Human-Centered Design Lessons for Implementation Science: Improving the Implementation of a Patient-Centered Care Intervention

Supplemental Digital Content is Available in the Text.


INTRODUCTION
Although today's public health response to HIV has a robust set of evidence-based tools with which to address the global epidemic [eg, antiretroviral therapy (ART), pre-exposure prophylaxis (PrEP), and voluntary medical male circumcision], implementation has failed to achieve the tools' full preventive and therapeutic potential. 1 Failures of implementation often result from inadequate fit between available innovations and the people, processes, and contexts in which they are delivered. Although an emergency-based response focusing on access largely drove HIV service delivery strategies over the past three decades, future success depends on more effectively engaging end-users with appropriate, desirable, and accessible services. 1,2 For example, although ART is lifesaving, both treatment initiation and retention remain suboptimal, leading to onward transmission, morbidities, and mortality. 3 These gaps demand alternative and innovative treatment delivery models-drawing from interdisciplinary perspectives-for advancing the public health response. 4,5 Human-centered design (HCD) is an emerging approach with roots in industrial design, engineering, psychology, anthropology, business, computer science, ergonomics, and design that hold promise for improving implementation of evidence-based interventions for health. HCD brings end-users and developers together to cocreate health products, services, or delivery strategies that identify, prioritize, and address barriers to usability. [6][7][8] Traditionally, HCD focused on product development using participatory activities emphasizing researcher and user interaction to improve intervention utility, uptake, sustainability, and effectiveness. [7][8][9] Although no single definition of HCD in health exists, 6 there are hallmarks present across HCD applications [7][8][9] (Table 1). HCD uses methods likely familiar to social scientists working in health 8,10-13 but emphasizes bringing the researchers and users together in a more empathetic way, generating breadth and flexibility in the investigation and prioritizes action over furthering scientific knowledge ( Table 1,  Tolley 8 ).
Despite a growing literature on HCD in health, 6 no synthetic appraisal of HCD in the public health HIV response yet exists. In this article, we seek to advance the conceptualization of the use of HCD to address HIV through both a literature review and a case study of our own experience using HCD methods to advance patient-centered care (PCC) in HIV treatment in Zambia. [14][15][16][17] The literature review summarizes the following: (1) outcomes to which HCD has been applied, (2) methods used, (3) results and effectiveness, and (4) lessons learned. We present HCD-derived case study insights that could be informative to others seeking to optimize the delivery of PCC HIV interventions in transferable contexts. In addition, we hope this article will call attention to opportunities to advance HCD as a tool for adapting implementation strategies to particular contexts and end-user populations to strengthen the public health response to HIV.

Literature Review
We conducted a narrative literature review 18,19 of published articles and the grey literature on HCD in the global HIV response. Article inclusion criteria are as follows: (1) published through the search date of May 22, 2019, (2) related to HIV, and (3) presented data on HCD. Articles were excluded if they did not describe a design process or if they explicitly attributed their methods to non-HCD methodology (see Appendix 1, Supplemental Digital Content, http://links.lww.com/QAI/B389. PubMed search strategy). To identify the grey literature, we conducted targeted searches of websites from 3 HCD design leaders 6 and another organization using HCD for health known to the authors, including any HIV-related cases on the website as of the search date. Case studies not articulating HCD methods or results were excluded. LB screened identified titles, abstracts, and case summaries using the inclusion criteria. LB and AB reviewed relevant full-text articles, abstracting the study author, participants and setting, HCD methods described, results, strengths, limitations, and lessons learned.

Human-Centered Co-Design Workshop
We supplement the literature review with our experience applying HCD to shape a PCC intervention. The "Person Centred Public Health for HIV Treatment in Zambia Study" (PCPH) is a stepped wedge, cluster-randomized trial of a PCC The investigative goal is to understand the user context and the holistic user experience with the product/process/ intervention of interest. This is performed through active collaboration between developers and end-users to generate context-relevant insights and solutions. User-focused research proactively supports identification of user ideas, assets on which to build, and capacity for change instead of focusing exclusively on challenges or reactively identifying feedback and preferences. It intends to promote trust, cooperation, and ownership of solutions among users. Engaging a broad range of user and influencers perspectives, as well as openness to exploration throughout the design process, promotes essential breadth of investigation.

