Human-centered design as a guide to intervention planning for non-communicable diseases: the BIGPIC study from Western Kenya.

Background Non-communicable disease (NCD) care in Sub-Saharan Africa is challenging due to barriers including poverty and insufficient health system resources. Local culture and context can impact the success of interventions and should be integrated early in intervention design. Human-centered design (HCD) is a methodology that can be used to engage stakeholders in intervention design and evaluation to tailor-make interventions to meet their specific needs. Methods We created a Design Team of health professionals, patients, microfinance officers, community health workers, and village leaders. Over 6 weeks, the Design Team utilized a four-step approach of synthesis, idea generation, prototyping, and creation to develop an integrated microfinance-group medical visit model for NCD. We tested the intervention with a 6-month pilot and conducted a feasibility evaluation using focus group discussions with pilot participants and community members. Results Using human-centered design methodology, we designed a model for NCD delivery that consisted of microfinance coupled with monthly group medical visits led by a community health educator and a rural clinician. Benefits of the intervention included medication availability, financial resources, peer support, and reduced caregiver burden. Critical concerns elicited through iterative feedback informed subsequent modifications that resulted in an intervention model tailored to the local context. Conclusions Contextualized interventions are important in settings with multiple barriers to care. We demonstrate the use of HCD to guide the development and evaluation of an innovative care delivery model for NCDs in rural Kenya. HCD can be used as a framework to engage local stakeholders to optimize intervention design and implementation. This approach can facilitate the development of contextually relevant interventions in other low-resource settings. Trial registration Clinicaltrials.gov, NCT02501746, registration date: July 17, 2015.

used to engage stakeholders in intervention design and evaluation to tailor-make interventions to meet their 23 specific needs. 24 Methods: We created a Design Team of health professionals, patients, microfinance officers, community health 25 workers, and village leaders. Over 6 weeks, the Design Team utilized a four-step approach of synthesis, idea 26 generation, prototyping, and creation to develop an integrated microfinance-group medical visit model for NCD. 27 We tested the intervention with a 6-month pilot and conducted a feasibility evaluation using focus group 28 discussions with pilot participants and community members. 29 Results: Using human-centered design methodology, we designed a model for NCD delivery that consisted of 30 microfinance coupled with monthly group medical visits led by a community health educator and a rural clinician. 31 Benefits of the intervention included medication availability, financial resources, peer support, and reduced 32 caregiver burden. Critical concerns elicited through iterative feedback informed subsequent modifications that 33 resulted in an intervention model tailored to the local context.
(Continued on next page) 124 Teaching and Referral Hospital, and a consortium of 125 North American academic medical centers [23]. 126 AMPATH established a system of HIV care in western 127 Kenya and has since expanded its clinical scope to in-128 clude population health and NCDs [24]. At the time that 129 this study was conducted, the Chronic Disease Manage-130 ment program at AMPATH had enrolled over 2000 131 patients with diabetes and 40,000 patients with hyper-132 tension, who were being cared for at nine rural health 133 centers and 30 rural dispensaries. The program dis-134 patches clinicians to rural clinics monthly, which are 135 otherwise staffed by nurses. This study was conducted as 136 part of the Bridging Income Generation and Group Inte-137 grated Care (BIGPIC) study which aims to evaluate the 138 combination of microfinance and group medical visits 139 for cardiovascular risk reduction in the AMPATH catch-140 ment area across four counties in western Kenya [25]. 141 Intervention design 142 In this project, we adapted a pilot BIGPIC model that 143 consisted of microfinance coupled with monthly group 144 medical care visits with a rural clinician [26]. In this 145 pilot model, participants with diabetes or hypertension 146 are recruited to join the group and consist of at least 147 50% of group members. Group medical visits with a 148 clinician occur immediately after each microfinance 149 meeting. In order to further refine and adapt this ap-150 proach to the local context, we utilized a HCD frame-151 work consisting of four steps -Discover, Design, Test, 152 and Refinewith an emphasis on community and end-153 user engagement at each stage ( Fig.   F1 1). As HCD is an 154 iterative process, the steps are described in sequence in 155 the Methods and Results sections. 156 Step 1. DISCOVER: understanding the community 157 The first step in refining the BIGPIC model was to 158 understand the strengths and needs of the local commu-159 nity. We utilized a combination of qualitative research 160 methodologies to explore community and individual 161 perspectives. The primary goals were to identify existing 162 barriers to NCD care, and to identify contextual factors, 163 barriers, and facilitators that could impact intervention 164 design, implementation, and sustainability. We held 165 mabaraza, traditional East African community gather-166 ings, to discuss community perspectives in an open for-167 mat. We also hosted focus group discussions (FGDs) 168 consisting of 10-15 individuals with common character-169 istics (rural clinicians, community health workers, pa-170 tients with NCDs, and microfinance group members) to 171 explore individual perspectives. All qualitative studies 172 were led by local team members utilizing a semi-173 structured guided interview in local languages.

