CoChat

The impact of COVID-19 and strategies for mitigation and suppression in low- and middle-income countries

Open access

The DOI matches its Crossref record; 57 of the 57 references it cites were found in a registry or an index; 7 of the 28 sources its claims cite were checked for support.

Abstract

Global prospects for COVID-19 control: Lower-income countries have recognized the potential impact of coronavirus disease 2019 (COVID-19) from observing ongoing epidemics. Many have intervened quickly and early with measures to slow viral transmission, which may partly explain the low rates observed so far in these countries. Walker et al. calibrated a global model with country-specific data (see the Perspective by Metcalf et al. ). Despite the potentially protective effects of younger demographics, the closer intergenerational contact, limitations on health care facilities, and frequency of comorbidities in lower-income countries require sustained nonpharmaceutical interventions (NPIs) to avoid overwhelming health care capacity. As a result of strict NPIs, the protective effects of immunity will be reduced, and it will be important to improve testing capacity. Ensuring equitable provision of oxygen and—when they are ready—pharmaceutical interventions should be a global priority. Science , this issue p. 413 ; see also p. 368

Overview

CoChat Summary

TL;DR

Walker et al. model COVID-19 in low- and middle-income countries using demographic, contact, and health-system data, finding that younger populations reduce per-capita risk but limited health care capacity and quality can offset this advantage. They conclude sustained nonpharmaceutical interventions, especially suppression, are needed to avoid overwhelming health systems, with major testing and oxygen gaps.

Key findings

  • In an unmitigated epidemic with R0 = 3.0, attack rates are similar across settings, slightly higher in LICs due to more homogeneous age mixing; assuming the same comorbidity profile and China-level care, lower-income settings have lower per-capita hospitalization, critical care, and mortality because of younger demography.
  • Health care capacity is strongly income-graded: LICs have a median 1.28 hospital beds per 1000 population versus 4.68 in HICs; ICU beds are 1.47% of hospital beds in LICs versus 3.30% in HICs. Under mitigation, peak critical care demand exceeds supply by a factor of 30.7 (95% UI 14.7–48.8) in a typical LIC versus 7.8 (95% UI 3.6–13.0) in a typical HIC; LMICs and UMICs show factors of 17.5 and 10.9, respectively.
  • Under mitigation, modeled deaths per 1000 population in LICs are 2.1 (95% UI 1.0–3.3) with unlimited HIC-quality care, 4.6 (95% UI 1.4–10.0) with capacity limits, and 6.5 (95% UI 2.7–12.3) with poorer-quality care; in LMICs the corresponding estimates are 2.4 (1.1–3.9), 3.9 (1.5–8.9), and 6.2 (2.8–11.9). Excess deaths due to poorer quality and lack of capacity are 4.4 (1.7–9.0) and 3.8 (1.7–8.0) per 1000 in LICs and LMICs, respectively.
  • For R0 = 3.0, optimal mitigation reduces infections by 30–38% (median 33%) and mortality by 19–55% (median 39%); adding enhanced social distancing of elderly individuals raises mortality reduction to 23–67% (median 49%). Proportional mortality reductions are lower in LICs (median 25.3% vs 49.9% in HICs for optimized mitigation including social distancing; 32.6% vs 60.1% with elderly-enhanced distancing). Suppression with a 75% contact reduction gives Rt = 0.75 for R0 = 3.0; to keep ICU demand below capacity, LICs are predicted to spend 77% of an 18-month period in suppression versus 66% in HICs. Prior to suppression, median reported cases per death were 8.6 in LICs, 19.3 in LMICs, and 72.6 in HICs.