Identification of actionable insights
The investigation is generally guided by open-ended questions (such as those in the "how might we.." format 9 ) related to the design topic of interest and informed by existing knowledge or research. The investigation is conducted using participatory activities, some of which are familiar to SBR and many of which are discussed in examples of HCD toolkits 9 and reports. 14 Rapid documentation of learning throughout the HCD process using notes, visual representations, and other media. Rapid reflection on documented insights (1) informs new questions to guide the investigation, (2) 24 In response, the PCPH study's intervention was conceived to comprise (1) training of health care workers (HCW) in the principles and practices of patient-centeredness; (2) collecting and sharing data with HCW on the patient experience at the health facility; (3) HCW coaching; (4) supporting facility-level quality improvement; and (5) incentives for improved practice. Before trial implementation, we undertook a pilot study implementing elements 1-3 of the planned intervention in 2 facilities to understand the context and test and refine the intervention. We conducted formative research on intervention components using interviews and focus-group discussions (FGDs). We then held a 5-day HCD workshop to engage HCWs who experienced the pilot in co-design activities to further refine the study intervention. Key HCD workshop strategies included the following: (1) developers cultivating empathy with HCWs and learning from insights and experiences, (2) developers and HCWs collaborating in investigation and creative problem-solving, and (3) defining actionable approaches for intervention improvement. Drawing on previous research and pilot findings, we defined 3 "How Might We.?" questions 9 (HMW) to guide the workshop: • Coaching: How might coaches be best positioned in health facilities to guide and support HCWs in delivering PCC according to best practices and in ways that are appropriate to facility context?; • HCW support and motivation: How might we foster a workplace culture that empowers and motivates health care providers to provide PCC?; • Information management: How might we make new and existing information on patient experience and patient   To inform the creation of the management intervention to improve service delivery efficiency 4 phases: (1) "empathize" with the user, (2) "define" the issue, (3) "ideate" or brainstorm solutions, and (4) "prototype" 1) collect and review management journals of activities and challenges; conduct key informant interviews; and non-participant observations at the CBO sites 2) reviewed and organized findings; distilled them into "insights"; mapped challenges, opportunities and insights to program planning and implementation 3) brainstormed .20 "prototype" intervention components 4) tested prototypes in 3 focus group discussions (FGDs); used FGDs to finalize intervention  outcomes accessible, desirable, and useable for facility staff and other key users?
The design workshop included 31 purposefully 25 invited HCWs (users) from the pilot facilities (see Appendix 2, Supplemental Digital Content, http://links.lww.com/QAI/ B389), 6 district health management team representatives (influencers) and 12 research team members (developers). The workshop agenda (see Appendix 3, Supplemental Digital Content, http://links.lww.com/QAI/B389) included common HCD insight gathering activities such as "journey maps" 9 and personas 26 to realize the workshop strategies. Workshop facilitation was led by an external HCD expert and cofacilitated by research team members.
To synthesize insights, participants generated visualized activity outputs common to HCD (see Appendix 4, Supplemental Digital Content, http://links.lww.com/QAI/ B389). The developers took notes during each workshop session and, each day, reviewed outputs and notes, and dialogued to identify key questions, emergent insights, and direct feedback on intervention components. Developers then categorized the insights through mapping and rapid thematic analysis 8 and proposed intervention revisions. Critical insights and themes were discussed with HCWs.

Ethics
Study activities were conducted under a health facilitylevel waiver of consent, approved by the University of Zambia Biomedical Research Ethics Committee and University of Alabama at Birmingham IRB.

