The cost of inaction to strengthen the resilience of primary health care in Latin America and the Caribbean: a modelling study

Introduction

More than five years since the COVID-19 pandemic declaration, the evidence is clear: COVID-19 had a significant impact on Primary Healthcare Systems (PHC) worldwide.¹ In addition to the millions of deaths directly caused by COVID-19 globally between 2020-2021, the COVID-19 pandemic led to substantive disruptions to the coverage and quality of healthcare services. For example, there was an observed increase in maternal deaths (odds ratio [OR]=1.37), stillbirths (OR=1.28) and neonatal deaths (OR=1.01) globally during versus before the pandemic^{2, 3} due to reductions in antenatal and delivery care services and reduced healthcare seeking behaviours among pregnant women. Disruptions to healthcare services were most prominent in low- and middle-income countries (LMICs)⁴, related to weaknesses in the health systems⁵.

Latin American and Caribbean (LAC) countries were significantly affected by COVID-19, with excess mortality rates well above the Organisation for Economic Co-operation and Development (OECD) average.⁶ The pandemic exposed weaknesses in health systems and added further strain to countries already facing significant structural challenges in their healthcare systems. Across 14 countries in LAC, an average of 20.4% of households reported experiencing disruptions in healthcare in 2020 (with the largest disruptions reported in Ecuador at 44.9%).⁷ In Peru, 30-40% of health care workers stopped working due to fears for their own health.⁷ In Brazil, all medical procedures underwent significant reductions, including a 42.6% reduction in screening, a 28.9% reduction in diagnostic procedures, and a 42.5% reduction in appointments with physicians.⁸ There were also reductions in health care services targeting children: in 2020, the number of consultations for children reduced by 59-63% in Mexico and 46-69% in Argentina compared to the 2018-19 average,⁷ and in 2021, only 10 of 33 LAC countries met the minimum recommended immunisation levels for DTP (90% target), although 16 countries had achieved this target during 2017-19.⁷ Care management programs for NCDs were also impacted, with a 39% reduction in diabetes blood tests in Argentina and a 41% reduction in Chile in 2020, and a 63% reduction in breast cancer screenings in Brazil compared to the 2018-19 average.

The majority of the burdens associated with health emergencies fall within the remit of PHC.^{9, 10} A robust PHC system, which is people- and community-oriented and offers comprehensive, continuous, and coordinated care, can significantly enhance preparedness and resilience during and after health emergencies.⁶ Regional resolutions by Pan American Health Organization (PAHO) Member States emphasize a PHC approach to address population needs, accelerate pandemic recovery, and move towards universal health care.¹¹ The World Bank (WB) has also called for investments in PHC-based systems to prepare for future public health emergencies, protect lives, and foster human capital. The WB-PAHO Lancet Regional Health Americas Commission on PHC and Resilience seeks to guide future development of PHC and resilience in the region. Despite efforts to date, there remains a critical need to transform PHC systems in LAC to effectively prevent, respond to and recover from health system shocks, particularly as the region faces risks compounded by weak health systems, fragmentation, constrained budgets, and long-standing inequalities.¹¹

Quantifying the health and economic impacts of potential shocks to PHC services can provide rationale for investing in PHC resilience. Previous work has modelled the impact that shocks to specific health care services could have on health outcomes. For example, one study estimated that a 9·8–18·5% reduction in high-impact maternal and child health interventions and a 10% increase in wasting for a six month period could result in 253,500 additional child deaths and 12,200 additional maternal deaths across 118 LMICs.¹² Another study estimated that reductions to family planning services of different length (3-12 months) and magnitude (5-40%) in 114 LMICs could lead to 325,000 to 15 million additional unintended pregnancies.¹³ In general, while some modelling studies are available that estimate the impact of shocks to health services, they have estimated health impacts only or have estimated health and economic impacts but for a limited subset of interventions. To our knowledge, studies are not available quantifying the health impact of shocks to PHC service coverage, or the economic outcomes associated with these shocks.

