Synthesizing qualitative and quantitative evidence on non-financial access barriers: Implications for assessment at the district level

Introduction: A key element of the global drive to universal health coverage is ensuring access to needed health services for everyone, and to pursue this goal in an equitable way. This requires concerted efforts to reduce disparities in access through understanding and acting on barriers facing communities with the lowest utilisation levels. Financial barriers dominate the empirical literature on health service access. Unless the full range of access barriers are investigated, efforts to promote equitable access to health care are unlikely to succeed. This paper therefore focuses on exploring the nature and extent of non-financial access barriers. Methods: We draw upon two structured literature reviews on barriers to access and utilization of maternal, newborn and child health services in Ghana, Bangladesh, Vietnam and Rwanda. One review analyses access barriers identified in published literature using qualitative research methods; the other in published literature using quantitative analysis of household survey data. We then synthesised the key qualitative and quantitative findings through a conjoint iterative analysis. Results: Five dominant themes on non-financial access barriers were identified: ethnicity; religion; physical accessibility; decision-making, gender and autonomy; and knowledge, information and education. The analysis highlighted that non-financial factors pose considerable barriers to access, many of which relate to the acceptability dimension of access and are challenging to address. Another key finding is that quantitative research methods, while yielding important findings, are inadequate for understanding non-financial access barriers in sufficient detail to develop effective responses. Qualitative research is critical in filling this gap. The analysis also indicates that the nature of non-financial access barriers vary considerably, not only between countries but also between different communities within individual countries. Conclusions: To adequately understand access barriers as a basis for developing effective strategies to address them, mixed-methods approaches are required. From an equity perspective, communities with the lowest utilisation levels should be prioritised and the access barriers specific to that community identified. It is, therefore, critical to develop approaches that can be used at the district level to diagnose and act upon access barriers if we are to pursue an equitable path to universal health coverage. The literature reviews focused on MNCH in four countries: Bangladesh, Vietnam, Ghana and Rwanda. Their selection was based upon inclusion in an earlier UNICEF-Rockefeller Foundation project to assess national health insurance in Africa and Asia [13], which noted the importance of non-financial barriers to reduced access to health interventions. Each country was seen to have sufficient data sets including nationally standardised health and expenditure surveys and literature on health service access. In each country, UNICEF supports the use of Multi-Cluster Indicator Survey (MICS), Demographic and Health Survey (DHS) and various standardised household expenditure surveys that provide key sources of data. The Rockefeller Foundation also includes these four countries in their Transforming Health Systems (THS) initiative to advance an equity-focused approach to strengthening national and sub-national health systems. Building on the platform of meta-ethnography, the reviews used thematic synthesis [14] so as to conduct reviews that addressed questions relating to intervention need, appropriateness, acceptability and effectiveness, without compromising the key principles developed in systematic reviews [15]. A statistical meta-analysis seeks to locate every relevant study since failure to do so may undermine the statistical models that underpin the analysis, thereby biasing the results. In meta-ethnography and thematic synthesis, however, it is not necessary to locate all available studies, but to search iteratively [14]. As Dixon-Woods et al. [16] conclude, “rather than aiming for … identification and inclusion of all relevant literature, as would be required under conventional systematic review methodology, we saw the purpose of the searching phase as identifying potentially relevant papers to provide a sampling framework”. As ‘saturation of findings’ is used in primary qualitative research, so ‘conceptual saturation’ can be used in thematic synthesis, because the purpose is interpretative explanation rather than prediction. Thematic synthesis focuses on the development of concepts and theory rather than a meticulous summary of all data [17–19]. We therefore devised ‘structured’ literature reviews, derived from the methodology for systematic reviews with rigorous study selection and defined inclusion and exclusion criteria, but without seeking to incorporate every paper published [20]. Rather, both reviews focused on being a comprehensive corpus of literature inclusive of all themes that could have been identified. Links to the structured reviews are provided in the web annex to this paper. The search strategy sought to identify studies that addressed non-financial barriers to access for MNCH services in the four countries. We searched for studies that included qualitative data derived from a range of methodologies including in-depth interviews, focus group discussions and survey-based studies with open-ended responses. All electronic searches used keywords covering the main search domains and a refined search string per country was used as the basic search tool for PubMed and ProQuest (incorporating the International Bibliography of the Social Sciences (IBSS); Applied Social Sciences Index and Abstracts; American Economic Association’s electronic bibliography (EconLit); and Sociological Abstracts). We used a five-phase search and inclusion/exclusion strategy. Phase 1: The initial search included three parameters: the date of publication was 1 January 2000 to 31 December 2012; the language was restricted to English (search terms in both UK and US English were used); and the publication type in PubMed was ‘journal article’ and in ProQuest was ‘peer reviewed scholarly journal’. Results from the searches in PubMed and ProQuest resulted in 3,643 articles. Phase 2: Two researchers independently analysed each paper by title and abstract and attributed a score: 2 for definite inclusion; 1 for potential inclusion; 0 for definite exclusion. The scores per article were added together and 262 articles totalling 4, 3 and 2 points were included in the next round. Thus 3,381 articles were excluded because they failed to satisfy one or more of the selection criteria: they were not in English; they did not focus on one or more of the four target countries; they were not qualitative; or they were not on-topic. Phase 3: The full texts of the remaining 262 articles were reviewed for inclusion by two researchers. Any inconsistencies were discussed and resolved, and inclusion agreed through consensus. A further 174 papers were removed after the texts had been reviewed because: they were off-topic (103 papers); quantitative (60 papers); or methodologically inappropriate (11 papers). In addition, eight papers were discounted because the full text could not be accessed despite attempts to contact the primary author. The remaining 80 articles were carried forward for detailed review. Phase 4: The full texts of the remaining 80 articles were read in detail in preparation for data extraction. During this phase the team decided to focus on two tracer interventions: skilled delivery and immunisation. Articles relating to other aspects of MNCH (breastfeeding, infant nutrition, prevention of mother-to-child transmission, etc.) were excluded (12 papers). A further 15 papers were discounted because they were found to be off-topic or predominantly quantitative. Finally, 48 papers were included in the full analysis. Phase 5: The bibliographies of all included articles were searched for additional references, including those that might present contrary views. UNICEF country offices were contacted and a number of universities and schools of public health in each country, but these sources did not yield significant new material. A very small selection of grey literature was identified, but due to time limitations it was decided to focus the structured review entirely on scholarly articles published in peer-reviewed journals. No new literature was added at this stage. A customised appraisal form was developed drawing on several previously published tools for qualitative research and equity [21–23]. Appraisal was undertaken in order to understand each study on its own terms, including potential bias [24], with the intention to exclude any study in which the depth and breadth of the reported data were insufficient to suggest that the findings were trustworthy [25]. No study was excluded due to assessment of conduct (validity and robustness), assessment of reporting (transparency) or assessment of content and utility of findings [21]. 48 papers were analysed in full (18 focusing on Bangladesh; 19 focusing on Ghana; 11 focusing on Vietnam; and 0 focusing on Rwanda). Line-by-line coding was applied to each paper included in the review. As each subsequent paper was coded, the emerging trends were grouped together, clustering like with like and creating new codes when required. The second phase of analysis was to analyse the grouped trends and build thematic constructs per country. In this way comparisons were made within and across similar studies. In the final stage of analysis, the thematic constructs per country were synthesised into analytic themes articulating the overarching barriers to access for MNCH services. Each analytic theme contained various sub-themes, under which in-country specifics were clustered for comparison and contrast. The lead researcher for the qualitative review (JB) undertook the coding and organisation of themes by hand and no software programme was used. A second researcher then analysed the papers deductively, going back through the material and labelling each identified thematic construct as it appeared. Any inconsistencies or areas for further analysis were highlighted and discussed by the research team until all the material was resolved into appropriate themes. Through thematic synthesis of the literature reviewed, six analytical themes were identified that articulate non-financial barriers preventing access to MNCH services in the focus countries: i) perception of the condition; ii) home management and local treatment; iii) influence of family and community; iv) lack of autonomy and agency to act; v) physical accessibility; and vi) health facility and biomedical deterrents. Several authors have commented on difficulties associated with locating qualitative studies, mainly because of the poor indexing of qualitative research [26, 27]. We attempted to mitigate this limitation by supplementing the database searches with citation searching in retrieved papers, generic web searches and directly consulting experts. Although this process did not yield any new or significant material, it is possible that relevant articles may still have been inadvertently omitted. The initial search parameters, particularly restricting the language to English, may also have prevented the identification of relevant papers. This is likely to have contributed to the lack of papers from Rwanda, as a proportion of the literature focusing on MNCH in that country is published in French. Importantly, however, because of the methods employed, the synthesis of themes is comprehensive for Bangladesh, Ghana and Vietnam. Retaining the context of the research is an important aspect of synthesising qualitative literature. In our review, we were aware of the obvious contextual differences between Bangladesh, Ghana and Vietnam, but also the contextual differences found within each country. Including characteristics of the study population (e.g. their socio-economic status, ethnicity, religion, geographic location etc.) were important facets of the review and were particularly relevant in the secondary phase of the project, when the results were combined with those of the quantitative review (see below). Unfortunately, however, not every paper adequately reported the context of the qualitative research, nor the impact of the context on findings. Where possible, we systematically recorded the various contextual factors and highlighted the remaining gaps in reporting. Whilst direct findings are not necessarily generalisable, theoretical insights arising from the synthesis of the included studies should be transferrable to other similar settings. The search strategy was designed to identify studies that used household survey data to analyse and quantify non-financial barriers to access for MNCH services in the four focus countries. The literature search was conducted through online databases: PubMed; EconLit; PsycInfo; Sociological Abstracts; and the IBSS. The initial inclusion criteria stipulated that: the publication language was English; the study was conducted in one of the focus countries; the publication reported quantitative data (derived from surveys or comparable sources); the paper was published in a peer-reviewed academic journal, book or publicly available report (e.g. postgraduate dissertation); and the date of publication was between 1st January 2000 and 31st December 2012. These criteria constituted the basis of the search strings used for each database. Key terms such as ‘cross sectional’ and ‘healthcare surveys’ were embedded in the search to better carve out the applicability of existing survey data within the context of the study. The search strings were refined for each source depending on the database requirements. A three-step screening method identified the relevant publications. Studies were excluded if: (i) an inappropriate methodology had been used (e.g. a predominantly qualitative study); (ii) if the research focused exclusively on finance, financial access barriers or health insurance; and (iii) if the main focus was not on a health service. Step 1: Papers from the initial database searches were combined and duplicates were excluded. 