Methods

  • Collated global demographic projections by age and country and social mixing data by age and country-level income; used a simple SIR model to estimate final attack rates and applied age-specific hospitalization, critical care, and infection fatality ratio estimates, with central R0 = 3.0 (range 2.3–3.5).
  • Modeled health care supply using a boosted regression tree with World Bank covariates to predict hospital beds per 1000 population, combined with ICU beds per 100 hospital beds from a systematic review; integrated expert clinical opinion on COVID-19 outcomes under scenarios of unlimited, limited, and limited/no mechanical ventilation/poorer-quality care.
  • Used an age-structured SEIR model to explore mitigation (Rt > 1) and suppression (Rt < 1) under health-system capacity constraints, generating 500 parameter sets with 10 stochastic realizations each; reviewed the ACAP COVID-19 Government Response Measures dataset to summarize when countries implemented suppression measures.

Limitations

  • Stated: the paper cannot predict exact cases, mortality, or disease burden for any country, and full understanding will only be available retrospectively; it does not quantify wider societal and economic impacts, nor does it address the challenge of intensive suppression in LICs/LMICs where informal labor limits adherence.
  • Stated: durable immunity to reinfection remains uncertain; all symptomatic cases are assumed to seek care; hospital stay durations are based on UK data and likely shorter in LIC/LMIC settings, making results conservative; HIC ICU capacity estimates are drawn almost exclusively from a review of Asian countries and may not reflect HICs worldwide; health care quality estimates rely on expert clinical opinion with likely heterogeneity not captured.
  • Evident: the model does not specify the combinations of NPIs needed to achieve contact reductions, and it does not include pharmaceutical interventions except through an assumed 18-month time window; income-stratified representative settings and contact surveys may not capture within-country heterogeneity.

Model-written summary. Check it against the paper before relying on it.

Integrity signals

Registry

Crossref · 10.1126/science.abc0035

Cites

57 of 57 references verified0 citation errors · 0 not checkable

Statistics

Nothing to recompute.

