Abiiro et al. BMC Health Services Research 2014, 14:235 102443 http://www.biomedcentral.com/1472-6963/14/235 RESEARCH ARTICLE Open Access Developing attributes and attribute-levels for a discrete choice experiment on micro health insurance in rural Malawi Gilbert Abotisem Abiiro1,2*, Gerald Leppert3, Grace Bongololo Mbera4, Paul J Robyn5 and Manuela De Allegri1 Abstract Background: Discrete choice experiments (DCEs) are attribute-driven experimental techniques used to elicit stakeholders’ preferences to support the design and implementation of policy interventions. The validity of a DCE, therefore, depends on the appropriate specification of the attributes and their levels. There have been recent calls for greater rigor in implementing and reporting on the processes of developing attributes and attribute-levels for discrete choice experiments (DCEs). This paper responds to such calls by carefully reporting a systematic process of developing micro health insurance attributes and attribute-levels for the design of a DCE in rural Malawi. Methods: Conceptual attributes and attribute-levels were initially derived from a literature review which informed the design of qualitative data collection tools to identify context specific attributes and attribute-levels. Qualitative data was collected in August-September 2012 from 12 focus group discussions with community residents and 8 in-depth interviews with health workers. All participants were selected according to stratified purposive sampling. The material was tape-recorded, fully transcribed, and coded by three researchers to identify context-specific attributes and attribute-levels. Expert opinion was used to scale down the attributes and levels. A pilot study confirmed the appropriateness of the selected attributes and levels for a DCE. Results: First, a consensus, emerging from an individual level analysis of the qualitative transcripts, identified 10 candidate attributes. Levels were assigned to all attributes based on data from transcripts and knowledge of the Malawian context, derived from literature. Second, through further discussions with experts, four attributes were discarded based on multiple criteria. The 6 remaining attributes were: premium level, unit of enrollment, management structure, health service benefit package, transportation coverage and copayment levels. A final step of revision and piloting confirmed that the retained attributes satisfied the credibility criteria of DCE attributes. Conclusion: This detailed description makes our attribute development process transparent, and provides the reader with a basis to assess the rigor of this stage of constructing the DCE. This paper contributes empirical evidence to the limited methodological literature on attributes and levels development for DCE, thereby providing further empirical guidance on the matter, specifically within rural communities of low- and middle-income countries. Keywords: Discrete choice experiment, Attribute and attribute-levels development, Qualitative study, Micro health insurance, Rural communities, Malawi * Correspondence: gilbiiro@yahoo.com 1 Institute of Public Health, Medical Faculty, University of Heidelberg, Heidelberg, Germany 2 Department of Planning and Management, Faculty of Planning and Land Management, University for Development Studies, Wa, Ghana Full list of author information is available at the end of the article © 2014 Abiiro et al.; licensee BioMed Central Ltd. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly credited. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated. Abiiro et al. BMC Health Services Research 2014, 14:235 Page 2 of 15 http://www.biomedcentral.com/1472-6963/14/235 Background As an attribute-based experiment, the validity of a There is a growing interest in discrete choice experi- DCE largely depends on the researchers’ ability to ments (DCEs) as a means of eliciting stakeholder prefer- appropriately specify attributes and their levels [10]. A ences for healthcare interventions and policy reforms misspecification of the attributes and attribute-levels [1-5] to support the prioritization, design and imple- has great negative implications for the design and mentation of such interventions [6,7]. DCEs are an implementation of DCEs and a risk of producing erro- attribute-driven quantitative technique used to elicit neous DCE results, which can misinform policy imple- stated preferences for new products and interventions mentation. To reduce the likelihood of researcher bias, that are yet to be introduced into the market [8-11]. attribute development has to be rigorous, systematic, In DCEs, potential products or interventions are usu- and transparently reported [34]. Various methods have ally described by their characteristics, referred to as been applied to the development of DCE attributes. attributes, and each attribute is assigned a range of These include literature reviews, existing conceptual defined dimensions called attribute-levels [12]. The attri- and policy relevant outcome measures, theoretical butes of the interventions and their assigned levels are arguments, expert opinion review, professional recom- usually combined using experimental designs to produce mendations, patient surveys, nominal group ranking a set of hypothetical choice alternatives [12,13]. Res- techniques and qualitative research methods [2,34,35]. pondents are then presented with a sequence of two or A recent review by Coast et al. [34] casts doubts on more of these competing choice alternatives and are whether the process of attribute and attribute-levels asked to choose which alternative they prefer [1,2]. The development for DCEs is always rigorous, leading to attribute-levels determine the utility respondents will at- the identification of credible attributes, given the brev- tach to a particular characteristic of an intervention, and ity with which it has been reported in existing studies. hence, their choices or preferences [2]. Acknowledging the limitations of deriving attributes In low- and middle-income countries (LMICs), par- from the literature, Coast et al. [34] argue that qualita- ticularly in Sub-Saharan Africa (SSA), DCEs have been tive studies are best suited to derive attributes, since applied within the health sector to elicit job preferences they reflect the perspective and experiences of the of health workers [14-17], hospital quality assessment potential beneficiaries. They insist on the need to [18], priority setting in resource allocation [19], maternal accurately describe such qualitative studies and other health issues [20,21] and health system reforms [22]. In approaches used in deriving attributes and levels, to general, only a few DCEs, none of which are from allow the reader the possibility of judging the quality of LMICs, have elicited community preferences for a health the resulting DCE. There is, however, paucity of such insurance product as an intervention in its entirety [23-30]. descriptions in the existing literature, in high and low Specifically, the DCE methodology has not been used to income countries alike [35,36]. elicit community preferences for micro health insurance Our study aimed at filling this gap by documenting a (MHI), an innovative health care financing strategy which rigorous process of developing attributes and attribute- has received substantial attention in LMICs [31-33]. levels for the design of a DCE, to elicit community pref- MHI refers to any voluntary health insurance system erences for a potential MHI product in rural Malawi. that pools funds and risks from members of a commu- nity, or a socio-economic organization, to ensure that its Methods members have access to needed care without the risk of Study setting financial consequences [32]. MHI schemes are often The study was conducted in the rural districts of Thyolo implemented at the local level, targeting low-income and Chiradzulu in Southern Malawi. Malawi is a low- households who work in the informal sector [33]. The income country in SSA with a population of about 15 premiums paid by MHI members are usually community- million [37]. The two districts include approximately rated and the schemes often adopt participatory manage- 6.7% of the national population [38]. ment approaches, which allow for community invo- In Malawi, over 60% of all health services are pro- lvement in decision making [32,33]. The relevance of vided by the government in public health facilities; 37% applying a DCE to configure micro health insurance by the Christian Health Association