Clinics
Cost Retail clinics have generally been able to offer cost savings over traditional providers largely because of less expensive staffing models (Chang et al., 2015). The median cost of retail clinic visits was $88.10 compared to $126.30 for similar services at traditional providers (Mehrotra & Lave, 2012;Rohrer, Angstman, & Bartel, 2009).
Average savings have been estimated to be approximately $50-55 per episode and some research suggests that an estimated 13-27% of all ED visits could be handled in retail clinics resulting in a potential savings of $4.4 billion dollars annually (Thygeson, Van Vorst, Maciosek, & Solberg, 2008;Weinick et al., 2010).
Ahmed and Fincham conducted a discrete choice experiment that found that despite a preference to be treated by a physician, cost remained a key factor in deciding where to be treated and by whom (Ahmed & Fincham, 2011). Specifically, they found that it would take an average savings of $31.42 for patients to seek care from a nurse practitioner at a retail clinic rather than a physician at a private office.
They also found that it would require an average savings of $83.20 to wait an additional day to seek care. These data support the success and continued growth of retail clinics as point of care options that offer reduced costs and increased convenience that are appealing to modern healthcare consumers. While the data on episodic costs highlights consistent savings, more research is needed to better understand the overall impact of retail clinics.

Access
The benefits of costs in retail clinics seem to go beyond simple, episodic cost savings. Most retail clinics accept insurance but also have pricing systems in place that are appealing to those needing or willing to pay out of pocket (Ahmed & Fincham, 2011;Rudavsky, Pollack, & Mehrotra, 2009). Their flat fee pricing is prominently displayed, which is generally not the case in traditional settings. This level of transparency can increase access for those who are without insurance or who are underinsured (Chang et al., 2015).
Similar to cost savings and transparent pricing, convenience has consistently proven to be a positive driving factor in the success of retail clinics. Retail clinics generally offer afterhours care on weekdays and access throughout weekends, which many physician offices do not (Mehrotra & Lave, 2012). Their locations in large retail settings also provide free, accessible parking in areas that patients already frequently travel to and from. Further, most retail clinics are co-located with or nested in retail pharmacies, allowing for prescriptions to be filled on-site (Dalen, 2016).
Some retail clinics will accept scheduled appointments, but their current business model continues to be based on walk-in services. Despite this, they are able to keep wait-times shorter than most traditional providers (Chang et al., 2015;Dalen, 2016). In fact, most retail clinics view what would be considered a modest traditional wait-time of 20 minutes to be far too long and are constantly trying to innovate ways to decrease wait-times. Such immediate access is of extreme importance to today's healthcare consumers as 75% of Americans report that it is difficult to make timely doctor's appointments, get phone advice, or obtain care after hours without seeking care from an emergency department (Levine & Linder, 2016).

Quality of Care
Despite offering lower costs though less expensive staffing models, the quality of care received continues to receive marks similar to traditional care in physician offices, urgent care, and emergency departments. Concerns about quality of care will be discussed later in this paper, but it should be noted that there is substantial evidence that quality of care by Nurse Practitioners is high (Horrocks, 2002 (Shrank et al., 2014).

Geographic Location
Geographically, access to retail clinics has been somewhat limited with 88% located in major metropolitan areas (Martsolf et al., 2017). With the unprecedented growth of retail clinics, access to retail clinics remains limited for many Americans. A subsequent study in 2012 noted that 43% of retail clinics were located in the south, 31% in the Midwest, and nearly half of all retail clinics were located in just 5 states, FL, CA, TX, MN, and IL (Mehrotra & Lave, 2012). People in these regions, especially those in and around urban settings, are likely to have access to a retail clinic within a 10-minute drive of their home.
Distribution of clinics across areas of high and low socioeconomic status presents another factor limiting access. According to a 2009 study, counties that had a retail clinic had lower Black population percentages, lower poverty rates, higher median incomes, and were less likely to be medically underserved (Craig Evan Pollack & Armstrong, 2009). Retail stores that had health clinics were also less likely to be located in medically underserved areas compared to stores without clinics.
Indeed, subsequent research has found similar results with only 12.8% being located in medically underserved areas and more likely to be located in metropolitan areas with lower poverty rates and higher median incomes (Mehrotra & Lave, 2012). These findings suggest that retail clinics and their benefits are not equally accessible for those with the greatest need. Increasing access to care could help to increase health equity and reduce demonstrated health disparities in low income areas if clinics were distributed in ways that improved access across communities.

