Department of Industrial Engineering and Management, JSS Academy of Technical Education, Bengaluru, India
© 2021 Korea Disease Control and Prevention Agency
This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).
Country | Cases reported | Death | Recovered case |
---|---|---|---|
United States | 35,689,184 | 629,072 | 29,652,042 |
India | 31,619,573 | 423,965 | 30,781,263 |
Brazil | 19,880,273 | 555,512 | 18,595,380 |
Russia | 6,265,873 | 158,563 | 5,608,619 |
France | 6,103,548 | 111,824 | 5,696,559 |
United Kingdom | 5,856,528 | 129,654 | 4,508,650 |
Turkey | 5,704,713 | 51,253 | 5,449,253 |
Argentina | 4,919,408 | 105,586 | 4,557,037 |
Colombia | 4,776,291 | 120,432 | 4,567,701 |
Spain | 4,447,044 | 81,486 | 3,711,200 |
Criteria | Inclusion | Exclusion |
---|---|---|
Document type | Published documents | Under review, unpublished or upcoming documents |
Domain | Prediction models of COVID-19 | Other than prediction models of COVID-19 |
Language | English | Other than English |
Database | Initial | Screened | Accepted |
---|---|---|---|
Google Scholar | 910 | 33 | 21 |
Scopus | 210 | 19 | 4 |
Web of Science | 76 | 10 | 5 |
Total | 1,196 | 62 | 30 |
Duplicates | 47 | 0 | 0 |
Total selected | 1,149 | 62 | 30 |
No. | Study | Objective | Type of model | Result | Quality assessment |
---|---|---|---|---|---|
1 | Yang et al., 2020 [28] | To forecast COVID-19 patterns in China using a SEIR and AI model | SEIR model and AI model | · The model was effective in forecasting COVID-19 cases. | 95% CI |
2 | Liang et al., 2020 [29] | To forecast the risk of critical illness at hospital admission and identify survival of COVID-19 patients | Statistical software: LASSO, logistic regression model | · The score gives an estimation of the probability of critical disease progression for a hospitalized patient with COVID-19. | AUC (accuracy) was 0.88, 95% CI. |
3 | Yan et al., 2020 [30] | Relieving clinical burden and potentially reducing the mortality rate of COVID-19 | Machine learning tool: XGBoost | To predict patients with higher risk and potentially reduce mortality rate | Overall accuracy was 0.90 |
· Survival prediction accuracy was 100%. | |||||
· Mortality forecast accuracy was 81%. | |||||
4 | Gong et al., 2020 [31] | To predict the early detection of cases at high risk for progression to serious COVID-19 | Statistical analysis | · Results helped in COVID-19 patient identification for effective management. | Training cohort: |
· AUC was 0.912, 95% CI. | |||||
Validation cohort: | |||||
· AUC was 0.853, 95% CI. | |||||
5 | Chatterjee et al., 2020 [32] | To develop a stochastic mathematical model to predict COVID-19 cases | SEIR | · To help in healthcare preparedness and in allocations of resources. | R0 was 2.28, growth rate of the epidemic in India was 1.15. |
· The model suggested that herd immunity may be achieved when 55% to 65% of the population is infected. | |||||
6 | Hu et al., 2020 [12] | To predict confirmed COVID-19 cases and group cities into clusters according to transmission pattern | AI | · AI-based prediction showed significant accuracy and may act as a powerful tool for helping healthcare planning and policymaking. | Average errors: |
• 6-Step (1.64%) | |||||
• 7-Step (2.27%) | |||||
• 8-Step (2.14%) | |||||
• 9-Step (2.08%) | |||||
• 10-Step (0.73%) | |||||
7 | Tomar & Gupta, 2020 [33] | To predict new COVID-19 cases using LSTM based techniques | LSTM | · Prediction corresponded to the original information with a reasonable CI. | ±5% CI |
