Analytic hierarchy process: An innovative technique for culturally tailoring evidence‐based interventions to reduce health disparities

Abstract Latinos in the United States represent a disproportionate burden of illness and disease and face barriers to accessing health care and related resources. Culturally tailored, evidence‐based interventions hold promise in addressing many of these challenges. Yet, ensuring patient voice is vital in the successful development and implementation of such interventions. Thus, this paper examines the application of analytic hierarchy process (AHP) to inform the augmentation and implementation of an evidence‐based chronic disease self‐management programme for underserved Latinos living with both minor depression and chronic illness. The process of AHP allows for direct input from the individuals that would utilize such a programme, including afflicted individuals, their family members and the health educators/promotores that would be responsible for implementation. Specifically, 45 participants, including 15 individuals with chronic disease, 15 family members/caregivers and 15 promotores, partook in the Stakeholder Values Questionnaire, which elicited preferences and values regarding major goals, processes and content for the intervention. AHP was employed to analyse pairwise comparison ratings and to determine differences and similarities across stakeholder groups. This analytical technique allowed for the adaptation of the EBI to stakeholders' specific priorities and preferences and facilitated complex decision‐making. Findings not only shed light on similarities and differences between stakeholder groups, but also the magnitude of these priorities and preferences and allowed the intervention to be driven by the participants, themselves. Applying AHP was a unique opportunity to optimize the decision‐making process to inform cultural adaptation of an EBI while considering multiple viewpoints systematically.


| INTRODUC TI ON
The 2012 National Healthcare Disparities Report asserts that, despite efforts to enhance access to care, access has not improved among most racial and ethnic groups. 1 In fact, racial and ethnic minorities fare worse in terms of access to health care and associated outcomes as compared to their white counterparts. 2

Among
Latinos, the largest minority population in the United States, contributing factors to poor health outcomes include lack of access to and utilization of preventive care, lack of health insurance and linguistic and cultural barriers. 3,4 Further, Latinos continue to face a disproportionate burden of illness and disease. In fact, diabetes rates among Latinos are nearly double that of non-Hispanic whites. 4 Culturally tailored, evidence-based interventions (EBI) hold much promise in overcoming these challenges. 5 EBIs such as Tomando Control de su Salud, a chronic disease self-management programme developed by Stanford University, have been touted for their ability to enhance chronic disease self-management practices, including improvements in health behaviours, health status, enhanced self-efficacy and fewer emergency room visits. 6 Such programmes have been disseminated globally to diverse populations and have shown positive results. 7 Chronic disease self-management programmes (CDSMP) have several features that make them worthy of adaptation, particularly the format used to provide health education, the utilization of peer group members and lay leaders, and the use of multidimensional techniques to address nutrition, physical activity, problem-solving, sleep, fatigue and patient empowerment through the enhancement of self-efficacy and positive behaviour change. 8,9 Yet, questions remain over how to ensure that EBIs are culturally tailored to local needs. The complex interplay between chronic illness and the host of factors that impact access to and utilization of health services by underserved Latinos requires EBIs to be responsive to local situation reality faced by Latinos. The adaptation of such interventions is costly and time-consuming and requires considerable resources. One approach to guiding adaptation includes seeking input from the drivers or potential end-users of the programme, an approach often employed in consumer marketing, new product development and assessing business risk levels. 10 Analytic hierarchy processing (AHP), a technique developed by Thomas Saaty,11,12 is one such approach, allowing the human drivers and key end-users to guide primary decision-making. AHP has been shown to be effective in guiding multi-attribute decision-making, and the process allows decision-makers to model complex problems using a hierarchical structure. AHP is used to prioritize criterion, in this case programme objectives and content. The resulting prioritization ranks items within the model ratio scale where priorities or weights are derived for each objective or subobjective, allowing the researcher to select the objectives that will have the most impact and helping to guide 'best fit' decision-making.

| Value of analytic hierarchy process
AHP is a 'science of scaling based on math, philosophy and psychology' in which a complex decision is broken down into factors that are arranged by the researcher in an ordered structure to allow weights to be assigned to each factor. 11 Rather than focusing on a single criterion, AHP takes into account all of the applicable criteria concurrently, encompassing a more systematic and transparent approach. 13 The decision-makers (in our case those for which the programme was being tailored) were asked pairwise comparison questions, deciding the importance of one criterion relative to another. 13 Traditionally, AHP has been used in the field of business as a technical and managerial group decision-making process where one seeks to find the partialities of differing groups from a macro-level view. 14 The value of AHP is its flexibility and ability to be precisely customized to each individual challenge. 10 In addition to product screening and development, those in the business sector also employ AHP as a tool for determining cost-effectiveness and how to appropriately allocate finite resources. 13 AHP successfully allows complex decisions to be more easily made with consideration of multiple criteria.
With the effectiveness and value of AHP evident, it is reasonable to establish that it can be translated to complex issues related to health programme decision-making as well. AHP has been used in health-related fields to assess patient satisfaction in services, determine liver transplantation patient priority setting, understand performance of intensive care units, accompany geographical information system (GIS) data in understanding the health needs of communities, assess applicability of telehealth programmes and help patients decide the specific course of treatment that best suits their needs. [15][16][17][18][19][20][21][22][23][24] Groups are calling for the engagement of patients in research, including patient voice in research, and patient centred health care. [25][26][27] AHP is gaining attention as useful methodology to engage patients.
However, application of AHP to the development of health promotion or health education interventions or in tailoring health education models to the needs of beneficiaries of the programme is scarce. 28 Thus, AHP as a tool to customize the objectives and content prioritization of an existing evidence-based programme to a specific target population would be of immense benefit to all stakeholders involved.
Before launching any health promotion campaign, it is imperative to ensure the relevance of the programme and feasibility of adaptation among the target population. Thus, the multiphase parent study of this paper 19 sought to adapt Stanford University's CDSMP, Tomando Control de su Salud (Tomando), to the needs and preferences of underserved Latinos in the Tampa Bay area suffering from both minor depression and a chronic illness and to determine whether the adapted intervention would be suitable for the community. This paper discusses Phase II of the parent study, which elicited preferences and values from key end-users for major goals, processes and content of Tomando using Stakeholder Values Questionnaire and AHP.

| ME THODS
This study employed AHP to inform the development and implementation of an EBI to enhance chronic disease self-management among underserved Latinos living with both minor depression and chronic illness. The approach built upon initial, formative research findings, 29 which assessed barriers and facilitators to chronic disease self-management, ultimately allowing for a robust assessment of needs, preferences and priorities among the target population and the implementation of an intervention driven almost entirely by the population it intended to serve. University institutional review board approval was sought from the research institution prior to implementation of the study (#107512).

| Sample population
Almost 30% of the population of Hillsborough County, the area of focus for this study, identify as Hispanic/Latino. 30 This population, now the largest minority population in both the county and the United States, lacks access to resources and services, particularly intensive, comprehensive and specialized services, and faces linguistic and cultural barriers to accessing care. 4 To help fill this gap, this study targeted underserved Latinos living with a chronic illness and minor depression, noted in this study as individuals with chronic disease (ICDs), their family members (FMs) and the promotoras (P) who would be responsible for delivering the intervention. Fliers at local community outreach events, the local library, clinics and at community partner sites were used to recruit participants. Eligibility requirements for ICDs

| Development of the Stakeholder Values Questionnaire
Findings from qualitative data collected through Phase I semistructured interviews (n = 37) and structured surveys (n = 35) with promotoras, nurses, physicians and CBO leaders, and focus groups (N = 9; n = 42) with ICDs and FMs were used to guide the development of our Stakeholder Values Questionnaire. 29 While findings from this phase are published elsewhere, this Stakeholder Values Questionnaire was developed based on emergent themes, key priorities and needs which arose from the formative research stage. 29,32 This included challenges with managing chronic illness, unmet needs and the importance of support and education for those living with chronic illness. Additionally, the questionnaire was designed to also evaluate the core elements of the Tomando programme and to elicit preferences and values from stakeholders regarding major goals, processes and content for the intervention. 33 Specifically, the questionnaire evaluated the important elements of the intervention (ie skill-building or informational and educational materials), the structure of the programme (ie the number of sessions and the content of those sessions), who should lead the programme (ie medical professionals, promotoras), and attendance (ie either alone or with a partner) at the programme.
The questionnaire was developed using the AHP method and presented respondents with a series of paired comparisons for each objective and subobjective for their input.

| Data collection
Prior to completing the questionnaire, 45 participants, 15 from each key stakeholder group (ie ICDs, FMs and promotoras/outreach workers) received a presentation on Tomando by trained research staff and watched a video developed to provide additional in-depth information about the programme. Participants were then asked to offer input on the Tomando programme and provide guidance on priority needs and preferences through the Stakeholder Values Questionnaire. Each questionnaire was orally administered in Spanish by a member of the research team, which allowed participants to select between paired comparisons, while also assigning a weight to their selection. A sliding scale tool was used to allow participants to select their response and the desired weight.
Questionnaires took 10-20 minutes to complete, and participants received a $40 stipend for their participation.