of Malawi (CHAM); products in LMICs emanates from the absence of and the rest by individual private for-profit health markets for health insurance products in many such practitioners and traditional healers/herbalists [39]. In settings. This makes alternative product design and principle, healthcare is provided free of charge at point preference elicitation approaches that rely on market- of use in public facilities (tax-funded) and subsidized oriented strategies, less feasible in generating timely in CHAM facilities, while private providers rely on data to support the design and implementation of user payments [40]. In practice, however, the provision MHI interventions in such contexts [2]. of free healthcare is constrained by constant shortages Abiiro et al. BMC Health Services Research 2014, 14:235 Page 3 of 15 http://www.biomedcentral.com/1472-6963/14/235 of drugs and health personnel, and poor infrastructure attributes and levels derived from such a qualitative and equipment, resulting in poor quality, which in turn study are considered demand-driven [2], reflective of reflects poor health outcomes [40,41]. A considerable local perspectives, understandable to respondents and proportion of healthcare is still being financed through thereby, plausible within the study context [34]. Deriving direct out-of-pocket payments [40]. attributes from a qualitative study can, therefore, improve The average total healthcare expenditure stands at US the content validity of a DCE study [10]. A qualitative $34 per capita, equivalent to 11.7% of Gross Domestic study is also capable of picking up other context- Product (GDP) [42]. There is no nationwide social specific and policy relevant attributes which might not health insurance scheme, and only very limited coverage exist in the literature, and hence, can potentially of private and employer-based insurance schemes [39]. reduce the risk of omitting relevant attributes and Due to inadequacies in the current tax-funded system and attribute-levels. Lastly, the context specific attributes limited coverage of existing health insurance schemes, and attribute-levels must be framed in a manner that private not-for-profit institutions, including microfi- allows for efficient elicitation and analysis of preferences, nance institutions (MFIs), have expressed increasing according to random utility theory, which is the theoret- interest in becoming active agents for the development ical foundation of DCE [8]. In this case, DCE attributes of MHI, with the aim of increasing social health protec- (and most particularly levels) must be exhaustive and tion for informal sector workers and rural populations. measurable [2]. The attributes and their levels must The absence of evidence on community preferences be unambiguously framed [27] and appear both cogni- for an MHI product, within a predominantly tax-funded tively (perceptually) and statistically uncorrelated in healthcare context like Malawi, provided the rationale the choice sets [44]. Additionally, attributes must be for our overall DCE study. The intention of the largest experimentally manipulatable [44], and defined in a MFI in the country, the Malawian Union of Savings manner that gives room for trading between attribute- and Credit Cooperatives (MUSCCO), to introduce MHI level alternatives [34]. To ensure these, expert opinion through its Bvumbwe Savings and Credit Cooperative and additional pilot studies within the study area are (SACCO), in the Southern Region, provided the policy also recommended [10,34]. context for our study. Guided by the above conceptual reasoning, we adopted a multi-stage attribute development process, whereby we Conceptual framework for developing attributes and initially identified policy relevant conceptual attributes attribute-levels from a literature review. We used these conceptual attri- There is a growing consensus in the literature that cred- butes and potential attribute-levels as a basis for designing ible attributes and attribute-levels for a DCE must be a qualitative study to identify context-specific attri- policy relevant, important to the study population, and butes, as those deemed directly by respondents to be consistent with the random utility theoretical founda- most important. To scale down the context-specific tion of DCE [2,10,34,43]. Policy relevant attributes and attributes to a number manageable within a DCE and attributes-levels are those that adequately reflect the to ensure that the final attributes and levels conformed essential dimensions or characteristics of the product, to the theoretical postulations of a DCE, we elicited or intervention, that will be evaluated by potential bene- expert opinion and further validated our results through a ficiaries in the DCE [8]. This implies that the identifi- pilot study. cation of such attributes and levels should be guided by appropriate conceptual and theoretical explanatory Study design models and empirical literature on the policy issue. A The overall DCE study adopts the instrument develop- rigorous literature review on the policy topic can, there- ment variant of an exploratory sequential mixed methods fore, lead to the identification of a comprehensive list of design [45], cognizant of the systematic stage-wise nature conceptual attributes, which can potentially, but not of a DCE process [12]. In line with the methodological necessarily, be included in a relevant DCE. According to prescriptions of the exploratory mixed methods design, Coast et al. [34], identifying attributes and their levels a qualitative design informed by an initial literature re- exclusively on the basis of a literature review may be view was used in the first phase of the study, to elicit easier to implement, but may also lead to the non- the relevant attributes and attribute-levels to construct inclusion of some important attributes. To be included the DCE, and an actual DCE was used to collect and in the DCE, the conceptual attributes must be consid- analyze quantitative data in the second phase (see ered important by the target population, whose prefer- Figure 1 for illustration). As described above, in relation ences will be elicited in the final DCE, and reflect the to our conceptual framework, this paper focuses exclu- needs of their local context. This requires a rigorous sively on the first phase of the study, describing the qualitative study within the local context [34,36]. The qualitative component in detail. Abiiro et al. BMC Health Services Research 2014, 14:235 Page 4 of 15 http://www.biomedcentral.com/1472-6963/14/235 Figure 1 The mixed methods design of the DCE. Initial literature review purchasing, and service provision [46], provided a In line with recent methodological recommendations broad framework for attribute identification; [4,10,11,34], the attribute development process began 2. Berki & Ashcraft’s framework, which identified direct with a review aimed at identifying conceptual attributes insurance policy characteristics (benefit package, relevant to an MHI product in the available published premium price and cost-sharing provisions such as literature. PubMed, Google scholar, ScienceDirect, deductibles, copayment, coinsurance and benefit EMBASE and EBSCOhost databases were searched ceilings) and delivery system characteristics (quality, using as first level search terms: discrete choice experi- spatial access, comprehensiveness and continuity) as the ment, conjoint analysis, best worst scaling, preferences most important features that influence consumer elicitation, perceptions, and design features/enrollment/ choice when purchasing insurance [47], provided a retention/dropout, which were variously combined with complementary framework for understanding second level search terms such as: health insurance, mu- consumers’ health insurance choice behavior; tual health organizations, health (care) financing, universal 3. Existing legislations and policy documents on health (health) coverage and Health Maintenance Organizations insurance in SSA [48-53] and empirical literature on (HMOs). Only empirical papers or reviews, policy docu- community perceptions about MHI product ments and theoretical/conceptual frameworks on health- characteristics, and their relationship to enrolment care financing systems and consumer choice behavior, in MHI in SSA [31,32,54-63], provided evidence on published in English between 1980 and 2013 were con- how MHI is currently being implemented within sidered. For the sake of space, this paper does not discuss SSA; and