Designing for Clinical Change: Creating an Intervention to Implement New Statin Guidelines in a Primary Care Clinic

Background Recent clinical practice guidelines from major national organizations, including a joint United States Department of Veterans Affairs (VA) and Department of Defense (DoD) committee, have substantially changed recommendations for the use of the cholesterol-lowering statin medications after years of relative stability. Because statin medications are among the most commonly prescribed treatments in the United States, any change in their use may have significant implications for patients and providers alike. Prior research has shown that effective implementation interventions should be both user centered and specifically chosen to address identified barriers. Objective The objectives of this study were to identify potential determinants of provider uptake of the new statin guidelines and to use that information to tailor a coordinated and streamlined local quality improvement intervention focused on prescribing appropriate statins. Methods We employed user-centered design principles to guide the development and testing of a multicomponent guideline implementation intervention to improve statin prescribing. This paper describes the intervention development process whereby semistructured qualitative interviews with providers were conducted to (1) illuminate the knowledge, attitudes, and behaviors of providers and (2) elicit feedback on intervention prototypes developed to align with and support the use of the VA/DoD guidelines. Our aim was to use this information to design a local quality improvement intervention focused on statin prescribing that was tailored to the needs of primary care providers at our facility. Cabana’s Clinical Practice Guidelines Framework for Improvement and Nielsen’s Usability Heuristics were used to guide the analysis of data obtained in the intervention development process. Results Semistructured qualitative interviews were conducted with 15 primary care Patient Aligned Care Team professionals (13 physicians and 2 clinical pharmacists) at a single VA medical center. Findings highlight that providers were generally comfortable with the paradigm shift to risk-based guidelines but less clear on the need for the VA/DoD guidelines in specific. Providers preferred a clinical decision support tool that helped them calculate patient risk and guide their care without limiting autonomy. They were less comfortable with risk communication and performance measurement systems that do not account for shared decision making. When possible, we incorporated their recommendations into the intervention. Conclusions By combining qualitative methods and user-centered design principles, we could inform the design of a multicomponent guideline implementation intervention to better address the needs and preferences of providers, including clear and direct language, logical decision prompts with an option to dismiss a clinical decision support tool, and logical ordering of feedback information. Additionally, this process allowed us to identify future design considerations for quality improvement interventions.


Background
There has been a dramatic shift with respect to how guidelines recommend that American medical providers should prescribe commonly used cholesterol-lowering statin drugs [1]. In 2013 and 2014, the American College of Cardiology (ACC) and American Heart Association (AHA) and the United States Department of Veterans Affairs (VA) and Department of Defense (DoD) released new clinical practice guidelines on the treatment of blood cholesterol to reduce cardiovascular risk in adults [2][3][4]. Where previous guidelines had focused on giving increasing doses of statins until a patient's cholesterol level dropped below a specific target, both new guidelines recommend fixed doses of medicine based on the patient's atherosclerotic cardiovascular disease (ASCVD) risk, the chance that the patient will develop cardiovascular disease (CVD) [2,[5][6][7][8][9].
The new guidelines present challenges to adoption. First, moving away from cholesterol target-based treatment models represents a conceptual change in clinical practice. Second, in many cases, risk-based guidelines would require providers to calculate a patient's ASCVD risk, which could substantially alter a provider's workflow. Third, the discrepancies between the VA/DoD and ACC/AHA guidelines could cause confusion [10].
Effective implementation of new guidelines should recognize the existing barriers to adoption [11], including providers' knowledge, attitudes, and behaviors about those guidelines [12]. They must also address those barriers in ways that are effective, accurate, and user centered. Existing strategies, including provider education, clinical decision support, and audit and feedback must address the barriers and the providers' needs [12]. This requires a strong framework for designing an intervention and for making sure the intervention is effective.
Effective interventions fit the needs of the end users. To this end, user-centered design focuses on understanding the physiological, cognitive, and social aspects of the intended user that could alter how someone will use a tool or system [13]. In a health care setting, user-centered design can be employed to create or adapt tools that are consistent with the physiological, cognitive, and social needs of providers to address challenges to adoption and increase the likelihood of their use.

