COVID-19: Forecasting mortality given mobility trend data and non-pharmaceutical interventions

Mark J. Willis (3); Oscar Andres Prado-Rubio (2); Victor Hugo Grisales Diaz (1)

REVIEW 4 major objections 3 minor 20 references

COVID-19: Forecasting mortality given mobility trend data and non-pharmaceutical interventions

T0 review · 4 major / 3 minor · reviewed 2026-08-27 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read Mobility restrictions account for roughly 47% of the suppression of COVID-19 transmission, and a hybrid SIRD model with mobility and NPI signals forecasts four-week-ahead cumulative deaths within about 3.5–3.8% MAPE.

desk verdict A credible short-term mortality forecasting model undermined by an unidentified NPI input that makes the central 47% attribution non-identifiable. read the letter →

arxiv 2009.12171 v2 pith:IMHJUE3K submitted 2020-09-25 q-bio.PE math.CAmath.OCq-bio.QM

Victor Hugo Grisales Diaz (1) , Oscar Andres Prado-Rubio (2) , Mark J. Willis (3) This is my paper · ORCID

classification q-bio.PEmath.CAmath.OCq-bio.QM MSC 92D30

keywords COVID-19mobilityrestrictionsnon-pharmaceuticalinterventionsSIRDcompartmentalmodeltransmissionratesuppressionmortalityforecastingAppledatacalibration

verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

This paper sets out to separate the effect of mobility restrictions from the effect of all other non-pharmaceutical interventions (NPIs) on COVID-19 transmission. Using a compartmental SIRD model whose transmission rate responds to an Apple mobility signal and a separate NPI signal, the authors calibrate the model for 49 countries and 51 U.S. states. They estimate that mobility restrictions contribute about 47% of the total reduction in the transmission rate, with other NPIs contributing the rest. If the model is right, short-term cumulative-death forecasts can be made with a mean absolute percentage error of 3.5% for the U.S. and 3.8% worldwide, which is accurate enough to inform planning.

What carries the argument

The key mechanism is equation (7), a first-order differential equation for the deviation of the transmission rate: $\tau\, d\beta'/dt + \beta' = -k_a u_a - k_m u_m$, with $\beta(t) = \beta_I + \beta'(t)$. The signal $u_m$ is a five-slope piecewise-linear model fitted to Apple's reported mobility indices, normalized to lie between zero and one, and $u_a$ represents all other NPIs. The gains $k_m$ and $k_a$ are obtained by per-region least-squares calibration together with the other model parameters, and the time constant $\tau$ measures how many days are required for most of the suppression to take effect. This formulation is what lets the model attribute separate shares of disease suppression to mobility versus other interventions.

What would settle it

Re-estimate the model for a single country twice, once with $u_a$ as a simple Boolean lockdown indicator and once with a graded stringency index, and compare the resulting $S_m$ values; if the mobility share moves outside the reported confidence intervals, the attribution is not robust to how 'other NPIs' is encoded. A second check is to test a region where mobility dropped sharply but no other NPIs were introduced, and verify that the fitted $k_m$ alone reproduces the observed decline in transmission.

Watch

Extended reading notes

Core claim

The central claim is that the time-varying transmission rate $\beta(t)$ can be written as a constant baseline $\beta_I$ plus a first-order filtered response to two control inputs: $u_m$, a mobility-reduction signal built from Apple's walking, driving and transit indices, and $u_a$, a signal for all other NPIs such as masks and school closures. The fitted gains $k_m$ and $k_a$ then enter the share $S_m = k_m/(k_a + k_m)$, which the paper estimates as approximately 0.47 both worldwide and for U.S. states. The same calibrated model, applied to data up to 13 August 2020, forecasts cumulative deaths four weeks ahead with a mean absolute percentage error of 3.5% (U.S. states) and 3.8% (worldwide), and the paper reports an average $R^2$ of 0.98 for cumulative confirmed and deceased cases over this later calibration window. The authors present this as the first demonstration across a large set of regions that mobility trend data and NPI signals together are needed to capture the dynamics and to forecast mortality.