the detailed results (e.g.: summaries of single papers), from 4. Attributes and attribute-levels used in previous the literature review, as would be the case in a system- DCEs on consumer preferences for health insurance atic review, but focuses on the insights from the litera- also gave insights into what features of health insurance ture that guided our identification of the conceptual can potentially be implemented within a DCE [23-30]. attributes and attribute-levels. In light of the specific circumstances of the Malawian context, a list of con- Guided by these insights from the literature, three of ceptual attributes was developed on the basis of four the authors (GAA, GL and MDA) derived a comprehen- main inputs from the literature: sive list of conceptual attributes and potential attribute- levels as illustrated in Table 1. The conceptual attributes 1. Kutzin’s framework, which defines the four main and their potential levels were used to guide the design components of any healthcare financing system as of data collection tools for the qualitative component of revenue collection, fund/risk pooling, service the study. http://www.biomedcentral.com/1472-6963/14/235 Abiiro et al. BMC Health Services Research 2014, 14:235 Table 1 Conceptual attributes and potential levels compiled from literature (adopted to the Malawian context) Functions [46] Based on the frameworks of Kutzin [46], Berki and Ashcraft [47], health insurance policy documents [48-53] literature on community perceptions on MHI characteristics in SSA [31,32,54-63] and attributes and levels defined in previous DCEs[23-30] Policy attribute Plausible levels definition (citations only provided for previous applications in DCEs) Revenue mobilization Who pays the premium Household members, employers [30], Government Unit of charging premium Individual, household [26], full family [23,27] Structure of premium • Flat rate [23,27] • Differential based on: income, employment, age, urban–rural Premium price (level) • Based on real cost of healthcare • Based on proposed/existing insurance premiums [23,29,30] • Based on WTP or qualitative studies [25-27] Forms of premium payment • Cash [23-27,29,30] • Material (farm produce) or both Premium payment mechanisms • Deduction from bank or payroll [23], institutional membership (MFI) account, salary • Pay through community agents • Pay directly to insurance office Premium collection modalities • Pay during wet, dry or all seasons • Pay weekly, two-weekly [26], monthly [23], yearly [29], installment Fund and risk pooling Unit of enrolment Individuals [26], households, families [23], microfinance institutional or occupational groups Dependents eligibility None, plus spouse, plus spouse and children [23] Extent of pooling Family/kin, community, Institutional(MFI) level, district, region, nation Nature of cross-subsidization • None • Based on income, employment, risk or geographical location status • Exemptions for poor and indigents Pooled fund Management and administration Who manages the pooled funds • Names of insurance provider [26,27] • Community committees, • Microfinance Institutions, • NGOs, Health providers, Governmental organization Quality of customer services Good, bad [25] Insurance information communication Not provided, weekly, monthly [26], yearly Enrollment procedure (paper work involved) • No forms to complete, few forms, lots of forms [26] Services purchasing Benefit package Comprehensive, medium, basic packages Page 5 of 15 Low cost vs. high cost events Low risk vs. high risk events Frequently occurring or rare events http://www.biomedcentral.com/1472-6963/14/235 Abiiro et al. BMC Health Services Research 2014, 14:235 Table 1 Conceptual attributes and potential levels compiled from literature (adopted to the Malawian context) (Continued) a. Specific services coverage • Hospitalization due to medical treatment or surgery [26] • Medical Consultation (by phone) [26] • Pharmaceuticals/drugs prescribed [25-27] • Preventive care, wellness and education [27] • Vision and hearing care [26,27] • Emergency services [26] • Mental health services [26,27] • Dental services [26,27] • Alcohol and substance abuse [26] • Treatment abroad or out of town emergency • Laboratory, x-ray and imaging • Maternal care • Consultations of traditional healers • Transportation • Loss of income when ill • Time loss of care giver b. Cost sharing arrangements Coverage ceiling (maximum liability) [28] benefits within specific facilities, communities, district, national, international Co-payments levels • None • Flat rate [23,30] • A percentage of cost (10%, 25%, 50%) [26,27] Deductibles [24,28] • Out-of-pocket payment for first visit • Insurance pays only at a certain quantum of cost Benefit delivery Cashless and re-imbursement Provision Type of providers Public, private, faith-based or all Choice of provider (facility) Choose any [27], limited to some, limited to one in the community [26], gatekeeper model Location of contracted provider • Defined in terms of distance from home or average travelling time to provider [23,26] • Defined setting: urban, rural Quality of care • Bad, moderate, good, very good, excellent [25-27] Reputation of affiliated providers Outstanding, average, below average [23] Waiting time for care Defined in terms of hours and minutes [26,29] Opening hours of health facility Only week days, weekends as well, nights and 24 hours [26] Page 6 of 15 Availability of providers Yes/no [23] Involvement in treatment decision making Yes/no [25] Abiiro et al. BMC Health Services Research 2014, 14:235 Page 7 of 15 http://www.biomedcentral.com/1472-6963/14/235 Identification of context-specific attributes through the Data collection qualitative study The list of conceptual attributes (Table 1), developed on Study population and sampling the basis of the existing literature, served as the basis for Qualitative data for the development of context-specific the development of one single interview/discussion attributes and attribute-levels was collected in August/ guide used to conduct all FGDs (see Additional file 1). September 2012, using 12 FGDs with community mem- The guide was adjusted to conduct the interviews with bers and 8 key informant interviews with health workers. health workers. The use of a guide was necessitated by Community residents (both MFI-members and non-MFI the limited familiarity of the concept of MHI among the members) were included as potential target clients of study participants and, hence, a need to provide modera- the future MHI product in the concerned districts. tors/interviewers with a common instrument, as a means Health workers were included as key informants because of ensuring uniformity in the topics to be discussed they observe the challenges communities currently face across all groups. The interview/FGD guide was semi- to access care. Triangulating community and providers’ structured around a list of open ended questions, views enhanced the study’s credibility [45]. Since the study including adequate probes. The guide was comprised deals with a non-sensitive topic, FGDs were deemed of two main sections. The first section aimed at deriv- appropriate for deriving attributes from community ing attributes, and hence, it contained broad questions residents, because of the potential of FGDs to yield on: how participants experience the healthcare system large amounts of consensual information from a broad and provision gaps; how participants would like an range of opinions on a specific topic over a relatively MHI scheme to be designed; the product attributes shorter period of time [8]. Moreover, it was relatively they would value as important when deciding whether easier to organize community residents for FGDs than or not to join; and the motivations for their responses. health workers, who were scattered all over the study Respondents were initially allowed to openly discuss area, and hence, could only feasibly be studied through the above topics. Afterwards, to ascertain their import- individual interviews [64]. ance, moderators probed for MHI characteristics that Stratified purposive sampling was used to select both were identified in the literature, but not spontaneously community residents and health workers, and the overall mentioned by the respondents during the FGDs. The sample size was determined by expected saturation point second section aimed at deriving specific attribute-levels. [64]. For community residents, we applied purposive Hence, using the comprehensive list of potentially relevant segmentation to achieve maximal variation, taking into attributes as a guide (Table 1), participants were asked to consideration possible diversity in opinions across geo- identify probable options for each attribute. graphic location, MFI membership status, and sex [64]. All FGDs were conducted in the local language First, five traditional authorities (TAs) were purposely (Chichewa) by the two research assistants; one serving as sampled to ensure geographical spread across the two facilitator and one as note-taker. Before the discussion, the districts. Second, one rural community from each TA facilitator provided respondents with a detailed explan- was selected, relying on evidence of the presence of ation of the MHI concept, using as illustrations locally sufficient MFI members. Third, in each selected commu- appropriate expressions and images (see Additional file 1). nity, adult (18+) individuals were selected to participate All FGDs were tape-recorded, transcribed, and trans- in one of two FGDs, one including MFI members ran- lated into English for analysis. FGDs lasted, on average, domly selected from the MUSCCO-MFI membership list 2 hours. All FGDs were conducted in secured, enclosed (with sex being held as sole purposive sampling criteria) places, such as schools or churches, free from external and one including non-MFI members sampled from the distraction. community. Men and women were separated into different All interviews with health workers were conducted in groups. Though women are generally more involved as mem- English, directly by GAA, tape-recorded, and later tran- bers in the local MFI than men, a total of 6 women’s groups scribed. Each interview lasted between 45 minutes and and 6 men’s groups were formed. Community leaders assisted one hour. the data collection team (GAA and two research assistants) to recruit 8 to 12 participants for each FGD. Ethical approval Health workers from health facilities in the concerned Ethical approval for the study was obtained from the areas were purposely selected to represent public, faith- Ethical Committee of the Faculty of Medicine of the based (CHAM), and private-for-profit providers. In University of Heidelberg in Germany and from the each sampled facility, the most experienced (senior) National Health Science Research Committee (NHSRC) health worker was selected for interview, resulting in a in Malawi. Before data collection took place, permission sample where almost all the 8 health workers were facil- was also obtained from the district commissioners, the ity heads. district medical officers, and the local authorities of the Abiiro et al. BMC Health Services Research 2014, 14:235 Page 8 of 15 http://www.biomedcentral.com/1472-6963/14/235 concerned communities. Written informed consent was two. This last step allowed for one last collective cred- obtained from all study participants. All sampled re- ibility and reality check on the list of retained attri- spondents consented to and participated in the study. butes and levels. Using the list of attribute and levels To enhance confidentiality, all FGD participants were retained at this stage, a quantitative DCE pilot study encouraged not to discuss each other ’s opinions out- was designed and administered to 49 respondents. The side the FGD setting. Also, to make it less possible for aim was to derive the parameters for the actual DCE respondents’ opinions to be easily linked to their personal design, to test other components of the DCE design identities, names of respondents were not recorded. We and to assess the clarity of the wording, as well as have adhered to the RATS guidelines for qualitative appropriateness of defined levels and local translations, research modified for BioMed Central instructions to and comprehensibility of attributes and levels within authors. the choice sets [10]. The last element is of specific rele- vance to the concepts and experiences described in this Data analysis paper. The interviewers working on the pilot were To ensure inter-researcher reliability, analysis began with specifically instructed to observe and document the an independent reading, coding, and categorizing of the respondents’ reactions and comments on the attributes qualitative transcripts by three different analysts [64]. GAA and attribute-levels used during the pilot. Their obser- analyzed the entire material using the computer assisted vations were discussed within the framework of an qualitative data analysis software NVivo (version 9). He FGD, bringing together all the interviewers. relied on a pre-established coding scheme developed on the basis of the FGD/interview guide and the concep- Results tual attributes identified in the literature, but allowed Qualitative analysis of the transcribed material and initial for new codes and categories to emerge as he proceeded attribute identification through the reading. MDA and GBM manually analyzed In total, 127 residents participated in the FGDs. These two-thirds of the material. They approached the mater- included: 64 from Thyolo and 63 from Chiradzulu dis- ial inductively, letting codes and categories emerge as tricts; 64 males and 63 females; and 61 SACCO and 66 they worked their way through the transcripts. At a later non-SACCO members. The eight health workers were stage, the three analysts compared the results of their comprised of two medical doctors, one from a CHAM analysis to obtain one single list of all elements identi- hospital and the other from a public district hospital; fied by community, and by providers, as attributes and two nurses/midwives, one from a CHAM hospital and relevant levels. Discrepancies in interpretation were the other a public district hospital; two medical assis- reconciled by returning to the text, “questioning” the tants/clinicians from the two public clinics; and a clin- transcribed material to identify which elements really ician and a paramedic from the two private health reflected an attribute and which ones did not. centers. The health workers from the private sector and the medical doctor from the CHAM facility had previ- Expert opinion ously worked in the public sector, while two of the pub- This step was aimed at reducing the attributes to a number lic sector workers had also previously worked in CHAM manageable within a DCE, by discussing the list of facilities. The health workers who participated in the context-specific attributes derived from the qualitative study had experience within the Malawian health system analysis with two sets of “informed” people, purposively ranging from 2 to 48 years. selected based on their experience with the DCE meth- Table 2 displays the complete list of all attributes and odology. These discussions served the purpose of en- attribute-levels identified by consensus among the three suring that the selected attributes were consistent with analysts during the initial triangulation process. They the methodological postulations of DCE. The list was include: premium level, premium collection modalities, also discussed in a group setting with five purposively premium structure, unit of enrolment, geographical level selected researchers familiar with Malawi and with of pooling, management structure, health services bene- MHI. This was to further ensure that the selected con- fit package, transportation coverage, copayment levels, structs not only appeared credible and realistic in the and provider network (i.e. the type of health facilities Malawian context, but also adequate to answer import- to be contracted by the MHI). To give voice to the ant pending research questions on community prefer- respondents’ views on attributes and their levels, direct ences for MHI in SSA. quotations, poignantly selected, from the qualitative transcripts are included in Table 2. Self-reflection and additional insights from a pilot study Attribute-levels were extracted directly from the tran- In this stage, the research team gathered to revise the list scripts, as illustrated by the relevant citations (Table 2). of attributes in light of the feedback received during step Only the three most relevant attribute-levels were defined http://www.biomedcentral.com/1472-6963/14/235 Abiiro et al. BMC Health Services Research 2014, 14:235 Table 2 Derivation of final list of DCE attributes and plausible levels (ordered from most preferred to least) Attribute label Lay terminology Key quotations from qualitative data (mostly FGDs) Labels of plausible levels Final inclusion Unit of enrollment How many family members will benefit • “If everybody in my family will benefit from this basket… it will be a good idea, … but • Entire extended family Yes from enrollment into the MHI scheme if I am the only person to benefit since I will be the one contributing into the basket, then it is not a good idea since I will still be paying hospital bills for my dependents” (Non-SACCO men) • “The head of the family should pay on behalf of the whole family” (SACCO women) • Core nuclear family • “If it offers a package covering them and their children, they will be more than happy • Individual to go for it” (Health worker at district hospital) Management The managers of the common basket • “Sometimes, just seeing the leaders who are managing this thing can make one to • Community committee Yes join or not” (SACCO men) • “There should be an elected committee to run the basket and trusted people” • An external NGO (SACCO women) • “I will be happy if this basket is managed by the community for easy monitoring and • Bvumbwe SACCO accessibility” (Non-SACCO men) • “If the basket can be managed by the NGOs it can be a good thing because if it is managed by people of this community…. if they buy chicken with their own money, people might think that they are misusing the money from the basket” (Non-SACCO men) • “I think the SACCCO can manage it but there should be a committee from the community …. linked to the SACCO, if it is managed by only SACCO there will be no trust” (SACCO-Men) Health service The health services that the MHI will • “There are some drugs which cannot be found at public hospitals except private hospitals, • Comprehensive: Drugs, lab test/ x- Yes benefit package pay for so this basket should cover these situations” (non-SACCO men). ray, and surgical operations • “(It should cover) x-ray and drugs, no more things (services) because we can’t manage • Medium: Drugs, lab tests/x-rays to pay” (Non-SACCO men) • “We have all agreed that medicine should be included in this basket” (SACCO women). • Basic: Drugs only • “They have to be sure that once they are putting money into this insurance, they are going to be covered properly” (health worker at private clinic) Copayment The proportion of health service bill that • “The basket should be assisting with half of the bill not the whole bill” (SACCO women) • None Yes a MHI member is expected to pay • “25% (from the patient) is fair ….. because we should think of others who will also need • 25% (quarter) out-of-pocket the basket” (non-SACCO men) • “It can happen that you are sick but you don’t have a single coin … the committee is • 50% (half) telling you, you will only get 50% of your charge from the basket, the other half will be paid by yourself…it will mean the basket will be of no use” (Non-SACCO men ) Transport Transport • “I will join …… if I fall sick and this basket will cover transport to the hospital“ • Always from home to the health Yes (SACCO Men). facility any time sic • “Private hospitals are very far from here so we need transport from this community to these private hospitals” (SACCO women) Page 9 of 15 • “Transport, because we have problems mainly in times of referral to Thyolo hospital” • Only during referral and (district hospital) (Non-SACCO Men) emergencies • “If they package involves offering transport to people from where ever they are to here, • none they will be more than happy to join” (health worker in public health center) http://www.biomedcentral.com/1472-6963/14/235 Abiiro et al. BMC Health Services Research 2014, 14:235 Table 2 Derivation of final list of DCE attributes and plausible levels (ordered from most preferred to least) (Continued) Premium per Membership contributions • “If the contributions will be unaffordable then I cannot join” (SACCO women) • MWK100 Yes person per month • “We will manage MWK100 per month, if they charge more than that; people will not • MWK300 be able to pay” (Non-SACCO-women) • ”We should agree on MWK500 per month” (Non-SACCO men) • MWK500 • “The amount of money to be contributed whether is it monthly or how often” (health worker, private clinic) Premium payment Frequency of premium contribution • “Here, most of us find money on a seasonal basis, so I think it would be ideal if we • Once-off annual payment No modalities contribute at the beginning of each and every year” (SACCO women) • Monthly payment • “Monthly contribution will help to have more money in the basket than annually” (non-SACCO men). Provider network Contracted healthcare facilities for • “When a person falls sick and goes to private hospital, he should use the money from • Private –for-profit No service provision by the MHI the basket to settle the bills because there is a difference between public and private hospitals in terms of treatment“ (non-Sacco men) • Faith-based (CHAM )facilities • “They will like to go to private facilities” (Health worker, public facility) • Public health facilities Pooling level Extent of geographical pooling • “Each and every village has to have its own basket” (non-SACCO Women) • Community level No • “I cannot be happy with district level” (non-SACCO Men) ”… there will be no trust and • Traditional Authority some will benefit from it while others will not benefit ……. unless it is at district level and managed by NGOs” (Non SACCO men) • District Premium structure Extent of dependency of contributions • “It should be one figure because everyone whether one earns more or less can fall sick • Flat rate contributions No on earnings so it should be one figure” (SACCO Men). • Contributions based on earnings Page 10 of 15 Abiiro et al. BMC Health Services Research 2014, 14:235 Page 11 of 15 http://www.biomedcentral.com/1472-6963/14/235 for each attribute, to ensure design simplicity and easy for premium collection modalities will depend on the recognition by respondents [10]. Only two attributes, premium amount – see Table 2. Second, attributes for premium level and health service benefit package, which clear preference was established in the FGDs for deserve further explanation. certain levels were dropped to avoid dominance. There In line with existing methodological recommendations was clear preference for: private-for-profit and CHAM [44], levels for the premium were set to reflect the facilities (as a proxy for quality of care); fixed rate pre- complete range of amounts agreed upon in the FGDs. mium payments; and pooling at the community level. Fi- The assumption was that the later DCE should elicit a nally, attributes were dropped if, in the FGDs, they had realistic marginal willingness-to-pay (WTP) value, rather been identified as elements of secondary importance, than reflecting the actual cost of the MHI product (which such as pooling level, which entered the discussions only needs to be subsidized in any case). Levels for the health after persistent probing. However, fixed levels were service benefit package were derived by combining the defined for all discarded attributes as part of the intro- single services frequently mentioned during the FGDs duction to the choice exercise. This reduces the ten- (drugs, laboratory tests, surgery) into meaningful in- dency of respondents inferring levels for such attributes cremental clusters. FGD participants mainly argued which can potentially introduce unobservable biases into that the benefit package should only include services the final DCE estimates [8]. for which they identified a current lack of effective coverage through public provision. Some services were Step three: Final attribute selection and revision in the mentioned as important, such as maternity care, but light of results from the pilot study recognized as adequately provided by governmental After the reduction and revision process of step two had facilities. These were excluded from the benefit pack- taken place, the research team once again discussed the age, with the rationale that MHI will be set to fill gaps in relevance of the selected items, their feasibility, and com- coverage and not to substitute existing public service prehensibility in the local context. Only minor changes in provision [31]. terminology were applied to the attribute levels. The core team agreed that all attributes and levels selected during “Maternal care should not be in the basket because; step two satisfied the essential characteristics of a DCE such complications are in the hands of the public attribute, i.e., they reflected the characteristics of an MHI hospitals. Any time there are such cases, the hospital product; were deemed important by the community; were calls the ambulance to assist by taking the patient to understandable; and mutually exclusive in nature [34], and the district hospital, so no need for antenatal mothers retained them for the final DCE. to be included in the basket” (Non-SACCO men). The analysis of the final DCE pilot results (run primar- ily to generate prior parameters for the DCE design) Step two: Selecting relevant attributes in the light of confirmed the theoretical validity of the defined attri- experts’ feedback butes and levels, since all had the expected signs, though The iterative process of discussion with additional scien- few were significant; probably due to insufficient sample tists led to the retention of 6 out of the initial 10 attributes size (n = 49). The FGD with the four research assistants identified in the qualitative material. The discussion was who administered the pilot study revealed that respon- oriented to limit the number of attributes to between 4 dents did not raise any major concerns relating to the and 8, in order to later allow the DCE to contain a man- appropriateness of the defined attributes and levels. Only ageable number of alternatives, that would not overwhelm a few minor revisions were made to the local transla- respondents [1]. The last column of Table 2 indicates tions of the attributes and attribute-levels. The pilot, whether an attribute identified during step one was therefore, enabled the confirmation and validation of the retained in step two. The discussion with additional scien- final framing of the attributes and attribute-levels, as tists also allowed the team to redefine the language used illustrated in Table 2. The pilot also indicated that par- to describe both the attributes and the relevant levels, ticipants had no cognitive difficulties in identifying and often requiring a return to the original text to identify the understanding the attributes and their levels. The inter- specific terminology used by the community. This was viewers argued that this result was achieved due to the meant to ensure consistency with the Malawian context. fact that attributes and their levels were illustrated to Multiple criteria guided the choice of attributes to be respondents using context-specific pictures. dropped. First, attributes and/or levels that overarched/ overlapped other attributes were discarded in order to Discussion avoid cognitive inter-attribute correlation [44]. For in- This paper contributes to the literature on DCE attribute stance, pooling levels overlapped management structure and attribute-level development [35,36], by explicitly since both had a geographical dimension; or preferences reporting on the systematic process of deriving attributes Abiiro et al. BMC Health Services Research 2014, 14:235 Page 12 of 15 http://www.biomedcentral.com/1472-6963/14/235 and attribute-levels for a DCE to elicit preferences for generally preferable for attribute derivation, because of an MHI product in rural Malawi. This study built on the its ability to constantly adopt the research questions in initial identification of conceptual attributes from the the light of emerging findings. Within the particular literature to develop a detailed interview/discussion context of our study, however, such an approach would guide used to gather primary qualitative data at the have not been feasible for a number of reasons. Geo- community level in a systematic manner. A rigorous graphical distance between the research team and the analytical process, characterized by three sequential concerned communities, as well as obvious language steps, allowed for the identification of relevant attri- barriers, made it impossible for the researchers them- butes and their levels. selves to engage in a constant iterative process during all Basing the interview guide on the results of the initial phases of data collection and analysis. Feasibility con- literature review, spanning from conceptual to applied cerns dictated the organization of the data collection studies, allowed the research team to identify a prelimin- and analysis phases. An iterative constant comparative ary broad series of attributes and attribute-levels that approach, however, was applied within an analytical reflected all possible important, and hence policy rele- process, also supported by the rigorous application of vant, components of an MHI product. Directly engaging the triangulation principle. Had the analysis revealed with communities and health workers allowed the that saturation had not been reached, however, the research team to work through this initial conceptual research team would have returned to the field to gather and very comprehensive list, to select context-specific more data [64]. The experience reported in this paper attributes that were understandable and important in indicates that in the event of feasibility constraints of the eyes of the potential beneficiaries of the insurance adopting a fully iterative approach to data collection and scheme [34]. The citations that accompany the attributes analysis, other rigorous qualitative approaches can yield and the relevant levels, in Table 2, offer a clear indica- equally relevant results for the development of credible tion of how decisions on attribute and levels selection attributes and attribute-levels. were rooted in the voices of the potential beneficiaries. Most prior qualitative studies aimed at deriving attri- The qualitative process also provided a clear understand- butes were conducted among people who had experienced ing of the likely order of preferences (most to least pre- the phenomenon under consideration [26,36]. The limited ferred) for the various attributes levels. This enabled the exposure of our participants to health insurance schemes design of DCE packages to actually compel respondents represented a major challenge. This compelled us to seek to make trade-offs in their choices [34]. out innovative ways of explaining the concept of MHI This initial qualitative phase, and the attribute valid- using appropriate local images and diagrams, and adjust- ation pilot study, also offered the research team the ing MHI social marketing concepts and illustrations from added benefit of framing the final DCE choice sets in other SSA settings to fit local socio-cultural constructs line with local concepts and terminology. This has the (see Additional file 1) [31]. The concern that the original potential of maximizing response efficiency in our DCE, framing of the FGDs might have influenced the partici- thereby enhancing the content validity of the study pants’ responses, however, was dissipated by the fact that [1,2,10]. The qualitative process also offered the oppor- findings from the individual interviews with health pro- tunity to identify and exclude attributes and levels that viders largely confirmed findings from the FGDs. Since are potentially dominant, less tradable, less important, MHI represents one of the many financing options being and perceptually correlated, from the choice sets, in discussed at a higher policy level, health workers, unlike order to fully satisfy the credibility criteria of DCE attri- communities, had already been exposed to the concept at butes and levels [2,34,44]. the time of the study and could not have been influenced Four of the final attributes derived - premium level, by our framing. management structure, health service benefit package, Based on the experience of this team, the analysis of and copayment levels - reflect what had been used in the data generated from this type of qualitative study is prior DCEs exploring preferences for health insurance often challenging. This is because while qualitative products in high income settings [23-30]. However, unit studies often generate large volumes of data, attribute of enrollment, as defined in our study, and transpor- development requires only little information on what tation coverage might not have been included had we community members see as important attributes and relied only on the literature review. This supports the levels. Given the amount of time and resources that are relevance of conducting qualitative studies to enhance often spent collecting data, researchers could develop the the contextual appropriateness of DCE attributes and impression that not all the data, such as the detailed illu- levels development [8,10,34]. minations and explanations of points provided by the Coast et al. [34] argue that an iterative constant com- study subjects during the FGDs and interviews, have been parative approach to data collection and analysis is adequately used. Moreover, it is a common tradition in Abiiro et al. BMC Health Services Research 2014, 14:235 Page 13 of 15 http://www.biomedcentral.com/1472-6963/14/235 public health that scholars cherish results that are statisti- Competing interests cally representative of the study subjects [34]. A qualitative The authors declare that they have no competing interests. study is not always able to generate this “representative” Authors’ contributions information, since such studies aim at illuminating com- GAA, GL and MDA conceptualized and designed the study and its data plexities and revealing similarities and differences, instead collection tools. GBM supported the design of the data collection tools. GAA administered and transcribed the interviews with health care workers, and of counting opinions [64]. Selecting attributes and levels supervised the data collection. GBM supervised the transcription of the FGDs. based only on qualitative studies, as in our case, could All authors participated in the data analysis. GAA wrote the first draft of the attract criticisms from quantitatively biased researchers, manuscript. GBM, PJR, GL and MDA revised the draft. All authors read and approved the final manuscript. who may argue that at least basic quantitative tools, such as best-worst scaling and nominal group ranking tech- Acknowledgement niques, should be included within the qualitative approach This study was supported by the German Research Society (DFG). We would in selecting attributes [35]. Therefore, it could be a good like to thank Assistant Professor Aleksandra Torbica, Bocconi University; Dr. Nasir Umar, London School of Hygiene and Tropical Medicine; Dr. Aurelia idea to use such simple quantitative tools, after the rigor- Souares, and the Health Financing Group of the Institute of Public Health, ous qualitative exercise, to support the scaling down of the University of Heidelberg for their scientific support in the design and potentially numerous attributes and levels, that will be implementation of the study. We are grateful to the staff and field research assistants of Reach Trust, Malawi, in particular Mr. Helecks Mtengo and Mrs. generated from the qualitative study, to a number man- Miriam Matengula, for their support during data collection. We are also ageable within the DCE. In this case, it must still be guar- grateful to our professional proof-reader, Amy Rue. anteed that the final attributes and levels selected are Author details capable of being used within the DCE, and this would still 1 Institute of Public Health, Medical Faculty, University of Heidelberg, require qualitative reasoning and deductions. Heidelberg, Germany. 2Department of Planning and Management, Faculty of Planning and Land Management, University for Development Studies, Wa, Ghana. 3Department for Cooperative Studies, University of Cologne, Cologne, Conclusion Germany. 4Research for Equity and Community Health Trust (REACH Trust), This study complements existing literature on DCE Lilongwe, Malawi. 5The World Bank, Washington, DC, USA. attribute development, by providing a detailed account Received: 9 December 2013 Accepted: 6 May 2014 of the scrupulous application of recently recommended Published: 22 May 2014 approaches to attribute and attribute-level development and reporting [10,34]. Our applied approach is based on References 1. De Bekker-Grob EW, Ryan M, Gerard K: Discrete choice experiments in the adoption of literature as the starting point, to inform health economics: a review of the literature. Health Econ 2012, 21:145–172. comprehensive field qualitative data collection, followed 2. Kjær T: A review of the Discrete Choice Experiment-with Emphasis on its by a rigorous analytical approach, supported by a series of Application in Health Care. Denmark: Syddansk Universitet; 2005. 3. Mengoni A, Seghieri C, Nuti S: The Application of Discrete Choice Experiments triangulation and validation exercises. As such, our study in Health Economics: A Systematic Review of the Literature. Working Paper n. provides additional empirical guidance on the methodo- 01/2013. Scuola Superiore Sant’Anna di Pisa. Istituto di Management; 2013. logical processes of developing attributes and attribute- [http://www.idm.sssup.it/wp/201301.pdf]. Accessed on 28-07-2013. 4. Bridges JFP, Hauber AB, Marshall D, Lloyd A, Prosser LA, Regier DA, Johnson levels for DCEs specifically within rural communities in FR, Mauskopf J: Conjoint analysis applications in health–a checklist: a LMICs. A transparent description of the attribute devel- report of the ISPOR Good Research Practices for Conjoint Analysis Task opment process of DCEs provides useful grounds for Force. Value Health 2011, 14:403–413. 5. Lagarde M, Blaauw D: A review of the application and contribution of the assessment of the rigor of this process in DCEs [34], discrete choice experiments to inform human resources policy and hence, should receive more attention in future DCE interventions. Hum Resour Health 2009, 7:62. studies. The potential of DCEs to support the design 6. Van Helvoort-Postulart D, van der Weijden T, Dellaert BG, De Kok M, Von Meyenfeldt MF, Dirksen CD: Investigating the complementary value and implementation of interventions, therefore, largely of discrete choice experiments for the evaluation of barriers and depend on the credibility of the attributes and attribute- facilitators in implementation research: a questionnaire survey. levels used in the experimental design. Implement Sci 2009, 4:10. 7. Farley K, Thompson C, Hanbury A, Chambers D: Exploring the feasibility of Conjoint Analysis as a tool for prioritizing innovations for Additional file implementation. Implement Sci 2013, 8:56. 8. Louviere JJ, Hensher DA, Swait JD: Stated Choice Methods: Analysis and Additional file 1: Data collection instruments. Applications. Cambridge: University Press; 2010. 9. Lancsar E, Louviere J: Conducting discrete choice experiments to inform healthcare decision making. Pharmacoecon 2008, 26:661–677. Abbreviations 10. Mangham LJ, Hanson K, McPake B: How to do (or not to do) … Designing CHAM: Christian Health Association of Malawi; DCE: Discrete choice a discrete choice experiment for application in a low-income country. experiment; FGD: Focus Group Discussion; GDP: Gross Domestic Product; Health Policy Plan 2009, 24:151–158. HMO: Health Maintenance Organization; LMICs: Low – and Middle-income 11. WHO: How to Conduct a Discrete Choice Experiment for Health Workforce countries; MFI: Micro finance Institution; MHI: Micro Health Insurance; Recruitment and Retention in Remote and Rural Areas: A User Guide with Case NHSRC: National Health Science Research Committee; MUSCCO: Malawi Studies. Geneva: World Health Oragnisation; 2012. Union of Savings and Credit Cooperatives; SACCO: Savings and Credit 12. Louviere JJ, Carson R, Pihlens D: Design of discrete choice experiments: a Cooperatives; SSA: Sub-Saharan Africa; TAs: Traditional Authorities; US: United discussion of issues that matter in future applied research. J Choice States; WTP: Willingness to pay. Model 2011, 4:1–8. Abiiro et al. BMC Health Services Research 2014, 14:235 Page 14 of 15 http://www.biomedcentral.com/1472-6963/14/235 13. Johnson RF, Lancsar E, Marshall D, Kilambi V, Mühlbacher A, Regier DA, 35. Hiligsmann M, Van Durme C, Geusens P, Dellaert J, Dirksen CD, van der Bresnahan BW, Kanninen B, Bridges JFP: Constructing Experimental Weijden T, Boonen A: Nominal group technique to select attributes for Designs for Discrete-Choice Experiments: Report of the ISPOR Conjoint discrete choice experiments: an example for drug treatment choice in Analysis Experimental Design Good Research Practices Task Force. Value osteoporosis. Patient Preference Adherence 2013, 7:133–139. Health 2013, 16:3–13. 36. Coast J, Horrocks S: Developing attributes and levels for discrete choice 14. Blaauw D, Erasmus E, Pagaiya N, Tangcharoensathein V, Mullei K, Mudhune experiments using qualitative methods. J Health Serv Res Policy 2007, S, Lagarde M: Policy interventions that attract nurses to rural areas: a 12:25–30. multicountry discrete choice experiment. Bull World Health Organ 2010, 37. World Bank: Malawi Country Data Profile. Country report Malawi. The World 88:350–356. Bank; 2014. [http://www.worldbank.org/en/country/malawi]. 15. Chowdhury ME, Johnson JC, Gyakobo M, Agyei-Baffour P, Asabir K, Kotha Accessed on 9/05/2014. SR, Dzodzomenyo M: Rural practice preferences among medical students 38. National Statistical Office: Population and Housing Census: Peliminary Report. in Ghana: a discrete choice experiment. Bull World Health Organ 2010, Malawi: National Statistical Office; 2008:2008. 