1,092 were taken forward. Step 2: Two researchers independently analysed the 1,092 papers by title and abstract. The researchers graded every publication against five criteria as strong inclusion (5 of 5), weak inclusion (4 of 5) or (3 or less) exclusion, and each were discussed until consensus was achieved. This excluded 1,009 articles. Step 3: The full text of the remaining 83 publications was reviewed for eligibility and inclusion agreed by consensus between researchers, with 36 articles carried forward for detailed review. Snowballing and forward citation did not yield any additional studies. All 36 studies remaining were well conducted with a methodological approach rated as either high or acceptable quality, characterised by the presentation of a well-defined underlying model or conceptual framework plus methodological rigor that acknowledged limitations and biases. No study was removed during the appraisal phase. Data extraction from the 36 eligible papers employed a tabular synopsis to consolidate relevant information. The tables included for each paper: author; year of publication; type of health service in focus; target population; and the main study question. Study design information included the data and methodology used, data sources (type and name of survey, date of survey, study population) and the specific analytic approach used. A second table for each focus country included the types of barriers identified or the different determinants of treatment seeking behaviour reported by each study. Findings were categorised according to the determinants of health services utilisation based on individual, household and community characteristics and referred to ‘predisposing’, ‘enabling’ and ‘need’ factors, following Andersen and Newman [28]. The majority of the included 36 research papers analysed subsets of data from large multipurpose surveys, such as DHS. The description of specific methodological approaches and the accommodation of survey design features influencing estimation and inference were often too brief to assess the relative impact of different determinants of utilisation across different studies. Between the studies there was a significant variation of research objectives implying a risk that the individual study focus was obscured in the joint assessment as the findings were subsumed under alternate study questions. As a variable, access to health services cannot be measured directly. Although uptake is often used as a proxy for access, access does not always translate into service utilisation [9]. Aday and Andersen [29] characterise utilisation (along with satisfaction) as an ‘outcome indicator’ of access. Neither the underlying need, nor the degree of autonomy in the decision underlying utilisation, is satisfactorily captured by survey variables. Further, the barrier itself, as a manifestation of the lack of fit between the underlying need and the service, is not reflected in household survey variables that only measure uptake. Rather, these variables, conceptualised as broad predisposing and enabling factors of healthcare utilisation, describe particular aspects of the socio-economic and socio-cultural background that promote barriers, and thus only provide one side of the access paradigm. In reviewing studies using household survey data, we adopted a pragmatic approach based on the understanding that access precedes utilisation. The multivariate analyses of individual and community associated factors as they relate to service utilisation imply a focus on ‘determinants’ rather than ‘barriers’, even if there is a straightforward inverse relationship between the two concepts. Distinguishing different categories of determinants, such as ‘predisposing’ and ‘enabling’ factors, added complexity to the interpretation of findings, particularly if the assignment of factors to categories was ambiguous across studies. The interpretation of all variables in the respective survey requires a good understanding of the context of service provision, and the interpretation of a particular variable may change with context, e.g. the significance of education in rural vs. urban settings. Another limitation was the substantial time lag between data collection and publication in many of the publications, which may reduce the potential relevance of survey analysis in a health policy context. Finally, the literature search was restricted to English which may have prevented the identification of relevant papers in local languages. At the conclusion of both literature reviews, we (JB, MT) undertook a conjoint iterative analysis that aimed to synthesise the key qualitative and quantitative findings. Using a similar method to that developed for the review of qualitative studies, key findings were thematically analysed, highlighting points of complementarities and contradictions between the qualitative and quantitative data. This was not a reductionist exercise, rather the synthesis aimed to produce a more detailed and textured analysis than was possible by using either qualitative or quantitative data alone. This demonstrates the value of integrating qualitative and quantitative data collection and analysis to achieve an optimal mixed-methods approach from the outset, rather than synthesising the data retrospectively. The limitations outlined above for the qualitative and quantitative reviews remained valid in the synthesis of findings, which also exposed the risks of information loss and misinterpretation. The two reviews addressed fundamentally different approaches to tackle similar research questions, and the findings could not simply be matched. Whilst the qualitative studies generally elicited barriers to access, the quantitative studies emphasised factors that promoted access or were predictors of access, depending on how the survey variables were typically framed. This difference in perspective resulted in findings being formulated or presented in different ways. In the synthesis, definitions had to be refined and cross-checks conducted to ensure that like was being compared with like.