Disclaimer: I’m not an expert on this field. I do this casually and just for the sake of curiosity. Ask people with the right expertise for more accurate estimation.
Edited: Mar 17, update the prediction chart
Edited: Mar 26, added concerns
COVID-19 started infecting more people in Indonesia this week, exponentially. Few schools and universities are closed. Few companies put a mandatory WFH exercise. COVID-19 just got real.
I tried to plot the cumulative confirmed cases for the top eight countries (except China) and three Indonesia’s neighboring countries. I said I tried because it’s probably not a good idea to plot eleven lines in one graph. But I did it anyway. I use data where the number of cumulative >= 22 and date <= March 13.
Again, I’m no expert on the field, but I do see two trends: the slow-exponential-growth like our neigboring countries and the rapid-blow-up like the rest.
Looking exponential curves on linear scale is not satisfying. I want to exploit the nice property of exponential curve where the log transform of the curve gives linear curve.
In the plot above, I can group the lines into three cohorts based on its gradient/slope; Singapore is in the low gradient, Malaysia and Australia are in the medium gradient, while the rest are in the high gradient group. Unfortunately, I think Indonesia belongs to high gradient group. And that is not a good news.
Now, to see the comparison more clearly, let’s see the day-to-day progression of the virus. I only take 16 days since the number of cumulative confirmed cases >= 22. We can indeed see that Indonesia is not in a good trajectory.
Lastly, I tried to extrapolate how many Indonesia’s cumulative confirmed cases will be in the next few days, considering its current trajectory, neighboring countries trajectory, and other countries (including China) trajectory. Since the trends after log transform is nicely linear, I used a linear model.
Moving from log scale to linear scale, we have exponential growth of the cumulative confirmed cases. Here’s the daily estimate.
The trend can change anytime and it’s up to us. It will be worse if we all are going out but it will be better if we stay at home.
Hopefully, it won’t be worse. Stay safe everyone.
Concerns over Low Number of Testings
Written on Mar 26.
The estimation above was off because the trend slows down since last weekend. While this could be a good news for us, I’m a little bit skeptical with the result. Especially, if we look at our number of tests and fatality rate below.
I cherry picked these countries due to the number of tested they ran and reputation. I also included Diamond Princess because almost all people were tested.
Fatality rate is death/confirmed. Population tested is tested/population.
|Country||Confirmed||Fatality Rate (%)||Tests||Population Tested (%)|
Anyway, let’s estimate Indonesia’s number of confirmed cases given different fatality rate.
In this Nature paper, COVID-19’s fatality rate is estimated at 1.4%, which is similar to Diamond Princess’. Given that, there are 5.5K confirmed cases today March 26, compared to just 893.
At best, if we assume the fatality rate is 2.5%, there are 3.1K confirmed cases.
At worst, if we use Singapore’s fatality rate, there are 24K confirmed cases, 25 times more than the current number.
Here’s the complete plot of cumulative cases versus death. Indonesia (large dark blue) is on the bad confirmed-death ratio area (bottom). The others dark blue dots are the number of estimated confirmed cases in Indonesia given different fatality rate.
The original growth is exponential. However, I did a transformation and plot the linearized version of the plot to make the lines more comparable. It will be a little bit messy because I tried to squeeze many lines in one graph.
The intention is to help the reader to estimate, visually, how many cumulative confirmed cases in Indonesia will be. There are multiple groups you can compare Indonesia’s trend with. Use the dropdown to select them.
This chart will be updated regularly.