Load-bearing premise

The estimate of 47% rests on the assumption that the mobility signal and the other-NPI signal are measured correctly, are not correlated with each other, and combine additively in their effect on transmission; if the two effects overlap or interact, the split between them is not identifiable.

Editorial extensions

If this is right

If mobility restrictions account for about half of transmission suppression, lifting them without compensating measures should produce a detectable rise in transmission of a size set by $k_m$.
Four-week-ahead forecasts of cumulative deaths with MAPE below 4% are usable for mortality planning, while forecasts of confirmed cases (MAPE about 10–11.5%) are less reliable.
The same calibration yields an effective reproduction number over time; the paper reports initial values near 3.0 worldwide and 3.2 in U.S. states, and finds many regions still above one as of 13 August 2020.
Extending the calibration window from 30 June to 13 August 2020 reduces forecast error, indicating that the method improves as more data accumulate.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

A natural extension is to replace the single 'other NPIs' signal with separate signals for masks, school closures, and gathering bans; the 47% share may change if those effects are modeled individually.
Because the model forecasts deaths more accurately than confirmed cases, deaths could serve as the primary calibration target in regions with patchy case reporting.
The mobility forecast itself is a weak point for policy simulation: new, unexpected lockdowns (such as Melbourne's) break the four-week mobility projection, so the model would need a behavioral rule for how mobility responds to new restrictions before it can be used for counterfactual policy.

Share X Bluesky LinkedIn Reddit HN

Formalized claims in Lean

Claim #1: The central claim is that the time-varying transmission rate $\beta(t)$ can be written as a constant baseline $\beta_I$ plus a first-order filtered response to two control inputs: $u_m$, a mobility-reduction signal built from Apple's walking, driving and transit indices, and $u_a$, a signal for all other NPIs such as masks and school closures. The fitted gains $k_m$ and $k_a$ then enter the share

/-- @claim 1 The central claim is that the time-varying transmission rate $\beta(t)$ can be written as a constant baseline $\beta_I$ plus a first-order filtered response to two control inputs: $u_m$, a mobility-reduction signal built from Apple's walking, driving and transit indices, and $u_a$, a signal for all other NPIs such as masks and school closures. The fitted gains $k_m$ and $k_a$ then enter the share  -/
def central_claim : Prop :=

Signed reviews

No signed human review yet.

Request a human review

A listed scientist reviews the paper for a fee and the review publishes here regardless of verdict. See the reviewers or get listed.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Desk Editor's Note

read the letter

Colleague,

The honest headline: this paper has a real, verified forecasting result and a central attribution claim that outruns the evidence. Four-week-ahead mortality forecasts with MAPE around 3.5% on 100 regions, evaluated on data not used for calibration, is a genuine out-of-sample result. That part deserves serious attention. The 47% mobility-attribution number, though, is not identified by the data as presented. Eq. (7) splits the transmission-rate reduction between a mobility signal u_m and an 'all other NPIs' signal u_a, but u_a is never defined—no data source, no Boolean/continuous specification, no construction recipe. Since the equation is linear in the two inputs, the least-squares split between k_m and k_a depends directly on the covariance of those two signals. Pick a different u_a and S_m moves. The ±10–14% intervals are local covariance bounds around one unspecified specification; they don't test the specification. The paper gives no sensitivity analysis, no alternative model with interactions or separate time constants, and no code. So the 47% is a modeling assumption expressed as an empirical result.

What is actually good: the SIRD-MC extension is straightforward but useful. The authors calibrate to a large set of regions, compare against IHME and Youyang Gu, and report MdAPE in a comparable way. The forecast validation is true out-of-sample, with errors that are competitive. The mobility-signal preprocessing (piecewise linear fit to Apple data) is described in some detail, and the forecasting of future mobility is honest about its limitations (Melbourne lockdown breaks the trend).

The soft spots beyond the attribution issue: the compartment bookkeeping is sloppy—R(t) is defined as recovered plus deceased, then C(t)=I+R is called cumulative confirmed cases, which double-counts deaths; Eq. (9) doesn't clarify. Minor, but it obscures what the model actually tracks. No code or data release also makes independent replication harder, and the forecast comparison table is not apples-to-apples across models (different calibration dates, different region sets), though the authors do match MdAPE conventions.