8 | IHME COVID-19 Health Service Utilization Forecasting Team & Murray, 2020 [34] | To predict deaths and requirements of total beds for hospitals due to COVID-19 | Statistical model | · The model estimated that the number of COVID-19 deaths would range from 81,114 to 162,106 over the next 4 mo. | Not available. |
9 | Chimmula & Zhang, 2020 [35] | To track COVID-19 cases and to help government and policymakers prepare | LSTM, R0 method | · ARIMA | RMSE (45.70) |
10 | Pandey et al., 2020 [36] | To create a predictive model to assess the need for clinical treatment for patients | Machine learning models: SEIR, regression model | · Predictions will help check supply and medical assistance and help policymakers prepare. | RMSLE: |
· SEIR model was 1.52. | |||||
· regression model was 1.75. | |||||
R0 between the 2 models was 2.02. | |||||
11 | Jehi et al., 2020 [37] | To develop a model for risk prediction for patients testing COVID-19 positive | Statistical prediction model: chi-square test | · Predictions could help direct healthcare preparedness. | C-statistic: |
· Development cohort was 0.863. | |||||
· Validation cohort was 0.840. | |||||
12 | Ardabili et al., 2020 [38] | To forecast the outbreak of COVID-19 using machine learning soft computing | Machine learning: logistic model. | Correlation coefficient | RMSE |
· Italy (0.997) | · Italy (3358.1) | ||||
· China (0.994) | · China (2524.44) | ||||
· Iran (0.997) | · Iran (628.62) | ||||
· USA (0.999) | · USA (350.33) | ||||
· Germany (0.997) | · Germany (555.32) | ||||
13 | Sujath et al., 2020 [39] | To forecast COVID-19 pandemic using machine learning | Machine learning: LR, MLP | · 95% CI with LR and MLP | 95% CI |
14 | Qi et al., 2020 [40] | To predict the hospital stay of COVID-19 patients | Machine learning: logistic regression, RF | · Predictions exhibited feasibility and accuracy for hospital stay for patients with pneumonia associated with COVID-19 infection. | LR model: |
· Sensitivity was 1.0. | |||||
· Specificity was 0.89. | |||||
RF model | |||||
· Sensitivity was 0.75. | |||||
· Specificity was 1.0. | |||||
15 | Ghosal et al., 2020 [41] | To forecast the number of deaths due to COVID-19 in India | Multiple regression and LR, auto-regression technique | · The estimated mortality rate (n) at the end of the 5th and 6th weeks was 211 and 467. | Multiple R was 0.9903. |
R squared was 0.9807. | |||||
Adjusted R squared was 0.9700. | |||||
Standard error was 234.1358. | |||||
16 | Hoertel et al., 2020 [42] | To develop a prediction model to identify patients needing professional care | Statistical analysis: Kaplan-Meier method, R Foundation for statistical computing | · Cox model predicted with a high accuracy (p<0.05). | · AUC was 0.97. |
· Overall C-statistic was 0.963 (95% CI, 0.936-0.99). | |||||
17 | Arora et al., 2020 [43] | To forecast the number of COVID-19 positive cases in 32 states and union territories of India using deep learning-based models | Deep learning: LSTM, RNN | · Model was highly accurate for short-term predictions (1–3 days) ahead. | · MAPE range <3% · Weekly forecast 4%–8% |
18 | Salgotra et al., 2020 [44] | To forecast COVID-19 outbreaks in India and use time series study and model on CC and DC in 3 states of India, Maharashtra, Gujarat, and Delhi | GEP model | · The model was highly effective in forecasting both reported cases and deaths around India. | · Lowest R value: 0.9881, DC in Delhi, |
· highest value was 0.9999, RC in India | |||||
19 | Dutta and Bandyopadhyay, 2020 [45] | To validate the predicted outcome of COVID-19 cases using machine learning | LSTM, GRU | Accuracy level | RMSE |
· Confirmed cases: 87% | · Confirmed cases: 30.15% | ||||
· Negative cases: 67.8% | · Negative cases: 49.4% | ||||
· Deceased cases: 62% | · Deceased cases: 4.16% | ||||
· Released cases: 40.5% | · Released cases: 13.72% | ||||
20 | Zhao et al., 2020 [46] | To develop risk ratings based on clinical categories and to forecast COVID-19 ICU admission and mortality | Logistic regression: multivariable regression model | · Predictions will significantly assist the flow of COVID-19 patients and distribute resources accordingly. | · ICU admission: AUC was 0.74, 95% CI. |
· Predicting mortality: AUC was 0.82, 95% CI. | |||||
21 | Hernandez-Matamoros et al., 2020 [47] | To predict COVID-19 behaviors in order to make future plans and hence to forecast the progress of the virus | ARIMA | · The model was able to predict the behavior of spread of COVID-19 infection. | RMSE average of 144.81. |
22 | Alazab et al., 2020 [48] | To predict COVID-19 cases across the world using an AI-based technique | PA, ARIMA, LSTM | · PA delivered the best performance. | Accuracy: |
· The model predicted COVID-19 cases and achieved an F-measure of 99%. | · Australia was 94.80%. | ||||
· Jordan was 88.43%. | |||||
23 | Parbat and Chakraborty, 2020 [49] | To predict the total number of deaths, recovered cases, cumulative number of confirmed cases, and number of daily cases | Vector regression model | The model: | RMSE: |
· Functioned well in fitting the total cases | · Total deaths: 0.092142 | ||||
· Poor fit for the daily number of cases | · Total recovered: 0.174036 | ||||
· Daily confirmed: 0.330830 | |||||
· Daily deaths: 0.361727 | |||||
24 | Zhao et al., 2020 [50] | To predict COVID-19 confirmed cases using 6 rolling grey Verhulst models | Rolling Grey Verhulst model | · Predictions exhibited good accuracy. | MAPE: training stage |
· Six models predicted S-shaped change characteristics consistently. | · Max (4.74%) | ||||
· Min (1.80%) | |||||
Testing stage | |||||
· Max (4.72%) | |||||
· Min (1.65%) | |||||
25 | Achterberg et al., 2020 [51] | To evaluate a diverse range of forecast algorithms for COVID-19 | Network-based forecasting | · The algorithm performed well in predicting COVID-19 cases and was superior to any other prediction algorithm. | NIPA |
· Hubei was 0.122. | |||||
· The Netherlands was 0.038. | |||||
26 | Fernandez et al., 2021 [52] | To develop a forecasting algorithm to consider patient survival | Logistic regression: multivariate logistic regression | · Patients that would be able to survive were classified by age, CRP, platelet count, and number of lung consolidations. | AUC was 0.8129. |
GOF: Hosmer and Lemeshow test, p=0.018; 95% CI (0.773–0.853, p<0.001) | |||||
27 | Li et al., 2020 [53] | To develop a prediction model for identifying patients at an increased risk of COVID-19 death | Machine learning: autoencoder model, logistic regression, SVM, RF | · The model exhibited specificity and accuracy above 0.9. | Logistic regression, SVM, RF |
· Sensitivities below 0.4. | |||||
· Autoencoder scores above a sensitivity value of 0.4. | |||||
28 | Siwiak et al., 2020 [54] | To develop a global model for COVID-19 in terms of the number of infected cases | GLEAM | · Presented a percentage difference over time between the number of reported, confirmed cases and CI limits for different modeled predictions. | 95% CI |
29 | Bhandari et al., 2020 [55] | To predict the progression of COVID-19 in India using ARIMA | ARIMA | · The COVID-19 forecast helps the government and policy makers to optimize resources and make decisions. | 95% CI |