| Data analysis
Following data collection, participant decisions were entered directly into Expert Choice ©, 34 a software programme designed to facilitate and analyse choices through the collaborative decision-making process. AHP was employed to analyse stakeholders' pairwise comparison ratings and to determine differences and similarities given by the three stakeholder groups. Key priorities regarding the refinement of the educational programme as well as differences and similarities across stakeholder groups were analysed and ranked. This ranking highlighted the differences between the stakeholder groups and allowed the research team to tailor the new educational health intervention to their specific priorities and preferences.

| Participant demographics
The mean age for the total population was 49. 8 Table 1.
Overall, participants were asked questions regarding import-

| Content assessment
One of the most vital aspects of tailoring this intervention was determining priorities in managing illness, including ensuring appropriate content and strategies for sharing needed information. Accordingly, participants answered questions regarding the importance of skillbuilding and educational materials in improving health and enhancing the management of chronic illness. In the combined model (C), which included all stakeholder groups, enhancing skill-building was ranked first (C = 0.59) compared to educational-and informationalbased elements (C = 0.41) (see Figure 1).
However, variations between group models were noted. Rank orderings differed in the promotoras model, which ranked educational components first (P = 0.52), followed closely by skill-building (P = 0.48). Both ICDs and FMs placed more weight on skill-building (ICDs = 0.76 and FM = 0.51), with ICDs placing greater importance on this element than on information and education (ICD = 0.24). To delve further into the content required for either skill-building or educational sessions, each area was probed individually.

| Educational components
Educational components identified in the formative research stage as being important to self-management, including nutrition and stress management, were analysed. Components that already existed through the Tomando programme, such as medication management, were also analysed for appropriateness. In the combined model, information on nutrition was ranked first (C = 0.30), followed closely by stress management (C = 0.29), then managing symptoms Similarly, promotoras ranked stress management first (P = 0.32), followed closely followed by nutrition (P = 0.29). Educational information regarding medications ranked lowest across all three groups overall (C = 0.08).

| Types of skills
Participants were also asked about specific strategies for enhanc-

| Assessing structure
An important consideration of any intervention is its structure.
Without an amenable structure, participants may not attend nor benefit from the content presented. Therefore, a portion of the questionnaire focused on designing a structure for Tomando that met the needs of participants, particularly concerning the number of sessions and the timing of those sessions.
The combined model ranked the incorporation of both priming and sustaining sessions to Tomando as the preferred 'augmentation'.

| Attending with a partner
There was consensus across groups that attending the programme with a companion was optimal. However, there was variation between groups regarding who would be the best option as a Tomondo companion. For example, the combined model ranked someone with a chronic illness first, followed by a family member, and then a friend.

| Priming participants for an intervention
The Stakeholder Values Questionnaire also sought to further elucidate information on potential additions to the intervention, includ-  were asked to weigh options to inform the development of a potential introductory session(s), including the number, length, content and who should lead the session. Ranks and associated weights for each option are presented in Figure 2.
Overall, there was consensus across models regarding the number of sessions that would be optimal to prepare individuals for participating in a CDSMP programme, with one session as the preferred option (C = 0.73). Consensus also existed across groups regarding the length of these sessions. While 90-minute sessions were ranked as the preferred option among all groups (C = 0.77), family members placed less weight on this option (FM = 0.60) than ICDs (ICD = 0.80) and promotoras (P = 0.85).
Moreover, participants were asked to rank potential topics that might be covered in a priming session. Variation existed between groups in both the rank ordering of topics as well as their weight.
Specifically, the promotora model placed more weight on having an overview of the programme (P = 0.36) that described key features of Tomando and its elements, followed by testimonials from recent graduates (P = 0.28). Less emphasis was placed on educational materials and information (P = 0.15). In contrast, both ICDs and FMs ranked having educational materials and informational supplements about heart disease, diabetes, hypertension and other chronic illness as their primary choice (ICD = 0.37 and FM = 0.32). All groups placed considerably less weight on educational materials about depression (C = 0.11).
Participants were also asked who should lead priming sessions.

| Sustaining positive outcomes
The

| D ISCUSS I ON
This study illustrates the utility of employing a novel approach for assessing the needs and preferences of multiple stakeholders for ultimately informing the augmentation and implementation of an evidence-based chronic disease programme for Latinos affected by chronic illness and co-occurring minor depression. While, overall, the core elements of the Tomando programme were well received, findings from the formative research phase 29 and results of the present study demonstrate that additional adaptations and tailoring the programme for the specific target population may enhance outcomes.
Employing AHP allowed for a detailed and rigorous exploration of these potential additions, while also ensuring consideration of the critical components of the Tomando programme.