88:333–341. 39. Phiri I, Masanjala W: Willingness to pay for micro health insurance in 16. Mangham LJ, Hanson K: Employment preferences of public sector nurses Malawi. In Handbook of Micro Health Insurance in Africa. Edited by Rösner in Malawi: results from a discrete choice experiment. Tropical Med Int H-J, Leppert G, Degens P, Ouedraogo L-M. Berlin: Lit Verlag; 2012:285–308. Health 2008, 13:1433–1441. 40. Zere E, Walker O, Kirigia JM, Zawaira F, Magombo F, Kataika E: Health 17. Rockers PC, Jaskiewicz W, Wurts L, Kruk ME, Mgomella GS, Ntalazi F, financing in Malawi: evidence from national health accounts. BMC Int Tulenko K: Preferences for working in rural clinics among trainee health Health Human Rights 2010, 10:27. professionals in Uganda: a discrete choice experiment. BMC Health Serv 41. Abiiro GA, Mbera GB, De Allegri M: Gaps in Universal Health Coverage in Res 2012, 12:212. Malawi: a qualitative study in Rural Communities. BMC Health Serv Res 18. Hanson K, McPake B, Nakamba P, Archard L: Preferences for hospital 2014, 14:234. quality in Zambia: results from a discrete choice experiment. Health Econ 42. Ministry of Health (Malawi): Malawi Health Sector Strategic Plan 2011–2016 - 2005, 14:687–701. Moving towards equity and quality. Lilongwe: Ministry of Health; 2011. 19. Baltussen R, Stolk E, Chisholm D, Aikins M: Towards a multi-criteria approach 43. Ryan M, Scott DA, Donaldson C: Valuing health care using willingness to for priority setting: an application to Ghana. Health Econ 2006, 15:689–696. pay: a comparison of the payment card and dichotomous choice methods. 20. Kruk ME, Paczkowski MM, Tegegn A, Tessema F, Hadley C, Asefa M, Galea S: J Health Econ 2004, 23:237–258. Women’s preferences for obstetric care in rural Ethiopia: a population- 44. Hensher DA, Rose JM, Greene WH: Applied Choice Analysis: A Primer. based discrete choice experiment in a region with low rates of facility Cambridge: University Press; 2005. delivery. J Epidemiol Community Health 2010, 64:984–988. 45. Creswell JW, Clark PVL: Designing and Conducting Mixed Methods Research. 21. Van Rijsbergen B, D’Exelle B: Delivery Care in Tanzania: A Comparative 2nd edition. Thousand Oaks: SAGE Publications; 2010. Analysis of Use and Preferences. World Dev 2013, 43:276–287. 46. Kutzin J: A descriptive framework for country-level analysis of health care 22. Kruk ME, Rockers PC, Tornorlah VS, Macauley R: Population preferences for financing arrangements. Health Policy 2001, 56:171–204. health care in liberia: insights for rebuilding a health system. Health Serv 47. Berki SE, Ashcraft MLF: HMO enrollment: who joins what and why: a Res 2011, 46:2057–2078. review of the literature. Milbank Memorial Fund Quarter Health Soc 1980, 23. Akaah IP, Becherer RC: Integrating a consumer orientation into the 58:588–632. planning of HMO programs: an application of conjoint segmentation. 48. Government of Ghana: National Health Insurance Act. 650. Accra: J Health Care Mark 1982, 3:9–18. Government of Ghana; 2003. 24. Becker K, Zweifel P: Age and choice in health insurance: evidence from a 49. Federal Government of Nigeria: National Health Insurance Scheme Decree No discrete choice experiment. Patient 2008, 1:27–40. 35 of 1999. Abuja: Federal Govergment of Nigeria; 1999. 25. Van den Berg B, Van Dommelen P, Stam P, Laske-Aldershof T, Buchmueller T, 50. Department of Health: National Health Insurance in South Africa, Policy paper. Schut FT: Preferences and choices for care and health insurance. Soc Sci Med South Africa: Department of Health; 2011. 2008, 66:2448–2459. 51. Ministry of Health: Rwanda National Health Insurance Policy. Kigali: Ministry 26. Chakraborty G, Ettenson R, Gaeth G: How consumers choose health of Health; 2010. insurance. J Health Care Mark 1994, 14:21–33. 52. Ministry of Health: National Social Health Insurance Bill. Nairobi: Ministry of 27. Gates R, McDaniel C, Braunsberger K: Modeling Consumer Health Plan Health; 2004. Choice Behavior To Improve Customer Value and Health Plan Market 53. Government of Tanzania: The Community Health Fund ACT, 2001. Dar es Share. J Bus Res 2000, 48:247–257. salaam: Government of Tanzania; 2001. 28. Hershey JC, Kunreuther H, Schwartz JS, Williams SV: Health insurance under 54. Criel B, Waelkens MP: Declining subscriptions to the Maliando Mutual competition: would people choose what is expected? Inquiry 1984, Health Organisation in Guinea-Conakry (West Africa): what is going 21:349–360. wrong? Soc Sci Med 2003, 57:1205–1219. 29. Vroomen JM, Zweifel P: Preferences for health insurance and health 55. De Allegri M, Kouyaté B, Becher H, Gbangou A, Pokhrel S, Sanon M, status: does it matter whether you are Dutch or German? Eur J Health Sauerborn R: Understanding enrolment in community health insurance Econ 2011, 12:87–95. in sub-Saharan Africa: a population-based case–control study in rural 30. Wellman GS, Vidican C: Pilot study of a hierarchical Bayes method for Burkina Faso. Bull World Health Organ 2006, 84:852–858. utility estimation in a choice-based conjoint analysis of prescription 56. De Allegri M, Sauerborn R, Kouyate B, Flessa S: Community health benefit plans including medication therapy management services. insurance in sub-Saharan Africa: what operational difficulties hamper its Res Social Adm Pharm 2008, 4:218–230. successful development? Tropical Med Int Health 2009, 14:586–596. 31. Matul M, McCord MJ, Phily C, Harms J: The landscape of micro health 57. Twikirize JM, O’Brien C: Why Ugandan rural households are opting to pay insurance in sub-Saharan Africa. In Handbook of Micro Health Insurance in community health insurance rather than use the free healthcare services. Africa. Edited by Rösner H-J, Leppert G, Degens P, Ouedraogo L-M. Berlin: Lit Int J Soc Welf 2012, 21:66–78. Verlag; 2012:59–87. 58. Mulupi S, Kirigia D, Chuma J: Community perceptions of health insurance 32. Basaza R, Pariyo G, Criel B: What are the emerging features of community and their preferred design features: implications for the design of health insurance schemes in east Africa? Risk Manage Healthc Policy 2009, universal health coverage reforms in Kenya. BMC Health Serv Res 2013, 2:47–53. 13:474. 33. Criel B, Atim C, Basaza R, Blaise P, Waelkens MP: Community health 59. De Allegri M, Sanon M, Bridges J, Sauerborn R: Understanding consumers’ insurance (CHI) in sub-Saharan Africa: researching the context. Tropical preferences and decision to enrol in community-based health insurance Med Int Health 2004, 9:1041–1043. in rural West Africa. Health Policy 2006, 76:58–71. 34. Coast J, Al-Janabi H, Sutton EJ, Horrocks SA, Vosper AJ, Swancutt DR, Flynn 60. Chankova S, Sulzbach S, Diop F: Impact of mutual health organizations: TN: Using qualitative methods for attribute development for discrete evidence from West Africa. Health Policy Plan 2008, 23:264–276. choice experiments: issues and recommendations. Health Econ 2012, 61. Royalty AB, Hagens J: The effect of premiums on the decision to 21:730–741. participate in health insurance and other fringe benefits offered by the Abiiro et al. BMC Health Services Research 2014, 14:235 Page 15 of 15 http://www.biomedcentral.com/1472-6963/14/235 employer: evidence from a real-world experiment. J Health Econ 2005, 24:95–112. 62. Jehu-Appiah C, Aryeetey G, Agyepong I, Spaan E, Baltussen R: Household perceptions and their implications for enrolment in the National Health Insurance Scheme in Ghana. Health Policy Plan 2012, 27:222–233. 63. Onwujekwe O, Onoka C, Uguru N, Nnenna T, Uzochukwu B, Eze S, Kirigia J, Petu A: Preferences for benefit packages for community-based health insurance: an exploratory study in Nigeria. BMC Health Serv Res 2010, 10:162. 64. Patton MQ: Qualitative Research & Evaluation Methods. 3rd edition. Thousand Oaks: SAGE Publications; 2002. doi:10.1186/1472-6963-14-235 Cite this article as: Abiiro et al.: Developing attributes and attribute- levels for a discrete choice experiment on micro health insurance in rural Malawi. BMC Health Services Research 2014 14:235. Submit your next manuscript to BioMed Central and take full advantage of: • Convenient online submission • Thorough peer review • No space constraints or color figure charges • Immediate publication on acceptance • Inclusion in PubMed, CAS, Scopus and Google Scholar • Research which is freely available for redistribution Submit your manuscript at www.biomedcentral.com/submit