174
Step 2. DESIGN: designing the intervention  from HCD methodology [21]. An initial agenda was de-192 veloped, however, timing for each step was flexible.

193
The specific objectives and activities utilized in the De-    All ideas were noted and subsequently evaluated in 218 group discussion for pros, cons, and feasibility in Proto-219 type meetings. In the latter example above, we consid-220 ered gender-specific groups, targeted screening locations 221 when men tend to congregate, and community educa-222 tion, particularly through male leadership involvement. HCD steps (Discover, Design, Test, and Refine). As HCD is an iterative process, the arrows describe how the results of each step impact the next f1:3 240 Feasibility pilot study Implementation strategies and 241 the prototype, as defined in Step 2 Results, were piloted 242 in one rural community in western Kenya. Adults who 243 screened "positive" were those with elevated blood pres-244 sure or elevated fasting blood glucose. Inclusion criteria 245 of those who screened positive were newly screened 246 adults, previously screened adults who had never linked 247 to care, or existing patients who had linked to care in 248 the last 6 months. Those with diabetes or hypertension 249 who did not meet the inclusion criteria were able to join 250 the group as non-study participants to a maximum 251 group size of 30 members. Participants and local CHWs 252 subsequently received training in microfinance, hyper-253 tension, and diabetes. Qualitative feedback was elicited 254 from the participating rural clinician, CHWs, and group 255 participants at one, three, and 6 months using guided in-256 terviews and FGDs. 257 Step 4. REFINE: intervention refinement approaches.

284
Step 3. TEST: assess community acceptance and pilot For example, one concern alleviated by the initial 305 prototype regarded gender and group dynamics. In 306 qualitative studies in Step 1 and Step 3, some commu-307 nity members expressed concern that differences in age 308 and gender could affect group participation, given a 309 strong culture of both gender and age-based hierarchy. as not to interfere with their annual harvest schedule,  Persisting concerns shared by pilot participants in-330 cluded apprehension that participation would be limited 331 by stigma associated with illness such as HIV. Partici-332 pants emphasized the importance of anticipatory com-333 munity education to enhance community receptiveness, 334 and suggested strategies to facilitate this. Based on their 335 prior experiences with brief lifecycles of programs due 336 to limitations in funding and service delivery, partici-337 pants and local leaders also expressed concern regarding 338 the sustainability of the program, and noted these prior 339 experiences may discourage some from joining. They 340 also reported that income generally is low among their 341 community and requested seed money or incentives to 342 jump start their savings. 343 Step 4. REFINE: intervention refinement 344 The final BIGPIC model consists of an integrated group 345 care and microfinance model, with specific changes to 346 the prototype model described in Table 3. These include 347 expanded access to common medications, a reduction in 348 maximum group size, and clarification of protocols with 349 CHWs to coordinate changes in meeting times during 350 the harvest season.   Fig. 4 Benefits and Concerns related to the BIGPIC model f4:2 383 medication adherence, and financial barriers to accessing 384 healthcare in rural Kenya. We gathered insights and opin-385 ions from the community and formed a transdisciplinary 386 Design Team of health professionals and community 387 members to evaluate our data and create an initial proto-388 type. This prototype was tested over six months and fine-389 tuned through community feedback to enhance accept-390 ability and sustainability. The resulting BIGPIC model 391 combines the benefits of microfinance with the peer sup-392 port available through group medical care to enhance 393 management of hypertension and diabetes. Key insights 394 that developed through the HCD process informed both 395 prototype features and implementation strategies and can 396 be mapped directly to the strengths, needs, and concerns 397 elicited from the community (Fig.   F5   surrounding circles represent the unique milieu that has informed BIGPIC's development. These include community strengths (green text), f5:3 barriers to care (red text), and concerns regarding the BIGPIC model (blue text) elicited from community and pilot participant feedback, as f5:4 described in Fig. 1 (Steps 1, 3, and 4). The surrounding descriptors in black text are key features and implementation strategies of the BIGPIC f5:5 model. Each can be mapped to a community-driven strength, barrier, or concern. The text highlighted in yellow represents changes that were f5:6 made during the Design Team Re-evaluation (Fig. 1, Step 4) in response to participant feedback f5:7 419 weight management, and tobacco use [28,31]. In limited 420 resource settings, HCD is a process that can comple-421 ment and support existing approaches to shaping NCD 422 control such as the World Health Organization STEP-423 wise approach [32]. HCD is one approach to optimize 424 stakeholder engagement, and as in the example of BIG-425 PIC, it can propel an understanding of local factors into 426 the development of a contextualized intervention. In this 427 study, our approach utilizing HCD for complex interven-428 tion design aligns with the Medical Research Council 429 guidelines for developing complex interventions, utiliz-430 ing a systematic approach to a development-evaluation-431 implementation process that is tailored to local circum-432 stances [33]. Similar approaches to design-thinking have 433 been described with other disease processes in low-434 resource settings [34][35][36].  t2:20 Key insights elicited from the design process can be mapped directly to prototype features and implementation strategies. CDM -Chronic disease management 451 critical insights from end users throughout our HCD 452 process. For example, multiple community members 453 voiced concern for a sustainable intervention that 454 would engage with local leaders and not be 455 dependent on external funding. The use of an 456 economic-based intervention as well as many of the 457 features of this microfinance and group care model 458 are in response to these lessons learned.   t3:5 Group education on NCDs at the time of t3:6 group formation and before every monthly t3:7 meeting.