Figures

Fig. 1 Demographic, societal, and mixing patterns relevant to SARS-CoV-2 transmission and burden. (A) Aggregated demographic patterns within 2020 World Population Prospects projections across countries within each 2018 World Bank GDP per-capita decile. (B) Average household (HH) size within Demographic Health Surveys of individuals aged 65 and over by 2018 World Bank GDP per capita. For reference, the average household size of contacts in the UK is also provided as an example for a HIC. (C) Final proportion of population infected in an unmitigated epidemic for an age-structured SIR model with R0 = 3.0 and age-specific social mixing based on contact surveys identified in HICs. (D and E) Equivalent figure for surveys identified in UMICs and LMICs/LICs, respectively. (F to I) Output from simulations across countries of an unmitigated pandemic with R0 = 3.0. (F) Attack rate in terms of number of individuals infected per 1000 population. (G) Equivalent rates of infection leading to illness requiring hospitalization. (H) Illness requiring critical care. (I) Mortality assuming a health system functioning at the level of China throughout the pandemic. LIC, low-income country; LMIC, low- and middle-income country; UMIC, upper–middle-income country HIC, high-income country.
Fig. 1 Demographic, societal, and mixing patterns relevant to SARS-CoV-2 transmission and burden. (A) Aggregated demographic patterns within 2020 World Population Prospects projections across countries within each 2018 World Bank GDP per-capita decile. (B) Average household (HH) size within Demographic Health Surveys of individuals aged 65 and over by 2018 World Bank GDP per capita. For reference, the average household size of contacts in the UK is also provided as an example for a HIC. (C) Final proportion of population infected in an unmitigated epidemic for an age-structured SIR model with R0 = 3.0 and age-specific social mixing based on contact surveys identified in HICs. (D and E) Equivalent figure for surveys identified in UMICs and LMICs/LICs, respectively. (F to I) Output from simulations across countries of an unmitigated pandemic with R0 = 3.0. (F) Attack rate in terms of number of individuals infected per 1000 population. (G) Equivalent rates of infection leading to illness requiring hospitalization. (H) Illness requiring critical care. (I) Mortality assuming a health system functioning at the level of China throughout the pandemic. LIC, low-income country; LMIC, low- and middle-income country; UMIC, upper–middle-income country HIC, high-income country.CC BY · PubMed Central
Fig. 2 Estimates of hospital bed and ICU capacity, and the potential impact of health care quality on the IFR. (A) Comparison of BRT model prediction and empirically observed numbers of hospital beds per 1000 population. Each point represents a country, with the x axis indicating the observed number of hospital beds per 1000 population for that country and the y axis indicating the model-predicted number of hospital beds per 1000 population. Coloring of the points indicates which World Bank income strata the country belongs to. (B) Boxplots of the number of hospital beds per 1000 population, stratified by World Bank income group. Points are modeled estimates of hospital beds per 1000 population obtained from the model. (C) Results from a systematic review describing the percentage of all hospital beds that are in ICUs, stratified by World Bank income group. Error bars indicate the interquartile range of the median. (D) Age-stratified scenarios for the IFR under different health care quality. The baseline age groups are estimates based on data for high-income settings. “No MV” denotes not being able to access an ICU unit with mechanical ventilation available. “Poorer outcomes” represents a higher risk of mortality from severe pneumonia in an LMIC setting if only limited or poor-quality oxygen support is available. “No Oxygen” represents the outcomes if hospitalized patients do not receive oxygen support. The stacked bars represent the cumulative increase in IFR at each stage. Note that the final stage “No MV and No Oxygen” represents the additional IFR due to increasing mortality rates from 20% in the presence of limited or poor-quality oxygen support to 60% in the absence of any oxygen support. (E) Estimated representative IFR averaged across age groups in different settings under a range of health care quality assumptions. The differences between LIC, LMIC, UMIC, and HIC at baseline reflect the demography and social contact patterns but otherwise assume the same health care quality. Lower health care quality is not shown for UMIC and HIC as these settings are likely to have the quality of health care incorporated in the baseline estimates.
