|Home | About | Journals | Submit | Contact Us | Français|
By using the Global Fund as a case example, we aim to critically evaluate the evidence generated from 2002 to 2009 for potential negative health system effects of Global Health Initiatives (GHI).
Systematic review of research literature.
All interventions potentially affecting health systems that were funded by the Global Fund.
Negative health system effects of Global Fund investments as reported by study authors.
We identified 24 studies commenting on adverse effects on health systems arising from Global Fund investments. Sixteen were quantitative studies, six were qualitative and two used both quantitative and qualitative methods, but none explicitly stated that the studies were originally designed to capture or to assess health system effects (positive or negative). Only seemingly anecdotal evidence or authors’ perceptions/interpretations of circumstances could be extracted from the included studies.
This study shows that much of the currently available evidence generated between 2002 and 2009 on GHIs potential negative health system effects is not of the quality expected or needed to best serve the academic or broader community. The majority of the reviewed research did not fulfil the requirements of rigorous scientific evidence.
The factors that have undermined and eroded health system performance in many low- and middle-income countries have been debated extensively ever since the emergence of the major Global Health Initiatives (GHI),1–4 with the assertion that they undermine the performance of already weak national health systems by bypassing them.5–8 It has been argued, however, that this criticism would be mainly based on pre-existing assumptions, impressions and beliefs about health systems in developing countries, stakeholder interviews, descriptive cross-sectional case studies, and commentaries and opinion pieces.5,9,10
The purpose of this systematic review is to collate and critically evaluate the available scientific evidence on the negative health system effects of GHIs. We focus on negative health system effects because these have been a source of criticism for GHIs and if true, have important implications for policy-makers. We will use the Global Fund as a case example, because it is currently one of the largest international financing institutions supporting disease-specific programmes in low- and middle-income countries.11,12 These results are expected to apply, to a large extent, to other GHIs as well, because we do not have any reason to believe that the research assessing the Global Fund would be, in general, systematically different in quality than the research conducted on the other GHIs.
This review aims to add to the current debate presented in recent comprehensive reviews,5,9 by critically assessing various aspects of methodological quality affecting the interpretation and application of the evidence base generated by current research, and which were not covered in detail in earlier reviews. We assess the evidence and how the evidence is presented, as uncritical repetition of anecdotal evidence carries the risk of generating a ‘socially constructed reality’, where unsubstantiated claims and perceptions of health system effects could eventually be accepted as a valid representation of the objective reality.13,14 Therefore, to understand the arguments and concerns expressed by the stakeholders and other actors in the field, we explore the current discourses and bring them under critical evaluation.
All interventions were required to be funded by the Global Fund, and the interventions had to be related to at least one of the six health systems building blocks as defined by the World Health Organization (WHO): service delivery; health workforce; health information; medical products, vaccines and technologies; financing; and leadership and governance.15 We did not set specific criteria for study designs or methods of data analysis. We used the following inclusion criteria when assessing studies for eligibility: papers must clearly state the Global Fund's involvement; relate results to health systems; be published in peer-reviewed scientific journals and use original data, either in the form of primary data or secondary data used as a basis for new analysis.
We identified relevant original studies using a comprehensive list of electronic bibliographic databases, with a highly sensitive search strategy and without language restrictions, to avoid both selection bias of published articles and language bias of publications. We limited our search to peer-reviewed academic journals and studies published between 2002 (coinciding with the founding of the Global Fund) and 2009 to capture the evidence generated during the early years of Global Fund-financed interventions. For MEDLINE/Ovid SP we used the following search syntax ‘global fund.af. OR gfatm.af.’ to identify all studies related to the ‘Global Fund’. The search syntax for other databases is available upon request. The list of electronic databases searched in August 2009 is provided in Appendix A. We identified additional relevant literature by searching the reference lists of included studies and other reviews. Documents available at the Global Fund website were also searched to identify studies meeting the eligibility criteria. ISI Web of Science was searched for articles that cited the studies included in the review.
Two review authors (MC, TP) independently screened all references to assess which studies met the inclusion criteria. Any potential disagreements were resolved through discussion between the authors. Figure 1 shows the study selection process. After screening 2207 references, 24 studies were included in this review.
None of these studies were explicitly designed to study health system effects, but eight studies reported or commented on perceived negative health system effects. We decided to include all the 24 studies to the review for a more detailed evaluation, because it was clear that the eight studies referred to negative health system effects more by chance than with a predefined purpose in study design. For all the 24 studies we evaluated the generation of evidence base related to the Global Fund in general, and for the eight studies, the generation of evidence base specifically to perceived negative health system effects.
Two review authors (AK, TP) independently extracted the relevant data from the included studies using standardized data extraction sheets. For data extraction we used an adjusted version of the EPOC data abstraction form.16 We developed a modified checklist for assessing methodological quality of reporting using checklists provided by the EPOC Review Group,17 the STROBE group,18 the Clinical Appraisal Skills Programme19 and Quality Framework,20 adjusted for each study type and design. The template used to assess study quality is provided in Appendix B. Given the lack of generally accepted standards in the appraisal of qualitative research,21,22 and the observed large variability in the methods and quality of reporting, we used the quality assessment framework only for evaluating the strengths and weaknesses of the body of evidence, and not to categorize studies according to predefined thresholds or exclude studies from the analyses.22
Our analysis was based on producing structured summaries and narrative tables, and then contrasting and highlighting similarities, differences and common factors across the studies. The purpose of the analysis was to critically evaluate the processes that generated the study results relating to health system effects and the conclusions presented by the authors. While our approach acknowledges that the requirements and use of evidence in policy-making contexts may have different priorities from clinical decision-making, it underlines the requirement to address the limitations of a given methodology and to acknowledge the appropriate conclusions each study design can optimally support in relation to causality.
There were no experimental studies assessing the effects of health system interventions. Of the 24 studies included, 16 were quantitative studies, six were qualitative and two used both quantitative and qualitative methods. Seven of the quantitative studies were descriptive and did not use any explicit statistical methods to analyse their data. The remaining nine quantitative studies used a variety of study designs: one study was an uncontrolled before–after study,23 one uncontrolled study reported data before and during implementation,24 one study utilized time-series data,25 two were cohort studies,26,27 two were cross-sectional studies,28,29 one study used Global Fund grant data for modelling30 and one study modelled economic costs of a national insecticide-treated bed net (ITN) voucher scheme.31 Only one of the six qualitative studies used an explicitly defined qualitative method of analysis.32 The remaining studies did not specify which methods they used.33–37 The two studies using a mix of quantitative and qualitative data were descriptive, without explicit methods of analysis.38,39
One study assessed Global Fund's performance-based funding and was therefore determined to address all health system components, as the Global Fund performance-based funding framework includes assessment of ‘system effects’ of its investments.40 Interventions were most often determined to be related to service delivery (n = 14), medical products, vaccines and technologies (n = 9), and financing (n = 6). Service delivery often overlapped with other health system components. Three studies addressed health workforce-related issues, and three studies also addressed leadership and governance-related issues. None of the included studies explicitly addressed interventions aimed at improving health information systems. Five studies did not target a specific disease, but were addressing wider issues such as countries’ absorptive capacity,30 Global Fund's performance-based funding approach,40 Global Fund-supported programmes’ contribution to international health targets,41 an innovative financing scheme used by the Global Fund (Debt2Health Conversion Scheme)37 or analysing stakeholder opinions and expectations using the Global Fund 360° Stakeholder Survey.29
None of the studies explicitly evaluated interventions aimed at strengthening health systems. Twelve studies reported interventions that were originally targeted at individuals. The results for these interventions were reported either at individual level,23,24,26,27,39 national level,25,31,33,42 district level,43 clinic/hospital level44 or at household level.28 Five studies used aggregate data from several countries worldwide, related to overall health systems.29,30,40,41,45 Two studies reported interventions targeted at community levels. Of these, one reported results at clinic level38 and the other at community level.34 The remaining four studies reported national-level interventions.32,35–37,46
Of the 10 studies addressing HIV/AIDS, seven studies were directly related to provision of antiretroviral treatment (ART),23,26,27,33,42,44,45 of which one analysed global prices of antiretroviral drugs.45 Of the five studies relating to malaria, four studies were directly related to distributing ITNs.25,28,31,43 Interventions targeting tuberculosis ranged from national programmes to improving case detection strategies.32,36,39,46 Of the 24 included studies, seven reported at least some data related to health outcomes.23–27,44,46 Of these seven studies, three had a study design that enabled them to study the effect of the target intervention. Two of these were ART efficacy studies from Haiti, and were conducted at the same clinic.26,27 The third was an ART efficacy study conducted in northern Malawi.23
Five of the studies used and analysed data directly related to the Global Fund, either because the Global Fund-financed programme collected the data or the data collection was commissioned by the Global Fund.29,30,40,41,45 One study was designed to explore stakeholder experiences with Global Fund's impact at local level.32 The Global Fund was often reported to support national disease programmes, but without clearly specifying the role of funding, recipients of the funding, range of interventions that were implemented using the Global Fund funding and other sources of funding. Overall, Global Fund involvement in the interventions described in the studies was expressed imprecisely and in various different parts of the articles. Some studies reported that Global Fund had financed the study reported.28
One study analysed several unfulfilled stakeholder expectations and found that the second largest group of unfulfilled expectations were related to impact.29 These unfulfilled expectations were related to interventions being able to reach target populations, health systems being strengthened through disease-specific approaches and effectiveness of performance-based funding. The authors did not provide explanations as to why they perceived these expectations as unfulfilled, but they found that the more respondents involved with the Global Fund, the fewer unfulfilled expectations stakeholders had. Stakeholders from sub-Saharan Africa were reported to have often unfulfilled expectations.
Table 1 outlines the negative health system effects referred to in the papers and the health system components that these effects relate to. Given the lack of identified studies directly assessing the impact of Global Fund investments on health systems, only seemingly anecdotal evidence or authors’ perceptions/interpretations of circumstances could be extracted from the included studies, which often repeated the commonly expressed concerns over potential negative health system effects of disease-specific programmes. While one of the included studies explicitly noted the lack of reliable evidence on the positive and negative impacts of Global Fund investments on health systems,30 none of the studies assessed the implications of this evidence gap. Studies consistently identified performance-based funding as a factor potentially having negative effects on health systems during all stages of the implementation process. The studies identified the burden placed on countries with the funding application process. Onerous requirements for preparing and presenting grant applications were noted as a disincentive if applicants lacked the capacity to respond and fulfil the criteria set by the Global Fund.34 Concerns were also expressed about sustainability of funding, given the large volumes of external financing and reduced funding of poorly performing grants without due consideration on the impact of the decisions on country programmes and the epidemics.35
Frequent reporting was seen as a burdensome ‘donor requirement’, with negative effects on programme implementation.32,34,44 For example in Malawi, after a successful scale up of ART programme, the clerical staff noted the challenges of compiling reliable quarterly reports for donors, including the Global Fund.44
The Global Fund expects grant recipients to adjust programme implementation following assessment of grant performance. However, lack of capacity to adjust programmes during implementation threatens sustainability.34 In Kenya, the Global Fund's concerns over grant performance led to delays in releasing funds,35 which negatively affected programme implementation.
A study which explored stakeholder experiences of Global Fund's local impact, suggested that using solely quantitative performance indicators could ignore significant performance related factors.32 The study used two case studies as examples to highlight the discrepancy between quantitative and qualitative performance indicators. In Nicaragua, the numerical target for training community health workers was exceeded by more than two-fold, but the quality of training and resources provided for the community health workers were considered to be poor. Similarly, the success of establishing ‘Tuberculosis Clubs’ was measured against the number of patients affected by tuberculosis attending these Clubs. This indicator ignored the negative experiences expressed by the participants.
Until recently, the Global Fund Board initially approved funding for a two-year period (Phase 1 of funding). The grant performance is evaluated against agreed targets and a decision is made on the funding for a further three years (Phase 2). One study analysed the potential negative effects of the Global Fund's performance-based funding in countries with low national income or with weak health systems,40 to conclude poor grant performance was not related to low country income, weak health systems, state fragility or limited human resources for health.
Thirteen studies (58%) had considerable inadequacies in reporting the data used in analysis, the methods or both. Of these studies, five had very little or no description of data. Assessing the quality of studies was particularly challenging in studies using qualitative approaches, but also in the descriptive quantitative studies. For example, inadequacies in transparency and documentation led to difficulties in establishing the level of scientific rigour of the included studies. Only two studies clearly indicated measures taken to avoid bias or sources of error.32,35 Four studies indicated a risk of selection bias.27,29,40,45 Overall, the quality of reporting was suboptimal for most included studies.
None of the identified studies explicitly stated that the studies were originally designed to capture or to assess health system effects (positive or negative). Only seemingly anecdotal evidence could be extracted from the included studies. Scientifically sound, high-quality research must be conducted before generalizations can be made on the negative (or positive) health system impacts of Global Fund investments.
In view of the absence of experimental studies directly assessing health system effects, the strength of our approach was that we were not limited to a particular study design. Our search strategy was sensitive for detecting the ‘Global Fund’ regardless of the actual projects and interventions, but was limited to studies making formal reference to the Global Fund in the published articles. Some potentially relevant studies may not have been identified, if the published articles did not make a reference to the Global Fund. As we were unable to estimate studies that might have been missed due to lack of referencing to the Global Fund, the representativeness of our sample in relation to all interventions remains unknown.
Our search strategy, however, enabled us to identify all studies that explicitly contribute to the debate on the health systems effects of Global Fund financing of disease-targeted programmes in low- and middle-income countries. Given that none of the included studies were explicitly designed to study health system effects and that there are no uniform guidelines for reporting health system effects, some authors of the original papers may have omitted reporting relevant health system effects alongside their results.
The assessment of study eligibility was often complicated because the authors of the identified articles did not use consistent approaches when referring to the Global Fund. For example, the Global Fund was often indicated to support national programmes, but the link between Global Fund-supported national programmes and the interventions described in the study was not always clearly established. In some cases, the reference to the Global Fund could have easily been omitted or replaced with some other donor organization. Some authors referred to the Global Fund financing of the interventions studied in the acknowledgements section, but not in other parts of the article such as in the introduction or methods, with many studies making the connection between the Global Fund, the interventions described in the study and the relevance to health systems in the discussion sections of the studies. Several discussions had to take place at this stage to clarify decisions to reach a transparent agreement between the review authors – a process, which undeniably involved a certain level of subjectivity by the review authors when determining eligibility. Assessment of eligibility was also significantly affected by the generally suboptimal quality of reporting in the screened studies.
Several studies, both quantitative and qualitative, omitted significant parts of describing data and methods that would have facilitated the assessment of eligibility. Given the methodological challenges faced and the certain level of subjectivity involved in assessing eligibility, it is worth considering potential effects of reviewer bias. The field of evidence synthesis addressing complex adaptive systems, such as health systems, is still in its infancy, and therefore reviewers are forced to make subjective decisions. We aimed to control this current methodological shortcoming by transparently describing each step of the review process and stating our rationale for all decisions so that potential sources of bias would be visible to the reader. Furthermore, the purpose of this review was to assess the current evidence base specifically in relation to type and quality of evidence. Our main results and conclusions are therefore related to general principles of scientific quality, and are thus less affected by subjectivity.
Studies addressing health system effects of the Global Fund investments have been published after the literature search of this review was conducted in 2009.47–51 Due to financial resource restrictions, we were not able to extend the analysis to cover years after 2009. Including more recent evidence into this review would undeniably add to the overall picture provided by this review, particularly in relation to observed health system effects, but it would not change the findings and conclusions on the evidence generation during the period studied.
None of the identified studies published between 2002 and 2009 explicitly and rigorously assessed effects of funding by the Global Fund on health systems. The evidence on effects of funding by the Global Fund currently arises from study designs with higher levels of uncertainty in relation to causality and potential sources of bias. Current discourses around GHIs, including the Global Fund, seem to form a significant part in generating the evidence on the potential negative effects of disease-specific programmes. In line with the previous major reviews,5,9 much of the current debate also specifically around the Global Fund was found to be based on anecdotal evidence and assumptions of perceived negative effects of disease-specific programmes in general.
The review shows the considerable gap between the optimal study designs and the actual study methods used to analyse health system effects of Global Fund investments. The use of anecdotal evidence is undeniably important in some situations, for example when drawing attention to potential adverse effects. But the persistent use and generation of anecdotal evidence when evaluating health system impacts is not scientifically justifiable. More importantly, in situations where anecdotal evidence is the only evidence, it should always be accompanied with careful and critical break down and assessment of attribution. This was not found to be the case in the studies we reviewed.
Compared with the evidence-base for effective health interventions, the current evidence-base for effective implementation of inherently complex health system interventions is very weak,8,52,53 requiring high-quality monitoring and evaluation as well as rigorously designed and executed studies to address the evidence gap – given the quantum of investment by the Global Fund which is essentially funded by tax payers of donor countries.
Limited theoretical understanding of models of causality at health system level further handicaps efforts to establish plausible or probable relationships between interventions targeted to individual health system components and system-level impacts. The lack of rigorous scientific evidence, however, complicates the assessment of observed health system impact and restricts conclusions that could be drawn on system level performance from information derived from lower levels (e.g. individual health system building blocks).
A recent comprehensive assessment on Global Fund's health impact (Global Fund 5-year evaluation) showed that evaluating health system effects at country level faces significant methodological challenges and problems, e.g. in terms of data availability and quality.54 Strengthening country health information systems is therefore a prerequisite in improving evidence base through high-quality research.
This study shows that much of the currently available evidence generated between 2002 and 2009 on Global Fund's potentially negative health system effects is not of the quality expected or needed to best serve the academic or broader community. The current evidence used in scientific literature seems to rely on personal views and anecdotal evidence. While this insight into the field is valuable in informing short-term decision-making, it should only serve as an initial step before acquiring more rigorous research.
The weight of the current debate around the GHIs should move away from non-peer reviewed materials, such as organizational reports, commentaries and ‘descriptive’ discussion papers without verifiable data. The lack of methodological standards for reporting health system effects of complex interventions in developing countries is likely to contribute to the subsequent suboptimal level of quality of reporting observed in this review.
RA was Director of Strategy, Performance and Evaluation Cluster between 2008 and 2012 at The Global Fund to Fight AIDS, Tuberculosis and Malaria. Global Fund had a role in designing the scope of the study, but had no role in the collection, management, analysis and interpretation of the data, or in the preparation, review or approval of the manuscript. RA conceived the study and contributed to the writing of the article by commenting on the drafts of the manuscript and providing additional insight. These comments did not alter the conclusions of the article
The review received a partial financial contribution from the Global Fund
JC designed and coordinated the study. AK, MC and TP collected and analysed the data. JC and TP together drafted the first versions of the manuscript and edited the paper for publication. AM provided methodological assistance and critically reviewed the manuscript. RA conceived the study and contributed to the writing of the article by commenting on the drafts of the manuscript
The Department of Primary Care & Public Health at Imperial College London is grateful for support from the NIHR Biomedical Research Centre scheme and the NIHR Collaboration for Leadership in Applied Health Research & Care (CLAHRC) scheme. We thank Rebecca Thompson for her assistance in conducting this review