Literature Review
The search strategy identified 77 published articles, of which 8 were relevant to the research question, representing 5 studies. The grey literature searching identified 4 studies, of which 2 had sufficient information for inclusion (Fig. 1). Studies came from Nigeria, 27 Kenya, 28 South Africa, 26,29,30 Uganda, 26 and the United States 31-36 and included adults and youth. Four studies designed an mhealth tool, 27,31-36 1, a management process, 28 and 2, PrEP delivery approaches with young women. 26,29,30 All studies specified a phased process, including each of the elements described in Table 1. The order, structure, and intensity of those elements differed by the study. All but one included study 36 articulated plans for a feasibility or effectiveness evaluation of the outcome resulting from the HCD process. However, although planned for 1 case study, 37 no studies estimated the effect of the HCD process itself by comparing the implementation of the outcome designed to either (1) outcome implementation not informed by users or (2) informed by another formative research approach (Table 2).
In each study, the HCD application was guided by the current state of research and the user behaviors desired by the developers. Most of the studies sought to create and optimize a specific outcome (ie, a management intervention to improve efficiency, 27 a decision support tool to integrate HIV and TB clinical care, 28 and a mobile application to improve HIV management or patient decision-making 31,32,36 ), the basic form of which was justified by previous research. Two studies, 29,38 both focused on a newer product and a less well-understood user group (PrEP for young women), sought more broadly to understand experiences and approaches relevant to increasing and sustaining PrEP use and then proceeded to design and refine specific PrEP products and processes. Articles reflect that even when specific products are of interest, HCD processes require sufficient flexibility during the user-driven investigation stages to allow for user priorities to guide the focus and form of the final solutions designed. For example, although based on an electronic medical record system, the final HIV-TB integration support tool design utilized paper-based messaging to clinicians as the most feasible form of communication. 28   How can interventions consider extreme patient behaviors and difficult circumstances that do exist and consider practical solutions to such problems? If VIPs represent a minority of patients, how can interventions help HCWs to consider these circumstances as the exception, instead of the rule that would always prevent patient-centered practices? The worst patients When thinking about patients who need help, HCWs often identified every possible negative trait and challenge and mentally assigned them to a single patient. As approaches were identified to deal with a challenge, the next challenge would arise, precluding the acceptance of the possible utility of patient-centered practices by always having a "next problem" at the ready.
Mentally assigning all problems to one person makes PCC seem overwhelming and impossible. How can interventions highlight more realistic example patient who have several challenges each when working through cases, instead of allowing every possible challenge to be mentally assigned to one patient? To avoid dismissal of PCC, how can interventions help HCWs respond to the patient at hand, realizing that not every patient is very, very difficult patient (even if some truly are)? How can interventions brainstorm with HCWs about true worst patient scenarios and how they handled the situations, how they can be supported by their superiors and facilities to do so? This may help to avoid dismissal by acknowledging the existence of real difficult patients and thinking through what is possible.
All published studies identified beneficial changes to the product/process being designed. This represents the poten6al for HCD application. Potential publication bias may limit available data on HCD processes that failed to yield meaningful, ac6onable insights. Ramos et al, 32 however, conclude that the HCD-improved HIE user interface still resulted in poor HIE comprehension and suggest that human interaction may be necessary for understanding. This may demonstrate that 1 cycle of iteration and testing may be insufficient to identify an optimized implementation design, indicating that implementers planning to use HCD need to allow time, resources, and/or strategies to support sufficient re-design and testing.
An HCD approach may itself be an implementation strategy, in addition to providing formative research to optimize and intervention. Catalani et al 28 conducted their HCD work with stakeholders in the same health system where their intervention would be implemented. The authors reflect that the user engagement required for HCD facilitated the trust and acceptance required to intervene in a complex health system. An extant literature suggests that HCD may be particularly apt for improving the usability of a specific product or tool for a relatively well-defined user group; identifying and addressing relevant concerns in a complex system or process; and identifying and avoiding implementation barriers for a new product or product access by a poorly understood user group. The limitations of HCD raised by study authors were poor generalizability of the designed outcomes 28,31,34 and the inability of HCD as an approach to estimate effectiveness. 36

Human-Centered Co-Design Workshop
Our application expanded HCD to a behavioral service delivery target: changing HCW patient-centered beliefs, attitudes, and practices. Like several published studies, we aimed to refine an evidence-based intervention with predetermined components based on extensive previous and formative research. The depth of the interactions between the HCWs and research team members during the workshop, as well as the creativity encouraged by the facilitator, allowed Although the stories shared during training were true, HCWs said they wanted to share "real" stories. We understood this to mean their stories of patientcentered experiences. Several HCWs were able to articulate stories of implementing patient-centered practice in the few weeks between the coachdelivered PCC curriculum training and the HCD workshop. HMW #2: How might we foster a workplace culture that empowers and motivates health care providers to provide PCC? Formally incorporate health care worker experience in the study theory of change.
Changes in the HCW experience may mediate intervention effects.
Expand coverage of study sensitization and curriculum participation to include a minimum of 75% of the facility, with ideally 100% of staff members reached.
HCWs who supported PCC practices needed a broader foundation built at their facility to allow them to encourage others to adopt PCC practices. More and greater access to initial information can help reduce suspicion and encourage wider access to perceived advantages study participation may convey. Orient the PCC intervention to be both patient and provider-centered in its implementation. Begin PCC intervention activities with HCWs by recognizing the HCW perspectives on care delivery instead of beginning with the challenges patients face.
PCC interventions should lead to care ie, optimized for patients. However, it is important to recognize that HCWs are the users of an intervention encouraging them to adopt more patient-centered practices. Patient-centered care interventions are better received, and likely more effective, if presented and implemented in a provider-centered way. This must balance patient needs, but reduce common feelings of blame and HCW judgment. Introduce quantitative survey measuring HCW satisfaction and experience to expand and anonymize HCW feedback beyond qualitative focus group discussion-based data collection.
While focus group discussions were planned with HCWs, they only offered limited participation opportunities and were not anonymous. Because patients were being asked about HCWs, HCWs wanted more opportunities to share their experiences, including identifying what positive and negative feedback they receive to do their work. HMW #3: How might we make new and existing information on patient experience and on patient outcomes accessible, desirable, and useable for facility staff and other key users? Instead of sharing average patient satisfaction scores or other aggregate indicators of facility performance, share disaggregated data by question to help HCWs to identify specific areas of high performance or those needing improvement.
HCWs were interested in the specifics of what is happening to understand the source of a problem and what may be needed for change, not averages or summaries. For example, an average patient satisfaction score was of much less interest than scores for each question used to measure satisfaction.
for a breadth and openness of conversation not achieved during the formative focus group discussions. The insights gained during our workshop may be partially explained by the open relationships established and intervention feedback received during the pilot. HCD revealed that patient-centeredness principles (eg, understanding the whole person, 2-way communication) resonated with HCW but met resistance with some HCWs such as time constraints and beliefs that punitive measures improve patient compliance. During HCD activity dialogues, HCW participants shared stories of using kindness to help struggling patients to re-engage in care, highlighting the ability to implement PCC under less-than-ideal conditions and the benefits of positive interactions. The explicit encouragement of HCD to offer multiple perspectives and dialogues and embrace a co-creative environment allowed for the interactions, which revealed these various perspectives. The key points of resonance with PCC (Table 3) and resistance (Table 4) were voiced directly or indirectly by HCWs and codified through rapid thematic analysis. While an unanticipated HCD workshop result, they present both likely challenges and promising solutions to promoting PCC practice. They would not apply to every situation; however, they offer guidance for the trial intervention coaches to anticipate and may address barriers to PCC adoption in transferable settings.
Intervention components that will be tested in our stepped-wedge trial were revised to improve acceptability, appropriateness, and feasibility of PCC practices, including better-using intervention mentoring and data collection to help HCWs feel visible, appreciated, and accountable. For example, HCD insights led us to augment the planned HCW FGD data collection with a quantitative HCW survey to expand and anonymize HCW feedback (additional examples in Table 5). The limited workshop days allowed for little formal iteration on proposed revisions beyond sharing ideas and eliciting additional HCW feedback. Our study, however, used further pilot period implementation to test some of the proposed changes. The rigor of our HCD process may have improved with extended time for formal iteration and testing.

DISCUSSION
Our review suggested that HCD offers an important and emerging tool for adapting strategies to enhance ART services. While traditionally applied to products or mHealth practices, we and others have also begun to apply these principles to shape and refine the service delivery context and patient experience. Overall, HCD has the potential to improve the HIV response by more closely aligning the implementation of evidence-based products and processes with user priorities and context. In an HIV response that has to date rightly prioritized scale-up and standardization, but which now must shift to a more targeted efforts to continue improvements, HCD offers a theoretically based, robustly mapped set of practices that seek to improve outcomes through active engagement and respect for end-user views; encouragement of broad, creative inquiry; and support for iterative idea refinement in response to testing and feedback.
Rigorous, successful HCD outcomes are no panacea however, and likely depend on relationships, time, and other resources required for authentic engagement and responsiveness, which may not always be available.
Extant evidence lacks an assessment of the comparative effectiveness of HCD and non-HCD processes in implementation optimization. Recent research, however, offers interesting insight into potential HCD gains. A study of the national scale-up of PrEP in Kenya identified that in addition to other stigma types, product stigma, including the similar appearance of PrEP bottles and ARV bottles, was a barrier to PrEP use. 39,40 The HCD approach in EMOTION 29,30,37 identified that the medical appearance of PrEP was associated with illness, creating a likely barrier to young women's uptake and use. EMOTION 29 thus prototyped PrEP starter kits that resemble make-up bags and stickers to disguise standard labels. User engagement lessons may facilitate use in similar populations.
HCD approaches are not the only user-engagement methods that may support improved implementation. A 2017 landscape mapping of end-user research in HIV prevention for young women in sub-Saharan Africa identified 53 projects, of which 3 were explicitly HCD. 41 There is often overlap in objectives and methods between HCD, more traditional qualitative methodologies 42,43 ; participatory research 11 ; engineering approaches 44,45 ; and discrete choice experiments as formative research. 46 Although efforts have been made to compare and contrast approaches, 8 distinctions are not concrete. Authors struggled to adjudicate use of HCD in some studies in our literature review. The Schnall et al 34 study discussed a user-centered design process but grounded their research in the Information System Research Framework and described some of their methods as participatory action research. We erred on the side of inclusion, but the multidisciplinary nature of user-centered approaches requires reviewers' interpretation of methods. HCD distinguishes itself in by focus on empathy; flexibility in inquiry as directed by users; iterative, rapid testing cycles; and emphasis on practical outcomes. Our own case study included both traditional formative research and HCD. Although our more traditional formative research raised the issue of HCW context and the need to avoid blame to increase receptivity of HCWs to PCC, the empathy established and nuance understood through the HCD workshop elevated the HCW experience to formally include improvement in the HCW experience in the study Theory of Change.
The limitations of HCD identified in the literature review including a lack of generalizability and inability to estimate effectiveness are not dissimilar to other qualitative methodologies. Instead of limitations, they are features of the approach. HCD findings may be transferable, 47 a metric of qualitative rigor that considers the applicability of findings from one setting to a different setting with a similar context, instead of generalizable. Potential for transferability was noted in several literature review studies, 26,28,29 as well as our own HCD case study. Critical reflection on HCD limitations would enhance future use. For example, thin preliminary knowledge of the evidence-based intervention may lead to poor design questions. An HCD process that is too rapid to allow for sufficient iteration or diversity of user engagement may produce results that lack rigor. Based on the importance of open discussion in our workshop, we believe that an absence of mutual respect and trust between users and developers could prevent empathetic interactions and produce inauthentic HCD outcomes.
An important HCD insight from our workshop was that effective implementation of an intervention to increase the patient-centeredness of care through HCW behavior change needs to be implemented in a provider-centered way. This does not diminish the central role of patients in their care processes and outcomes. It highlights that just as the lived realities of the patients' need to inform HCW engagement with patients as care partners, the lived realities of the HCWs demand that interventions to change HCW practice prioritize feasible, acceptable, and appropriate implementation strategies. HCD may be particularly appropriate in complex settings that need to consider patient, provider, and systems realities. The compendium of resonance and resistance to PCC (Tables 3 and 5) provide concrete illustrations of how using HCD can reveal challenges and opportunities that may influence intervention effectiveness. Consistent with the reflections of Catalani et al, 28 our team felt that the communication, collaboration, and engagement between planners and users during HCD can support successful intervention implementation subsequent to the design phase.

Limitations
Our review may have missed relevant case studies given the limited grey literature search and the practical outputs of many HCD studies, which are less likely to be published in peer-reviewed manuscripts. Similarly, we focused on HIV-related HCD, excluding possible lessons from other health areas. Owing to time and resource constraints, our HCD process did not include patients, the intended intervention beneficiaries, and key actors in their own care and treatment. Inclusion of patients would almost certainly have identified additional insights and intervention revisions. Our choice to focus on the HCWs allowed us to engage with the primary users of the planned intervention, to conduct our HCD approach in limited time, and to manage identified power dynamics within HCW cadres based on shared knowledge and vocabulary from the pilot phase.

CONCLUSIONS
HCD has the potential to improve effective implementation of the HIV response by tailoring implementation strategies to enhance evidence-based interventions in particular service delivery settings. Although HCD is a promising and increasingly advocated 48,49 approach to bridge this gap; more studies that document and critically reflect on the use and impact of HCD methods in the global HIV response are needed to guide their effective use. Those seeking to promote PCC may improve implementation success by seeking out the resonance and anticipating and defusing the challenges our HCD process identified.