Tracking development assistance for health and for COVID-19: a review of development assistance, government, out-of-pocket, and other private spending on health for 204 countries and territories, 1990–2050

Summary Background The rapid spread of COVID-19 renewed the focus on how health systems across the globe are financed, especially during public health emergencies. Development assistance is an important source of health financing in many low-income countries, yet little is known about how much of this funding was disbursed for COVID-19. We aimed to put development assistance for health for COVID-19 in the context of broader trends in global health financing, and to estimate total health spending from 1995 to 2050 and development assistance for COVID-19 in 2020. Methods We estimated domestic health spending and development assistance for health to generate total health-sector spending estimates for 204 countries and territories. We leveraged data from the WHO Global Health Expenditure Database to produce estimates of domestic health spending. To generate estimates for development assistance for health, we relied on project-level disbursement data from the major international development agencies' online databases and annual financial statements and reports for information on income sources. To adjust our estimates for 2020 to include disbursements related to COVID-19, we extracted project data on commitments and disbursements from a broader set of databases (because not all of the data sources used to estimate the historical series extend to 2020), including the UN Office of Humanitarian Assistance Financial Tracking Service and the International Aid Transparency Initiative. We reported all the historic and future spending estimates in inflation-adjusted 2020 US$, 2020 US$ per capita, purchasing-power parity-adjusted US$ per capita, and as a proportion of gross domestic product. We used various models to generate future health spending to 2050. Findings In 2019, health spending globally reached $8·8 trillion (95% uncertainty interval [UI] 8·7–8·8) or $1132 (1119–1143) per person. Spending on health varied within and across income groups and geographical regions. Of this total, $40·4 billion (0·5%, 95% UI 0·5–0·5) was development assistance for health provided to low-income and middle-income countries, which made up 24·6% (UI 24·0–25·1) of total spending in low-income countries. We estimate that $54·8 billion in development assistance for health was disbursed in 2020. Of this, $13·7 billion was targeted toward the COVID-19 health response. $12·3 billion was newly committed and $1·4 billion was repurposed from existing health projects. $3·1 billion (22·4%) of the funds focused on country-level coordination and $2·4 billion (17·9%) was for supply chain and logistics. Only $714·4 million (7·7%) of COVID-19 development assistance for health went to Latin America, despite this region reporting 34·3% of total recorded COVID-19 deaths in low-income or middle-income countries in 2020. Spending on health is expected to rise to $1519 (1448–1591) per person in 2050, although spending across countries is expected to remain varied. Interpretation Global health spending is expected to continue to grow, but remain unequally distributed between countries. We estimate that development organisations substantially increased the amount of development assistance for health provided in 2020. Continued efforts are needed to raise sufficient resources to mitigate the pandemic for the most vulnerable, and to help curtail the pandemic for all. Funding Bill & Melinda Gates Foundation.


Introduction
How much countries spend on health has long been of interest for relevance to a range of social and economic objectives, including the goal of providing essential health services and universal health coverage. 1- 6 The COVID-19 pandemic has renewed interest in the past, present, and future of global health financing, in part because responding to the COVID-19 pandemic has been and continues to be tremendously costly. Governments around the world implemented restrictions on travel and mass gatherings; required masks and quarantines; and rolled out and ramped up access to COVID-19 testing, contact tracing, and, when possible, COVID-19 treatment. 7,8 Governments have fought to secure access to the first round of approved COVID-19 vaccines, with vaccination campaigns beginning in more than 30 countries in 2020, including China, Israel, Russia, Mexico, the USA, and the UK. [9][10][11] Funding the pandemic response has been complicated by global economic recession, which has not spared lowincome and middle-income countries. 12 It was estimated that the global economy shrank by 3·3% in 2020, and the economies of low-income and middle-income countries (excluding China) shrank by 4·3%. 13 Uncertainty is expected to linger as third and fourth waves sweep the globe and new COVID-19 variants spread with increasing speed. 14,15 Unemployment increased globally, and extreme poverty is estimated to have increased by between 88 and 115 million in 2020. 16 Moreover, the economic consequences of the health crisis are leading to long-standing reductions in economic development in some countries, and further indirect adverse effects on health. 4,[17][18][19][20][21] The response to the dual health and economic crises caused by COVID-19 generated substantial costs across countries. High-income countries and many middleincome countries have been able to finance their government programmes with financial reserves, reallocation of government resources, and by borrowing. The International Monetary Fund (IMF) estimates that globally, government debt increased by $20 trillion between Sept 1, 2019, and Sept 1, 2020. 22 Although some low-income countries have also borrowed resources (obtaining loans at market rates) to fund their responses to the 2020 crises, development assistance plays a unique role in funding health systems during emergencies in many countries and can be a catalyst for rapidly scaling up novel health services. 23 We estimated development assistance for health for COVID-19 in 2020, and assessed how that assistance compared with broader trends in health financing. We first present retrospective estimates of domestic spending on health and development assistance for health to provide context on the broader health financing landscape. Then we focus on development assistance for health and how it was used to target COVID-19 in 2020, and ramifications for development assistance for health funding in other essential health areas. We disaggregate estimates for development assistance for health for COVID in 2020 by funding sources, disbursing agencies, recipients, and programme areas to enable comparison with other key focus areas such as HIV/AIDS, health system strengthening, and pandemic preparedness.
