Early Forecasts of the COVID-19 Pandemic: History
View Latest Version
Please note this is an old version of this entry, which may differ significantly from the current revision.
Contributor:
Rahul Pathak
,
Daniel W Williams

This entry reviews early forecasting of the COVID pandemic in the context of forecast accuracy and epidemic and pandemic forecasting. 

  • COVID 19
  • Forecasting accuracy
  • Pandemic forecasts
  • SEIR models
  • Time series forecasting
  • Short term accuracy
  • Long term accuracy

The COVID-19 Pandemic has been a unique experience for many people in the current era. While there have been epidemics and pandemics, they have been regionally isolated, relatively mild, or primarily associated with particular demographic groups. The AIDS pandemic and previous SARs pandemics, while world-wide, affect a limited population. The Ebola outbreaks, while severe, have affected a limited geographical area. Living adults in much of the world may remember the poliomyelitis epidemics of the mid-20th century, but even it is comparatively less severe when considering the combination of health severity and number of people affected. The most likely comparative pandemic is the 1918 influenza pandemic. COVID-19 was first identified in late 2019 and became recognized as a worldwide pandemic in early 2020. By March 2020 it was a source of concern for the entire planet. It also became the frequent subject of forecasts, some of which soon became questioned.[1][2][3]

There has been extensive research on forecast accuracy across many domains – although heavily weighted toward finance and related matters.[4][5][6][7][8][9][10][11][12][13][14][15][16][17][18][19] These studies often focus on technical issues such as the precise error measure to consider or the comparative accuracy of various approaches. One well-known set of studies, the M-competitions, has compared the accuracy of numerous time series forecast methods.[20][21][22][23] A well-established practice is to compare multiple forecasts and examining variation by forecaster characteristics.[24][25][26][27] In recent years, there has been evaluation of short-term disease forecasting, such as influenza and dengue.[28][29][30]

The COVID pandemic led to a variety of forecasting challenges. First, there was a poor understanding of factors such as the transmission rate, the virus's method of spread, and the amount of time the virus survives in different environments. Second, any significant precedence on social distancing outcomes and the general population’s compliance with shutdowns was not available. The forecasting models now have relatively much more information and robust assumptions than they had access to in the initial months of the coronavirus outbreak. Table 1 summarizes the major forecasts for the United States that were being aggregated as part of the early ensemble forecasting efforts of the Center for Disease Control and Prevention (CDC). Some of these forecasts were epidemiological, and some were purely statistical forecasts, based on exponential smoothening and time series of observed trends.

The epidemiological projections relied on a variety of methods, including using simulations based on the behavior of earlier strains of the coronavirus family, like SARS-CoV-1 and MERS, using the transmission data from China, Italy, and Spain, and increasingly the information gained from the U.S. transmission behavior and social distancing compliance. Kissler, Tedijanto, Goldstein, Grad, and Lipsitch[31] used a medical model of transmission behavior, based on the immunity, cross-immunity, and seasonality for Hcov-OC43 and HcoV-HKU1, to suggest that a seasonal resurgence is the most likely scenario, requiring intermittent social distancing through 2022 and resurgences as late as 2024. Another study used early data from Hubei province in China. between 11 January 2020 and 10 February 2020 and predicted that the cumulative case count in China by 29 February could reach 180,000 with a lower bound of 45,000 cases.[32] The reported cases were less than the forecasted possibly because of underreporting, effective non-pharmaceutical interventions, or flawed model assumptions. The epidemiological models rely on various scientific assumptions related to the behavior of viruses, underlying health conditions and immunity, availability of health infrastructure, etc., which may pose a significant challenge for forecasting in the initial days of a new virus, such as COVID-19. However, as data becomes available, employing time series or purely statistical forecasting models could also be effective. For example, Petropoulos and Makridakis[33] used exponential smoothening with multiplicative error and multiplicative trend components to forecast the trajectory of COVID-19 outcomes. The forecasts released and their subsequent follow-up on social media, typically showed over-forecasting in cases and deaths, but the actual values were within the 50% prediction level, except in the first round.[34]  Castle, Doornik, and Hendry[35][36] used local averaged time trend estimation that assumed no seasonality, and they argue that, in the short run, their forecasts outperformed the epidemiological forecasts, such as the ones from Imperial College. A probabilistic model from the Los Alamos National Laboratory in New Mexico compared the forecasts to the actuals and reported fairly robust coverage of the forecasts for three-week periods following their releases.[37]

Though the statistical forecasts have immense appeal. since they can be created in real-time and can forecast micro-level patterns, for example, state or county-level trends, they have also come under criticism. Jewell, Lewnard, and Jewell[38] raised several concerns that warrant a careful approach toward statistical models in the case of epidemics. They highlighted that epidemic curves may not follow a normal distribution, and curves may fit early data in various ways, which may change as the epidemic progresses, for example, a second wave may occur and change things. They suggested that such models can be helpful for short-term predictions, but, otherwise, extreme caution has to be exercised, a point that is reaffirmed by Holmdahl and Buckee.[1] They argued that, for long-term outcomes, only mechanistic models, like SEIR (Susceptible–Exposed–Infectious–Recovered) models, are reliable—many of the forecasts listed in Table 1 used SEIR models.

