Jakarta Breakout: can 700,000 deaths across Indonesia be averted?
by Danendra Argyaputra Sakti, Maurice Glucksman and Alvin Cheng
This is the third article in a series. Today we have focused on two surprises:
- Here we are reporting on what can only be described as a potentially terrifying situation in Indonesia.
- In a companion article: an update of our New York City case. There is some potentially better news in the face of a tragedy.
- Every year, a mass migration occurs in Indonesia at the end of Ramadan. We are concerned this migration could spread COVID-19 across Indonesia. If the traditional migration from Jakarta to hometowns at the end of Ramadan happens, this could cause up to 700,000 deaths across Indonesia.
- We have localised three copies of our COVID 19 simulation model to reach this appalling conclusion. First, we have set up one model to represent Jakarta. The Jakarta model reveals that 350,000–400,000 people in the city may currently be infected — almost all of them without significant symptoms, so they are unaware they are infected but can infect others and spread the virus. By contrast, the current official data reports just 0.6% of the Jakarta population infected. At the end of Ramadan, the Jakarta model shows the number of infected people in Jakarta could be 1 million, with nearly all infections, around 96%, being unreported.
- We estimate 3 million of Jakarta’s 10 million population could travel home to visit their families at the end of Ramadan. The Jakarta model reveals 250,000–300,000 of these travellers. When they go home, they will catalyse infections in their home villages, towns, and cities.
- We set up two more copies of the simulation model to represent Medan and Tegal. We modified the models to accept travellers from Jakarta, and they project the arrivals from Jakarta could lead to 2,300 deaths in Medan and 4,000 deaths in Tegal.
- Medan has a younger population compared to the rest of the country, while Tegal has an older population. Consequently, Medan and Tegal’s death rates offer the lower and upper bound for the death rate in the rest of Indonesia. Scaling up from Medan and Tegal and adding in 25,000 deaths from Jakarta yields a range of 250,000 to 700,000 deaths.
- What can be done to avert this potential catastrophe? One example: A similar prospect was projected in Wuhan, China. Ahead of the annual mass migration for Chinese New Year, a lockdown on all travel out of Wuhan prevented the otherwise probable catastrophe. What about travellers who have already left Jakarta? Our sources indicate 200,000 workers may have already left ahead of the lockdown last weekend….
Urgent action is needed. The models we have used can help by testing different measures to assess which actions will have the greatest impact.
Modelling the impact of Lebaran migration on COVID-19 cases throughout Indonesia. See Appendix 1 for links to the models
Indonesia faces an impending crisis. In past years, around 30 million people travel to their hometown at the end of Ramadan during a period known in the country as Lebaran. In Jakarta, out of a population of over 10 million, as many as 3 million people participate in a mass exodus each year.
The simulation Dashboard (Exhibit 1) shows some key events to note about the situation in Jakarta today. One validity test we have used is to see if the model can explain the path of the minimal historical data (see blue circle 1 in Exhibit 1). Some historical data does not match because of definitions. For example, the reported recoveries are far less than the simulation because the model simulates a vast number of asymptomatic recoveries in addition to those in the records (see blue circle 2 in Exhibit 1).
Today our sources have estimated that already, 200,000 people have left Jakarta because many lack the means of support during the Jakarta Provincial Government lockdown implemented in the capital last weekend (See blue circle number 3 in Exhibit 1). In addition, currently, the Indonesian central government affirms that a lockdown will not be implemented to prohibit the migration at the end of Ramadan. Over the coming weeks, the total COVID-19 infections are steadily rising and posing an ever-growing risk if the infected people leave Jakarta (see blue circle 4 in Exhibit 1 and the detail in Exhibit 2)
Independently of any travel plans by people who want to leave during Lebaran, Jakarta is facing a very serious situation all on its own. The deaths are set to keep rising (see blue circle 5 in Exhibit 1) and accumulate to over 20,000 in the city.
Exhibit 2 shows a larger chart of the infections in Jakarta with the dates that people would leave. Our simulation reveals that 350,000–400,000 people have already been infected by the virus, accounting for 3.5% of the city’s population. By the end of Ramadan, the number of infections will likely increase to around 1 million people or over 9% of Jakarta’s population. Exhibit 2 shows how COVID-19 infections grow in Jakarta.
The simulation is a dynamic model using a stock and flow framework. Stocks are the accumulation of people in a specific state, and flows allow people to move from stock to stock depending on the progress of their condition. It is an adaptation of the standard SIR framework used by the Imperial College Analysis. Susceptible people (1) become infected and incubate, not able to infect others yet (2) and then pre-symptomatic infectious (3). From there they move to one of four states. (4) are a substantial amount of people who do not know they are infected and dangerous to others — they do not change their behaviour and continue to circulate freely. (5) are people who have mild symptoms and isolate themselves to protect others. (6) are severe cases who regrettably cannot get hospital treatment. 80% of these people die. (7) are people in hospitals split between the Wards and the ICU.