Quality of Care
The American College of Physicians and others in the medical field have expressed concern about the rise of retail clinics and their impacts on the healthcare system (Daniel & Erickson, 2015;Rohrer et al., 2009). The core concern often centers on the implications for long-term care and they argue for a balance of accessibility and convenience with the importance of longitudinal care. The issues of patient care coordination are supplemented by additional concerns related to over-utilization, overprescribing of antibiotics, perceived lack of preventive care and the potential for eroding relationships with PCPs and medical homes. Also, there is some concern about public awareness related to providers in retail clinics with some patients being treated by NPs reporting beliefs they are being treated by "doctors" (Hunter, Weber, Morreale, & Wall, 2009).
The concern about patient care coordination and subsequent impacts is supported by a few studies. ). There is also evidence suggesting that patients who visit retail clinics make fewer subsequent visits to their PCPs and as a result, may have less continuity of care (Reid et al., 2013). Fewer interactions with PCPs could lead to less knowledge of the patient and for those without PCPs, the availability of retail clinics may impact their motivation to seek one (Craig E. Pollack, Gidengil, & Mehrotra, 2010). However, this seems to be a part of the system that can and is being continuously improved upon.
In an article published in the New England Journal of Medicine, Cassel highlights three ways to improve the coordination of care in retail settings (Cassel, 2012 Two specific concerns stemming from the continued growth of retail clinic usage are the potential for treatment over-utilization and over-prescribing of antibiotics. Over-utilization is mostly limited to the emerging area of telemedicine in retail settings, which consumes valuable physician resources and can generate unnecessary follow-up appointments (Chang et al., 2015;Levine & Linder, 2016).
More research is needed to better observe and understand the potential for treatment overutilization in retail clinic settings.
Concern about the over-prescribing of antibiotics is better researched and findings suggest this concern is overstated with rates of prescriptions in retail clinics being similar to or better than those in physician offices, urgent care, and emergency rooms (Mehrotra & Lave, 2012). Specific findings have shown that 99.75% of patients in a retail clinic received an appropriate antibiotic prescription and that 99.05% of cases appropriately withheld antibiotic prescriptions. Of the remaining 0.95% where antibiotics were prescribed, half were supported with documentation of clinical concerns justifying the prescription as reasonable (Woodburn, Smith, & Nelson, 2007). In fact, antibiotic prescribing has been shown to be more guideline concordant in retail clinics and thus, more diagnostically appropriate than one might find in primary care practices and emergency rooms (Mehrotra, Gidengil, Setodji, Burns, & Linder, 2015).
Concerns about a lack of preventive care in retail clinics have also been raised.
These concerns stem from the advantage in cases where a patient presents at their PCP for an acute episode. The PCP knows the patient and their ongoing medical risks and despite an unrelated presenting problem, has the opportunity to check in and follow up on ongoing or chronic conditions. Despite these seemingly valid concerns, the limited research to explore the impacts of retail clinic visits on preventive care have found no significant differences compared to primary care and urgent care (Mehrotra & Lave, 2012;Reid et al., 2013).

Utilization of Retail Clinics
The three largest retail clinic operators reported 8.9 million visits between 2007 and 2009 and predict that total retail clinic visits will exceed 11 million per year by 2017, highlighting the rapid growth of utilization (Accenture, 2015;Mehrotra & Lave, 2012;Uscher-Pines, Harris, Burns, & Mehrotra, 2012). It's believed that as many as 1 in 5 PCP visits and 1 in 10 emergency room visits can be treated in retail clinics in more cost-effective ways. With these data in mind, a better understanding of who is using retail clinics, for what presenting problems, and why they are choosing retail clinics is important.

Patient Characteristics
A few trends have emerged from the limited research about the characteristics of patients utilizing retail clinics. Generally speaking, utilization has been higher among women and those younger in age. They also tend to be patients who either lack a regular healthcare provider or do not have insurance (Ashwood et al., 2011;Leppel, 2010;RAND Corperation, 2016). Some evidence suggests that patients with concerns about misdiagnosis and provider qualifications are less likely to utilize retail clinics (RAND Corperation, 2016). In a study limited to commercially insured patients the top predictors of retail clinic use were distance to retail clinic, age, chronic illness, income, and gender (Ashwood et al., 2011 (Hunter et al., 2009).

Presenting Problems Treated
As indicated earlier, presenting problems are generally limited to acute issues with well-established treatment guidelines. Indeed, 95% of all presented cases fall into categories of upper respiratory infections, sinusitis, bronchitis, sore throat, immunizations, inner ear infections, swimmer's ear, conjunctivitis, urinary tract infections, and screening blood tests with the other 5% being referred to other providers (RAND Corperation, 2016). This is in notable contrast to rates seen for these issues in primary care (18%) and in emergency rooms (12%). Approximately 40 percent of all visits to retail clinics are for immunizations, which seem driven by customer demand, convenience, and profitability. However, more research is needed to better understand these services and how well they are integrated into health department immunization registries (Arthur et al., 2015;Uscher-Pines et al., 2012).