Finally, we generate estimates of future health spending to enable an assessment of the implications for future health financing. 24,25 Comprehensive health spending estimates are important for examining potential gaps in resource needs versus available funding.

Overview
Health spending can be disaggregated into four key financing sources: government, out-of-pocket, and prepaid private health spending, which collectively make up domestic health spending; and development assistance for health, which includes international disbursements for health to low-income and middleincome countries. Government health spending includes social health insurance and government public health programmes. Out-of-pocket health spending includes health-care spending by a patient or their household but excludes insurance premiums. Prepaid private health spending includes private insurance spending and spending by non-governmental agencies on health.
To estimate total health spending for 204 countries and territories, we estimated each of the four financing sources separately for the years that underlying data were available, and used past trends and relationships to forecast each financing source from the point where retrospective estimates end to the end of 2050. The primary data source for domestic spending financing streams was the WHO Global Health Expenditure Database (GHED), and retrospective estimates were made from 1995 to 2018. Development assistance for health was estimated by use of a diverse set of project and agency expenditure and revenue statements, and estimates extend from 1990 to 2020, with additional project-level databases used to generate special estimates of development assistance for health for COVID-19 in 2020 (appendix pp 26,91). Forecasts for each financing source begin in 2019 or 2021. All estimates are inflation-adjusted and are mostly reported in 2020 US$, although some were adjusted for national prices and are thus reported in purchasing-power parityadjusted US$ or relative to gross domestic product (GDP).

Estimating domestic health spending, 1995-2018
For government, out-of-pocket, and prepaid private health spending, we downloaded data from the GHED for all available countries in current national currency units. 26 We adjusted these estimates for inflation, converted to 2020 US$, modelled estimates to ensure consistency over time and comprehensiveness across countries and territories, and estimated uncertainty (appendix p 21). We also converted these estimates into 2020 purchasing-power parity-adjusted US$. We used deflator series and exchange rate data based on data from the IMF World Economic Outlook. 14 For each of the three domestic financing sources, we used the metadata provided by WHO to qualitatively assess the reliability of data extracted from the GHED. We assigned a weight to each downloaded datapoint according to the documented source information included in the metadata, completeness of metadata, and documented methods of estimation (appendix p 13). We then used a spatiotemporal Gaussian process model to generate a complete time series of data from 1995 until 2018 for each country, and calculated 95% uncertainty intervals (UIs). 27,28 Estimating development assistance for health, 1990-2020 Development assistance for health refers to the financial and non-financial resources that are disbursed through international development agencies to maintain or improve health in low-income and middle-income countries. We tracked these disbursements from their originating sources through their disbursing agencies to the health focus areas that these resources were designed to target in recipient countries. Originating sources were typically the national treasuries of donor governments or private philanthropies from which development assistance funds are transferred. The funds from originating sources are channelled through international development agencies, which are here referred to as the disbursing agency, before being disbursed to the recipient country. We relied on project-level disbursement data from major international development agencies' online databases, including the Organisation for Economic Cooperation and Development's Creditor Reporting System (OECD CRS), the Global Fund to Fight AIDS, Tuberculosis and Malaria (the Global Fund), and annual financial state ments and reports for information on income sources. Data were not yet available for some more recent years, and we relied on budget data to generate these estimates. Detailed explanation of how the disbursements were estimated for each disbursing agency is provided in the appendix (pp 47-80) and published elsewhere. 