key: cord-0953031-n0fcf9y9
authors: Tyagi, R.; Bramhankar, M.; Pandey, M.; M, K.
title: COVID 19: Real-time Forecasts of Confirmed Cases, Active Cases, and Health Infrastructure Requirements for India and its Majorly Affected States using the ARIMA model.
date: 2020-05-22
journal: nan
DOI: 10.1101/2020.05.17.20104588
sha: 9f70d160f6e37f346b7a58aa1d2e454178688f54
doc_id: 953031
cord_uid: n0fcf9y9

Background: COVID-19 is an emerging infectious disease which has been declared a Pandemic by the World Health Organization (WHO) on March 11 2020. This pandemic has spread over the world in more than 200 countries. India is also adversely affected by this pandemic, and there are no signs of slowing down of the virus in coming time. The absence of a vaccine for COVID-19 is making the situation worse for the already overstretched Indian public health care system. Objective: This study is forecasting the confirmed and active cases for COVID-19 until June, using time series ARIMA model. In further analysis, based on predicted active cases, we estimated the requirement of isolation beds, ICU beds and ventilators for COVID-19 patients until June. Methods: We used ARIMA model, and Auto ARIMA model for forecasting confirmed and active cases till the end of June month using time series data of COVID-19 cases in India from March 14, 2020, to May 3 2020. We estimated requirement of ICU beds as 10%, Ventilators as 5% and isolation beds as 85% of the active cases predicted from our calculations. Results: We expect that India will be having 441896 confirmed cases (95% CI: 210240, 673552), 124712 active cases (95% CI: 68481, 180944) at the end of June based on our forecasts. Maharashtra, Punjab, Gujarat and Delhi (UT) will be the most affected states, having the highest number of active and confirmed cases at the end of June while Kerala is expected to have less than 1000 confirmed cases and no active cases at the end of June. We expect that India has to prepare 106006 isolation beds (95% CI: 58209, 153802), 12471 ICU beds (95% CI: 6848,18094) and 6236 ventilators (95% CI: 3424,9047) to accommodate the patients at the end of June. Discussion and Conclusion: Our forecasts show a very alarming situation for India in coming days and, the actual numbers can go higher than our estimates of confirmed cases as India is observing partial lockdown currently. In future, lockdown might be lifted, and in that case, there will be a surge in the number of daily confirmed and active cases. The requirement of isolation beds, ICUs and ventilators will also be increased in that scenario. Migrants returning to their homes due to loss of livelihood and income in the lockdown period may lead to a rise in the number of cases, which could not be accounted for in our projections. We suggest a Public-Private Partnership (PPP) model in the health sector to accommodate COVID-19 patients adequately and reduce the burden of the already overstretched Indian public health care system.

COVID-19 is an emerging infectious disease which has been declared a Pandemic by world health organization. The World Health Organization (WHO) has declared the outbreak of the novel Coronavirus (COVID-19) as a pandemic on March 11 2020. It is caused by severe acute respiratory syndrome Corona Virus 2 (SARS-CoV-2). This pandemic has spread over the world in more than 200 countries, and developed countries like Italy, China, Australia, the USA are the most affected by this virus. The origin of the virus is yet to be confirmed, but the first person tested positive is from Wuhan, China. It is spreading very quickly throughout the world, and the number of confirmed cases is about to reach 4 million, and the recovery rate is around 34% in the world. (Worldometer, May 7 2020). In India, the first positive case of COVID-19 was detected on January 30 2020, in Kerala. On May 7, the country has 35,902 active cases, with All India Institute of Medical Sciences (AIIMS) director Randeep Guleria stating that according to data released by experts and going by the current trend, the cases are likely to peak in June-July. Of the total 35,902 active cases, 4.8 percent of patients are in ICU, 1.1 per cent on ventilators, and 3.3 per cent are on oxygen support. Currently, on May 8, India had 821 dedicated COVID hospitals with 1,50,059 beds (1,32,219 isolation beds and 17,840 ICU beds) and 1,898 dedicated COVID Health Centres with 1,19,109 beds (1,09,286 isolation beds and 9,823 ICU beds) along with 7,569 quarantine centres (PIB-Updates on COVID 19, May 8) . However, if infections continue to rise at the same rate in May as they have done so far in April, India could be facing a deficit in isolation beds, intensive care unit (ICU) beds and ventilators in the coming time.

The infrastructure stress will be especially acute in eight high-burden states, led by Gujarat, Maharashtra, Delhi and Tamil Nadu as they have more than 4000 active cases currently. It is really crucial in this situation to be prepared for healthcare infrastructure when the infections will be at its highest and accordingly the requirement for isolation beds, intensive care unit (ICU) beds and ventilators will also be at its peak.

According to the report of World Population prospects (2019), India has a population of more than 1.36 billion and most of the population of urban areas and cities are under the risk of contracting the virus. So, it is important to forecast numbers of confirmed and active cases at the national and state level for India to see which states will be having the highest burden of the increasing confirmed and active cases of COVID-19 by the end of the June. Another objective of this study is to forecast the requirements of healthcare infrastructure in the coming months based on the numbers of active cases projected. This will be helpful to do proper future planning for the health facility in India. This will be helpful for the government to provide health facilities like isolation beds, ventilators and ICU units in time to time without any overburden on the health system. Tiwari et al. (2020) made their prediction for India based on the pattern of China using a machine learning approach. They predicted that the peak of the cases for India would be attained between the third and fourth weeks of April 2020 in India. This outbreak is predicted to be controlled around the end of May 2020. The total number of predicted confirmed cases of COVID-19 might reach around 68,978, and the number of deaths due to COVID-19 are predicted to be 1557 around April 25, 2020, in India. However, this prediction does not seem to fit in the current situation as we are observing around 4000 daily confirmed cases and (AIIMS) director Randeep Guleria also stated that according to data released by experts and going by the current trend, the cases are likely to peak in June-July. Another study by Singh et al. (2020) to identify the top 15 countries with the spatial mapping of the confirmed cases. A comparison was made between the identified top 15 countries for confirmed cases, deaths, and recoveries, and an (ARIMA) model was used for predicting the COVID-19 disease spread trajectories for the next two months. Their predicted values showed that the confirmed cases, deaths, and recoveries would double in all the observed countries except China, Switzerland, and Germany. Chakraborty (2020) also used a hybrid approach based on Autoregressive integrated moving average model (ARIMA) and Wavelet-based forecasting model that can generate short-term (ten days ahead) forecasts of the number of daily confirmed cases for Canada, France, India, South Korea, and the UK. Guan et al. (2020) in his study on data regarding 1099 patients with laboratory-confirmed COVID-19 from 552 hospitals in 30 provinces, autonomous regions, and municipalities in mainland China through January 29, 2020, found that 5.0% who were admitted to the ICU, 2.3% who underwent invasive mechanical ventilation, and 1.4% who died. Another study by Wu et al. (2020) from China suggests that 15-20% of COVID-19 cases require hospitalization, with around 15% of cases presenting with severe symptoms and 5% requiring intensive care. A study by Lazzerini et al. (2020) for Italy and Spain suggested that 40-55% of COVID-19 positive cases have been hospitalized, with 7-12% requiring admission to intensive care units. Based on the COVID-19 Cases in Italy shows that 10-25% of patients will require ventilation, and some patients will need ventilation for several weeks. Remuzzi et al., (2020) find that the percentage of patients in intensive care reported daily in Italy between March 1 and March 11, 2020, has consistently been between 9% and 11% of patients who are actively infected.

In the Indian scenario, a press release by PIB on May 8 stated that of the total 35,902 active cases, 4.8 per cent patients are in ICU, 1.1 per cent on ventilators and 3.3 per cent are on oxygen support (PIB Press release).

It is quite evident from the above studies that ICU admission rates for COVID 19 patients vary for countries depending on the severity of the disease in the patients of that country. For India, as of now, patients requiring ICU admission for COVID 19 is almost 5% which is lower than in other countries. However, this proportion might increase in the coming times as the severity of infection might increase as the infection will be at its peak in the month of June-July and therefore the percentage of COVID 19 patients requiring ICU facility and Ventilation support will be at its peak.

For our study, the required data of daily total confirmed cases and total active cases of COVID-19 infection collected for India as well as its selected states from the (https://www.covid19india.org/), and excel of the patient database is used to build a time-series database for confirmed and active cases. In this study, forecasting is done based on the data from March 14, 2020, to May 3 2020. This data is being used to build Forecast models for a, particularly short duration.

In the past few months, an increasing number of research related to forecasting the trend of pandemic COVID-19 cases are being published using different approaches in various part of the world. Benvenuto et al., 2020) . In this study, the well-known Autoregressive Integrated Moving Average (ARIMA) time-series model used for the further forecasting purpose. ARIMA model is one of the generalized forms of an autoregressive moving average (ARMA) model among the time series forecasting. We fit both models to understand the data better or to predict future points in the series (Forecasting) . ARIMA model depends or always represented with the help of some parameters, and the model has expressed as ARIMA (p, d, q): p, d and q are non-negative integers. The parameters have their usual meaning, here, p stands for the order of auto-regression, d represents the degree of trend difference (the number of times the data have had past values subtracted) for the stationary of the trend and q signifies the order of moving average. This model combines auto regression lags under the stationary trend and moving average and predict better future values based on past and recent data. For this model, the degree of parameters p, d and q determine based on the partial Auto-correlation function (PACF) graph, The Augmented Dickey-Fuller Test to test the stationary of the time series observations and Complete Auto-Correlation Function (ACF) graph respectively 

We have applied the ARIMA model and Auto ARIMA model using R, to our considered time series data of COVID-19 cases in India for the forecasting the total confirmed and active cases for India and its majorly affected states. We selected states based on the criterion that chosen states should have at least 100 confirmed cases till May 3 2020. By using this selection criterion, India and 17 other states selected which are Andhra Pradesh, Bihar, Delhi, Gujarat, Haryana, Jammu & Kashmir, Odisha, Karnataka, Kerala, Madhya Pradesh, Maharashtra, Punjab, Rajasthan, Tamil Nadu, Telangana, Uttar Pradesh and West Bengal. The cases are forecasted under the assumption that people will be maintaining condition similar to the full or partial lockdown situation. After fitting the model, the built model is used to forecast confirmed and active cases COVID-19 cases for the next 58 days, i.e. from May 3, 2020, to June 30 2020. The model for forecasting future confirmed and active cases of COVID-19 cases is represented as,

Here, Xt is the predicted number of confirmed and active COVID-19 cases at t th day, α1, α2, β1 and β2 are parameters whereas Zt is the residual term for t th day. The trend of forthcoming incidences can be estimated from the previous cases, and a time series analysis is performed for this purpose .

In our study, the forecasted cases are mainly used for preparing the government for the health infrastructure such as the number of isolation beds, ICU beds and ventilators etc.

In further analysis, based on predicted active cases, we estimated the requirement of isolation beds, ICU and ventilators for COVID-19 patients in the coming days. Based on the literature, the requirement of ICUs and ventilator support increases as the infection hit its peak, which India may get in the month of June-July said by the director of AIIMS. So, based on the previous experiences of China and Italy, estimates for health infrastructure requirements on projected active cases should be made at 10% requirement for ICUs, 5% for ventilation support and 85% active cases requires isolation beds.

The requirement of ICU, Ventilator support, and beds, based on active forecast cases Required number of beds = Forecast active cases at ith day*(85/100) Required number of ICU = Forecast active cases at ith day*(10/100)

Our health infrastructure requirement is estimated based on the active cases as our projections are made on the basis on data till May 3 when our country was observing the complete lockdown. However, India is observing partial lockdown currently and might remove lockdown in the future, so for being prudent, we will estimate health infrastructure requirements based on the estimates of the upper confidence interval of active cases.

. CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted May 22, 2020. . https://doi.org/10.1101/2020.05.17.20104588 doi: medRxiv preprint Results:

In Fig 1(a) shows the previous trend of confirmed cases in India. Based on that previous data, we have forecasted the trends for confirmed cases in India which can be seen in the second subfigure. The 3 rd and 4 th subfigures for the ACF and PACF plot for the determine the value of q and p for the model.

. CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted May 22, 2020. . Fig.1(a) , conveys about the forecasting of confirmed cases by the ARIMA model. The model fit based on the three parameters (p, d, q) which found based on Augmented Dickey-Fuller Test, ACF and PACF plot which given in Fig 1(a) which reveals the best model fit with ARIMA (1,3,1) having lower Akaike Information Criterion (AIC). This model predicts the future confirmed cases with a 95% confidence interval based on the previous data, which shaded by blue in the above Correlogram figure till June 30 2020. Similarly, from Fig 1(b) at the India level, we have forecasted total active cases with the best fit with model ARIMA (0,2,1), which estimates the predicted active cases for the same duration.

The analysis result for India level from Table1 and Fig 2, shows that the total confirmed and active cases at the India level will increase in the future. In India, based on our predictions, total confirmed cases may cross 4.4 lakhs, whereas total active cases will be close to 1.25 lakhs by the end of June 2020. We can see that the gap between confirm and active cases is also increasing as the rate of recovery and death also increases with the increasing cases in the coming time. In the mid of June. From our estimates, we expect that India will be having 294858 confirmed cases (95% CI: 173917, 415800). At the mid of June, we expect that Kerala and Odisha will be having less than 1000 confirmed cases based on our projections from data till May 3. However, states like Maharashtra (46611 confirmed cases (95% CI: 27513, 65709), Punjab (31172 confirmed cases (95% CI: 5310, 57034), Delhi (19245 confirmed cases (CI: 7135, 31355) & Gujarat (21510 confirmed cases (95% CI: 8448, 34572) will have a high number of confirmed cases at the mid of June.

. CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted May 22, 2020. . https://doi.org/10.1101/2020.05.17.20104588 doi: medRxiv preprint At the end of June, from our estimates, we expect that India will be having 441896 confirmed cases (95% CI: 210240, 673552). At the end of June, we expect that Kerala and Odisha will be having less than 1000 confirmed cases based on our projections from data till May 3. However, states like Maharashtra (58330 confirmed cases (95% CI: 28654,88005)), Punjab (42041 confirmed cases (95% CI: 667, 83416)), Delhi (24372 confirmed cases (CI: 5674, 43070) & Gujarat (27120 confirmed cases (95% CI: 6748, 47492)) will have a high number of confirmed cases at the end of June. At the end of June, from our estimates, we expect that India will be having 124712 active cases (95% CI: 68481, 180944). At the end of June, we expect that Kerala and Madhya Pradesh might not be having any active cases based on our projections from data till May 3. However, states like Maharashtra (47209 active cases (95% CI: 22106, 72312)), Punjab (36690 active cases (95% CI: 702, 72677)), Delhi (19315 active cases (CI: 0, 43592) & Gujarat (15137 active cases (95% CI: 3531, 26744)) will face a high burden of active cases at the end of June.

. CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted May 22, 2020. . In Fig.3(a) , we see different zones according to the prevalent cases by their shaded colour and we can compare the confirmed COVID-19 cases, how changes will occur in forthcoming days at three points of time. It indicates that some states have changes occur over time, the states like Gujarat and West Bengal showing less than 20 thousand confirmed cases at the end of May 2020, but till mid of June 2020 these states will infect cases more than 20 to 30 thousand. Further, if we compare map for confirmed cases in Uttar Pradesh and Andhra Pradesh also has increases cases from mid of June to the end of June. Similarly, in fig.3(b) shows the map for the active cases at three-point of time. It indicates that the Maharashtra and Delhi, Gujrat, Punjab and other states will have been increasing number of active cases over the three-point of time.

From these maps, we can understand some states like Maharashtra, Delhi, Gujrat, Tamil Nadu, Punjab, West Bengal, Andhra Pradesh and some other showing more affected states in countries by the COVID-19 Cases. Table 4 and fig.4(a) shows that the forecast value of the required number of beds for patients that will be suffering from COVID-19 in the coming time. At the end of May, the total number of beds required for India 64082 (95% CI: 47598, 80567). In some states, Kerala and Madhya Pradesh will cure all COVID-19 patients, and another way in some states as Maharashtra 23916 (CI: 16535, 31297), Panjab 15193 (CI: 5403, 24983), Delhi 9299 (95% CI: 2149, 16449), Gujarat 8017 (95% CI: 4546, 11489) and Tamil Nadu 3816 (95% CI: 1666, 5966) will require the highest number of beds. Fig.4(a) represents the number of beds in the middle month of June. The table suggests that in India, the required number of active beds is 85044 (95% CI: 54243, 115835) and Maharashtra will require 32022 isolation

. CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted May 22, 2020. . https://doi.org/10.1101/2020.05.17.20104588 doi: medRxiv preprint beds (95% CI:18263, 45781), Punjab will require 23189 isolation beds (95% CI: 3958, 42421), Delhi will require 12858 isolation beds (95% CI:0,26172), and Gujrat will require 10442 isolation beds (95% CI:4044,16840) will continue to maintain the top four position with an increasing trend in terms of the requirements. The numbers of beds required in Kerala, Karnataka, Odisha, and Karnataka will be minimum.

As the study suggested that the number of COVID-19 patients will be high in June; therefore, the number of beds also require high in June month. Table 6 shows the forecast number of beds require to admit the patients who suffer from COVID-19 at the end of June. India has to prepare 106006 isolation beds (95% CI: 58209, 153802) to accommodate the patients. The required numbers of beds have increased compared to the end of May.

Maharashtra will require 40128 isolation beds (95% CI:18790,61465), Punjab will require 31186 isolation beds (95% CI:597,61776), Delhi will require 16418 isolation beds (95% CI: 0, 37053), and Gujarat will require 12867 isolation beds ( 95% CI:3001, 22732), whereas Kerala and Madhya Pradesh would be removed from the table due to the zero forecasted active cases at the end of June. : 126-968) ventilators. It is noteworthy that Kerala will require 0 (95% CI:0,53) ICU and 0 (95% CI: 0,27) ventilators, and MP will require 2 ICU (95% CI: 0, 364), and 1 ventilator (95% CI: 0,182)) and are likely to overcome the situation.

. CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted May 22, 2020. . Table 5 & 4(b) shows the forecasted value of the required number of ICU and ventilators to admit the patients who suffer from COVID-19 at the mid of June. India has to prepare 10005 (95% CI: 6383,13628) ICU and 5003(95% CI: 3193,6814) ventilators to accommodate the patients. The required numbers of the facility (ICU and ventilators) are increased compared to the end of May.

Maharashtra will require 3767 ICU (95% CI: 2149,5386) and 1884 ventilators (95% CI: 1074,2693), Punjab will require 2728 ICU beds (95% CI: 466,4991) and 1364 ventilators (95% CI:233,2495), Delhi will require 1513 ICU beds (95% CI: 0,3079) and 756 ventilators (95% CI: 0,1540) and Gujarat will require 1228 ICU beds (95% CI:476,1981) and 614 ventilators (95% CI:238,991), whereas Kerala and Madhya Pradesh will not be considered in the calculations due to the zero forecasted active cases at the mid of June. Table 6 represents the forecasted value of the required number of ICU and ventilators to accommodate the COVID-19 active patients at the end of June. As the number of active cases increases the requirement of the facilities also increase. The required number of ICU will be 12471 (95% CI:6848,18094) and 6236 (95% CI:3424,9047) ventilators. Maharashtra will require 4721(95% CI:2211,7231) ICU and 2360 (95%CI:1105,3616) ventilators, Punjab will require 3669(95% CI: 70, 7268) ICU and 1834 (95% CI:35,3634) ventilators. Delhi will Fig. 4 

. CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted May 22, 2020. . https://doi.org/10.1101/2020.05.17.20104588 doi: medRxiv preprint require1932(95% CI:0,4359) ICU and 966(95% CI:0,2180) ventilators. Gujarat will require 1514 (95% CI:353,2674) ICU and 757(95% CI:177,1337) ventilators will continue to maintain the top four position with an increasing trend in terms of the requirements. The numbers are increasing very slowly in Odisha.

The world is going through a pandemic, and almost every country is affected by it. A country needs to know how much burden of active and confirmed cases it will have to bear in the coming time. It will help the country in taking pro-active measures to prepare adequate health infrastructure for the coming time based on future needs. We used ARIMA model and Auto ARIMA model on the time series data of COVID-19 cases in India for forecasting the total confirmed and active cases till June end. Based on our forecasted values of active cases, we calculated the healthcare infrastructure required in the coming time. Based on our forecasts, confirmed cases for India at the end of May are expected to be 178387 (95% CI: 128806, 227968). India will be having 294858 confirmed cases (95% CI: 173917, 415800) in the mid of June from our estimates. We expect that India will be having 441896 confirmed cases (95% CI: 210240, 673552) at the end of June. Our current estimates for confirmed cases are in line with the actual number of confirmed cases as on May 14, India has 82,047 confirmed cases, while our estimates suggested 83,336 confirmed cases (95% CI: 74752, 91921). AIIMS also suggested that peak will be achieved in months of June-July. Our results also show that daily confirmed cases are increasing at a faster pace even at the end of June with around 10,000 daily confirmed cases, so it is likely that peak will be attained afterwards. By the end of the June, Maharashtra (58330 confirmed cases (95% CI: 28654,88005)), Punjab (42041 confirmed cases (95% CI: 667, 83416)), Delhi (24372 confirmed cases (CI: 5674, 43070) & Gujarat (27120 confirmed cases (95% CI: 6748, 47492)) will be the most affected states. While Kerala and Odisha will be least affected having less than 1000 confirmed cases based on our projections from data till May 3

Based on our forecasts of active cases for India, we are expecting 75391 active cases (95% CI: 55997, 94785) at the end of May. India will be having 100052 active cases (95% CI: 63827, 136277) in the mid of June from our estimates. We expect that India will be having 124712 active cases (95% CI: 68481, 180944) at the end of June. Our current estimates for active cases are also in line with the actual number of active cases as on May 14, India has 51,388 active cases, while our estimates suggested 47442 active cases (95% CI: 42271, 52614). Maharashtra (47209 active cases (95% CI: 22106, 72312)), Punjab (36690 active cases (95% CI: 702, 72677)), Delhi (19315 active cases (CI: 0, 43592) & Gujarat (15137 active cases (95% CI: 3531, 26744)) will face a high burden of active cases at the end of June. While Kerala and Madhya Pradesh might not be having any active cases based on our projections from data till May 3.

In terms of health Infrastructure requirement at the end of June, we expect that India has to prepare 106006 isolation beds (95% CI: 58209, 153802) to accommodate the patients. When it comes to states, Maharashtra will require 40128 isolation beds (95% CI:18790, 61465), Punjab will require 31186 isolation beds (95% CI: 597,61776), Delhi will require 16418 isolation beds (95% CI: 0, 37053), and Gujarat will require 12867 isolation beds (95% CI:3001, 22732). For critically ill patients, there is a requirement of ICU beds and ventilators. Based on our forecasts for the end of June month, the required number of ICU beds for India will be 12471 (95% CI: 6848,18094) and 6236 ventilators (95% CI: 3424,9047). Maharashtra will require 4721 ICU beds (95% CI: 2211,7231) and 2360 ventilators (95% CI: 1105,3616), Punjab will require 3669 ICU beds (95% CI: 70, 7268) and 1834 ventilators (95% CI:35,3634), Delhi will require 1932 ICU beds (95% CI:0,4359) and 966 ventilators (95% CI:0,2180) and Gujarat will require 1514 ICU beds (95% CI:353,2674) and 757 ventilators (95% CI:177,1337). At present, India seems to be well prepared for the challenge as on May 8, India had 821 dedicated COVID hospitals with 1,50,059 beds (1,32,219 isolation beds and 17,840 ICU beds).

. CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted May 22, 2020. . According to our forecasts, it is a very alarming situation for India in coming days. However, the actual numbers can go higher than our estimates of confirmed cases, active cases and healthcare infrastructure as we made this forecast based on the data till May 3, when India observed complete lockdown. Currently, India has a partial lockdown with restrictions varying for three zones (red, orange and green zone) based on the current assessment of the situation in there. In future, lockdown might be lifted, and in that case, there will be a surge in the number of daily confirmed and active cases. The requirement of isolation beds, ICUs and ventilators will also be increased in that scenario. The migrants are returning to their homes due to loss of livelihood and income in the lockdown period, which may lead to a rise in the number of cases, and could not be accounted for, in our projections. So, India and its majorly affected states like Maharashtra, Gujarat, Tamil Nadu and Delhi need to be well prepared for the pandemic challenge in coming time and focus on increasing their healthcare infrastructure, and other states should also remain alert till the pandemic completely recedes. We suggest a Public-Private Partnership (PPP) model in the health sector to accommodate COVID-19 patients adequately and reduce the burden of the already overstretched Indian public health care system.

The forecasting of COVID-19 cases is done based on the data under the lockdown duration. So, the forecasted cases in future will be showing the same trend as India would have observed, had it been observing complete lockdown. Since May 4, India is observing partial lockdown, and which might be removed in the coming time so that actual cases will be more than the forecasted cases. For some states like Punjab showing more COVID-19 infection and Madhya Pradesh showing decline trend in future, but the situation may not occur in future because of the nature of the previous trend-pattern is different from now. Forecasted cases based on ARIMA model in our study for some states having lower bound for the 95% CI comes negative values which we have considered zero cases in that situation. In our study, the seasonality factor did not consider which is important in India, and maybe it affects our forecasting in future because of monsoon diseases.

. CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted May 22, 2020. . https://doi.org/10.1101/2020.05.17.20104588 doi: medRxiv preprint CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)

The copyright holder for this preprint this version posted May 22, 2020. . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)

The copyright holder for this preprint this version posted May 22, 2020. .

. CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)

The copyright holder for this preprint this version posted May 22, 2020. . https://doi.org/10.1101/2020.05.17.20104588 doi: medRxiv preprint CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)

The copyright holder for this preprint this version posted May 22, 2020. . https://doi.org/10.1101/2020.05.17.20104588 doi: medRxiv preprint . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)

The copyright holder for this preprint this version posted May 22, 2020. . https://doi.org/10.1101/2020.05.17.20104588 doi: medRxiv preprint CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)

The copyright holder for this preprint this version posted May 22, 2020. . https://doi.org/10.1101/2020.05.17.20104588 doi: medRxiv preprint

. CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)

The copyright holder for this preprint this version posted May 22, 2020. . https://doi.org/10.1101/2020.05.17.20104588 doi: medRxiv preprint

Estimation of the final size of the COVID-19 epidemic

Application of the ARIMA model on the COVID-2019 epidemic dataset

Real-time forecasts and risk assessment of novel coronavirus (COVID-19) cases: A data-driven analysis. Chaos, solitons, and fractals, 135, 109850. Advance online publication

analysis and forecast of COVID-19 spreading in China

A SIDARTHE Model of COVID-19 Epidemic in Italy. ArXiv preprint

Clinical characteristics of coronavirus disease 2019 in China

Trend Analysis and Forecasting of COVID-19 outbreak in India

Forecasting the dynamics of COVID-19 Pandemic in Top 15 countries in April 2020 through ARIMA Model with Machine Learning Approach

Application of an autoregressive integrated moving average model for predicting the incidence of hemorrhagic fever with renal syndrome

An ARIMA model to forecast the spread and the final size of COVID-2019 epidemic in Italy

Updates on COVID-19

COVID-19 and Italy: what next? The Lancet

A statistical analysis of the seasonality in pulmonary tuberculosis

Pandemic for the Top 15 Affected Countries: Advanced Autoregressive Integrated Moving Average (ARIMA) Model. JMIR public health and surveillance

Coronavirus (COVID-19): ARIMA based time-series analysis to forecast near future

Coronavirus (COVID-19): ARIMA based time-series analysis to forecast near future

Outbreak Trends of Coronavirus Disease-2019 in India: A Prediction. Disaster medicine and public health preparedness, 1-6. Advance online publication

Applications and Comparisons of Four Time Series Models in Epidemiological Surveillance Data

Note 1) *** shows active decline cases to zero in the respective duration and states.2) '0' show the minimum case or lower bound.