Objectives
In this study, we developed and tested a multicomponent guideline implementation intervention (hereafter referred to as the intervention) to improve statin prescribing. Our intervention was developed with semistructured qualitative interviews, an established theoretical framework, and principles of user-centered design. This paper describes the intervention development process with providers, which was conducted to (1) illuminate knowledge, attitudes, and behaviors and (2) elicit feedback on intervention prototypes developed to align with and support the use of the VA/DoD guidelines. Our aim was to use this information to design a local quality improvement intervention focused on statin prescribing that was tailored to the needs of primary care providers.

Intervention Background and Development
We sought to develop and test a multicomponent guideline implementation intervention to improve statin prescribing. The processes were designed to find new, but practical, components for the intervention and help us improve the processes that we already had planned based on the literature and existing practices. For example, research demonstrates that a multicomponent intervention is often more effective than a single approach alone [14].
In the following section, we describe the components of the intervention: educational program, clinical decision support tool, and performance measurement with audit and feedback. Paper-based prototypes were created as working models to be tested for acceptability before investing in computerized systems. The prototypes were modeled after currently existing tools utilized in the VA health system to facilitate providers' ability to imagine how the prototypes would function in their current workflow. In our user-centered design process, we asked providers to identify their needs and preferences specific to the 4 prototypes described below.

Educational Program
In collaboration with providers, we developed an educational program about the new clinical practice guidelines to be delivered to all providers before the intervention began. The educational program lasted 15 min. It included a summary of the guidelines and 3 cases that demonstrated the differences between the new guidelines and the old. We also developed a single-page tool describing and comparing the VA/DoD and the ACC/AHA Clinical Practice Guidelines (see Multimedia Appendix 1). The single-page tool was designed to be a concise and convenient reminder of changes to the statin guidelines.

Clinical Decision Support Tool
The clinical decision support tool was designed to address 2 predicted quality gaps-the traditional reminder role of pop-ups and a need to make it easier to follow the guidelines. The new guidelines require providers to calculate the risk of ASCVD for some patients using an algorithm that incorporates risk factors (eg, age, sex, smoking, high-density lipoprotein cholesterol, total cholesterol, systolic blood pressure), which significantly complicates use of the guidelines. In our facility, an ASCVD risk algorithm has not yet been incorporated into the electronic medical record, and providers access ASCVD risk calculators through external websites. The paper-based prototype resembled the existing clinical decision support alerts, thus meeting the reminder role and automatic calculations of ASCVD risk of computerized clinical decision support (see Multimedia Appendix 2). The impact of not having the tool in the electronic health record was evaluated in our interviews.

Performance Measurement With Audit and Feedback
Traditionally, performance measurement is used in pay-for-performance programs within the VA. We worked with the VA's Center for Analytics and Reporting to create a novel performance measure that is aligned with the VA/DoD guidelines (hereafter referred to as the VA proposed performance measure; see Multimedia Appendix 3). In the VA proposed performance measure, providers would have stronger incentives for patients for whom treatment is more likely to be clinically important using a weighted point measurement system to create risk categorization of patient groups. For example, the VA proposed performance measure would award providers different points for prescribing a moderate dose statin to a patient with clinical ASCVD (5 points), a patient with diabetes (3 points), and a patient with a 10-year ASCVD risk greater than 12% (1 point). In distinction, other performance measures, such as those of the Healthcare Effectiveness Data Information Set, do not incorporate risk prediction in patient treatment recommendations. The weighted VA proposed performance measure was designed to emphasize prevention through risk calculation.
We designed an audit and feedback report template (see Multimedia Appendix 4), wherein providers would be informed of their individual performance on the VA proposed performance measure. The template includes 2 provider performance reports. The first includes breakdown of provider performance by patient risk categorization (eg, history of ASCVD; diabetes; low-density lipoprotein, LDL >190; high risk; low risk). The second displays performance by overall statin use across patients. Similar to the VA proposed performance measure, the audit and feedback report features risk prioritization.

Setting and Participants
This local quality improvement project was conducted in primary care at a single Veterans Affairs Medical Center (VAMC) between late October 2015 and June 2016. In total, 37 professionals across 5 Patient Aligned Care Teams (PACTs)-including 32 physicians with their own patient panel at the start of the project and 5 pharmacists-were invited to participate in qualitative interviews via email. Though 37 professionals were invited, data collection was designed to continue until thematic saturation was reached [15][16][17]. Invitees were presented with a project information sheet at the time of initial email contact, which was reviewed at the time of the interview.

Ethical Considerations
Veterans Health Administration (VHA) Handbook 1058_05 [18] provides guidance about authorization of manuscripts that have been developed through nonresearch activities (ie, without institutional review board approval under the authority of VHA operations). All VHA authors of this manuscript attest that the activities that resulted in producing this manuscript were not conducted as part of a research project but as part of the nonresearch evaluation conducted under the authority of the VA's Quality Enhancement Research Initiative.

Data Collection
We conducted semistructured interviews with providers to guide the development and testing of a multicomponent guideline implementation intervention. A qualitative approach was selected to explore user knowledge, attitudes, and behavior to improve the adaptation and implementation of the intervention. Interviews were audiotaped, transcribed verbatim, and lasted an average of 49 min. One member of the research team (CR) conducted all interviews while a research assistant took notes. Providers were not compensated for their time, and participation was completely voluntary. We began by eliciting feedback on determinants of providers' guideline uptake. We then engaged providers in a user-centered design process to examine and improve prototypes for the 4 components of the intervention.

Analysis
After a review of the literature, we determined that existing frameworks could be used to understand implementation of clinical guidelines in our setting. We used in-depth qualitative research principles structured by the Clinical Practice Guidelines Framework for Improvement [12] to guide our understanding of the barriers and facilitators to use and Nielsen's Usability Heuristics [19] to guide the user-centeredness of our development process. The Clinical Practice Guidelines Framework for Improvement examines individual-level factors (knowledge, attitudes, and behaviors) of providers [20], whereas Nielsen focuses on elements of user-centeredness and design.
Using an initial codebook based on constructs from the Clinical Practice Guidelines Framework for Improvement and Nielsen's Usability Heuristics (see Table 1), we (JS, CR, and MD) used a deductive approach to apply descriptive codes to 3 transcripts and modified our codes based on the data. We then applied codes from the modified codebook to 3 more transcripts and discussed our codes to determine consensus. The remaining 9 transcripts were coded by 1 team member (MD). We used QSR International's NVivo version 11 data analysis software to apply codes to segments of text and to create code reports that grouped all text sharing the same code. Code reports were then summarized independently by the project team members (JS, CR, MD) and discussed to reach shared understanding of themes.
Project team members (JS, CR, CK, and BY) discussed interview notes in team meetings. Following a user-centered design approach, the team discussed provider needs related to the intervention and made changes to the wording and format of the prototypes as interviews progressed. When there was misalignment between providers' preferences and design decisions, we used an adapted consensus process [21] to decide which suggested modifications were feasible.

Participant Description
In total, 15 individuals-13 physicians and 2 clinical pharmacists-participated in interviews and represented all 5 PACT teams. In total, 9 providers did not respond to 3 email invitations and 13 declined to participate. There were no observable differences in gender, age, or participation in the educational seminar, between those we interviewed and those who did not participate.

Summary of Findings
Overall, providers were generally comfortable with the paradigm shift to risk-based guidelines but less clear on the need for the VA/DoD guidelines in specific. They preferred tools that helped them provide the care they wanted to provide without limiting their autonomy (see Table 2 for abbreviated list of changes made in response to interviews; see Multimedia Appendix 5 for detailed list of user-centered design changes).

Providers Accept the Paradigm Shift in Cholesterol Treatment but Some Question the Need for Separate Department of Veterans Affairs/Department of Defense (VA/DoD) Statin Guidelines
Most providers felt the risk paradigm was more closely aligned to their clinical perspective: We've moved away from focusing on LDL, this one just seems more compelling…here's the person's risk, it just seems more informative and like a compelling reason to treat. [Participant #13] Others highlighted the benefit of providing patients with more precise, tailored risk estimates using risk-based guidelines.
One core distinction between the guidelines is that the VA/DoD guidelines are generally less aggressive than the ACC/AHA guidelines: they recommend treatment for fewer people, permit use of less-intense statin regimens, and create a gray zone where treatment is neither recommended for nor against. A few providers stated their preference for the VA/DoD guidelines and felt the ACC/AHA guidelines encouraged overtreatment. One provider explained: A few developed approaches that incorporated aspects of both sets of guidelines, such as one who appreciated the deemphasis of routine cholesterol monitoring in the VA/DoD guidelines but preferred the risk cut points established in the ACC/AHA guidelines.
Due to their patient population, a few providers noted that the differences between the 2 guidelines would likely have a very small impact:

Within the Risk Paradigm, Providers Are Not Confident in How to Deal With Shifting Risk
Providers were generally comfortable with the role of risk prediction in the guidelines. They did express some confusion about how to address changing risk factors and the lack of consistency of risk prediction. For these problems, they felt that the guidelines were not responsive. One provider explained:

Providers Desire Clinical Decision Support Tools That Allow for Cognitive Ease of Use and Speed
Providers' interest in having a clinical decision support tool during the patient encounter was based around efficiency: If the reminder already calculated the risk, I'd love that. I hate having to go to the internet, or look on my smartphone, so I think the ideal reminder would calculate the risk for you. [Participant #6] A few providers indicated that the clinical decision support tool may be especially useful in patients whom the calculator estimates to be at high risk for ASCVD but have no history of heart attack or stroke:

Providers Want Clinical Decision Support Tools That Allow for Autonomy
When asked about the computerization of clinical decision support tools, most providers indicated a need for autonomy within the system, whereby providers can exit or cancel a clinical reminder when it is inappropriate or inaccurate for the particular visit or patient: Sometimes it seems like things come up that aren't supposed to, or they don't come up and they should…I think there's often circumstances where it's like, "How do you get out of this loop?" where this isn't right and it should go away, but you can't make it go away and so I like that there's an option for like, "This is wrong," and so you can get out of that.
[Participant #7] Providers said they generally appreciate being reminded when a patient is not meeting a guideline but want to be able to accurately state why the patient is not on a statin rather than bend the truth simply to disable the reminder.

Providers Want Clinical Decision Support Tools That Can Be Disabled
Providers wanted a clinical decision support tool that would not continue to alert after an issue has been addressed. However, there was some debate as to which clinical situations should lead to a reminder being disabled indefinitely and which would warrant a revisited conversation:

Some Providers Prefer Dichotomous Performance Measures, Whereas Others Prefer Performance Measures That Incorporate Risk Categorization of Patient Groups
The team proposed a new performance measure consistent with the new guidelines that would provide weighted performance assessment. In this system, patients for whom statin treatment was particularly likely to prevent a heart attack or stroke would be given more credit in evaluation. Providers had mixed feelings about the proposed VA performance measure, particularly the idea of weighting the performance measure to reflect risk categorization of patient groups (based on patients' ASCVD risk):

Providers Want More Credit for Shared Decision Making
Several providers were concerned with the lack of credit given for shared decision making in the proposed and existing VA performance measures. Most providers agreed that the high-risk patients, or the "no-brainers" as one provider put it, take less effort and time to convince to initiate and adhere to treatment because their risk is more palpable. Rather, it is the patients who: feel fine and they haven't had any negative outcomes yet [sometimes]

Providers Feel That Hierarchical Patient-Level Feedback Is Most Useful Within Audit and Feedback Reports
Providers regularly receive audit and feedback of their care within this clinic, usually in the form of printouts of tables of care provided. We attempted to understand how the new guidelines might alter the best way to provide audit and feedback. Providers generally preferred the audit and feedback report when broken down into component parts, indicating first, how the provider fares on each individual goal (ie, the percentage of the provider's patients with ASCVD that are on a moderate or high-dose statin) and second, broken down by patient fallout, with the highest risk patients listed first, and the lower-risk patients listed last. Providers indicated that listing out patients that did not meet the guideline by risk category would be more actionable than having a single list of patients not meeting guidelines, as members of the PACT team would then be better able to triage follow-up phone calls. As one provider explained: It does help you gauge again from the standpoint of, where do you least want to make mistakes, with the people that have significant disease already and if you had someone with very, very low risk taking a statin, it's not going to be the worst thing in the world. I mean, you're not happy about it, but I think that is important to see the breakdown. [Participant #9] Several individuals indicated that comparing providers by their percentile of measures met is not motivating, in part, because it can be difficult to distinguish who they are being compared with, whether it is providers at the local level, or providers at the system level. Another provider mentioned that delayed receipt of the report also decreases impact on provider behavior, stating:

Providers Would Value Audit and Feedback More if It Were Used to Help Their Care More Directly
Participants generally wanted performance measures and audit and feedback reports to be more tightly linked to ways to help the providers improve practice in response:

Principal Findings
We developed a system to create a multicomponent implementation intervention that was to be user-centered and evidence based. Our system helped us identify ways to improve aspects of the intervention and develop new ones (see Table 2). • Proposed response: Think of mechanism to account for complicated patients in performance measure and subsequently in the audit and feedback report In short, we found that providers were interested in changing their care but needed support in doing so.
Our team incorporated feasible design suggestions into the prototype intervention, particularly when there was general agreement among providers on a given design element and it aligned with design and user experience best practices. Consistent with previous research [22,23], providers overwhelmingly preferred simple information, clear and accurate decision prompts, and logical ordering of information that aligned with their values and needs, such as including highest risk patients first on audit and feedback report fallout lists. More specifically, providers wanted to be able to accurately and rapidly use clinical decision support tools during the patient encounter without any loss to their autonomy [24]. Many of these wording or formatting suggestions were addressed in the second iteration of the clinical decision support tool and audit and feedback report template.
Some providers found the shift to new guidelines difficult, even when the guidelines were more closely aligned to their clinical perspective. For example, providers also felt guidelines don't recognize the most difficult aspects of their work, particularly the time and resource demands of shared decision making and introducing the concept of risk, which is strongly emphasized in the new guidelines [25,26]. In addition, providers requested more evidence, education, and resources to make any clinical change. Educational and training resources for both providers and patients were thought to be essential in effective shared decision making and, as a result, adherence to statin guidelines. In response, we implemented an educational seminar during a primary care meeting whereby differences between the guidelines were highlighted by way of a pocket guide [27] and explained in detail before the commencement of the intervention phase of the project.
Our work adds to, but is supported by, existing research in implementation science on guideline implementation and how to change clinician habits. Our findings align well with our underlying framework, the Clinical Practice Guidelines Framework for Improvement [12]. As that framework and other research suggests, we found barriers and wide variation in providers' knowledge, attitudes, and behaviors about the new guidelines [12,28]. Previous work has also found that providers find guidelines and performance measures demotivating, especially when they are not user-centered or well-aligned with the providers' goals of care [29,30]. Similarly, decision support tools regularly impact patient care but details of usability also have large effects on provider satisfaction and uptake [31,32]. Our work is one of a few studies that have attempted to synthesize these diverse fields of research into a single intervention. Our findings were also unusual in noting the central divide between providers' desire for new guidelines for support and efficiency versus a sense that they are intended to remove providers' autonomy.

Limitations and Future Research
We sampled a small number of providers from one VAMC. Nonetheless, the providers who participated in the interviews for this project provided important insights that influenced both the type and content of the intervention later executed at this site. We expect our research design to be transferable to other sites, as user-centered design and qualitative methods both emphasize local context.
We were also limited in our ability to incorporate many of our findings into the intervention. At times, providers' opinions and preferences were at odds. Thus, our team needed to prioritize and rank feedback, accommodating feasible design suggestions with strong provider consensus, and vetoing design elements that were too provider-specific, acknowledging that a provider-specific interface is not feasible within the health system. Relatedly, there were requests for user-friendly features that were technologically infeasible. Consequently, we have identified future design considerations for each of the above domains (see Textbox 1) that were outside the scope of this project but could be considered in other projects.
Finally, the purpose of this study was to follow a user-centered design approach to capture the needs and preferences of providers in the final intervention design. Though beyond the scope of this study, future research should examine the effectiveness of similar multicomponent implementation interventions.

Conclusions
The guideline implementation planning process provided important insights about the refinement of the intervention plan. By combining qualitative methods and user-centered design principles, we could understand the needs and preferences of providers and modify prototypes to increase their acceptability and usability in practice. Our findings allowed us to target several factors providers reported as being important determinants to the uptake of and adherence to clinical practice guidelines. The qualitative process of working with providers also allowed us to identify future design considerations for multicomponent guideline implementation interventions.