5,[29][30][31][32][33][34][35][36] We disaggregated the estimates into ten health focus areas and 53 programme areas. This disaggregation captures the main programmatic areas to which development assistance for health contributions have historically been provided, and facilitates comparison with 2020 contributions and the international funding for the ongoing pandemic. We defined relevant health focus areas and programme areas for projects on the basis of a keyword search of the project descriptions downloaded from international agencies' online project databases. The specific keywords we used and their assigned health focus or programme area are detailed in the appendix (pp [35][36][37][38][39][40][41][42][43][44][45][46]. We leveraged information from available financial documents on revenue to remove double counting across disbursing agencies, so that each flow of funding is counted only once even if it was moved from one agency to another. Although OECD CRS completeness has improved over time, in earlier years (for this study, 1990 to 1996 especially) the data were less complete; thus, we used adjusted commitment data from the Development Assistance Committee tables to estimate disbursements. We also estimated the expenses asso ciated with administering loans and grants.

Estimating development assistance for health for COVID-19
We developed a separate method for estimating development assistance for health for COVID-19 because much of the project-level data used as input for the historical development assistance for health estimates do not extend to 2020, and therefore do not include resources disbursed in response to COVID-19. Likewise, budget and  We used the information available to determine whether projects were new grants or previous grants repurposed to COVID-19 projects. UNOCHA data designated whether resources were new or not. For IATI, we assumed that COVID-19 projects starting before 2020 were repurposed. For data received from correspondence or extracted from online databases, we contacted the respective agencies and searched their websites for more information. We included all research and development funding that went through international development agencies (appendix pp 91-92).
The approach used for each disbursing agency was specific to the data extracted and is detailed in the appendix (pp 92-106). Briefly, the general method was to use keywords to isolate COVID-19-relevant projects from each agency's database. We then examined these projects for completeness of information and adjusted the data to ensure that the estimate used was as precise as possible and could be compared between agencies. For instance, for projects with commitment estimates but without disbursement estimates, we calculated the average commitment-to-disbursement ratio for projects with complete data (for that development agency if possible) and multiplied the mean of those disbursement ratios by the commitment estimates for the projects without disbursement information. This commitment-todisbursement ratio captures the propor tion of committed funds that were estimated to have been disbursed in 2020. Next, we used a keyword search on project descriptions to disaggregate the estimated COVID-19 disbursements into eight COVID-19 programme areas: country-level coordination (planning, monitoring, and evaluation; risk communication; community engage ment; and travel restrictions); surveillance, rapid-response teams, and case investigation; national laboratories and testing; infection prevention and personal protective equipment; treatment; supply chain and logistics; main taining other essential health services and systems; and research and development for vaccines and other thera peutics (from a development Figure 1: Development assistance for health by health focus area, 1990-2020 HSS/SWAps=health system strengthening and sector-wide approaches. *Other captures development assistance for health for which we have health focus area information but is not identified as being allocated to any of the health focus areas listed. Health assistance for which we have no health focus area information is designated as unassigned. †2019 and 2020 disbursement estimates are preliminary.

Health focus area
Unassigned Other health focus areas* HSS/SWAps Other infectious diseases Non-communicable diseases Reproductive and maternal health Newborn and child health Tuberculosis Malaria HIV/AIDS 1 9 9 0 1 9 9 1 1 9 9 2 1 9 9 3 1 9 9 4 1 9 9 5 1 9 9 6 1 9 9 7 1 9 9 8 1 9 Year agency). These programme areas were developed based on a review of the data and literature. Finally, we reviewed the development agencies to identify any instances of double counting of the same resources across agencies. To do this, we reviewed income and recipient agency data for each of the disbursing channels that we tracked and excluded disbursements to recipient agencies that we tracked   separately as disbursing agencies. We typically kept the disbursed resources with the development agency that was closest in the process to the final recipient country. This review ensured that we counted each disbursement from the source agency, to ultimate disbursing agency, to the final recipient only once.