Fig. 2 Estimates of hospital bed and ICU capacity, and the potential impact of health care quality on the IFR. (A) Comparison of BRT model prediction and empirically observed numbers of hospital beds per 1000 population. Each point represents a country, with the x axis indicating the observed number of hospital beds per 1000 population for that country and the y axis indicating the model-predicted number of hospital beds per 1000 population. Coloring of the points indicates which World Bank income strata the country belongs to. (B) Boxplots of the number of hospital beds per 1000 population, stratified by World Bank income group. Points are modeled estimates of hospital beds per 1000 population obtained from the model. (C) Results from a systematic review describing the percentage of all hospital beds that are in ICUs, stratified by World Bank income group. Error bars indicate the interquartile range of the median. (D) Age-stratified scenarios for the IFR under different health care quality. The baseline age groups are estimates based on data for high-income settings. “No MV” denotes not being able to access an ICU unit with mechanical ventilation available. “Poorer outcomes” represents a higher risk of mortality from severe pneumonia in an LMIC setting if only limited or poor-quality oxygen support is available. “No Oxygen” represents the outcomes if hospitalized patients do not receive oxygen support. The stacked bars represent the cumulative increase in IFR at each stage. Note that the final stage “No MV and No Oxygen” represents the additional IFR due to increasing mortality rates from 20% in the presence of limited or poor-quality oxygen support to 60% in the absence of any oxygen support. (E) Estimated representative IFR averaged across age groups in different settings under a range of health care quality assumptions. The differences between LIC, LMIC, UMIC, and HIC at baseline reflect the demography and social contact patterns but otherwise assume the same health care quality. Lower health care quality is not shown for UMIC and HIC as these settings are likely to have the quality of health care incorporated in the baseline estimates.CC BY · PubMed Central
Fig. 3 The prevalence of different comorbidities across income settings and the proportion of SARS-CoV-2 infections co-occurring with them. The age distribution of comorbidities relevant as modifiers of COVID-19 disease severity was extracted from Global Burden of Disease 2017 estimates (12) and integrated with estimates of the predicted age distribution of infection in an unmitigated pandemic scenario. For (A) cardiovascular disease (CVD), (B) chronic obstructive pulmonary disease (COPD), (C) diabetes, (D) HIV/AIDS, (E) malnutrition, and (F) tuberculosis, the left heatmap shows the age distribution of these comorbidities across different income settings, expressed as the proportion of the population in that income setting that has the comorbidity. The bar charts (colored according to age group) show the number of infections per 1000 population that co-occur with the respective comorbidity.
Fig. 3 The prevalence of different comorbidities across income settings and the proportion of SARS-CoV-2 infections co-occurring with them. The age distribution of comorbidities relevant as modifiers of COVID-19 disease severity was extracted from Global Burden of Disease 2017 estimates (12) and integrated with estimates of the predicted age distribution of infection in an unmitigated pandemic scenario. For (A) cardiovascular disease (CVD), (B) chronic obstructive pulmonary disease (COPD), (C) diabetes, (D) HIV/AIDS, (E) malnutrition, and (F) tuberculosis, the left heatmap shows the age distribution of these comorbidities across different income settings, expressed as the proportion of the population in that income setting that has the comorbidity. The bar charts (colored according to age group) show the number of infections per 1000 population that co-occur with the respective comorbidity.CC BY · PubMed Central
Fig. 4 The impact of health care capacity and quality on COVID-19 mortality in different settings. (A) Representative epidemic trajectories for an unmitigated epidemic (gray line), an epidemic involving minimal social distancing (pale blue line, 20% reduction in social contacts), an epidemic involving extensive social distancing (teal, 45% reduction in social contacts), and an epidemic trajectory that involves extensive suppression (75% reduction in social contacts) followed by lifting of restrictions after 6 months, leading to resurgence (dark blue line). (B) The excess deaths associated with constraints on health care quality and quantity, including the deaths associated with a hypothetical setting with unlimited high-quality health care (green lines), settings where high-quality health care is available but limited (yellow lines), and settings where only limited, poorer-quality health care is available (orange lines). Pale lines show an unmitigated scenario, colored lines a mitigated scenario. (C) The multiple by which ICU demand exceeds capacity for each World Bank income strata for an unmitigated (gray) and mitigated (teal) epidemic. (D) The modeled IFR for different World Bank income strata under different scenarios of health care quality and quantity available, assuming a mitigated scenario in which baseline contacts are reduced by 45%. (E) The modeled deaths per million population for different World Bank income strata under different assumptions of health care quality and quantity available, assuming a mitigated scenario in which baseline contacts are reduced by 45%. Plots show medians (bars) and interquartile ranges (boxes), as well as points <1.5× the interquartile range (whiskers) and >1.5× (points) from 500 parameter draws.