Who should read this: people building hybrid compartmental models for epidemic forecasting, and modelers wanting a benchmark for short-term death forecasts. The forecasting contribution is citable. The attribution number should not be cited until u_a is disclosed and a sensitivity analysis is done.

Recommendation: send to peer review, but with a required major revision: define u_a exactly, run the model under alternative u_a specifications and interaction terms, and report how S_m changes. If the authors do that, the paper could be solid. If they can't, the 47% claim should be dropped or reframed as conditional on their input choice.

Referee Report

4 major / 3 minor

Summary. The paper develops a hybrid SIRD compartmental model (SIRD-MC) in which the time-varying transmission rate obeys a first-order equation driven by two additive inputs: an Apple mobility-based signal u_m and an unspecified 'other NPI' signal u_a. Seven parameters per region are calibrated for 49 countries and 51 U.S. states using least squares. The central attribution claim is that mobility restrictions account for S_m = k_m/(k_m+k_a) ≈ 47% of the suppression of the transmission rate. The forecasting claim is that four-week-ahead forecasts of cumulative deaths achieve MAPE of 3.5% (U.S. states) and 3.8% (worldwide) when calibrated with data up to 13/08/2020, with comparable or better MdAPE than IHME, Youyang Gu, and LANL benchmarks. The paper reports out-of-sample forecast validation, R² values, parameter tables, and effective reproduction number trajectories.

Significance. If the attribution result were identified, the paper would provide a policy-relevant decomposition of the effect of mobility restrictions versus other NPIs, and the out-of-sample four-week forecasts would be a useful benchmark for short-term mortality planning. The genuine strengths are the size of the calibration exercise (100 regions), the use of true out-of-sample forecast windows, and the explicit comparison with established forecast models. However, the central attribution estimate is not currently identified from the described data because the second input u_a is not defined, and the reported forecast accuracy depends on unstated assumptions for the future path of u_a. The paper also lacks the referenced appendices, so the model and calibration are not reproducible as submitted.

major comments (4)

[Non-Pharmaceutical Intervention policies, Eq. (7)] The term u_a(t) is never operationally defined: the paper states it represents 'all other NPIs' but does not specify the data source, whether it is Boolean or continuous, the timing of on/off events, or how its intensity is scaled. Because Eq. (7) is linear and additive in u_m and u_a, the least-squares split of the gains k_m and k_a, and hence the share S_m in Eq. (8), depends directly on the covariance between u_a and u_m. Without a definition of u_a and a sensitivity analysis over plausible alternative specifications, the 47% attribution is not identified; the reported nlparci intervals are local covariance intervals around one unspecified specification and do not address this model uncertainty.
[Model Forecasting predictive power] The four-week-ahead forecasts require future values of both u_m and u_a in Eq. (7). The paper explains how u_m is extrapolated with the piecewise linear mobility model, but it never states how u_a is set for the forecast horizon (constant at its last value, set to zero, or manually scheduled). Consequently the reported MAPE and MdAPE values are conditional on an unstated assumption, and the forecasting claim is not fully reproducible. In addition, the MAPE calculation that selects the lowest absolute percentage error among the mean, maximum, and minimum of the three simulated bounds is an optimistic error metric; the standard MAPE or MdAPE using the point forecast should also be reported.
[SIRD-MC model, Eqs. (3)-(6)] There is an internal inconsistency in the compartment definitions. The text states that R(t) is the summatory of recovered and deceased individuals, but Eq. (3) gives dR/dt = α·I·x, which draws only from the reported infectious compartment, while Eq. (5) gives dD/dt = μ·I_d with no corresponding contribution to R(t). If R(t) is meant to include deceased individuals, then the ODE should include a term for deaths; if not, the text is wrong. This ambiguity also affects the definition of cumulative confirmed cases C(t)=I+R and the interpretation of the fitted parameters, so it should be resolved rather than treated as a purely verbal issue.
[Appendices A-C] The manuscript repeatedly refers to Appendix A (mobility model), Appendix B (objective function), and Appendix C (parameter values), but these appendices are not present in the submitted text. Without Appendix A, the mathematical form of the mobility signal u_m is not fully defined; without Appendix B, the least-squares objective and constraints are not specified; without Appendix C, the per-region parameters that underlie the averages in Table 1 cannot be checked. Until the appendices are included, the calibration and attribution results are not reproducible.

minor comments (3)

[Results, Mobility trend model and forecasting] The sentence '37 of the 49 countries studied does not have restrictions of mobility' has a subject-verb agreement error and should read 'do not have.'
[Model Forecasting predictive power] The paper introduces MdAPE but never defines it explicitly; the definition should be given alongside the MAPE definition, and it should be made clear that Table 2 uses the point-forecast MdAPE rather than the minimum-of-bounds version.
[Results, Disease suppression with and without mobility restrictions] The text reports the effective reproduction number in Figure 3 but does not give the formula used to compute it from the model; a one-line definition would improve reproducibility.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the forecasting claim is validated on held-out data and the 47% attribution is a fitted parameter ratio, not an input by construction.

full rationale

The paper's central forecasting claim is tested on data not used for calibration: parameters are fitted to data up to 30/06/2020 or 13/08/2020 and four-week-ahead cumulative deaths are compared with later observations, so the reported MAPE/MdAPE values are genuine predictions rather than fitted values renamed as forecasts. The attribution S_m = k_m/(k_a + k_m) in Eq. (8) is an arithmetic ratio of two calibrated gains appearing in the stated differential equation (Eq. 7), not an input defined in terms of the output; the paper does not fit S_m directly, nor does it define u_a in terms of S_m. The use of the authors' earlier model [5,6] is transparent, and the model equations are restated in this paper rather than imported as an unverified uniqueness theorem, so the self-citation is not load-bearing in a circular way. External benchmarks (IHME, Youyang Gu, LANL) are used for comparison, and the model is self-contained in that all fitted parameters are estimated from the data described. The lack of a precise specification for the other-NPI signal u_a and the absence of a sensitivity analysis are identifiability and reproducibility concerns, but they do not amount to circularity because the reported value is not equal to an input by construction and the forecasting evaluation is genuinely out-of-sample.

Assumptions & free parameters 8 free parameters · 7 assumptions · 0 invented entities

The central 47% claim rests on the additive, independently measured NPI inputs in Eq. (7) and the fitted gains, while the forecasting claim rests on out-of-sample validation. The seven fitted dynamic parameters plus the mobility signal model are the main free inputs; no external entity is invented.

free parameters (8)

k_m (mobility gain) = 0.1863 (worldwide average, data to 13/08/2020)
Fitted per country/state via lsqnonlin; directly determines the 47% mobility share through Eq. (8).
k_a (other-NPI gain) = 0.208 (worldwide average, data to 13/08/2020)
Fitted per country/state; paired with k_m in the suppression share.
x (under-reporting factor) = 39.4 (worldwide average)
Scales the removal rate term to match reported recoveries; fitted.
beta_0 (initial transmission rate) = 0.3597 (worldwide average)
Determines initial R0 (approximately beta_0/alpha); fitted.
mu (fatality rate) = 0.2284 (worldwide average)
Converts the symptomatic infected compartment into deaths; fitted and highly variable across regions.
tau (time constant) = 35.06 days (worldwide average)
Time constant of the first-order transmission-reduction dynamics; fitted.
y (I_d/I proportionality) = 0.9026 (worldwide average)
Proportionality between the death-prone compartment and the infected compartment; fitted between zero and one.
Mobility signal piecewise linear parameters = not reported in main text
Intercept and up to five slope segments, plus changepoints, define u_m from Apple data; fitted or chosen per region in Appendix A.

assumptions (7)

domain assumption Susceptible-infected-removed-dead compartmental structure with homogeneous mixing (Eqs. 1-6).
Standard epidemic modeling assumption; ignores age, contact structure, and spatial heterogeneity.
domain assumption Constant removal rate alpha = 1/8 day^-1 from infected compartments.
Cited as reasonable from prior work; not fitted to the data in this paper.
ad hoc to paper Transmission rate beta(t) is driven by two independent additive first-order inputs u_m and u_a (Eq. 7).
Core modeling choice; no empirical test of additivity or independence is given, and the 47% attribution is a function of this structure.
ad hoc to paper Mobility signal u_m from Apple data is well approximated by a piecewise linear function with five slope changes.
Used to smooth noisy mobility data; the number of slope changes appears chosen for fit rather than data-driven.
ad hoc to paper The other-NPI signal u_a is Boolean (0 or 1) and correctly represents the timing and intensity of all non-mobility NPIs.
Construction of u_a is not detailed in the main text; errors propagate directly into k_a and the mobility share.
domain assumption Deaths accrue only from the I_d compartment with rate mu, and I_d is proportional to I via y.
Simplifies clinical progression; not validated against hospitalization or symptom data.
standard math Least-squares estimates are approximately normal so that nlparci confidence intervals are valid.
Standard asymptotic assumption; may be optimistic with seven fitted parameters and noisy data.

how reviews work

0 comments

Cite this review

Pith. "Pith review of COVID-19: Forecasting mortality given mobility trend data and non-pharmaceutical interventions." pith.science (2026). https://pith.science/paper/IMHJUE3K

@misc{pith2026200912171,
  author       = {Pith},
  title        = {Pith review of: COVID-19: Forecasting mortality given mobility trend data and non-pharmaceutical interventions},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/IMHJUE3K}},
  note         = {Machine review of arXiv:2009.12171}
}

read the original abstract

We develop a novel hybrid epidemiological model and a specific methodology for its calibration to distinguish and assess the impact of mobility restrictions (given by Apple's mobility trends data) from other complementary non-pharmaceutical interventions (NPIs) used to control the spread of COVID-19. Using the calibrated model, we estimate that mobility restrictions contribute to 47 % (US States) and 47 % (worldwide) of the overall suppression of the disease transmission rate using data up to 13/08/2020. The forecast capacity of our model was evaluated doing four-weeks ahead predictions. Using data up to 30/06/20 for calibration, the mean absolute percentage error (MAPE) of the prediction of cumulative deceased individuals was 5.0 % for the United States (51 states) and 6.7 % worldwide (49 countries). This MAPE was reduced to 3.5% for the US and 3.8% worldwide using data up to 13/08/2020. We find that the MAPE was higher for the total confirmed cases at 11.5% worldwide and 10.2% for the US States using data up to 13/08/2020. Our calibrated model achieves an average R-Squared value for cumulative confirmed and deceased cases of 0.992 using data up to 30/06/20 and 0.98 using data up to 13/08/20.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

20 extracted references · 19 canonical work pages

[1]

Hsiang, D

S. Hsiang, D. Allen, S. Annan-Phan, K. Bell, I. Bolliger, T. Chong, H. Druckenmiller, L.Y. Huang, A. Hultgren, E. Krasovich, P. Lau, J. Lee, E. Rolf, J. Tseng, T. Wu, The effect of large-scale anti-contagion policies on the COVID-19 pandemic., Nature. (2020). https://doi.org/10.1038/s41586-020-2404-8

work page doi:10.1038/s41586-020-2404-8 2020
[2]

https://doi.org/https://www.worldometers.info/coronavirus/country/us/

Wolrdometers, Coronavirus Cases, (2020). https://doi.org/https://www.worldometers.info/coronavirus/country/us/

work page 2020
[3]

Buckee, M.A

C.O. Buckee, M.A. Johansson, Individual model forecasts can be misleading, but together they are useful, Eur. J. Epidemiol. (2020) 731–732. https://doi.org/10.1007/s10654-020-00667-8

work page doi:10.1007/s10654-020-00667-8 2020
[4]

A., Wilder-Smith, A

Y. Liu, A.A. Gayle, A. Wilder-Smith, J. Rocklöv, The reproductive number of COVID-19 is higher compared to SARS coronavirus, J. Travel Med. 27 (2020) 1–4. https://doi.org/10.1093/jtm/taaa021

work page doi:10.1093/jtm/taaa021 2020
[5]

Willis, V.H.G

M.J. Willis, V.H.G. Díaz, O.A. Prado-Rubio, M. von Stosch, Insights into the dynamics and control of COVID- 19 infection rates, Chaos, Solitons & Fractals. 138 (2020) 109937. https://doi.org/10.1016/j.chaos.2020.109937

work page arXiv 2020
[6]

Willis, A.R

M.J. Willis, A.R. Wright, V. Bramfitt, V.H. Grisales Díaz, COVID-19: Mechanistic model calibration subject to active and varying non-pharmaceutical interventions, MedRxiv. (n.d.). https://doi.org/https://doi.org/10.1101/2020.09.10.20191817

work page doi:10.1101/2020.09.10.20191817 2020
[7]

https://covid19.apple.com/mobility

Apple, Apple Mobility Trends Reports, (2020). https://covid19.apple.com/mobility

work page 2020
[8]

Flaxman, S

S. Flaxman, S. Mishra, A. Gandy, H.J.T. Unwin, T.A. Mellan, H. Coupland, C. Whittaker, H. Zhu, T. Berah, J.W. Eaton, M. Monod, P.N. Perez-Guzman, N. Schmit, L. Cilloni, K.E.C. Ainslie, M. Baguelin, A. Boonyasiri, O. Boyd, L. Cattarino, L. V. Cooper, Z. Cucunubá, G. Cuomo-Dannenburg, A. Dighe, B. Djaafara, I. Dorigatti, S.L. van Elsland, R.G. FitzJohn, K.A...

2020

Show all 20 references

[9]

Delen, E

D. Delen, E. Eryarsoy, B. Davazdahemami, No Place Like Home: Cross-National Data Analysis of the Efficacy of Social Distancing During the COVID-19 Pandemic, JMIR Public Heal. Surveill. 6 (2020) e19862. https://doi.org/10.2196/19862

2020 doi
[10]

Greenstone, V

M. Greenstone, V. Nigam, Does Social Distancing Matter?, SSRN Electron. J. (2020). https://doi.org/10.2139/ssrn.3561244

2020 doi
[11]

Koo, A.R

J.R. Koo, A.R. Cook, M. Park, Y. Sun, H. Sun, J.T. Lim, C. Tam, B.L. Dickens, Interventions to mitigate early spread of SARS-CoV-2 in Singapore: a modelling study, Lancet Infect. Dis. 20 (2020) 678–688. 13 https://doi.org/10.1016/S1473-3099(20)30162-6

2020 doi
[12]

https://coronavirus.jhu.edu/map.htm

C.-19 Map, Johns Hopkins Coronavirus Resource Center, (2020). https://coronavirus.jhu.edu/map.htm

2020
[13]

M. Chen, M. Li, Y. Hao, Z. Liu, L. Hu, L. Wang, The introduction of population migration to SEIAR for COVID-19 epidemic modeling with an efficient intervention strategy, Inf. Fusion. 64 (2020) 252–258. https://doi.org/10.1016/j.inffus.2020.08.002

2020 doi
[14]

Almquist, M

J. Almquist, M. Cvijovic, V. Hatzimanikatis, J. Nielsen, M. Jirstrand, Kinetic models in industrial biotechnology – Improving cell factory performance, Metab. Eng. 24 (2014) 38–60. https://doi.org/10.1016/j.ymben.2014.03.007

2014 doi
[15]

Nadella, A

P. Nadella, A. Swaminathan, S. V. Subramanian, Forecasting efforts from prior epidemics and COVID-19 predictions, Eur. J. Epidemiol. (2020) 727–729. https://doi.org/10.1007/s10654-020-00661-0

2020 doi
[16]

Basson, P.J

L.M. Basson, P.J. Kilbourn, J. Walters, Forecast accuracy in demand planning: A fast-moving consumer goods case study, J. Transp. Supply Chain Manag. 13 (2019) 1–9. https://doi.org/10.4102/jtscm.v13i0.427

2019 doi
[17]