AHP stands out from other evaluation and planning techniques
in that it relies heavily on the population being affected by a problem or decision and allows their preferences and values to be translated into a scaled ranking, leading to the data being 'invariant to politics and behaviour'. 11 Thus, AHP allowed for a rigorous approach to programme adaptation by facilitating the comparison of priorities across stakeholder groups and as a combined group, thereby reducing bias in the reporting of overall group decisions and allowing the research team to elucidate varying preferences across subgroups.
When implementing EBIs, it is vital to ensure local acceptability in regard to the target population and their needs and potential challenges. However, it is also fundamental to ensure fidelity and maintain the original elements of the EBI that have been proven to be effective. 35,36 Researchers must often balance the demands of holding true to a validated programme while also meeting the needs of the local population, especially as culture and language influence perceptions of health, health behaviours and access to resources.
Utilizing the stakeholder-driven methods employed in this study allowed the research team to confirm the critical importance of the core Tomando elements, while also considering emergent stakeholder needs through the formative research stage to be assessed further, allowing their relative importance to drive decision-making. for the augmented programme, promotoras placed weight on educational materials, a choice that generally aligns with their training and expertise. 41 One of the greatest contributions of promotores is their knowledge of the community they serve and its specific needs. 41,42 The prioritization of the informational and educational materials for managing chronic illness by promotores reflects their nuanced perspective. In contrast, ICDs and FMs ranked skill-building as most important for programme content, which may be more immediately beneficial to them, as receiving the skills to manage a chronic condition may be considered more advantageous than learning about that condition. These contrasting prioritizations demonstrate the merit of both education and skill-building, while highlighting the need for certain elements based on the insight of key end-user groups.
Ultimately, these contrasting priorities resulted in the addition of role-playing activities that would put into practice both skill-building content as well as the informational and educational materials that allowed for experiential learning.
The application of AHP also allowed for the identification of congruence between groups, as well as the nuanced differences.  However, it is also important to note that these categories are not mutually exclusive as some family members in this study also had a chronic illness.
Despite contrasting priorities, through the application of AHP, this study was able to clarify the priority needs of the target population and better adapt the content to local needs and preferences.
The data generated allowed the study team to identify and negotiate varying, congruent and contrasting needs and priorities across the three stakeholder groups, providing valuable quantitative insight to inform augmentation.
This research demonstrates the utility of AHP for future health education and health promotion-related research. Accordingly, the authors recommend the incorporation of AHP methods into the research process, particularly the adaptation of validated EBIs for local settings, as such an approach can enhance the feasibility of resultant programmes as well as increase their adoption by the local community through the incorporation of stakeholder voices. The use of AHP in this study allowed researchers insight into various stakeholder groups, their priorities and needs, and the value they place on various aspects of the validated EBI. Moreover, through AHP, researchers were able to compare and contrast these different stakeholder groups' feedback in a rigorous, quantitative manner, a level of detail that is often difficult to elucidate. Through this approach, this study was able to breakdown the various elements of Tomando in order to focus on how to best tailor the programme to the needs of multiple stakeholder groups. Further, the use of AHP maintains the potential to enhance community-based, participatory research methods through its rigorous methodology and stakeholder engagement, allowing for stronger partnerships.

| Limitations
This study models an innovative technique for the adaptation of culturally tailored interventions. However, the majority of participants were female (80%). Additionally, nearly a quarter of ICDs were unemployed. This may have influenced views about timing, the length and number of sessions in the intervention and family member participation. Additional analysis with a larger sample size is recommended for future studies.

| CON CLUS ION
The varying responses presented by participants overall and across the three different stakeholder groups provide critical insight into the components of an EBI that may resonate with different participant groups. Understanding these varying needs and perspectives allows investigators to select the elements that best reach the target audience and their needs and are more likely to ensure success of an intervention. This is where the value of AHP is most evident.
As a research tool, AHP allows investigators to individually tease out the priorities and preferences for each stakeholder group, while also creating a combined score in a mathematically sound fashion.
Together, these scores (each individual score and the overall com- Ellington.

CO N FLI C T O F I NTE R E S T
The authors certify that they have NO affiliations with or involvement in any organization or entity with any financial interest or nonfinancial interest in the subject matter or materials discussed in this manuscript. There are no conflicts to report.

DATA AVA I L A B I L I T Y S TAT E M E N T
The data that support the findings of this study are available from the corresponding author upon reasonable request.