The objective of this modelling study was to estimate the health and economic cost of reductions in PHC service coverage due to potential future health system shocks for 33 countries in the LAC region. This evidence can be used as a proxy for the cost of inaction to strengthen PHC, and to provide a rationale to support investment in improving PHC resilience.

Methods

Overview

The potential impact of health system shocks (e.g., epidemiological events such as COVID-19, dengue, and Zika; demographic such as migration; natural disasters, including floods, hurricanes, fires, and earthquakes; and economic, such as recessions and economic crisis) on PHC were approximated as reduced service coverage, with the magnitude and duration of the coverage reduction dependent on the nature of the health system shock. The impact of coverage reductions were estimated for family planning, antenatal care, child health, and non-communicable disease (NCD) services domains (Figure 1), compared to a scenario with maintained coverage (“strengthened primary health care”).

• Download figure
• Open in new tab

Figure 1:

Framework used to estimate the health and economic costs of shocks to primary health care services. The Lives Saved Tool (LiST)¹⁴ was used to estimate the impact of reduced coverage of antenatal care and child health interventions and custom models were used to estimate the impact of reduced primary health care coverage on family planning and non-communicable disease (NCD) mortality and years lived with disability.

• Download figure
• Open in new tab

Figure 2:

Cumulative additional health outcomes for each primary health care shock scenario. Additional stillbirths (top; left), neonatal deaths (top, middle), NCD deaths (top; right), child deaths (bottom; left), maternal deaths (bottom; middle) and additional unintended pregnancies (bottom; right) over 2026-2030.

Results

Differences across countries

Out of the total economic cost of the PHC shock scenarios in the LAC, the largest proportion of costs was from Brazil (33%), followed by Mexico (25%), Venezuela (8%), and Argentina (7%), largely driven by population size (Error! Reference source not found.). The impact of each shock scenario as a proportion of GDP were more similar across countries (Figure 4), with the short and low magnitude scenario costing between 0.02-0.17% of countries’ GDPs, the short and high magnitude scenario costing between 0.04-0.36% of countries’ GDPs, the long and low magnitude scenario costing between 0.05- 0.44% of countries’ GDPs, and the long and high magnitude scenario costing between 0.11-0.89% of countries’ GDPs. The economic cost of PHC shocks was the highest as a percentage of GDP in Haiti (0.17%-0.89%) followed by Honduras (0.14%-0.72%).

• Download figure
• Open in new tab

Figure 3:

Cumulative societal economic cost of primary health care shock scenarios, disaggregated across countries (left), and across cause of mortality or unintended pregnancies (right). Costs are total over 2026-2030 across the 33 countries and subsets of service areas modelled, presented in 2023 US$ with 3% per annum discounting applied. The economic costs vary across countries due to relative population size and epidemiological indicators. FP = family planning; MNCH = maternal, newborn and child health; NCDs = non-communicable diseases.

• Download figure
• Open in new tab

Figure 4:

Economic cost of primary health care shocks as a percentage of the national GDP. 2023 GDP values were used; the economic cost in some countries may appear small relative to population size as not all intervention categories could be included for some of the smaller countries (Appendix A). The economic costs of the shocks occur spread over many years but totals are shown as a percentage of annual GDP for contextualisation.

Discussion

This study estimated the health and economic cost of potential future public health emergencies or large-scale shocks to PHC in the LAC region, as a proxy for the cost of inaction to strengthen PHC. The results suggest that events that result in 25-50% reductions in PHC coverage, with recovery time ranging from one to five years, could lead to an additional 31,000-157,000 deaths (stillbirths, neonatal deaths, child deaths, maternal deaths and deaths due to NCDs), and 2.7-14.1 million unintended pregnancies across the 33 countries in the LAC region. These health impacts translate to cumulative societal economic costs of US$6.9-35.4 billion due to years of life lost and reduced workforce participation.

It is unknown when or what type of health system shocks will occur in the future, and so the range of scenarios considered have been designed to reflect a range of potential shocks. To put these in context, the large magnitude PHC shock scenarios are roughly consistent with observations reported in the LAC region during the COVID-19 pandemic. For example, in Brazil, which represents nearly one third of the population in the region, there was a 42.5% reduction in appointments with physicians⁸ and in Mexico and Argentina there were 59-63% and 46-69% reductions in consultations with children, respectively.⁷ Natural disasters or mass population movements could also have large effects; in 2018, the government of Colombia granted nearly 477,000 special stay permits to Venezuelan citizens in response to a mass migration.⁴⁸ Some events could cross territorial borders: during economic and political crises in Venezuela, malaria cases resurged and neighbouring countries, such as Brazil, reported an increasing trend of imported cases .⁴⁹ It is also possible that multiple health system shocks could occur together and in sequence, which would put further strain on the health system if proper planning and strengthening has not occurred.

The estimated societal economic costs of the PHC shock scenarios represent a lower bound, as they are based on a selection of services delivered through PHC only, and many additional impacts were not accounted for. First, the NCD model only captures mortality and not morbidity. This is because while disutility weights are available for different NCDs, there is limited evidence on the relative reduction in these disutility weights when PHC is accessed and/or the conditions are controlled. In a sensitivity analysis, the total economic cost of the PHC shock scenarios increased to US$12.3-61.9 billion when a 5% reduction in NCD disutility weights was assumed with PHC access (Appendix Figure F.4). Second, only a subset of NCDs were considered, based on the availability of parameters to approximate the impact of PHC on mortality risk (Table 1), and in particular this did not include screening for cancers. Third, other communicable diseases such as HIV, hepatitis and tuberculosis can be managed through PHC and so would also be impacted by shocks to PHC but were not included. Fourth, the estimated costs of PHC shocks were intended as a proxy for the cost of inaction, however investment to strengthen PHC could have additional benefits even if shocks do not occur, which are not captured in this study. Fifth, in the main analysis excess stillbirths and neonatal deaths resulting from additional unintended pregnancies were not included; in a sensitivity analysis their inclusion more than doubled the total economic cost of the PHC shock scenarios (from US$6.9-35.4 billion to US$14.8-75.1; Appendix Figure F.3).

Most of the total impact came from countries with the largest population sizes; however, as a proportion of GDP the relative impacts were similar in each country. For example, Brazil, Mexico, Venezuela and Colombia represent 64% of the population from the 33 countries and accounted for 73% of the total economic costs of the PHC shock scenarios. As a proportion of GDP, the societal cost of these shock scenarios ranged from 0.02-0.17% in the short and low magnitude scenario to 0.11-0.89% in the long and high magnitude scenario. It is difficult to benchmark these estimates against data from COVID-19, where for example during the first year of COVID-19 pandemic aggregate GDP contracted by 7% in the LAC region,⁷ since this observed 7% includes the direct impacts of COVID-19, shocks to health services beyond PHC, and shocks to all other sectors.

Some components of this work can be compared against estimates relating service disruptions from COVID-19 to health outcomes. For example, Lara et al.⁵⁰ estimated that a 28% reduction in non-COVID-19-related hospitalisations in Brazil, Ecuador, Mexico, and Peru could lead to a 15% increase in mortality rates compared with pre-pandemic years, with 89% of the mortality increase due to NCDs. When considering our short and low magnitude PHC shock scenario (25% reduction in PHC for one year) in just these four countries, and only considering maternal, newborn and NCD-related deaths to align with Lara et al.⁵⁰, our model estimates an excess ∼18,000 deaths (an ∼0.25% increase compared to no shock). However, it should be noted that Lara et al. consider hospitalisations for which health system shocks are more likely to have an immediate impact of mortality, whereas this study is considering PHC-delivered interventions only. In another study, UNFPA estimated that across 33 countries in the LAC region, a six-month moderate reduction to public sector utilization of modern contraceptives services could have led to 1.7 million additional unintended pregnancies, 2,900 maternal deaths and 39,000 infant deaths.⁵¹ This is aligned with our modelled estimates of PHC shocks over a one to five year time frame leading to 2.7-14.1 million unintended pregnancies and 2,100-11,300 maternal deaths.