Reasons for Utilization
A 2005 survey completed by the Wall Street Journal and Harris examined retail clinic utilization to better understand why patients are choosing them over more traditional providers (Gullo, 2005). Not surprisingly, the results mirror many of the issues discussed in this review. At the time, only 7% reported that they had used a retail clinic, but interestingly 42% stated that they would if they had access to one. A study by Wilson et al. reported that 90% of those who had used retail clinics lived within 10 miles of a clinic (Wilson et al., 2010). Wang and colleagues (2010) also explored this question by directly asking patients, "what is it about this clinic that brought you in today?" (Wang, Ryan, McGlynn, & Mehrotra, 2010). The most commonly recorded responses were short travel distance, reasonable pricing, and fast service. These findings support the importance of availability, access, and cost for utilization (Hunter et al., 2009;Wilson et al., 2010).
The Wall Street Journal survey also reported that 92% of patients were satisfied with the convenience, 89% with the quality of care, 88% with the staff qualification, and 80% with the cost. Reasons cited for using a retail clinic were lack of a PCP, being uninsured, unable to schedule a convenient or timely appointment with their PCP, and a desire to avoid issues of wait times in emergency rooms related to triage. Other factors highlighted in this research were walk-in availability, short wait times, hours of operation and interestingly, a desire among some respondents to shop at the retail store in conjunction with their healthcare visit (Hunter et al., 2009;Mehrotra & Lave, 2012). The overall theme seems to be that retail clinics can provide at least adequate care as a cost effective, convenient solution to consumers' healthcare needs.

TTM Overview
The Transtheoretical Model (TTM) is an integrative model of intentional behavior change that describes why, how, and when people change their behavior (Prochaska & DiClemente, 1983;Prochaska & Velicer, 1997). The TTM frames behavior change as something that happens over time and across a series of stages referred to as the stages of change. These stages include Precontemplation (not ready), Contemplation (getting ready), Preparation (ready), Action (reached criteria for change) and Maintenance (criteria reached for 6 months or more) (Prochaska & DiClemente, 1983;Prochaska, Redding, & Evers, 2008). Movement through the stages is not always linear and it is common for individuals to relapse to earlier stages throughout the change process (Prochaska et al., 2008). Clinically, interventions to change behavior can be tailored and matched to stage of change, which has been shown to be effective across a range of different health behaviors (Krebs, Prochaska, & Rossi, 2010;Noar, Benac, & Harris, 2007;Prochaska et al., 2008;Velicer, Brick, Fava, & Prochaska, 2013).
A second construct of the TTM is Self-efficacy, which conceptualizes a person's perceived ability to perform a task as a mediator of performance on future tasks (Bandura, 1977). In the context of the TTM, this construct describes confidence individuals have to cope with situations that might be considered high risk for relapse. (Velicer, Diclemente, Rossi, & Prochaska, 1990). As one might imagine, self-efficacy generally increases as people move through the stages of change. Cross sectional studies have observed that people in Precontemplation have relatively lower selfefficacy that those in the later stages of Action and Maintenance (Prochaska, DiClemente, Velicer, Ginpil, & Norcross, 1985;Velicer et al., 1990) Based originally on the decision-making model of Janis and Mann (Janis & Mann, 1977), the Decisional Balance construct captures the relative weighing of pros (benefits) and cons of changing (Velicer, DiClemente, Prochaska, & Brandenburg, 1985). Decisional balance patterns vary with the stages of change and has been useful in predicting movement through the stages (Prochaska, 1994;Prochaska et al., 1994;Velicer et al., 1985). The cross-sectional relationship between the stages of change and the pros and cons typically shows a pattern with cons being greater than Pros in PC, tied in C, and Pros increasingly higher than Cons for PR, A, and then M. From PC to A, the pros increase 1 SD while from C to A the cons decrease by one half of a SD (Hall & Rossi, 2008;Prochaska, 1994;Prochaska et al., 1994).
The final core TTM construct is the processes of change. Process of change differs from the stages of change in that the stages describe shifts in the intent to change, while the processes of change are independent variables that describe how people implement progress from one stage to the next (Prochaska & Velicer, 1997).
The variables are covert and overt strategies and techniques people use to alter their experiences and environment to progress through the stages of change (Prochaska, Velicer, DiClemente, & Fava, 1988;Prochaska, Velicer, Guadagnoli, Rossi, & DiClemente, 1991). The TTM theorizes that there are ten processes of change, which are typically divided into the higher order constructs of experiential (5 processes) and behavioral (5 processes) (Prochaska et al., 1988). People who have been successful in changing behavior have been shown to utilize different processes at each individual stage of change (Prochaska et al., 1991).

Aims
There are no measures based on the TTM for the constructs of Stage of Change, Decisional Balance or Self-Efficacy for patient readiness to utilize retail health clinics. Using the TTM as a guide, this study conducted a survey to assess patients' readiness to utilize retail health clinics, including measures of core TTM constructs. Specifically, the aim was to develop TTM measures for the Stage of Change, Decisional Balance and Self-Efficacy for patients' readiness to utilize retail health clinics. The processes of change were not developed or included in this study due to concern about the amount of time participants may be willing to spend on the survey.
It was hypothesized that the Decisional Balance and Self-efficacy measures developed in this study would be structurally similar to other TTM measures. It was further hypothesized that the measures would vary across the Stages of Change in patterns predicted by the TTM. That is, the Pros and Cons would show typical patterns across the Stages of Change as seen in previous TTM research. Self-Efficacy was also hypothesized to predictably show higher endorsement across the Stages of Change.
The development of valid and reliable TTM measures for retail health clinic utilization can aid future research into understanding what drives patients to these clinics and towards a better understanding of healthcare consumerism in a consistently evolving healthcare environment.