Friedman, P

J. Friedman, P. Liu, E. Gakidou, * IHME COVID-19 Model Comparison Team, Predictive performance of international COVID-19 mortality forecasting models, MedRxiv. (2020). https://doi.org/https://doi.org/10.1101/2020.07.13.20151233

2020 doi
[18]

Gu, COVID-19 Projections Using Machine Learning, (2020)

Y. Gu, COVID-19 Projections Using Machine Learning, (2020). https://covid19-projections.com/

2020
[19]

https://covid- 19.bsvgateway.org/

Los Alamos Natinoal Laboratory COVID-19 Confirmed and Forecasted Case Data, (2020). https://covid- 19.bsvgateway.org/

2020
[20]

Miller, N.J

A.C. Miller, N.J. Foti, J.A. Lewnard, N.P. Jewell, C. Guestrin, E.B. Fox, Mobility trends provide a leading indicator of changes in SARS-CoV-2 transmission, MedRxiv. (2020) 2020.05.07.20094441. https://doi.org/10.1101/2020.05.07.20094441

2020 doi

Pith tools

Reviewed August 27, 2026 · model on record in the stance chip above.

[1] [1]

Hsiang, D

S. Hsiang, D. Allen, S. Annan-Phan, K. Bell, I. Bolliger, T. Chong, H. Druckenmiller, L.Y. Huang, A. Hultgren, E. Krasovich, P. Lau, J. Lee, E. Rolf, J. Tseng, T. Wu, The effect of large-scale anti-contagion policies on the COVID-19 pandemic., Nature. (2020). https://doi.org/10.1038/s41586-020-2404-8

work page doi:10.1038/s41586-020-2404-8 2020

[2] [2]

https://doi.org/https://www.worldometers.info/coronavirus/country/us/

Wolrdometers, Coronavirus Cases, (2020). https://doi.org/https://www.worldometers.info/coronavirus/country/us/

work page 2020

[3] [3]

Buckee, M.A

C.O. Buckee, M.A. Johansson, Individual model forecasts can be misleading, but together they are useful, Eur. J. Epidemiol. (2020) 731–732. https://doi.org/10.1007/s10654-020-00667-8

work page doi:10.1007/s10654-020-00667-8 2020

[4] [4]

A., Wilder-Smith, A

Y. Liu, A.A. Gayle, A. Wilder-Smith, J. Rocklöv, The reproductive number of COVID-19 is higher compared to SARS coronavirus, J. Travel Med. 27 (2020) 1–4. https://doi.org/10.1093/jtm/taaa021

work page doi:10.1093/jtm/taaa021 2020

[5] [5]

Willis, V.H.G

M.J. Willis, V.H.G. Díaz, O.A. Prado-Rubio, M. von Stosch, Insights into the dynamics and control of COVID- 19 infection rates, Chaos, Solitons & Fractals. 138 (2020) 109937. https://doi.org/10.1016/j.chaos.2020.109937

work page arXiv 2020

[6] [6]

Willis, A.R

M.J. Willis, A.R. Wright, V. Bramfitt, V.H. Grisales Díaz, COVID-19: Mechanistic model calibration subject to active and varying non-pharmaceutical interventions, MedRxiv. (n.d.). https://doi.org/https://doi.org/10.1101/2020.09.10.20191817

work page doi:10.1101/2020.09.10.20191817 2020

[7] [7]

https://covid19.apple.com/mobility

Apple, Apple Mobility Trends Reports, (2020). https://covid19.apple.com/mobility

work page 2020

[8] [8]

Flaxman, S

S. Flaxman, S. Mishra, A. Gandy, H.J.T. Unwin, T.A. Mellan, H. Coupland, C. Whittaker, H. Zhu, T. Berah, J.W. Eaton, M. Monod, P.N. Perez-Guzman, N. Schmit, L. Cilloni, K.E.C. Ainslie, M. Baguelin, A. Boonyasiri, O. Boyd, L. Cattarino, L. V. Cooper, Z. Cucunubá, G. Cuomo-Dannenburg, A. Dighe, B. Djaafara, I. Dorigatti, S.L. van Elsland, R.G. FitzJohn, K.A...

2020