Out of 1 million infections in Jakarta during Lebaran, based on current detection rates in Jakarta, 96% of infections will likely be unreported. If the same proportion of unreported infections within Jakarta applies to those who will leave the city during Lebaran, 250,000–300,000 infections will spread all over Indonesia.
The next step was to set up the simulation to show the path of the outbreak after the travellers return to their hometowns. As an example, in Medan, our simulation estimates about 2,300 people will die from this second wave of COVID-19 infections (see Exhibit 4), not including the outbreak that has already happened in Medan. Another simulation for Tegal Regency, an area with no known COVID-19 death so far, forecasts 4,000 deaths from the influx of people coming back home from Jakarta.
With their unique age demographics, Medan and Tegal are not representative of the country as a whole, but by combining them, they provide a foundation to scale up an estimate of deaths for areas in Indonesia outside Jakarta.
Medan has a younger population while Tegal has an older population in comparison to the rest of the country. Indonesia is made up of villages, towns, cities, and regencies like Medan and Tegal. We are assuming that the deaths in these two areas would be repeated across Indonesia. To scale up to the country as a whole, the population of these cities is scaled up to 250,000,000. If Indonesia ex Jakarta has the same demographics as Medan, total deaths would be a lower bound of 250,000. Scaling up from Tegal yields a much higher death toll of 700,000. Indonesia’s overall population as a whole is more like Tegal’s than Medan’s, so we believe 700,000 a possible outcome from COVID-19. We have also used a separate simulation using the demographics of Indonesia as a whole, and this results in a projection of 600,000 deaths. These figures do not account for any shortcomings in the healthcare system, and if there are insufficient medical staff and facilities, the outcome would be worse.
Note that the projected death toll is entirely based on the people leaving Jakarta during Lebaran, and does not account for the other 27 million Indonesians that might also consider travelling during Lebaran or the infections that have already occurred outside of Jakarta. Consequently, the number of deaths across the country is likely to exceed the upper bound of 700,000 if current policy continues.
The surge in cases and deaths could be mitigated by preventative measures such as [a] stopping infected people leaving Jakarta and [b] tracing infected people who have already left. We hope this analysis will stimulate innovations that avert such a catastrophe.
The imperative must be: to contain unseen infections as much as possible within Jakarta and identify and isolate them elsewhere.
What are the possible Solutions?
Indonesia might be able to emulate the strategies put in place in China, and particularly in Wuhan, around the Chinese New Year holiday. During this busy period, the local Chinese government enacted a strict lockdown of the city that involved limiting any travel whatsoever and initiated a “close-contact detector app” that alerts users if they have been near a person who has been confirmed or suspected of having COVID-19.
Workers whose incomes have been entirely lost while Jakarta is in lockdown are at especially high risk if they have no means of support and must travel home to survive. We understand that as of today, as many as 200,000 people in these circumstances may have already left Jakarta. It is vital to track and trace infected people who have left and provide a means to help other people stay in Jakarta to avoid spreading COVID-19.
This article is a summary of extensive work worldwide representing the efforts of more than 20 multi-disciplinary experts and their extended network in relevant fields of science, business and government policy over the past four weeks. We are hopeful that our analysis will help the Indonesian central government and the Jakarta provincial government prepare for the challenges that will arise during Lebaran. We are standing by to help with interpreting the underlying analysis and the search for solutions.
The analysis is based on the publicly available information we have access to and is extensively benchmarked internationally.
We believe the analysis is accurate, but if new information becomes available that might change the conclusions we are eager to incorporate further information.
Acknowledgements: This was a rapidly assembled team effort.
Special thanks to Dr Hans Schepers and Dr Kim Warren for modelling help and all the synthesis and insights.
Thank you to Bima Sakti Ediyono, Diah Kurniawati, Daniswara Arsyaldi Sakti, Nina Triantis, Markos Glucksman, their help with fact checking and editing this article, and sharpening the underlying analysis.
Thanks to all the others providing key inputs to these case studies article: Stephen Allott, Dr Andreas Coutras, Dr. Pantelis Katharios, Barbara Mester
And again we are really grateful to John Hill and Chris Spencer for creating the Sheetless modelling platform: we have learned how to use so rapidly and now we can seek feedback from anyone, anywhere, for free, on our live model.
Disclaimer and request
We believe these results are accurate, but we cannot guarantee that.
Due to the urgency of the COVID-19 Crisis we have used a fast prototyping approach and released results and models before they are fully tested in the hope that the will be used and any flaws that may exist will be identified and you will let us know so we can make improvements. We have also deliberately invited high school students to work with us so we can try to ensure the tool only requires that level of skill. We have no special access to information, everything we have used as inputs is publicly available and it is possible material facts may surface that would change the results. If you have relevant information you can share to help us improve our models and analysis we hope you will get in touch. If you are interested in localising the model for your area please get in touch.
Appendix 1 — Links to key reference materials: