The Cost of Cutbacks: How Reduction in Development Assistance for Health May Affect Progress Made in HIV/AIDS Control in Africa

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

Thank you for the opportunity to review such a timely manuscript. Given the healthcare funding uncertainty on the continent, the topic is relevant to policymakers, healthcare funders, and the general population. The manuscript has excellent scientific rigor. The title of the manuscript captures its content. The abstract is a correct summary of the manuscript.

Introduction

The introduction orients the readers to the topic, discusses what is known about the topic, identifies the literature gap, and clearly states the study's objective.

Methods

The authors clearly explain the conceptual framework and models used, which are all relevant to the study. The conceptual framework figure clearly shows the relationships. The authors stated the target region and countries, the variables, and the data sources. The data sources used are credible. Data analysis is clearly described to allow for the replication of the study. The simulation sensitivity analysis conducted adds credibility to the findings.

Results

The authors clearly present their findings with the help of tables, figures, and a narrative.

Discussion

The authors compare their findings with previous literature and explain the findings with the help of the literature. The recommendations provided are based on the study findings. The authors also discussed the strengths and limitations of the study.

Conclusion

The conclusions provided are based on the study findings.

References

Most references are recent and formatted according to the journal's requirements.

Language

There are no concerns in the use of the English language in the manuscript.

 

 

Author Response

REVIEWER 1

COMMENT: Thank you for the opportunity to review such a timely manuscript. Given the healthcare funding uncertainty on the continent, the topic is relevant to policymakers, healthcare funders, and the general population. The manuscript has excellent scientific rigor. The title of the manuscript captures its content. The abstract is a correct summary of the manuscript.

RESPONSE: Thank you for your insightful comment. We appreciate your positive evaluation.

COMMENT: The introduction orients the readers to the topic, discusses what is known about the topic, identifies the literature gap, and clearly states the study's objective.

RESPONSE: Thank you for this supportive remark. We are pleased that it was clear as intended.

Comment: The authors clearly explain the conceptual framework and models used, which are all relevant to the study. The conceptual framework figure clearly shows the relationships. The authors stated the target region and countries, the variables, and the data sources. The data sources used are credible. Data analysis is clearly described to allow for the replication of the study. The simulation sensitivity analysis conducted adds credibility to the findings.

RESPONSE: We appreciate your favourable and thoughtful comments.

COMMENT: The authors clearly present their findings with the help of tables, figures, and a narrative.

RESPONSE: We are encouraged by your appreciation of our work in this area. Thank you.

COMMENT: The authors compare their findings with previous literature and explain the findings with the help of the literature. The recommendations provided are based on the study findings. The authors also discussed the strengths and limitations of the study.

RESPONSE: We are sincerely grateful for this encouraging feedback and are pleased that this section of our manuscript was well received.

COMMENT: The conclusions provided are based on the study findings.

RESPONSE: We appreciate your comments and are glad that this part of the manuscript was clear and satisfactory.

COMMENT: Most references are recent and formatted according to the journal's requirements.

RESPONSE: Thank you for this positive feedback. We appreciate it. We have also updated our reference list in line with the journal’s guidelines.

COMMENT: There are no concerns in the use of the English language in the manuscript.

RESPONSE: Thank you for this observation. We are pleased that the manuscript was up to standards.

Reviewer 2 Report

Comments and Suggestions for Authors

Some comments.

This is clearly a very timely and important topic to address. I think the political sensitivity makes it even more important to ensure that the claims made match the evidence and don’t overstate any findings. For the most part, the paper does achieve this but there are a few cases, where more nuance is needed. This is particularly true with the forward-looking claims:

• Page 6. Line 173. “To test our hypothesis – that reductions in U.S. DAH will weaken the overall effectiveness of GHES…” It is not possible to test this forward looking claim. The conclusion is much more nuanced in the implications of this analysis ( as is section 3.3). That is, the results tell us about the past. If things continue to operate as they did in the past we would expect this to continue. But as the authors point out the implication of their results for the future depend on how other resources are allocated. I don’t think anything is lost by toning down the claimed hypothesis.
• Similarly – line 235 page 10 makes similar claims to tone down.
• Similarly – line 337 page 17 makes similar claims to tone down.
• Similarly – line 384 page 18, avoid use of “impact” this suggest a causal claim, whereas the study can only make associational claims.

I was wondering whether the incidence results also tell us something about targeting of DAH resources, are DAH resources increased essentially in response to some existing in-country success?

Magnitude of the coefficients: I understand using the SD units. This makes a lot of sense. That said, I think the paper would benefit from taking the reader through the magnitude of the effects at the mean in dollar terms and lives saved. This should be relatively easy to compute, using the information from the sample.

For the robustness check where the lambda varies, it would be good to cite a statistical reference that informs this procedure.

A couple of additional robustness checks:

1. Are the results sensitive to the cut-off of post COVID? I doubt they are. But, perhaps good to show that the results are robust if the sample period is cutoff end of 2019.
2. An additional robustness check that would be good given the underlying country heterogeneity is to run a drop one country at a time, and re-estimate the results. i.e. Jackknife estimation. Given the amount of funding to Kenya and South Africa, it would be good to know the results are robust to their exclusion in particular.
   1. Related to this, I was wondering how much the triple interaction between DAH, Other DAH and GHES depends on the set of countries. Is this interaction effect driven by a particular country?

Other minor things:

• Some of the outcomes have zero values as can be seen in Table 1, taking the natural log of this will be undefined. How is this addressed in the paper? Are these observations dropped? Is some alternative strategy used?
• Typo page 7 line 227, there is a space missing between “funds” and the (17.41%...)
• Consistency of capitalization: GHES, DAH etc. The paper switches back and forth.

Author Response

REVIEWER 2

This is clearly a very timely and important topic to address. I think the political sensitivity makes it even more important to ensure that the claims made match the evidence and don’t overstate any findings. For the most part, the paper does achieve this but there are a few cases where more nuance is needed. This is particularly true with the forward-looking claims:

COMMENT: Page 6. Line 173. “To test our hypothesis – that reductions in U.S. DAH will weaken the overall effectiveness of GHES…” It is not possible to test this forward looking claim. The conclusion is much more nuanced in the implications of this analysis ( as is section 3.3). That is, the results tell us about the past. If things continue to operate as they did in the past we would expect this to continue. But as the authors point out the implication of their results for the future depend on how other resources are allocated. I don’t think anything is lost by toning down the claimed hypothesis.

RESPONSE: Thank you for making this thoughtful observation. Prior to revising this sentence on page 6, line 173 according to your suggestion, we revised and toned down the initially forward-looking hypothesis, as indicated in paragraphs below, and used more cautious language on page 6, line 173.

Revised version (2.1 Conceptual Framework): “Within this framework, three critical questions were asked: (1) How were international donor funds for HIV/AIDS control disbursed across the African sub-regions and countries during the last 32 years, and what trends can be observed in the evolution of these allocation patterns? (2) How did the U.S. public sector-specific DAH interact with other donor funding and GHES in shaping HIV/AIDS control efforts in Africa? (3) What might these historical associations imply for future trends, assuming that resource allocation and disease patterns continue as they have in previous periods?”

Revised version (2.4 Data Analysis): “Data on HIV/AIDS-specific DAH funding allocation and distribution patterns in Africa were summarised using median, interquartile range (IQR), and changes over time. To answer our second and third research questions, we employed a multi-level linear mixed-effects model to test the associations between the set of predictors and the outcome (HIV/AIDS incidence and mortality), as shown in our conceptual model, accounting for clustering by sub-region and potential variability due to baseline HIV/AIDS incidence and mortality and other sources of heterogeneity at the sub-region and country levels. We nested 54 countries within the five sub-regions of Africa, according to the United Nations (UN) country groupings. In line with the best practice guidance for linear mixed-effects modelling, different models were developed and compared to determine the best-fitting model for the study data (Figure S1).”

COMMENT: Similarly – line 235 page 10 makes similar claims to tone down.

RESPONSE: Thank you for catching this up. We have applied the suggestion on page 10, line 235.

Revised version (3.2. Association Between DAH and HIV/AIDS Incidence and Mortality): “As illustrated in Figure 2A, in situations where U.S. DAH was low (-1 SD), the effect of other international DAH on HIV/AIDS incidence was negligible. However, at average levels of U.S. DAH (Mean = 0), an increase in other international donor DAH (+1 SD) was associated with a significant decrease in HIV/AIDS incidence, and this reduction was even more pronounced when U.S. DAH was high (+1 SD). Similarly, when U.S. DAH was high, increased GHES (+1 SD) was associated with a significant reduction in HIV/AIDS incidence. Interestingly, the predicted margins computed over 27 combinations of funding levels confirmed this synergy; for example, the combination of high U.S. DAH (+1 SD), high other-donor DAH (+1 SD), and high domestic spending (+1 SD) was associated with the lowest predicted log HIV/AIDS incidence (Figure 3).”

COMMENT: Similarly – line 337 page 17 makes similar claims to tone down.

RESPONSE: Thank you for highlighting this. We have adjusted the wording to provide a more balanced and cautious representation on page 17, line 337.

Revised version (4.2 Policy Implications and Recommendations): “Consistent with the literature, our study has shown that the U.S. public sector has been a driving force behind the global fight against HIV/AIDS, channelling large sums of money to the African region. Strategically, these funding initiatives have targeted critical areas of HIV/AIDS control, including counselling and testing, clinical care and support for patients, health system strengthening, human resource development, prevention of MTCT, support for orphans and vulnerable children, and prevention of new infections and drug resistance. As shown in the mixed-effects modelling, in addition to exerting a direct effect on HIV/AIDS incidence, U.S. DAH also enhances the effect of domestic government funding and other international donor DAH on HIV/AIDS incidence and mortality.”

COMMENT: Similarly – line 384 page 18, avoid use of “impact” this suggests a causal claim, whereas the study can only make associational claims.

RESPONSE: We appreciate this insightful observation. The sentence has been rephrased to temper the strength of the assertion on page 18, line 384.

Revised version (5. Conclusions): “Our study examined international aid disbursement for HIV/AIDS control and evolving patterns among African countries, from 1990 to 2022. It also investigated how U.S. DAH interacted with other donors and domestic general government health funding, and further explored what these historical associations might imply for future trends. As with previous studies, our findings indicate that the U.S. public sector has contributed largely to the fight against HIV/AIDS in Africa through its international DAH initiatives. Returning to the initial questions posed, it is now possible to state that in addition to exerting a direct effect, the U.S. public sector-specific DAH moderates the effect of other international donor funding and domestic government health spending on HIV/AIDS incidence and mortality.”

COMMENT: I was wondering whether the incidence results also tell us something about targeting of DAH resources, are DAH resources increased essentially in response to some existing in-country success?

RESPONSE: We appreciate you highlighted this aspect. Find below our critical analysis, as it appears in our manuscript.

Revised version (3.3 Additional Results): “On the issue of whether donors rewarded countries with lower past HIV/AIDS incidence, again, we focused exclusively on U.S. DAH, which was previously characterised by clearly defined, performance-based metrics and consistent reporting standards, contrary to Other DAH flowing from multiple sources with varied priorities, making them less directly comparable for assessing the targeted resource allocation. Under the premise that lower past incidence, a proxy of improved HIV/AIDS control, signalled programme effectiveness and merited further support, our primary specification (Option 1) found that a one-unit decrease in lagged HIV/AIDS incidence was associated with a 0.56-unit increase in subsequent U.S. DAH (p ≈ 0.058). This negative association, alongside a significant, persistent effect of U.S. DAH (coefficient 1.55, p < 0.001), is consistent with the notion that donors might reward better-performing countries. However, the extremely low Hansen test p-value (p < 0.001) raised concerns about instrument validity and potential overidentification. The alternative specification (Option 2), which addressed the instrument validity concern, did so at the expense of precision. Thus, our findings underscore the trade-offs inherent in dynamic panel estimation and highlight the need for a cautious interpretation.”

COMMENT: Magnitude of the coefficients: I understand using the SD units. This makes a lot of sense. That said, I think the paper would benefit from taking the reader through the magnitude of the effects at the mean in dollar terms and lives saved. This should be relatively easy to compute, using the information from the sample.

RESPONSE: Thank you for this insightful suggestion that enhances the clarity and quality of our manuscripts. We have applied the suggestions accordingly.

Revised version (3.2. Association Between DAH and HIV/AIDS Incidence and Mortality): “In real world terms, where GHES and Other DAH were held high—specifically, with Other DAH at roughly $100,000 and GHES at a value of $200 per capita, an increase in U.S. funding from approximately $56 to about $16,200 was associated with a decrease in HIV/AIDS incidence from roughly 181 to around 50 per 100,000 people, reflecting a 73% decline in incidence. In contrast, in situations where U.S. DAH was fixed at a low level (roughly $56) and other donor funding was maintained at a high level (about $100,000), the predictive margins revealed a rather strange results: as governments increased health spending from $11 per capita to about $200 per capita, the estimated HIV/AIDS incidence rate rose from roughly 53 to around 181 per 100,000 people, potentially reflecting reactive government spending.”

“A scenario where Other DAH is maintained at a high level (a mean‑centred value of 3.64, which back‑transforms to roughly $100,000), two patterns were observed: First, when GHES is also high (mean‑centred value of 1.43, or about $200 spending per capita), an increase in U.S. DAH from a low level (mean‑centred value of –2.83, approximately $56) to a high level (mean‑centred value of 2.83, roughly $16,200) was associated with a decline in predicted HIV/AIDS mortality from about 112 per 100,000 people to approximately 28 per 100,000 people—a reduction of roughly 75%. In a separate scenario, holding U.S. DAH low (–2.83, about $56 per capita) and Other DAH high, an increase in domestic government spending from low (mean‑centred value of –1.43, around $11 per capita) to high (mean‑centred value of 1.43, approximately $200 per capita) was associated with an increase in predicted HIV/AIDS mortality from roughly 27 to about 112 per 100,000 people.”

COMMENT: For the robustness check where the lambda varies, it would be good to cite a statistical reference that informs this procedure.

RESPONSE: We are grateful for this suggestion. The appropriate changes have been made and references have been added.

Revised version (2.5. Robustness Assessment): To further test the validity of our findings, we conducted a series of robustness assessments, including a Monte Carlo simulation-based sensitivity analysis. Specifically, we simulated a hypothetical confounder, U, drawn from a standard normal distribution, assuming it is associated with our primary predictors (U.S. DAH, Other DAH, and GHES) via a range of plausible lambda values (0.1 to 0.9), capturing the strength of these associations. For each plausible λ value, we generated new values for U, refitted the full mixed-effects model with U included as a covariate, and recorded the corresponding coefficient for U.S. DAH, Other DAH, and GHES. This procedure yielded a distribution of estimates that quantified the potential bias introduced by omitted variable confounding, allowing us to predict the robustness of our findings across a range of plausible λ values.

COMMENT: Are the results sensitive to the cut-off of post COVID? I doubt they are. But, perhaps good to show that the results are robust if the sample period is cut-off at the end of 2019.

RESPONSE: We are grateful for this suggestion and have provided more details to illustrate that the results are robust if the sample period is cut-off at the end of 2019.

Revised version (4.3 Robustness of the Results): “Moreover, comparisons between the full sample (1990–2021) and the pre‑COVID sub-sample (1990–2019) showed that the primary funding (U.S. DAH, Other DAH, and GHES) effects and interactions persisted across both periods, further reinforcing the robustness of our findings.”

COMMENT: An additional robustness check that would be good given the underlying country heterogeneity is to run a drop one country at a time, and re-estimate the results. i.e. Jackknife estimation. Given the amount of funding to Kenya and South Africa, it would be good to know the results are robust to their exclusion in particular. Related to this, I was wondering how much the triple interaction between DAH, Other DAH and GHES depends on the set of countries. Is this interaction effect driven by a particular country?

RESPONSE: We appreciate your observation and thoughtful reflection on this point. The necessary details have been provided to improve clarity as illustrated below.

Revised version (2.5. Robustness Assessment): “Also, we applied a jackknife (leave‑one‑country‑out) procedure in our primary mixed-effects model to ensure that key interaction effects were not driven by any single country, especially those receiving high funding. In addition, we computed and compared estimates across two sample periods—the full sample (1990–2021) and a restricted pre‑COVID sub‑sample (1990–2019)—to ensure the stability of the relationships between the funding streams and HIV/AIDS outcomes.

Moreover, we employed a comprehensive multi-method econometric strategy to address concerns about potential endogeneity and reverse causality. Our primary, complementary approaches included fixed‑effects Instrumental Variable (IV) estimation, dynamic panel data modelling using Arellano-Bond/System Generalised Method of Moments (GMM), and a control function approach, with endogenous variables instrumented using their lagged values and instrument validity assessed via Hansen tests”.

Revised version (3.2 Association Between DAH and HIV/AIDS Incidence and Mortality): “Our simulation‐based sensitivity analysis indicates that our main results are robust to unmeasured confounding. For HIV/AIDS incidence, the adjusted coefficients for U.S. DAH and other international donor funding remained consistently negative with only modest fluctuations under plausible confounder strengths (λ between 0.1 and 0.5), indicating a robust inverse relationship with incidence. However, GHES consistently shows a positive effect, reflecting reactive spending in response to higher epidemic burden. For HIV/AIDS mortality, whereas Other DAH retained a consistently negative association even under extreme confounding (λ = 0.9), GHES continued to show a stable positive relationship. Similarly, U.S. DAH was consistently associated with higher mortality. However, while our analysis supports the stability of this positive association between U.S. DAH and HIV/AIDS mortality, the fact that the primary model did not yield a statistically significant result means that this should be interpreted with caution. The simulation does not prove a causal relationship but rather suggests that the observed non-significant association is not merely an artefact of omitted variable bias within the tested range and assumptions.The jackknife analysis, in which each composite country–sex unit was sequentially removed, revealed that the key three-way interaction effect among U.S. DAH, other donor DAH, and GHES was remarkably stable, even when excluding high‐funding countries such as Kenya and South Africa.”

COMMENT: Some of the outcomes have zero values, as can be seen in Table 1, taking the natural log of this will be undefined. How is this addressed in the paper? Are these observations dropped? Is some alternative strategy used?

RESPONSE: Thank you for the observation. We acknowledge this point and have provided details of how this concern was handled as shown below.

Revised version (2.4. Data Analysis): “For positively skewed variables containing zero values (U.S. DAH, Other DAH, and PPP health spending), we applied a [log(x+1)] transformation to them”.

COMMENT: Typo page 7 line 227, there is a space missing between “funds” and the (17.41%...)

RESPONSE: Thank you for noting this. It has been corrected accordingly.

COMMENT: Consistency of capitalization: GHES, DAH etc. The paper switches back and forth.

RESPONSE: Thank you for highlighting this. The appropriate changes have been made.

Reviewer 3 Report

Comments and Suggestions for Authors

This paper addresses a highly timely topic – the impact of cutbacks in US aid on health outcomes focusing on HIV/AIDS in Africa.  Specifically, the study investigates whether US aid helps reduces incidence and mortality from HIV/AIDS, and whether it enhances the effectiveness of non-US aid and government expenditure on health. The main results are insightful.  But two results are counterintuitive, and they are not explained in the paper, which is an issue.

1. Increased domestic government health spending is associated with an increase in HIV/AIDS incidence. This is counterintuitive. What could drive this result?  Is it endogeneity? That is, could it be that governments facing high incidence of HIV/AIDS are forced to spend more on health to combat the infection? In that case the relationship would go from HIV/AIDS incidence to government health expenditure. The authors should explore this possibility using appropriate econometric methods.
2. When US health aid is low but other funding from other sources (non-US health aid and government health expenditure) is/remain high, HIV/AIDS incidence and mortality increase/are higher. The paper does not provide a convincing explanation of what lies behind this result.

 

The authors should attempt to provide a conceptual explanation of the two results above. Furthermore, they should explore econometrically the possibility of endogeneity of government spending and aid relative to HIV/AIDS incidence and mortality.

Author Response

REVIEWER 3

This paper addresses a highly timely topic – the impact of cutbacks in US aid on health outcomes focusing on HIV/AIDS in Africa.  Specifically, the study investigates whether US aid helps reduce incidence and mortality from HIV/AIDS, and whether it enhances the effectiveness of non-US aid and government expenditure on health. The main results are insightful.  But two results are counterintuitive, and they are not explained in the paper, which is an issue.

COMMENT: Increased domestic government health spending is associated with an increase in HIV/AIDS incidence. This is counterintuitive. What could drive this result?  Is it endogeneity? That is, could it be that governments facing high incidence of HIV/AIDS are forced to spend more on health to combat the infection? In that case the relationship would go from HIV/AIDS incidence to government health expenditure. The authors should explore this possibility using appropriate econometric methods.