Estimating future health spending, 2019-50
We estimated GDP; general government spending (across all sectors); government, out-of-pocket, and prepaid private health spending; and total development assistance for health provided and received until 2050. The methods used for these projections draw heavily from our previous research, with the key change being the updating of the retrospective estimates on which these projections are based. 5,32,33 Because this research draws from a diverse set of underlying input data, updates to these data have cascading effects and affect all of the projections. We generated these projections by use of ensemble modelling techniques, such that the estimates are the mean of 1000 estimated projections from a broad set of models. We defined model selection by out-of-sample validation. This selection was country-specific and yearspecific. We completed projections sequentially so that previously projected values could be used as covariates and for bounding other models. We forecasted GDP per working-age adult aged 20-64 years from 2022 to 2050, with 2020 and 2021 estimates drawing on methods that focused on estimating economic growth sensitive to COVID-19 projects (appendix p 8). We forecasted general government spending from 2020 to 2050 (as the retrospective estimates extend to 2019). We modelled development assistance for health as a proportion of the donor country's general government spending, or, for private donors, on the basis of AutoRegressive Integrated Moving Average modelling techniques from 2021 to 2050 (as the retrospective estimates extend to 2020).
We aggregated total development assistance for health provided across sources and used a separate model to project the proportion of total development assistance for health that each recipient was expected to receive from 2019 to 2050 (as these retrospective data extend to 2018). We also modelled when countries are projected to transition to being high-income and are no longer eligible to receive development assistance for health. We projected government health spending as a share of general government spending, prepaid private health spending as a share of GDP, and out-of-pocket health spending as a share of GDP from 2019 to 2050 (as the retrospective data extend to 2018). To capture increased government spending in response to COVID-19, we checked whether our 2020 estimates of government health spending increased by at least the estimates made by the IMF October 2020 Fiscal Monitor of previously unanticipated government health spending. Countries that had year-over-year spending increases in 2020 that were less than estimated increase reported in the Fiscal Monitor were adjusted upward proportionally.
All projections incorporated several types of uncertainty. We used ensemble modelling techniques to propagate model uncertainty. We took draws of the variance-covariance matrix of each estimate's model to propagate parameter uncertainty. Finally, we added a random walk residual to each projection to propagate fundamental uncertainty. We generated 95% UIs by taking the 2·5th and 97·5th percentile of the 1000 estimated random draws.

Reporting
We report all the historical and future spending estimates in inflation-adjusted 2020 US$ and 2020 US$ per capita, and in purchasing-power parity-adjusted US$ per capita and as a proportion of GDP. For the development assistance for health estimates, we adjusted for inflation by taking disbursements in nominal US$ in the year of disbursements and using US GDP deflators from the IMF World Economic Outlook database to convert the series to constant 2020 US$. For the historical and future global health spending estimates, we used countryspecific exchange rate data and deflator series from IMF to convert the series to constant 2020 US$. We report all  spending estimates by 2019 Global Burden of Diseases, Injuries, and Risk Factors Study (GBD) super-region and 2020 World Bank income groups. 37 For these aggregates (and the global aggregate) the reported estimates provide information about a group as a whole, rather than the mean of the countries included in that group. For all tables and figures, the country income classifications were held constant at the 2020 reported level, irrespective of whether they changed groups. The time periods for each of the financing sources differs relative to the availability of the underlying data (development assistance for health, 1990-2020; domestic health spending, 1995-2018; and future health spending, 2019-50). These time periods provide a time series of health spending data that makes sure to leverage all available data. We completed all the analyses using Stata (versions 13 and 15), R (versions 3.6.0 and 3.6.1), and Python (version 3.7.0).

Role of the funding source
The funder of this study had no role in study design, data collection, data analysis, data interpretation, or writing of the report.
Between 1995 and 2019, inflation-adjusted health spending per person increased in 182 of the 204 countries and territories considered. For health spending for every country and year from 1995 to 2050, see the appendix (p 142). We estimate that global health spending will reach $14·3 trillion (95% UI 13·7-15·0) by 2050.
Between 2012 and 2019, development assistance for health contributions plateaued at an annualised rate of 1·2%. However, in 2020, total development assistance for health (including development assistance for health for COVID-19) amounted to $54·8 billion, a $14·0 billion (34·6%) increase from 2019. The increase is largely (96·5%) attributable to disbursements for the health response to COVID-19, which amassed $13·7 billion of health assistance (or 24·9% of total development assistance for health) in 2020 ( figure 1). For year-on-year com parisons, we observed decreases in assistance targeted toward reproductive and maternal health (-6·8%), tuberculosis (-5·5%), and malaria (-2·2%). Meanwhile, development assistance for sector-wide approaches and health system strengthening increased (8·8%). Of note is the small proportion of health systems strengthening resources (14·9% in 2020) that have been targeted towards pandemic preparedness. Figure 2A shows the main sources of development assistance for health in 2020. Most of the funding came from the USA, the UK, and the Bill & Melinda Gates Foundation. The key disbursing agencies for these resources were USA bilateral organisations, nongovernmental organisations, and the World Bank ( figure 2B). Figure 3A shows the flow of development assistance for health contributions toward COVID-19 from the original source of funds, through the disbursing agency, and to the targeted programme area of focus as available for 2020. A total of $13·7 billion was disbursed in 2020 toward addressing the health-related effects of COVID-19 in low-income and middle-income countries. Of this total, the largest bilateral contributors were Japan ($2·3 billion), Germany ($1·3 billion), and the USA ($0·9 billion). Most of Japan's support ($1·4 billion) was disbursed through its own bilateral agencies, mostly the Japanese Interna tional Cooperation Agency, and was targeted to India ($360·6 million), Morocco ($204·7 million), and Indonesia ($194·4 million). Sub-Saharan Africa ($2·7 billion); south Asia ($2·2 billion); and southeast Asia, east Asia, and Oceania ($2·0 billion) were the main geographical super-regions that received COVID-19 funds. Both the UK and Germany primarily supported the Coalition for Epidemic Preparedness Innovations (CEPI; $64·3 million from the UK and $56·9 million from Germany). The Asian Development Bank, Gavi, and the Global Fund are the international development agencies that channelled most of the resources committed or disbursed for COVID-19. Figure 3B shows the allocation of development assistance for health for COVID-19 in the GBD superregions. 23·5% of donors' COVID-19 resources were for global initiatives and did not directly target a single country or region. Examples of such investments include vaccine development and procurement, and partner organisations' coordination. The specific region to receive the most development assistance for health for COVID-19 was sub-Saharan Africa (28·8%). This figure also illustrates the poor alignment between COVID-19 assistance for health and COVID-19 burden. In 2020, 34·3% of recorded COVID-19 deaths in low-income and middle-income countries occurred in Latin America, but countries in Latin America received 7·7% of all development assistance for health for COVID-19 allocated to any specific country.
The proportion of COVID-19 funds that were new resources versus those that were previously budgeted and repurposed to COVID-19 is shown in figure 4. Of the total $13·7 billion in development assistance for health for COVID-19, $1·4 billion (9·9%) was money from already existing projects that was redirected toward COVID-19-related response activities. $12·3 billion (90·1%) of this funding was additional, previously unbudgeted resources. Across the disbursing entities, the Asian Development Bank ($1·8 billion) and Gavi ($1·8 billion) disbursed most of their resources as new money. Relative contributions for health and non-health response are shown in table 2. The health response accounted for 9·8% of the overall response, with the non-health response making up the dominant share ($125·4 billion) of the disbursed response. The types of contributions made and disbursed across the various disbursing agencies as part of the health-related COVID-19 response are shown in table 3. Grants made up most contributions, at $8·7 billion (63·6%), compared with loans at $4·8 billion (34·9%).
Relative to global spending, development assistance for health for COVID-19 remained small in 2020 (0·1%), and it is too early to know how much domestic spending has focused on COVID-19. Although there is a great deal of uncertainly, future spending on health is expected to continue to climb, albeit at a slower pace than anticipated before the pandemic. We estimated that global spending on health will reach $9·9 trillion (95% UI 9·7-10·1) in 2030 and $14·3 trillion (13·7-15·0) in 2050. Health spending over time for each income group is shown in figure 5, which highlights that health spending disparities are expected to persist and that there is a great deal of variation in total spending within income groups. We estimate that in 2050, low-income countries will spend $46 (95% UI 44-47) per person, lower-middle-income countries will spend $150 (141-159) per person, uppermiddle-income countries will spend $1001 (922-1083) per person, and high-income countries will spend $8536 (8074-9032) per person.

Discussion
The context in which COVID-19 has spread globally is one of grave inequality in access to health services. 39,40 This research highlights how this inequality exists at the national level within the health sector. In 2019, national health spending ranged from $7 to $11 345 per person, with government health spending ranging from $2 to $6578 per person (table 1, appendix p 142). Development assistance for health reached $40·4 billion in 2019 and increased to $54·8 billion in 2020 because of the additional resources provided in response to COVID-19. Although development assistance for health makes up less than 1·0% of health spending in most middle-income countries, development assistance for health was 24·6% (95% UI 24·0-25·1) of total health spending in low-income countries. Health spending per person has grown almost universally between 1995 and 2019 and is expected to rise in 195 of 204 countries and territories to 2050, but expected spending remains Repurposed money New money Type of assistance highly unequal. The finding of rising global health spending before the onset of the pandemic is aligned with what is reported in the WHO 2020 Global Health Spending Report. 4 In both our study and WHO spending report, the pandemic is expected to perpetuate already existing vulnerabilities in global health spending, such as inequalities in spending across and within income groups.
Our estimates of development assistance for health for COVID-19 show that in 2020, a total of $13·7 billion was disbursed to aid the health response in low-income and middle-income countries. This figure constituted over a quarter of all development assistance for health, and is evidence of how resources can be scaled quickly when needed. Resources focused on country-level coordination, supply chain and logistics, case management and treatment, surveillance, rapid-response teams, case investigation, maintaining essential health services and systems, and infection prevention and control have received modest contributions.
Previous estimates of development assistance for health for COVID-19 vary greatly, primarily because each set of estimates focused on tracking different aspects of the international response. Our total estimates of resources committed is similar to that reported by the Economist Intelligence Unit COVID-19 Health Funding Tracker, which estimated $13·8 billion pledged as of December, 2020, although it excluded support for individual countries. Our estimates of resources committed are much smaller in magnitude, however, than what is reported by the other trackers. As of its last update in April 2020, the Kaiser Family Foundation estimated $19 billion had been pledged toward the health sector response. 24 As of Jan 20, 2021, the Centre for Disaster Protection online visualisation tool reported $100·7 billion had been committed in loans (concessional and non-concessional) and $14·24 billion had been Spending data are reported in millions of 2020 US$, where estimates were available. In-kind contributions such as personal protective equipment was assumed to be grants. $200·7 million in bilateral development agencies funds were not able to be identified as either grants or loans. *Commitments of at least $10 000 from USA foundations internationally and in the USA.  38 In the case where a grant was allocated to multiple sectors (health, as well as one or more non-health sectors), we calculated the funds attributed to health spending by multiplying the total grant amount by 1 divided by the number of sectors represented. Proportions of global commitments and disbursements by type of spending (health and non-health) were multiplied by IHME's estimate of health-related disbursements. IHME=Institute for Health Metrics and Evaluation. committed in grants. 25 A key difference between these estimates and our own is that most of the other research on this topic has not estimated disbursements, and instead focused on commitments or pledges. In practice, disbursements are generally less than com mitments. Moreover, in the case of unexpected crisis, disbursing resources can take much longer than the time span that we have predicted. As of Dec 13, 2020, the Devex database reported 987 programme announcements worth $193·1 billion, 810 grants worth $2·9 billion, 2171 tenders worth $4·7 billion, and 1133 contracts worth $1·7 billion. 26 Unlike our estimates, these have a much broader focus and include support for the humanitarian response (outside of the health sector).
Only a few studies have estimated the resources needed for an effective pandemic response. 3,12 Previously, the Commission on a Global Health Risk Framework for the Future, which was set up after the Ebola epidemic, estimated that annually $4·5 billion was needed globally for pandemic preparedness, by contrast with an expected annual loss of more than $60 billion from potential pandemics. One study 12 on a sample of 73 low-income and middle-income countries estimated that the additional cost of responding to the COVID-19 pandemic was between $33 billion and $62 billion.
One consequence of the COVID-19 pandemic that is different from other recent epidemics, such as the Ebola outbreak in 2014 or MERS-CoV in 2012, is the significant economic toll on high-income countries. Precisely how the economic recession will affect governments' willingness to maintain record development assistance resources for low-income and middle-income countries is not known. Some countries have already restructured their aid budgets to prioritise addressing domestic challenges. For instance, the UK has passed a bill to reduce the foreign aid commitment from 0·7% of GDP to 0·5% of GDP to enable management of the national budget. 41 In addition to this, Kobayashi and colleagues 42 suggest that the global scale of the COVID-19 pandemic might have weakened public sentiment about contributing public resources abroad, given the immense challenges at home. In the next decade, development assistance for health disbursements might grow at a substantially slower pace and could rely more heavily on private philanthropy and multilateral institutions.
Nonetheless, as the pandemic has evolved, so have calls for support for low-income and middle-income countries. These calls for support are not only to address immediate health needs but also to provide additional social safety net support and economic relief. Multilateral institutions such as the IMF, the World Bank, and regional development banks including the Asian Development Bank, the Inter-American Development Bank, and the African Development Bank have all responded promptly to these requests with various packages. However, much of this support consists of loans that will add to national debt, with potential for long-term effects on growth and the ability to increase social spending. 43 Additionally, there are calls for debt repayment suspension and a repurposing of such resources towards bolstering struggling economies.
The pandemic's effect on other essential service provisions has been notable, largely because addressing the urgent demands of the pandemic required a realignment of health resources, including personnel, hospital supplies, services, data, and funding, and has made accessing health services safely substantially more challenging. 44 At the beginning of the pandemic, when some health-care systems were overwhelmed by the unfolding crisis, the complete refocus of resources seemed to be the most logical response. However, over time, this burden on the health system has made evident the importance of strong health systems worldwide in maintaining global health security, and made it even more obvious that targeting development assistance  contributions to the strengthening of the broader health system is a crucial investment for global health security overall. 45,46 Although contributions toward health system strengthening have improved, the relative proportion of development assistance funds that are targeted toward such activities remains small, with an even smaller amount allocated toward pandemic preparedness. 47 The pandemic has shown how quickly a local epidemic can develop into a global crisis, and the need for global activity and cooperation to be commensurate with this challenge.
The COVID-19 pandemic has also had ramifications for the Sustainable Development Goals (SDGs) more broadly. According to the SDGs Report 2020, 48,49 globally, progress has been insufficient and the pandemic has only worsened the precarious progress made on most of the goals. For instance, in several countries in sub-Saharan Africa it seemed the health toll of the pandemic was less than expected in 2020, but the economic fallout due to public health measures such as lockdowns might have had much more severe consequences and pushed an additional 23 million people into poverty. 50, 51 The mitigation strategies used to reduce the impact of the pandemic have also led to a meaningful global response. So far, development assistance contribution that falls outside of the health sector has been much larger than health-related contributions ($125·4 billion vs $13·7 billion). This reflects how the effect of the COVID-19 pandemic has reverberated outside of the health sector, affecting the social, economic, and environ mental sectors in most countries.
Another challenge associated with raising and disbursing record-setting amounts of development assistance is the crucial task of allocating these scarce resources. Despite the differences in the burden of the pandemic across geographical regions, the patterns in the allocation of development assistance for health for COVID-19 remain largely the same as development assistance for health in the past, with Latin America and the Caribbean receiving little support despite their high burden. In future, key contributions must be aligned with where they are needed the most.