5, Tables
There is low interest in group education. Health education time is modified from didactic teaching to facilitated group discussions on selfmanagement and problem solving. CHWs receive training in group facilitation.
Large groups may overburden clinicians. Maximum group size is decreased to~20 participants.
t3:9 Village-based health screenings to recruit t3:10 intervention participants. There is low income generation among community members, particularly elderly and those with low education levels.
Agribusiness and financial trainings are incorporated.
Health education time is modified as above.
t3:19 Feedback and concerns elicited from pilot participant feedback informed key intervention modifications 504 consider shorter and faster cycles of prototyping for less 505 complex interventions, in order to efficiently evaluate 506 ideas and integrate lessons-learned for continuous im-507 provement and sustainability. Beginning with shorter cy-508 cles may also help gain stakeholder support for 509 subsequent longer cycles of more complex intervention 510 development. 511 We also recognize that HCD is a process that re-512 quires tolerance of ambiguity, pivots, and prototyp-513 ing-factors that can seem to be in opposition to 514 traditional hypothesis-driven research methodologies 515 [38]. However, we feel that HCD is a process for the 516 design and development of interventions and imple-517 mentation strategies that are both desirable and feas-518 ible in the local context, which can then be evaluated 519 with traditional hypothesis-driven statistical method-520 ologies. In our study, we have combined HCD with a 521 more traditional randomized controlled implementa-522 tion research trial to evaluate the effectiveness of the 523 intervention [25].

524
Finally, there is growing enthusiasm in both academic 525 medicine and global health spheres for social innovation 526 and design thinking as tools that are more capable and re-527 sponsive to the needs of end users [38-40]. However, there 528 is still limited evidence regarding the impact of design 529 thinking methodologies and related concepts on health out-530 comes [41]. Additional research is needed to evaluate the 531 impact of participatory methodologies such as design think-532 ing and social enterprises on health outcomes. 533 Application to other contexts 534 The development and implementation of BIGPIC is one 535 example of how HCD concepts can be used in resource-536 limited settings. Of particular relevance was the inclusion 537 of transdisciplinary community stakeholders on our

553
Recognizing that there are universal elements to care 554 that are common across geopolitical and financial land-555 scapes, we advocate for context-specific interventions 556 that can help to optimize care in these settings. How-557 ever, we recognize that potential unintended conse-558 quence is that such specificity may lead to variability and 559 inequities in care. For this reason, we urge caution with 560 planning for context-specific settings.

562
In this study, we describe how a four-step HCD frame-563 work was used to tailor NCD service delivery to address intervention package that is tailored to the local context f6:1 Fig. 6 Key themes were organized together to stimulate idea generation f6:2