30 | Muhammad et al., 2021 [56] | To forecast COVID-19 infection using machine learning | Machine learning: logistic regression, decision tree, support vector machine, naive Bayes, and artificial neutral network | · Decision tree model accuracy was 94.99%. | RMSE: LMST (27.187) |
· Support vector machine model sensitivity was 93.34%. | LR (7.562) | ||||
· Naive Bayes model has a specificity of 94.30%. |
No. | Study | Year | Country | Citation (January 2, 2021) | Model |
---|---|---|---|---|---|
1 | Yang et al. [28] | 2020 | China | 467 | SEIR and AI model |
2 | Liang et al. [29] | 2020 | China | 327 | Statistical software |
3 | Yan et al. [30] | 2020 | China | 194 | Machine learning |
4 | Gong et al. [31] | 2020 | China | 134 | Statistical analysis |
5 | Chatterjee et al. [32] | 2020 | India | 131 | SEIR |
6 | Hu et al. [12] | 2020 | China | 130 | Artificial intelligence |
7 | Tomar & Gupta [33] | 2020 | India | 129 | LSTM |
8 | IHME COVID-19 Health Service Utilization Forecasting Team & Murray [34] | 2020 | USA | 119 | Statistical model |
9 | Chimmula & Zhang [35] | 2020 | Canada | 99 | LSTM |
10 | Pandey et al. [36] | 2020 | India | 57 | Machine learning |
11 | Jehi et al. [37] | 2020 | USA | 45 | Statistical analysis |
12 | Ardabili et al. [38] | 2020 | Worldwide scenario | 41 | Machine learning |
13 | Sujath et al. [39] | 2020 | India | 41 | Machine learning |
14 | Qi et al. [40] | 2020 | Worldwide scenario | 41 | Machine learning |
15 | Ghosal et al. [41] | 2020 | India | 39 | Regression model |
16 | Hoertel et al. [42] | 2020 | France | 37 | Statistical analysis |
17 | Arora et al. [43] | 2020 | India | 34 | LSTM, RNN |
18 | Salgotra et al. [44] | 2020 | India | 34 | GEP model |
19 | Dutta & Bandyopadhyay [45] | 2020 | India | 33 | LSTM, GRU |
20 | Zhao et al. [46] | 2020 | China | 13 | Logistic regression |
21 | Hernandez-Matamoros et al. [47] | 2020 | Chile | 11 | ARIMA |
22 | Alazab et al. [48] | 2020 | Jordon | 9 | PA, ARIMA, LSTM |
23 | Parbat & Chakraborty [49] | 2020 | India | 9 | Regression model |
24 | Zhao et al. [50] | 2020 | China | 6 | Grey Verhulst |
25 | Achterberg et al. [51] | 2020 | China | 2 | Network-based forecasting |
26 | Fernandez et al. [52] | 2021 | UK | 2 | AI |
27 | Li et al. [53] | 2020 | Worldwide scenario | 1 | GLEM |
28 | Siwiak et al. [54] | 2020 | India | 1 | ARIMA |
29 | Bhandari et al. [55] | 2020 | UK | - | Logistic regression |
30 | Muhammad et al. [56] | 2021 | Mexico | - | Machine learning |
This data is as of July 29, 2021 from Worldmeter (
COVID-19, coronavirus disease 2019.
COVID-19, coronavirus disease 2019; SEIR, susceptible-exposed-infectious-removed; AI, artificial intelligence; CI, confidence interval; LASSO, least absolute shrinkage and selection operator; AUC, area under the curve; XGBoost, eXtreme gradient boosting; LSTM, long short-term memory; ARIMA, autoregressive integrated moving average; RMSE, root mean square error; RMSLE, root mean square logarithmic error; LR, linear regression; MLP, multilayer perceptron; RF, random forest; RNN, recurrent neural network; MAPE, mean absolute percentage error; CC, confirmed case; DC, death case; GEP, genetic evolutionary programming; RC, reported case; GRU, gated recurrent unit; ICU, intensive care unit; PA, prophet algorithm; NIPA, network inference-based prediction algorithm; CRP, C-reactive protein; GOF, goodness of fit; SVM, support vector machine; GLEAM, global epidemic and mobility framework.
SEIR, susceptible-exposed-infectious-removed; AI, artificial intelligence; LSTM, long short-term memory; RNN, recurrent neural network; GEP, genetic evolutionary programming; GRU, gated recurrent unit; ARIMA, autoregressive integrated moving average; PA, prophet algorithm; GLEM, global epidemic and mobility.