The distribution of the COVID-19 vaccine has generated much public discourse. 52 Although the initial discussion in popular media and among policy makers focused on national interests, the Access to COVID-19 Tools (ACT) Accelerator partnership was launched in April, 2020, by WHO, the European Commission, and the French Government to ensure pooled and equitable distribution of tools to fight COVID-19. ACT-Accelerator was estimated to need $33 billion for its work. Of ACT-Accelerator's four key pillars, the pillar that focuses on vaccines is the COVID-19 Vaccines Global Access Facility, which is led by Gavi, WHO, and CEPI. Leveraging its previous experience with managing advanced market commitment for immunisation, Gavi was tasked with facilitating the procurement of vaccines for 92 middle-income countries through the COVID advance market commitment. The objective was to raise $2 billion before the end of 2020 to enable it to support access to vaccines irrespective of income. 53 As of August, 2021, $17·9 billion had been committed to the ACT Accelerator partnership.
COVID-19 has highlighted the nuanced role health systems play in providing health security. Although robust health systems-with capacity to test, track, and treat those with the virus and ability to access and provide vaccines quickly and efficiently-are necessary for keeping deaths from COVID-19 at bay, it has also been made clear that a robust health system alone is not sufficient. 54 COVID-19 mortality rates have varied greatly across the globe, but there is little consensus on what is driving this variation, with key historical measures of pandemic preparedness highlighting capacity gaps, but not accurately predicting where the outcomes will be the best or worst. 55 To be successful in fighting the virus and preventing substantial loss, a constellation of characteristics that fall both inside and outside of the traditional health system is necessary. These characteristics include leadership, public health system capacity, social safety nets, and trust in the systems providing information and care. 56,57 A painful lesson of this pandemic is that global health security is only as strong as its weakest link. It is crucial that those working in the health sector ensure that the entire health system is in fact robust, and that policy makers outside of the health sector also be prepared for many potential crises that can accompany a pandemic.
This research highlights the context in which COVID-19 spread quickly across the globe-one of enormous variation in health-care spending. Robust health systems require more resources, and those working in government and advocacy outside of the health sector are encouraged to see COVID-19 and the staggering loss of life and broader setbacks in global development as indicators of the importance of investing in health in and outside the health system, and building governments and social systems that can also contribute towards health security.
This study has several limitations. The data were extracted from multiple sources with different reporting structures and components included. We kept definitions and methods as consistent as possible across data sources to ensure a reliable estimate of resources. Furthermore, given that the pandemic is ongoing and development projects take some time to reach implementation, disbursement could still be low even where commitments have been made. Moreover, global health resource tracking does not allow comparable tracking of how domestic resources are spent on health and does not allow for comprehensive estimates of domestic health spending on COVID-19. More information is needed to assess how governments, households, and private organisations are spending health resources, and what resources are available for responding to health emergencies. Health financing reporting, especially in many low-income and middle-income countries, but also in high-income countries, is constrained by how timely domestic health spending data are available. Health financing data are not collated in a manner that is shareable or comparable across time and country. As countries deliberate on ways to rebuild systems after the pandemic, the health financing architecture and equity are important aspects to consider. Finally, our quantification of uncertainty captures some types of uncertainty, but not all types of uncertainty. It is included here as a relative quantification of where we have the most confidence in our estimates.
Health spending in 2019, before the COVID-19 pandemic, ranged widely, and marks tremendous variation in the access to essential health services and universal health coverage. Much more money is needed to fully address the effects of the pandemic in most low-income and middle-income countries, along with access to key tools such as vaccines. Moreover, long-term projections suggest that health inequalities are likely to persist and that international efforts to invest in global public goods related to global pandemic preparedness are essential. Resources are needed urgently to mitigate the loss associated with the COVID-19 pandemic and fund systems that can prevent and respond quicker to the next global health crisis.

Contributors
Please see appendix (pp 1151-55) for more detailed information about individual author contributions to the research, divided into the following categories: managing the estimation or publication process; writing the first draft of the manuscript; primary responsibility for applying analytical methods to produce estimates; primary responsibility for seeking, cataloguing, extracting, or cleaning data; designing or coding figures and tables; providing data or critical feedback on data sources; development of methods or computational machinery; providing critical feedback on methods or results; drafting the manuscript or revising it critically for important intellectual content; extracting, cleaning, or cataloguing data; designing or coding figures and tables; and managing the overall research enterprise. All authors had full access to all the data in the study, and A E Micah and J L Dieleman had final responsibility for the decision to submit for publication.