Some recent studies have examined the effectiveness of different models and also compared the individual models to ensemble forecasts. (2022) evaluated the individual and ensemble forecasts and found that ensemble forecasts outperformed any individual forecasts, using the data from more than 90 different forecasting agencies at https://covid19forecasthub.org/. Marchant et al.[2] evaluated the forecast accuracy of IHME data and found that IHME data underestimated mortality, and the results did not improve over time. Perone[39] examined the efficacy of hybrid models against a range of technical forecasting models, using Italian Ministry of Health data, in early 2020, and found that hybrid models were better at capturing the pandemic’s linear, non-linear and seasonal patterns and significantly outperformed single time series models. Wang et al.[40] used the data from the second wave of the pandemic in India and the United States. They found that the ARIMA model had the best fit for India and the ARIMA-GRNN model had the best fit for the United States. Bracher et al.[41] undertook an evaluation of thirteen forecasts for Germany and Poland during the ten weeks of the second wave and found considerable heterogeneity in both point estimates and the forecasts concerning spread. Pathak and Williams[42] examined forecast errors for two SEIR models and two time series models and found suggestive evidence that the SEIR models performed better over the long run, but the time series models performed better over shorter horizons. Ioannidis, Cripps, and Tanner[43] argued that the COVID-19 pandemic highlighted the weakness of epidemic forecasting, and that when forecasts and forecast errors could determine the strength of policy measures, such as the implementation of lockdowns, they should receive closer scrutiny. This study adds to this growing literature and undertakes a comparative evaluation of two sets of statistical and epidemiological models, using trend-based comparison, and a set of forecast accuracy measures discussed in the next section. The findings have relevance for the practice of ensemble forecasting and the study of short-term forecasting of infectious diseases.

Table 1. Selected COVID-19 Forecasts, Methodology, Assumptions.

No.

Models

Method

Assumptions

Webpage

1.

Institute of Health Metrics and Evaluation (IHME)

Combination of Mechanistic transmission model and curve-fitting approach

Adjusted to differences in mobility.

https://covid19.healthdata.org/united-states-of-america

2.

Columbia University

Metapopulation SEIR model

Accounts for social distancing.

https://columbia.maps.arcgis.com/apps/webappviewer/index.html?id=ade6ba85450c4325a12a5b9c09ba796c

3.

Auquan Data Science

SEIR Model

No assumption about interventions.

https://covid19-infection-model.auquan.com/

4.

COVID-19 Simulator Consortium

SEIR Model

20% increase in contact rates after lifting statistics at home orders.

https://www.covid19sim.org/team

5.

Georgia Institute of Technology

Deep Learning

Assumes effects of interventions embedded in the data.

https://www.cc.gatech.edu/~badityap/covid.html

6.

Imperial College, London

Ensembles of mechanistic transmission models

No specific assumptions about the interventions.

https://mrc-ide.github.io/covid19-short-term-forecasts/index.html

7.

John Hopkins University

Stochastic Metapopulation SEIR model

Assumes reduction in effectiveness of mitigation after lifting shelter-in-place.

https://github.com/HopkinsIDD/COVIDScenarioPipeline

8.

Los Alamos National Laboratory

Statistical dynamic growth model accounting for population susceptibility

Assumes the NPIs would continue.

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

9.

Massachusetts Institute of Technology

SEIR Model

Assumes continuation of present interventions.

https://www.covidanalytics.io/projections

10.

Northeastern University

Metapopulation, age structured SLIR model

Assumes continuation of social distancing policies.

https://covid19.gleamproject.org/

11

Iowa State University

Nonparametric spatiotemporal model

No specific assumptions related to interventions.

http://www.covid19dashboard.us/

12

Predictive Science Inc.

Stochastic SEIRX model

Assumes that current interventions would not change.

https://github.com/predsci/DRAFT

13

US Army Engineer Research and Development Center

SEIR mechanistic model

Projections assume that interventions would not change.

https://www.cdc.gov/coronavirus/2019-ncov/covid-data/forecasting-us.html

14

University of California, Los Angeles

Modified SEIR Model

Projections assume that interventions would not change.

https://covid19.uclaml.org/

15

University of Texas, Austin

Nonlinear Bayesian hierarchical regression with a negative-binomial model

Estimate the extent of social distancing, using mobile phone geolocation data. Does not assume changes in social distancing during the forecast period.

https://covid-19.tacc.utexas.edu/projections/

 

References

  1. Holmdahl, I., & Buckee, C. (2020). Wrong but useful—what covid-19 epidemiologic models can and cannot tell us. New England Journal of Medicine, 383(4), 303-305.
  2. Marchant, R., Samia, N. I., Rosen, O., Tanner, M. A., & Cripps, S. (2020). Learning as we go: An examination of the statistical accuracy of COVID19 daily death count predictions. arXiv preprint arXiv:2004.04734. Retrieved from https://arxiv.org/abs/2004.04734 accessed on 11/18/2022.
  3. Piper, K. (2020). This coronavirus model keeps being wrong. Why are we still listening to it? Vox, Online. Retrieved from https://www.vox.com/future-perfect/2020/5/2/21241261/coronavirus-modeling-us-deaths-ihme-pandemic accessed on 11/17/2022.
  4. Ashton, A. H., & Ashton, R. H. (1985). Aggregating Subjective Forecasts: Some Empirical Results. Management Science, 31(12), 1499-1508.
  5. Cetin, B., & Yavuz, I. (2021). Comparison of forecast accuracy of Ata and exponential smoothing. Journal of Applied Statistics, 48(13-15), 2580-2590.
  6. Chung, I. H., Williams, D. W., & Do, M. R. (2022). For Better or Worse? Revenue Forecasting with Machine Learning Approaches. Public Performance & Management Review, 1-21. doi:10.1080/15309576.2022.2073551.
  7. Clemen, R. T. (1989). Combining forecasts: A review and annotated bibliography. International Journal of Forecasting, 5(4), 559-583.
  8. Fullerton Jr, T. M., & Molina Jr, A. L. (2010). Municipal water consumption forecast accuracy. Water Resources Research, 46(6).
  9. Hall, J. L., & Tacon, P. B. (2010). Forecast accuracy and stock recommendations. Journal of Contemporary Accounting & Economics, 6(1), 18-33.
  10. Hoover, J. (2009). How to track forecast accuracy to guide forecast process improvement. Foresight, 14, 17-23.
  11. Hyndman, R. J. (2006). Another look at forecast-accuracy metrics for intermittent demand. Foresight: The International Journal of Applied Forecasting, 4(4), 43-46.
  12. Hyndman, R. J., & Koehler, A. B. (2006). Another look at measures of forecast accuracy. International Journal of Forecasting, 22(4), 679-688.
  13. Kim, S., & Kim, H. (2016). A new metric of absolute percentage error for intermittent demand forecasts. International Journal of Forecasting, 32(3), 669-679.
  14. MacManus, S. A. (1992). Forecasting frustrations: factors limiting accuracy (Budget forecasting). Government Finance Review, 8(n), 7.
  15. Ramesh Babu, N., & Arulmozhivarman, P. (2013). Improving forecast accuracy of wind speed using wavelet transform and neural networks. Journal of Electrical Engineering and Technology, 8(3), 559-564.
  16. Syntetos, A. A., Nikolopoulos, K., & Boylan, J. E. (2010). Judging the judges through accuracy-implication metrics: The case of inventory forecasting. International Journal of Forecasting, 26(1), 134-143.
  17. Vallance, L., Charbonnier, B., Paul, N., Dubost, S., & Blanc, P. (2017). Towards a standardized procedure to assess solar forecast accuracy: A new ramp and time alignment metric. Solar Energy, 150, 408-422.
  18. Williams, D. W., & Kavanagh, S. C. (2016). Local Government Revenue Forecast Competition/Comparison. Journal of Public Budgeting, Accounting, and Financial Management, 28(4), 488-526.
  19. Zarnowitz, V. (1978). On the Accuracy and Properties of Recent Macroeconomic Forecasts. The American Economic Review, 68(2), 313-319.
  20. Makridakis, S., Andersen, A., Carbone, R., Fildes, R., Hibon, M., Lewandowski, R., & Winkler, R. L. (1984). The forecasting accuracy of major times series methods (Vol. null).
  21. Makridakis, S., Chatfield, C., Hibon, M., Lawrence, M., Mills, T., Ord, K., & Simmons, L. F. (1993). The M2-competition: A real-time judgmentally based forecasting study. International Journal of Forecasting, 9(1), 5-22. Doi: 10.1016/0169-2070(93)90044-N