Item Development
The preliminary steps in development of the measures began with defining the constructs for this application followed by the generation of a large pool of items for potential inclusion in the final scale (DeVellis, 2012). The current literature on the TTM and retail health clinic utilization in addition to previous TTM scales were used to develop the initial items for Stage of Change, Decisional Balance, and Self-Efficacy. Items were refined in consultation with experts in TTM scale development and edited for clarity based on focus group testing. The main objective of this step was to develop clear items that were also as concise as possible while accurately reflecting constructs. Other considerations included response format, scale length, and potential response bias (DeVellis, 2012;Noar, 2003;Redding et al., 2006).
An algorithm was determined to be the best way to assess Stage of Change.
Multiple versions of the algorithm were created utilizing the current literature on both healthcare utilization as well as the limited data on retail health clinic utilization. The final version (described below) was the result of multiple rounds of revisions in consultation with TTM experts. The items for Self-Efficacy and Decisional Balance were written with the goal of creating at least twice as many items as expected in the final scale (Comrey, 1988;DeVellis, 2012). All items for Self-Efficacy and Decisional Balance utilized Likert scales similar to previous TTM research.

Measures
Demographics: Single item assessment of age, gender, race, ethnicity, education level, and household income. Stage of Change: The TTM frames behavior change as a process that happens over time and across a series of stages referred to as the stages of change. These stages include Precontemplation (not ready), Contemplation (getting ready), Preparation (ready), Action (reached criteria for change) and Maintenance (criteria reached for 6 months) (Prochaska & DiClemente, 1983). An algorithm was used to stage participants in this study.
The nature of retail health clinic usage presents a unique challenge for the stages of change and there are currently no established criteria. The traditional usage of set time frames presents an issue due to health clinic usage being dependent on need. Thus, alternative criteria are needed. Americans are visiting a physician's office 3 times per year on average and it's estimated that 1 in 5 visits to a primary care office and 1 in 10 visits to an emergency department can be treated at retail clinics (Ashman, Hing, & Talwalkar, 2015;RAND Corperation, 2016). Given these data, it seems reasonable that a patient in the Action stage of retail health clinic utilization would have at least a single use in one calendar year. Patients with a history of utilization and plans for continued use would define Maintenance. For patients who have not used a retail health clinic in the past year, we would assess their intention to use one. If they planned to use one the next time they are in need they would be in Preparation, and if they did not intend to use one the next time, but open to using one in the future, they would be in the Contemplation stage. Patients showing no intention using a retail clinic at this time or in the future would be staged in Precontemplation.
Self-Efficacy: Self-Efficacy conceptualizes a person's perceived ability to perform a task as a mediator of performance on future tasks (Bandura, 1977). Measurement of self-efficacy focuses on the confidence one has to maintain a desired behavior change in situations that often lead to a return to previous behavior.
In this study, participants were asked to rate how confident they are that they would utilize a retail health clinic in certain situations. Responses were on a 5-point Likert scale including not at all confident, a little confident, moderately confident, very confident, or extremely confident. Items were developed from the existing literature relevant to the utilization of retail health clinics and TTM experts reviewed and refined the items prior to distribution to participants.
Decisional Balance: Based on the decision-making model of Janis and Mann (Janis & Mann, 1977), the decisional balance construct captures the relative weighing of pros and cons of changing . In this study, participants were asked how important specific issues are in their decision about whether or not to utilize a retail health clinic. Similar to the Self-Efficacy measure, items describing the pros and cons of utilizing a retail health clinic were developed based on existing literature on retail health clinics and subsequently reviewed and revised by TTM experts.
Retail Clinic Utilization: Single item assessment for number of visits; single item assessment for satisfaction with services; single item assessment noting the reason for their visit.
Medical Mistrust: Mistrust in healthcare is an important barrier to getting medical treatment (LaVeist et al., 2003). To assess this construct, we used The Medical Mistrust Index 2.1, which is a 7-item scale that uses Likert-type responses with the following response codes: "strongly disagree", "disagree" "agree", and "strongly agree" (Laveist, Isaac, & Williams, 2009). Items have a range of 1-4 and the range of the total score is 7-28. There is moderate agreement in the field that a sample of 300-500 is sufficient for measure development as it allows the sample to be randomly split in sufficiently large halves for exploratory (N=150) and confirmatory (N=150) samples (DeVellis, 2012;Noar, 2003). Given this, our goal was to recruit no less than 300 participants and our budget ultimately allowed for the recruitment of between 500 and 600 participants. The survey was distributed by Cint to a community sample and was accessible on PCs, laptops, tablets, and mobile phones. Cint also managed incentives for participants through their incentive points program and estimated that each participant's incentive was equal to less than $3.