This study did not consider actions required to strengthen PHC, but investment in PHC resilience will be necessary to reduce the impact of external shocks in the future, while also enhancing health services even in the absence of these events. It has been recommended that OECD countries invest 1.4% of GDP annually on average to strengthen health system resilience overall, based on 2019 expenditure, in health workforce, prevention, and key infrastructure.⁵² This includes supporting frontline health professionals (at 0.7% of GDP) and additional investments in preventive care (at 0.3% of GDP), which is broadly consistent with the recent WHO recommendation for countries to allocate extra 1% of GDP targeting PHC.⁵² Policy interventions to expand fiscal space, improve revenue collection, enhance budgeting practices, optimize pooling funds and resource allocations, and establish efficient purchasing and payment systems for providers will be critical.^{7, 52} Programmatically, system fragmentation will need to be reduced, and the benefits packages offered by PHC clearly defined. Infrastructure will also be required, including to strengthen the capacity of supply chains.^{6, 52} This will take coordinated efforts across various government levels and sectors. Finally, accountability, transparency, and equitable resource allocations will be vital to ensure the long-term sustainability of PHC.

There are several limitations to this study. First, demographic and epidemiological data to parametrize the models for each country and service area were taken from a variety of sources, each with their own bias. Where possible, primary sources such as Demographic and Health Services or other large surveys were used, but in many cases model inputs were taken from the LiST model or Institute of Health Metrics and Evaluation databases (Appendix D), which are imputed or are the outputs of other models. In addition, the relative Universal Health Coverage service index¹⁸ for each country was used as a proxy for both PHC coverage and the percentage of people with each NCD who had their disease “controlled” (Appendix Table D.6); it is unclear whether this is likely to be an underestimate or overestimate. Second, not all countries had all service domains considered in the analysis. For example, there was not enough data to model shocks to family planning, ANC, and vaccination services for 4/33 countries (12%), child health services for 21/33 countries (64%), and other pregnancy-related services for 18/33 countries (55%) (Appendix A); however, the missing data was mostly from countries with small populations and so is unlikely to have major impacts to the aggregate outcomes. Third, the cause, timing and magnitude of future shocks is unknown, and so the PHC shock scenarios are theoretical and meant to reflect a range of possibilities rather than being specific to any type of event. We also modelled the shocks to begin in 2026 because the most recent data for each country is more reflective of 2026 than of future years, however in the absence of major changes to epidemiological indicators and interventions, the shocks are likely to have similar impacts if they occurred in future years. Fourth, the cost of strengthening PHC were not included in this modelling because they are currently unknown but should be a subject of further work.

Conclusion

The COVID-19 pandemic led to reductions in health service coverage and highlighted weaknesses in PHC systems in the LAC region. If another health system shock occurred that led to a 25-50% reduction in PHC service coverage with recovery time ranging from one to five years, then across the 33 countries in the LAC region this could lead to an additional 31,000-157,000 deaths (stillbirths, neonatal deaths, child deaths, maternal deaths and deaths due to NCDs) and 2.7-14.1 million unintended pregnancies, corresponding to US$6.9-35.4 billion in societal economic costs. Therefore, substantive investment in PHC resilience would be warranted if it could limit or eliminate the impact of shocks on service coverage.

Funding

The World Bank

Author contributions

All authors conceived the study. TW and NS designed the model with inputs from PAVD, RMGS, MVU. TW implemented model scenarios. TW and NS analysed model outputs. PAVD, RMGS, MVU validated data inputs. TW and NS drafted the manuscript. All authors contributed to the final version of the manuscript.

Competing interests

None declared.