RESPONSE: We thank you for this thoughtful feedback regarding this aspect of the manuscript. We have examined this section and provided further details to explain the relationship between HIV/AIDS incidence and government spending.

Revised version (3.3. Additional Results): “For HIV/AIDS incidence, the dynamic system GMM results showed a statistically significant positive association between GHES and incidence, suggestive of reactive spending (an indication of reverse causality). The control function approach refined this picture: when the endogenous (‘reverse causal’) component of GHES was isolated via the residuals, that component was significantly negatively associated with incidence, implying that once reverse causality is addressed, higher government spending would correlate with lower incidence. In contrast, for HIV/AIDS mortality, both the fixed‑effects IV and control function methods consistently yielded non-significant effects of GHES, suggesting that reverse causality was likely not a driving factor in the relationship between GHES and mortality. Overall, the evidence points to a nuanced role of reverse causality in the incidence relationship, where reactive spending could obscure the true beneficial effects of GHES, while for mortality, there is little indication that reverse causality biased the estimates.”

COMMENT: When US health aid is low but other funding from other sources (non-US health aid and government health expenditure) is/remains high, HIV/AIDS incidence and mortality increase/are higher. The paper does not provide a convincing explanation of what lies behind this result.

RESPONSE: Thank you for making this observation. We have provided more details to improve clarity as indicated below.

Revised version (3.3. Additional Results): “Regarding the donor funding mechanism, the fixed‑effects IV approach showed no evidence that changes in HIV/AIDS mortality drove U.S. DAH funding, which seems counterintuitive in our primary mixed-effects results (Table 2). The dynamic panel data model showed that U.S. DAH was largely pre-determined by its past levels rather than current mortality rates, and the control function approach demonstrated that any unobserved component (residual) potentially capturing reverse causality was negligible. Taken together, these findings indicate that reverse causality was not a major concern in explaining U.S. DAH, reinforcing the view that funding levels were reactive to contemporaneous changes in HIV/AIDS mortality.”

COMMENT: The authors should attempt to provide a conceptual explanation of the two results above. Furthermore, they should explore econometrically the possibility of endogeneity of government spending and aid relative to HIV/AIDS incidence and mortality.

RESPONSE: We thank you for these insightful comments. They have been addressed accordingly.

Revised version (4.1. DAH Allocation and Its Effects on HIV/AIDS Incidence and Mortality): “These empirical findings can be understood by considering the comprehensive role of the U.S. health aid in the global HIV/AIDS control. Previously, when the U.S. public sector programmes like PEPFAR were running at full capacity, with adequate funding, well-trained personnel, and effective management, they provided not only financial support but also essential technical expertise, close monitoring, and coordinated programmatic support that ensured sustainable, systemic improvements in HIV/AIDS prevention and care. In its absence, alternative funding streams are likely to be deployed reactively, addressing immediate crises rather than underpinning long‑term capacity building, leading to fragmented service delivery and inefficiencies. For example, recent Joint United Nations Programme on HIV/AIDS (UNAIDS) case analyses and reports from high disease burden countries including Eswatini, South Africa, Nigeria, Kenya, and Zimbabwe, have linked reductions in U.S. aid to disruptions in antiretroviral supply chains and gaps in prevention services, thereby amplifying the epidemic despite increased levels of domestic and non‑U.S. donor expenditures.”

Revised version (4.2. Policy Implications and Recommendations): “Our findings also indicate that African governments tend to increase health spending reactively in response to rising HIV/AIDS incidence, rather than mortality. One possible explanation is that spikes in new infections generate considerable political and media pressure, leading to immediate initiatives and budget reallocations—as seen in parts of South Africa and Uganda, where governments rapidly scaled up treatment and preventive campaigns in the early 2000s when incidence numbers surged. In contrast, mortality data—often delayed or underreported—may fail to create the same urgency, and the substantial donor funding directed toward treatment may further diminish the domestic political will and government motivation to adjust spending in response to death rates.”

Additional Clarifications

The aforementioned revisions are equally reflected in the abstract. The list of abbreviations has also been updated to reflect changes made.

Round 2

Reviewer 3 Report

Comments and Suggestions for Authors

The revisions have addressed the 2 issues I had raised on the results.  I am happy to recommend ACCEPTANCE of the manuscript for publication.

Thank you.