A medical worker with a swab test for coronavirus at a drive-through testing facility in Tangerang. Modelling indicates tens of thousands of potential undetected cases in Indonesia. Photo by Fauzan/Antara

 

Covid-19 has moved quickly and quietly around the world – it’s a sneaky pathogen that in many cases has made its way into populations faster than it can be detected. Detection is one of the key sources of uncertainty in this pandemic, especially in countries where surveillance and response have been inadequate. This means that in many places, cases simply aren’t detected, even as populations suffer or die from the virus.

 

In Indonesia, numbers in the news have changed rapidly – from zero cases reported in February, to more than 2,700 cases this week. The country’s death rate from the virus is one of the highest in the world, underlining the possibility of a large number of undetected cases. A spike in the number of funerals in Jakarta in March is a stark indication of the cost of the virus’s spread, with local scientists predicting another 240,000 deaths from the virus by the end of this month. An urgent response based on what we know is needed to prevent such a crisis.

 

Modelling can help uncover the number of active cases that we might expect to see in a population of Indonesia’s size, and how fast the virus may be spreading. It can also be helpful to provide a future-looking perspective on the unfolding crisis, rather than responding only to the shock of daily news updates. Further, it can allow us to check on public health interventions, and what effect they may or may not have.

 

The data we have on hand can give us a 10-day forecast on the likely numbers of Covid-19 cases in each country around the world. My colleagues and I have spent the past few weeks constructing an interactive website that is updated daily and can do just that. Development of the site is ongoing as we refine the methodology and update the statistics.

 

Using the global standard Johns Hopkins University dataset, we use the data to make predictions about the growth rate of active cases, success in “flattening the curve” of the infection rate, and propose working estimates of the rate of detection. The idea is to present transparent data that can be used by people around the world to get a better grip on the scale of the pandemic and how fast it is moving in their countries. Without forecasting of this type, Covid-19 will catch many unprepared.

 

Source: Coronavirus 10-day forecast (covid19.science.unimelb.edu.au) as of 7 April 2020

 

As may be expected, our forecast shows the confirmed number of cases in Indonesia still steeply on the rise – this is not surprising, given that no cases were officially reported until March. However, with a total death rate of 221 among the raw number of 2,738 cases – reported as the world’s highest – we can also infer that many more cases are going undetected in the population.

 

This high apparent death rate is not because the virus is more dangerous to Indonesians, but because many milder cases have not been detected: there are many more cases out there than are reported.

 

Non-detected cases include those we call “undiagnosed”, where infected individuals have yet to present symptoms but are soon to be diagnosed through testing, and those we call “undetected”, where individuals present only mild symptoms, or are altogether missed by surveillance.

 

As of 7 April, our model estimates as many as 1,975 undiagnosed cases in Indonesia and 51,810 undetected cases, giving a possible total of 56,173 cases – much higher than the reported number of 2,738. This indicates an urgent need for increasing testing and medical treatment for tens of thousands of Indonesians.

 

Of course, these numbers aren’t fail-safe. Our method assumes a fairly high fatality rate of 3.3% among symptomatic cases, it assumes widespread community transmission, and a period of 17 days between infection and death. But given Indonesia’s very high death rate, we can confidently conclude that the actual number of cases is likely to be much higher than is currently being reported.

 

Source: Coronavirus 10-day forecast (covid19.science.unimelb.edu.au) as of 7 April 2020

 

Even if we look only at the active case numbers, the growth rate in Indonesia is still a cause for concern. This growth rate is the percentage change in case numbers each day. It is analogous to a compound interest rate, compounded daily. Progress in controlling the epidemic would be indicated by a growth rate that steadily declines, eventually moving into negative territory. Malaysia, for example, looks to have managed a rapid reduction in growth rate over the last week. In Indonesia, we see alarmingly high growth rates and only a gentle trend towards lower growth across this same 10-day period.

 

Some of the daily fluctuations may be attributed to reporting issues, for example, when countries miss a day of reporting and then include those numbers the following day. But overall, the numbers do indicate that public health interventions by individuals and the government have yet to make a big impact on the spread of the virus in Indonesia.

 

Bear in mind that in the case of coronavirus, public health interventions take about ten days to start affecting growth rates. Changes in growth we see now are due to changes made ten days ago, because growth in diagnosed cases is driven by people that are already infected, but are yet to present symptoms.

 


Source: Coronavirus 10-day forecast (covid19.science.unimelb.edu.au) as of 7 April 2020

 

We can also step back, beyond the last ten days. Our site reports a curve-flattening index that gives an idea, across the entire time-span of the epidemic, how well a country is ‘flattening the curve’. Positive values of the index indicate that growth rate is slowing at that particular time. Here, Indonesia may be seen to be making some progress, but we need to be aware that the index is sensitive to rapid changes in testing efforts, so this result needs to be set against Indonesia’s history of surveillance for coronavirus.

 

For example, no testing, followed by a sudden increase in testing as cases become apparent, can lead the index to trend negative, followed by positive, as we can see happened here between late February and early March in both Indonesia and Malaysia. Nonetheless, both countries do appear to now be slowing the spread.

 

This is the important part – by giving an idea of the scale of the pandemic and the speed at which new cases can appear, modelling can indicate where interventions need to be made, and how effective interventions have been so far.

 

Indonesia has been slow to recognise the outbreak within its borders, but since cases were confirmed and modelling has shown how fast the epidemic might spread, the government has begun to take measures such as calling on the public to self-isolate where possible, with President Joko Widodo asking Indonesians to “stay, work and pray at home” for two weeks. The decision to allow Jakarta to implement so-called “large-scale social restrictions” (known as PSBB in Indonesian) from 10 April is a vital step.

 

These measures are important, because they reduce the growth rate of the virus, reducing pressure on the medical system and buying Indonesia time to plan. Access to more reliable numbers on the true number of cases, and expected cases in ten days’ time can also help the public take the government’s directives seriously and avoid the advice of misinformation and hoaxes spread on social media.

 

As in other countries, social and economic inequality will continue to be an issue for implementing self-isolation measures in Indonesia, as many casual or informal workers don’t have the option of working from home, and living conditions aren’t always conducive to stopping the spread of virus. The next big challenge will be curtailing the usual mass movement of people from urban to rural areas at the end of Ramadan next month.

 

Our model indicates that cooperation by the public in Indonesia with government directives to self-isolate and to avoid going home for the holidays will be crucially important to help slow the spread of the virus throughout the archipelago. It also indicates that tens of thousands more Indonesians need access to testing and treatment to avoid more deaths from the virus. We hope that the numbers presented here can support early action and avoid a more severe crisis.

 

, ,

We acknowledge and pay respect to the Traditional Owners of the lands upon which our campuses are situated.

Phone:13 MELB (13 6352) | International: +(61 3) 9035 5511
The University of Melbourne ABN:84 002 705 224
CRICOS Provider Code:00116K (visa information)