Analyses
Multiple steps were conducted for the analysis and development of the TTM measures for retail health clinic utilization. First, the sample was randomly divided into two samples (exploratory and confirmatory) to allow for psychometric analyses.
Initial descriptive statistics were assessed in the exploratory half of the sample to understand the normality of the data. Next, we tested and confirmed the best fitting structural model for both the Self-Efficacy and Decisional Balance scales. The final step evaluated the hypothesized relationships between the scales and the Stages of Change using the entire sample.

Exploratory Analyses
After randomly dividing the sample, initial descriptive statistics were assessed in the exploratory half to understand the normality of the data. Next, item means, standard deviations, and frequencies were evaluated in the Decisional Balance and Self-Efficacy scales (Redding et al., 2006). This process was used to assist in the identification and removal of items that reduced alpha or did not discriminate well among participants.
Following the initial item analysis, the remaining items were entered into a principal component analysis (PCA) to determine the number of factors measured by each scale. Based on previous TTM research, decisional balance factors are expected to be orthogonal, suggesting the use of varimax rotation for that scale (Hall & Rossi, 2008;Harlow, 2014)To determine the final number of factors to be retained we employed a Parallel Analysis method (Horn, 1965;Lautenschlager, 1989)as well as Minimum Average Partial (MAP) (Zwick & Velicer, 1986). Factor loadings in the retained items were analyzed and those with loadings of less than .40 or that load greater than .40 on more than one factor were removed from the scale (Redding et al., 2006). This process was done in stages with one item removed at a time and both the PCA and item-level analysis were repeated to assess the new distribution of variance (Widaman & Floyd, 1995). Cronbach's coefficient Alpha was used to test the internal consistency reliability of each factor (Cronbach, 1951). Additional items were removed to avoid redundancy and create the shortest possible scale while maintaining statistical integrity. The final step in this process was to run an exploratory CFA (Noar, 2003).

Confirmatory Analyses
Structural equation modeling using confirmatory factor analysis (CFA) was completed on the confirmatory half of the sample using the lavaan package for 'R' (Rosseel, 2012) for the final Decisional Balance and Self-Efficacy scales. Several fit indices were used to evaluate the CFA including Chi-square, Comparative Fit Index (CFI), Root Means Square Error of Approximation (RMSEA) and Standardized Root Mean Square Residual (SRMR). If the models appeared to be a good fit based on these indices, coefficient alpha, factor loadings, and effect size estimates were evaluated as well as how well the models fit the theoretical predictions (Noar, 2003).
Also known as the Bentler Comparative Fit Index, CFI ranges from 0 to 1 is useful in evaluating the fit of a model with values closer to 1 indicating a better fit (e.g., .93 is acceptable, .95 is a great fit) (Bentler, 1990). Both RMSEA and SRMR also range from 0-1, but unlike CFI, values closer zero indicate a better fit.
Specifically, RMSEA values of .05 or less are considered a good fit, while values of .1 or greater are considered a poor fit (Bentler, 1990). For SRMR, a value less than .08 is generally considered a good fit (Hu & Bentler, 1999). Chi-square was utilized to evaluate the models with non-significant findings signaling an acceptable fit because the predicted covariance matrix does not differ from the observed. Chi-square will also be used to assess the differences between the correlated and uncorrelated models of the decisional balance scale.

External Validation
Expert reviewers and a detailed review of extant literature on retail health clinics were critical in developing the scales to ensure the scales were built on face and content validity. The process of replicating the factor from the exploratory sample with the confirmatory sample was used to demonstrate construct validity.
Decisional Balance Scale: The Decisional Balance scale exploratory factor loadings and final items are shown in Table 5. The initial decisional balance scale included a total of 19 items, 9 representing the Pros and 10 representing the Cons.  Note. Exploratory alpha α = .85.

Confirmatory Procedure
With the exploratory procedures completed, we sought to replicate the findings with the confirmatory half of the sample as means to cross-validate the factor structures. Only subjects with complete data were used for this procedure (n=236).
Decisional Balance Models. The two-factor correlated model including items and factor loadings is shown in Figure 1. Fit indices for the three comparison models can be viewed in table 7. Based on previous TTM research, we tested 3 models for the decisional balance scale: (1) null model, (2) two-factor correlated model, (3) twofactor uncorrelated model (Hall & Rossi, 2008;Prochaska, 1994 Factor loadings ranged from .63 to .88 and the internal consistency was good for both the Pros (α = .87) and Cons (α = .83). The two factors accounted for 70% of the total variance including 36% and 34% for the Pros and Cons respectively. The correlation between the pros and cons factors was .55.  Note: N=236, χ2 = chi square; df = degrees of freedom; AIC= Akaike's information criterion. *p<.001.
Self-Efficacy Models. The one-factor Self-efficacy model including items and factor loadings is shown in Figure 2. Fit indices for the comparison models are shown in Table 8. For the Self-Efficacy scale, we compared 2 models including the (1) null model and (2) the one-factor model based on previous TTM research. The 1-factor model was the best fit c 2 (5) = 5.406, p >.05, CFI=.999, SRMR= .019, RMSEA=.018.
Factor loadings were greater than .73 and coefficient alpha was α = .80.   Note. SE= Self-efficacy Figure 3. Pros, Cons, and Self-Efficacy T-scores by Stage of Change