Fig. 4 The impact of health care capacity and quality on COVID-19 mortality in different settings. (A) Representative epidemic trajectories for an unmitigated epidemic (gray line), an epidemic involving minimal social distancing (pale blue line, 20% reduction in social contacts), an epidemic involving extensive social distancing (teal, 45% reduction in social contacts), and an epidemic trajectory that involves extensive suppression (75% reduction in social contacts) followed by lifting of restrictions after 6 months, leading to resurgence (dark blue line). (B) The excess deaths associated with constraints on health care quality and quantity, including the deaths associated with a hypothetical setting with unlimited high-quality health care (green lines), settings where high-quality health care is available but limited (yellow lines), and settings where only limited, poorer-quality health care is available (orange lines). Pale lines show an unmitigated scenario, colored lines a mitigated scenario. (C) The multiple by which ICU demand exceeds capacity for each World Bank income strata for an unmitigated (gray) and mitigated (teal) epidemic. (D) The modeled IFR for different World Bank income strata under different scenarios of health care quality and quantity available, assuming a mitigated scenario in which baseline contacts are reduced by 45%. (E) The modeled deaths per million population for different World Bank income strata under different assumptions of health care quality and quantity available, assuming a mitigated scenario in which baseline contacts are reduced by 45%. Plots show medians (bars) and interquartile ranges (boxes), as well as points <1.5× the interquartile range (whiskers) and >1.5× (points) from 500 parameter draws.CC BY · PubMed Central
Fig. 5 The proportion of time that countries will need to spend in lockdown to remain within health-system critical care capacity. Scenarios are generated using the stochastic SEIR model (see materials and methods). (A) The time period between lockdowns for a representative LIC setting, and how it varies with the extent of suppression during lockdown. Gray-shaded area denotes time period of first suppression (triggered at a threshold of 60 ICU cases per day), and brown (75% reduction) and green (85% reduction) vertical lines indicate the next time point at which suppression would be implemented (using the same threshold). (B) Time under suppression over the next 18 months for triggering thresholds of 30 (pale pink) and 500 (brighter pink) ICU cases per day, respectively. Gray-shaded areas indicate time in suppression (a 75% reduction in R0). (C) The proportion of time required to be spent in lockdown over the next 18 months as a function of the maximum ICU demand for a representative LIC, LMIC, UMIC, and HIC (colored purple lines). Colored points indicate the median ICU capacity for each of these different income strata. (D) The proportion of time required to be spent in lockdown over the next 18 months as a function of the number of deaths caused by the COVID-19 epidemic for a representative LIC, LMIC, UMIC, and HIC assuming comparable quality (but not quantity) of health care across all settings (colored purple lines), and when assuming a reduction in the quality of health care available in LICs and LMICs (red and orange dashed lines, respectively). (E) Modeled COVID-19 epidemic trajectories over the next 18 months for a representative LIC, LMIC, UMIC, and HIC where suppression is implemented at ICU incidence trigger thresholds to keep the maximum ICU demand beneath 50% of ICU capacity. The first triggering of suppression has been determined on the basis of the actual patterns of suppression timing observed across LICs, LMICs, UMICs, and HICs.
Fig. 5 The proportion of time that countries will need to spend in lockdown to remain within health-system critical care capacity. Scenarios are generated using the stochastic SEIR model (see materials and methods). (A) The time period between lockdowns for a representative LIC setting, and how it varies with the extent of suppression during lockdown. Gray-shaded area denotes time period of first suppression (triggered at a threshold of 60 ICU cases per day), and brown (75% reduction) and green (85% reduction) vertical lines indicate the next time point at which suppression would be implemented (using the same threshold). (B) Time under suppression over the next 18 months for triggering thresholds of 30 (pale pink) and 500 (brighter pink) ICU cases per day, respectively. Gray-shaded areas indicate time in suppression (a 75% reduction in R0). (C) The proportion of time required to be spent in lockdown over the next 18 months as a function of the maximum ICU demand for a representative LIC, LMIC, UMIC, and HIC (colored purple lines). Colored points indicate the median ICU capacity for each of these different income strata. (D) The proportion of time required to be spent in lockdown over the next 18 months as a function of the number of deaths caused by the COVID-19 epidemic for a representative LIC, LMIC, UMIC, and HIC assuming comparable quality (but not quantity) of health care across all settings (colored purple lines), and when assuming a reduction in the quality of health care available in LICs and LMICs (red and orange dashed lines, respectively). (E) Modeled COVID-19 epidemic trajectories over the next 18 months for a representative LIC, LMIC, UMIC, and HIC where suppression is implemented at ICU incidence trigger thresholds to keep the maximum ICU demand beneath 50% of ICU capacity. The first triggering of suppression has been determined on the basis of the actual patterns of suppression timing observed across LICs, LMICs, UMICs, and HICs.CC BY · PubMed Central

Verification

Claims the paper makes

4 Valid · 1 Not valid · 16 Not checked · 64 Own results and interpretations

47 more claims

Create a free account to see all 85 claims and to verify the 20 references still open.

See all claims — free account
  • It may state two separate claims as one. 29
  • Probably not a substantial claim. 7
  • The quote only partly states this claim. 1

Create a free account to see these claims.

Claims other papers make about this one

Each is checked against this paper’s own text: does it say what the citing paper says it does?

Self-check

Percentages, sums and reported differences re-computed from the paper’s own numbers.

1 of 1

Citations

Cites

Show all 57 references

Cited by

Items 601–700 of 991

Metadata

Authors
Walker, P. G. T., Whittaker, C., Watson, O. J., Baguelin, M., Winskill, P., Hamlet, A., Djafaara, B. A., Cucunubá, Z., Olivera Mesa, D., Green, W., Thompson, H., Nayagam, S., Ainslie, K. E. C., Bhatia, S., Bhatt, S., Boonyasiri, A., Boyd, O., Brazeau, N. F., Cattarino, L., Cuomo-Dannenburg, G., Dighe, A., Donnelly, C. A., Dorigatti, I., van Elsland, S. L., FitzJohn, R., Fu, H., Gaythorpe, K. A. M., Geidelberg, L., Grassly, N., Haw, D., Hayes, S., Hinsley, W., Imai, N., Jorgensen, D., Knock, E., Laydon, D., Mishra, S., Nedjati-Gilani, G., Okell, L. C., Unwin, H. J., Verity, R., Vollmer, M., Walters, C. E., Wang, H., Wang, Y., Xi, X., Lalloo, D. G., Ferguson, N. M., & Ghani, A. C.
Affiliations
  • Imperial College London (GB)
  • Brown University (US)
  • London School of Hygiene & Tropical Medicine (GB)
  • MRC Centre for Global Infectious Disease Analysis, Department of Infectious Disease Epidemiology, Imperial College London, London, UK.
  • University of Oxford (GB)
  • Liverpool School of Tropical Medicine (GB)
Corresponding author
  • Walker, Patrick G. T.
  • Whittaker, Charles
  • Watson, Oliver J.
  • Ferguson, Neil M.
  • Ghani, Azra C.
Venue
Science
Published
2020-07-24
Volume
369
Issue
6502
Pages
413-422
ISSN
  • 0036-8075 (ISSN-L)
  • 1095-9203
OpenAlex
W3035706118
Cited by
993
References
54
Citations by year
2026: 18 · 2025: 42 · 2024: 69 · 2023: 122 · 2022: 215 · 2021: 281 · 2020: 245
FWCI
34.30
Type
journal article
Language
en
Open access
Locations
  • Science (publishedVersion, cc-by)
  • PubMed Central (publishedVersion, cc-by)
  • PubMed (publishedVersion)
  • Bristol Research (University of Bristol) (publishedVersion)
  • LSTM Online Archive (Liverpool School of Tropical Medicine) (acceptedVersion)
  • LSHTM Research Online (London School of Hygiene and Tropical Medicine) (acceptedVersion)
  • Spiral (Imperial College London) (submittedVersion, other-oa)
Funders
  • Wellcome (221350/Z/20/Z)
  • Medical Research Council
  • National Institute for Health Research
  • Community Jameel
Topics
  • Physical Sciences › Mathematics › Modeling and Simulation › COVID-19 epidemiological studies
  • Social Sciences › Sociology and Political Science › COVID-19 Prevention and Impact
  • Health Sciences › Medicine › Infectious Diseases › SARS-CoV-2 and COVID-19 Research
Keywords
COVID-19 pandemic, low- and middle-income countries, mitigation measures, suppression strategy, health system capacity, excess mortality, intergenerational interaction, disease severity, contact patterns, age structure, quality of care, pandemic response
MeSH terms
COVID-19, Developing Countries, Humans, Patient Acceptance of Health Care, Pneumonia, Viral, Poverty, Public Health, Global Health, Coronavirus Infections, Pandemics

About these checks

Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.

The claim verdicts this report counts are experimental: their accuracy is not yet measured.

A language model drew the claims from the paper on 2026-10-04, under version 3 of the claim definition (model deepseek-v4.1-flash, prompt extract-v3).

A second model labelled each claim on 2026-10-04 (model jev-1.13-20260917, definition revision gold-guideline-v0.2).

Cited sources are checked with a language model’s help (qwen3.8-omni-flash and gpt-6-sol).

Statistics and arithmetic are recomputed, not judged.

Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.

What counts as a claim ›How we measure confidence ›How we check this ›

Are you an author, or do you see a problem with this page? Request a review

Data sources

Bibliographic data: Semantic Scholar, OpenAlex, Crossref.

Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).

Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.