  22. Makridakis, S., & Hibon, M. (2000). The M3-competition: Results, conclusions and implications. International Journal of Forecasting, 16(4), 451-476.
  23. Makridakis, S., Spiliotis, E., & Assimakopoulos, V. (2018). The M4 Competition: Results, findings, conclusion and way forward. International Journal of Forecasting, 34(4), 802-808.
  24. Clements, M. P. (2022). Forecaster Efficiency, Accuracy, and Disagreement: Evidence Using Individual‐Level Survey Data. Journal of Money, Credit and Banking, 54(2-3), 537-568.
  25. Garcia, J., & Iskrev, N. (2019). Inflation expectations in the Survey of Professional Forecasters: An exploratory analysis. Economic Bulletin and Financial Stability Report Articles and Banco de Portugal Economic Studies.
  26. Lim, T. (2001). Rationality and analysts' forecast bias. The Journal of Finance, 56(1), 369-385.
  27. Williams, D. W. (2012). The politics of forecast bias: forecaster effect and other effects in New York City revenue forecasting. Public Budgeting & Finance, 32(4), 1-18. doi:10.1111/j.1540-5850.2012.01021.x
  28. Reich, N. G., McGowan, C. J., Yamana, T. K., Tushar, A., Ray, E. L., Osthus, D., . . . Gibson, G. C. (2019). Accuracy of real-time multi-model
  29. Viboud, C., & Vespignani, A. (2019). The future of influenza forecasts. Proceedings of the National Academy of Sciences, 116(8), 2802-2804.
  30. Yamana, T. K., Kandula, S., & Shaman, J. (2016). Superensemble forecasts of dengue outbreaks. Journal of The Royal Society Interface, 13(123), 20160410.
  31. Kissler, S. M., Tedijanto, C., Goldstein, E., Grad, Y. H., & Lipsitch, M. (2020). Projecting the transmission dynamics of SARS-CoV-2 through the postpandemic period. Science, 368(6493), 860-868.
  32. Anastassopoulou, C., Russo, L., Tsakris, A., & Siettos, C. (2020). Data-based analysis, modelling and forecasting of the COVID-19 outbreak. PLoS ONE, 15(3), e0230405.
  33. Petropoulos, F., & Makridakis, S. (2020). Forecasting the novel coronavirus COVID-19. PLoS ONE, 15(3), e0231236.
  34. Petropoulos, F., Makridakis, S., & Stylianou, N. (2020). COVID-19: Forecasting confirmed cases and deaths with a simple time series model. International Journal of Forecasting.
  35. Castle, J. L., Doornik, J. A., & Hendry, D. F. (2020a). Short-term forecasting of the Coronvirus Pandemic. Retrieved from https://forecasters.org/blog/2020/04/30/short-term-forecasting-of-the-coronavirus-pandemic/ accesed on 11/17/2022.
  36. Castle, J. L., Doornik, J. A., & Hendry, D. F. (2020b). Short-term forecasting of the Coronvirus Pandemic, 2020-W06. Nuffield College Economics Discussion Papers. Retrieved from https://www.nuffield.ox.ac.uk/economics/Papers/2020/2020W06_COVID-19_shortterm_forecasts.pdf accesed on 11/17/2022.
  37. Los Alamos National Laboratory. (2020). COVID-19 Confirmed and Forecasted Case Data. Retried from https://covid-19.bsvgateway.org/ on 11/18/2022/
  38. Jewell, N. P., Lewnard, J. A., & Jewell, B. L. (2020). Caution warranted: using the Institute for Health Metrics and Evaluation model for predicting the course of the COVID-19 pandemic. Annals of Internal Medicine, 173(3), 226-227.
  39. Perone, G. (2021). Comparison of ARIMA, ETS, NNAR, TBATS and hybrid models to forecast the second wave of COVID-19 hospitalizations in Italy. The European Journal of Health Economics, 1-24.
  40. Wang, G., Wu, T., Wei, W., Jiang, J., An, S., Liang, B., . . . Liang, H. (2021). Comparison of ARIMA, ES, GRNN and ARIMA–GRNN hybrid models to forecast the second wave of COVID-19 in India and the United States. Epidemiology & Infection, 149.
  41. Bracher, J., Wolffram, D., Deuschel, J., Görgen, K., Ketterer, J. L., Ullrich, A., . . . Bhatia, S. (2021). A pre-registered short-term forecasting study of COVID-19 in Germany and Poland during the second wave. Nature communications, 12(1), 1-16.
  42. Pathak, R., & Williams, D. W. (2022). Evaluating the Comparative Accuracy of COVID-19 Forecasts: Exploratory Analysis of the First-Wave Mortality Forecasts in the United States. Forecasting, 4, 798-818.
  43. Ioannidis, J. P., Cripps, S., & Tanner, M. A. (2022). Forecasting for COVID-19 has failed. International Journal of Forecasting, 38(2), 423-438.
More
©Text is available under the terms and conditions of the Creative Commons-Attribution ShareAlike (CC BY-SA) license; additional terms may apply. By using this site, you agree to the Terms and Conditions and Privacy Policy.
This entry is offline, you can click here to edit this entry!
ScholarVision Creations
Feedback