Differences between Precontemplators and Maintainers
To further explore possible differences between those who utilized retail health clinics and those who did not, we evaluated various participant characteristics for differences between Precontemplators and Maintainers. We chose to focus on the most extreme Stages of Change for these comparisons as a means to most easily identify differences between those utilizing and not utilizing retail health clinics. That is, Precontemplators represent the portion of the sample who have not utilized a retail clinic in the past year and do not plan to, while Maintainers represent those who have utilized a retail health clinic in the past year and plan to again the next time they need an available service. We conducted t-tests for continuous variables, Mann-Whitney U tests for ordinal, and chi-square for categorical. For these analyses, we used the full sample and included all participants staged in either Precontemplation (n=127) or Maintenance (n=184). Results are presented in Table 10.

Perceptions of mental health screening and treatment in retail health clinics
Patients were asked how likely they would be to utilize a retail health clinic for mental health screening and mental health services to gauge the acceptability and likelihood that patients would utilize retail health clinics for these services if offered.
Results are displayed in Table 11 and broken into 3 categories, those with negative PHQ2 and GAD2 screeners, those with a positive screen on either or both, and the full sample. The application of the TTM to retail health clinic utilization is novel in a number of ways. The TTM has largely been applied to health behavior change (Prochaska & Velicer, 1997), but has also been applied to more broad areas including provider populations (Blaney et al., 2018;Park et al., 2003) and consumer education (Xiao et al., 2004). The direct application to healthcare consumerism is novel and especially unique as applied to retail health clinic utilization. Unlike traditional applications to health behavior change, it's possible that consumers may not be aware of the Pros of Cons of retail health clinic utilization unless they have utilized them and personally experienced them. This is especially true when compared to health behavior areas like smoking cessation and increased exercise that have widely understood and accepted health benefits, regardless of one's experience with them. In fact, as a new addition to the healthcare marketplace, it's likely that there are pros and cons of utilization that are yet to be considered or even discovered. Another unique aspect of the application of the TTM for retail clinic utilization is that unlike health behavior change that can be initiated at any time once a person is ready, healthcare is something only sought when there is a need.

Demographics.
The utilization of a survey company for the recruitment of participants for this study allowed us to recruit nationally and target areas with known retail health clinic availability. This process ensured that the sample included participants who at least had the option of going to a retail clinic given the primary goal of measure development. The alternative, to recruit a general population sample, would have run the likely risk of including a high number of participants who would not have access to retail clinics or even know what they are. Indeed, 87.7% of the sample reported that they lived within 30 minutes of their nearest retail clinic. Because retail health clinics currently tend to be clustered in metropolitan areas, the sample is weighted to metropolitan statistical areas, which includes the metro area and surrounding suburbs.
Unfortunately, the inclusion of more rural populations was not feasible for this study due to a lack of retail health clinics in those areas.
The average age of the sample (45.8, sd=16.7) and the distributions of gender (48.8% female) and race were representative of a population sample based on the 2010-2015 American Community Survey (U.S. Census, 2016). Various education levels were broadly represented, ranging from less than high school to doctoral degrees. Income levels were generally distributed on a bell curve centered on $50,000 -$74,999/year with a slight right-skew do to 18.6% of the sample reporting an income of less than $20,000/year. Some of this may be explained by nearly 15% of the sample being of retirement age as well as the inclusion of current college students. Overall, our sample selected from a range of metropolitan statistical areas was largely representative of the general US population on demographic variables.

Health and Retail Clinic Related Variables.
Data on retail health clinic utilization is limited, with the majority of research focused on those already using retail health clinics making it difficult to know how our sample performs in terms of rates of use. Of those who have utilized a retail clinic at least once (69.5%), the average number of lifetime visits was 4.26 (sd = 12.24).
However, this distribution was highly skewed with a median number of visits of two.
Only a quarter of the sample reported utilizing a retail health clinic more than 6 times in their lifetime to date. This may be explained by the nature of retail clinics being a service often used when primary providers are unavailable. Retail clinics are also relatively new additions to the healthcare marketplace and we may hypothesize that lifetime utilization rates will increase as they become more established and people accrue more years of utilization.
Levels of chronic health conditions, smoking, and a range of BMIs were broadly represented in the sample. Interestingly, 82.6% of the sample reported taking at least one prescription medication and 40.6% reported taking 3 or more prescription medications, which is substantially higher than the 48.9% and 23.1% respectively reported by the CDC in 2016 based on data obtained from 2011-2014 (National Center for Health Statistics, 2017). It is unclear what accounts for this difference but may signal physical proximity to pharmacies and medical care (i.e. metropolitan sample) are related to the number of prescription medications a person takes. The high level of prescription medication use also provides added support for pharmacy-based retail clinics as regular healthcare points of contact for many individuals and highlights the potential for assessment and treatment of some population behavioral medicine needs in these settings (e.g. smoking cessation, chronic disease management, weight loss programs, exercise interventions, routine screenings, mental health screening, etc.).
Our sample was largely covered by insurance with 87% reporting that they had health insurance. Of those with insurance, 24.5% reported no deductible, 51.5% reported a deductible, and a surprising 24% reporting either not knowing if they had a deductible or the amount of the deductible if they had one. While a full understanding of the role insurance coverage plays in healthcare consumer decisions with regard to retail health clinics is outside the scope of this paper, future analyses of these data may provide additional insight in this area. Also noted is that 82.5% of the sample reported having a regular primary care provider.

Stage of Change
As previous described, this is the first study to apply the TTM to retail health clinic utilization, which is a novel application. While identifying patients who were not using retail clinics (Precontemplation) and those who reported using them regularly when needed (Maintenance) was intuitive, the intermediate stages were more difficult to conceptualize and discriminate between. The final algorithm developed following multiple consultations with TTM experts, resulted in a relatively good distribution across all stages. The majority of the sample were staged into Precontemplation and Maintenance representing 24.4% and 35.3% respectively, with Action being the least represented at just 5.8%. The staging for Action was difficult given that traditionally, this stage is defined within a time-frame (i.e. has made change for less < 6 months). As an alternative, we chose to ask about "the next time you need services provided by a retail clinic". Further research into staging for retail health clinics would be beneficial to test alternative algorithms, however, we believe the current algorithm largely captured the construct given our results.

Decisional Balance
This study replicated previous TTM research in demonstrating a two-factor Decisional Balance model representing the Pros and Cons of behavior change.
However, unlike previously validated TTM measures, we did not find the expected Maintenance. These differences also proved to be significant during the external validation MANOVA analysis.
There are several hypotheses to account for this. First, it's possible that those not utilizing retail clinics regularly have simply not experienced or may not even be aware of the Cons of utilization. For example, one concern for retail clinic utilization is the potential for poor communication between the clinic and a patient's regular provider. This may not seem important to someone who has never used a retail clinic, but may become very important for someone who has utilized them and encountered an issue related to information not being adequately communicated to their primary provider. Thus, as people utilize retail clinics more, they also increase their exposure to the negative aspects of retail clinic care. This stands in contrast to common health behaviors like smoking, where most smokers can readily identify the Cons of quitting without having to quit first to recognize them.
Second, this is a potential signal that there is on-going ambivalence among those utilizing retail clinics and might predict that people will not continue to use them. It's possible that some are using them only when their primary providers are unavailable. Thus, they may acknowledge and experience the Cons, but feel that the alternative of either waiting to receive treatment or to present at more expensive options (e.g. urgent care, emergency room, etc.) is less favorable. While the healthcare marketplace is constantly evolving, the current model of retail clinics is not to fully replace primary care providers, but rather offer a situationally more convenient option.
We might assume that these data suggest that people remain connected to their primary providers but are willing to accept the cons of retail clinic utilization in exchange for convenience in certain situations. focused on understanding why people utilize these clinics. As a result, there is little existing patient-level data describing why they aren't utilizing retail clinics. The majority of this previous data comes from industry insiders, providers, and policy makers, who may have different concerns than a healthcare consumer. For example, a consistent concern expressed in the literature by these stakeholders is the potential break in the continuum of care due to the systems implications. However, patients may not share the same concerns unless they have experienced a specific issue related to the continuum of care. These findings suggest that further research is needed to better understand the Cons of utilizing retail health clinics for patients. Qualitative studies addressing this may be of particular interest.

Self-Efficacy
This study replicated previous TTM research in demonstrating a one-factor Self-Efficacy model for retail health clinic utilization. The results also replicated the underlying structure found in previous TTM self-efficacy measure development studies (Velicer et al., 1990). Self-Efficacy generally varied across the stages as expected, consistent with previous TTM research Velicer et al., 1990). Patient's confidence to utilize retail health clinics was lowest for Precontemplators and increased through Contemplation to Preparation. There was a slight decrease in SE between Preparation (T-score = 51.65) and Action (T-score = 51.22), before reaching the highest levels in Maintenance (T-score = 53.51).

It is unclear what may explain the slight reduction in confidence between
Preparation and Action, although it may be the result of the staging algorithm. The question used for Action was to ask those who have used a retail clinic in the past year if they plan to use one "the next time they need services offered by a retail health clinic". The wording of this question may have unintentionally captured patients who used a retail clinic, but do not plan to use them going forward. Thus, some participants staged in Action, may have actually relapsed into earlier stages. However, the Pros scale did not find the same dip in the Action stage, which might be expected if this was the case. Participants in the Action stage constituted the smallest group in the analyses (n=26), increasing the likelihood that a small number of patients with lower SE scores (possibly those who relapsed to earlier stages) may have pulled down the average of the group.

Differences between Precontemplators and Maintainers
Exploring the differences between participants in the most extreme stages, Precontemplation and Maintenance may help to further identify what might impact the decision to utilize a retail clinic. Based on our results, Maintainers reported significantly more medical provider visits than Precontemplators and were significantly more likely to have insurance and a regular primary care provider. They also differed significantly in that Maintainers reported higher levels of education and income. Taken together, these findings suggest that these people may have better access to care and are more likely to utilize the care available to them. Interestingly, we did not find a difference on the medical mistrust scale and while there was a significant difference in the distance from retail health clinics, Maintainers actually reported being slightly further from their closest retail clinic, not closer.
When we examine the rates of common health conditions, Maintainers did not appear to be more "ill" than their counterparts in Precontemplation. In fact, of the eight health condition categories examined, prevalence rates were higher among Precontemplators for five of them including cancer, cardiac, arthritis, high cholesterol, and high blood pressure. This may partially be explained by an age discrepancy with Precontemplators having an average age of 52.49 and Maintainers significantly younger at 41.07 years of age, a trend consistent with previous research on retail health clinic utilization (Ashwood et al., 2011;Leppel, 2010;RAND Corperation, 2016). It's likely that some of the discrepancy in the prevalence of these health conditions is attributable to the Precontemplators being older and more likely to experience higher rates of these conditions. These data suggest a potential area for expansion of services into mental health screening and possibly mental health treatment. Integrating mental health treatment into primary care settings has received a lot of attention in recent decades and has been a goal for many primary care practices. Co-locating mental health services with other services have been shown to increase referral rates, reduced wait-times for appointments, and a reduction in the stigma associated with seeking mental health services from a specialty provider (Bartels et al., 2004;Blaney et al., 2018;Clement et al., 2015;Durbin et al., 2012;Hampton-Robb, Qualls, & Compton, 2003). Retail clinics offer a point of contact for mental health screening and treatment that is easily accessible.
Physical space is one of the largest barriers to integrating care, which may also be an issue for retail clinics, given their small physical space housed in retail settings.
However, advances in telemedicine options may be one way to provide these services without an onsite mental health provider. Some retail clinics are already set up for medical telemedicine visits with a large video screen in a small private room and medical devices such as a stethoscope and otoscope for the patient to use on themselves with the direction of a live physician on the screen. It seems reasonable to believe that if medical appointments can be conducted remotely via such technology, that there is potential to conduct mental health treatment, which requires no physical contact, in a similar manner. Future research should anticipate the potential for this expansion.

Limitations and Future Directions
Some limitations of this study should be noted. First, the data for the study was cross-sectional and future research would benefit from longitudinal data to evaluate change over time. Second, the development of these measures was largely built on limited research, that mostly focused on provider and systems level data. Many of the items, especially for the Pros and Cons scales, were not entirely based on data directly from patients or in some cases, even the perspective of a patient. Third, as previously discussed, this unique application of the TTM required adjustments to the common TTM staging algorithms that may require further refinement. Fourth, the Processes of Change (POC) were not included in this study due to concern about response burden on participants. Future studies should consider the development of a POC measure to further explore the covert and overt behaviors required to move through the Stages of Change. Taken together, this study should be viewed as an initial step in gaining a better understanding of how the TTM can be applied retail health clinic utilization and healthcare consumerism.

Summary
Overall, this study supports the application of the TTM to retail health clinic utilization and the initial development of specific TTM measures for Self-Efficacy and Decisional Balance. As retail health clinics continue to grow in numbers and expand in scope, learning about patients' perceptions about them, including their benefits and costs, will be vital information not only to these clinics and their operators, but to the healthcare system as a whole. While there remains a lot of debate in the healthcare field about the risks of the addition of these retail clinics, they are here to stay, and providers may benefit from understanding which of their patients are more likely to utilize them and why. The TTM provides one possible framework to assist in that understanding. Future research can expand on the application of the TTM to retail health clinic utilization to assist in this understanding.

James Prochaska, PhD Department of Psychology Transtheoretical Model Development for Retail Health Clinic Utilization
Thank you for your interest in participating in this study! Please read the following informed consent document before proceeding. You are being asked to take part in a research study. The purpose of the research study is to better understand the utilization of retail health clinics. Please read the following before agreeing to be in the study. If you agree to be in this study, it will take you approximately 20 minutes to complete this survey. Questions will be asked about your health, your healthcare providers, and attitudes about healthcare delivery. There are no known risks, benefits or compensation provided by the investigators or the University of Rhode Island. You may receive compensation in accordance with your agreement with CINT. Your responses will be strictly anonymous. The responses may be used in research papers and related presentations (e.g. posters, talks, etc). The decision to participate in this study is entirely up to you. You may refuse to take part in the study at any time without affecting your relationship with the investigators of this study or the University of Rhode Island (URI). Your decision will not result in any loss of benefits to which you are otherwise entitled. You have the right not to answer any single question, as well as to withdraw completely from the survey at any point during the process; additionally, you have the right to request that the researchers not use any of your responses. You have the right to ask questions about this research study and to have those questions answered by me before, during or after the research.