Recent Work on COVID-19

RECENT (2nd Wave) PROJECTIONS

Note: green line--test adjusted predicted cases; red line: actual cases (test adjusted). It is assumed that the tests will be around 15-20 thousand. So if the tests are 50K+ then reported predicted cases can be 15 thousand or more.

If you have any trouble in viewing the number either try desktop browser or use this link https://plotly.com/~shafiun.ihe/54/

Bangladesh may observe another peak in the Mid to end of July with the daily cases up to 15,000+ daily cases at peak . However, govt. test capacity stands as it is (35000 or so), we will not be able to detect so many cases and it may report up to 10-12 thousand. And the infection will continue until mid-Aug even with lockdown. Then the infection may go down slowly. However, the level of govt. intervention and vaccine drive may affect actual infection.

Assumption: no intervention after Aug 5, 2021

Note: The first graph is test adjusted. To capture increased transmissibility prediction curve is shift one-week left to what model predicted.

(Update July 27, 2021)

Model Credit: CoMo Consortium, Oxford University, Bangladesh Team

Rt (Effective reproduction rate is close to one, that infection continues at steady rate

If infection continues hospital things may out of control

Current Rt is greater than 1.4 which is very alarming

Projection using SIR Model (FIRST WAVE PROJECTIONS)

Last update: Jan 16, 2021; Since official cases reporting is likely to be underestimated, the actual cases will be much higher than the numbers shown here. Also note, since early July, we are not testing enough, so I used death figures to understand the infection. Even though it can also be under-estimated, it is relatively more reliable than case numbers

(IMPORTANT: NO-COUNTRY AND NOT MODEL CAN CORRECTLY PICK THE PEAK OR 100% ACCURATE PREDICTION SINCE IT DEPENDS ON SO MANY THINGS, ALL THESE PROJECTIONS POSTED HERE IS TO PROVIDE SOME TENTATIVE ESTIMATES BASED ON SOME FORECASTING MODELS)

PROJECTION USING DEATH NUMBERS & SIR MODEL

--Death number is used

--Last update: Jan 16, 2021

Interpretation of the SIR Model:

Top graph: Indicates the actual and projected cases. The top graph is illustrated with three colors. Pink color indicates the danger zone and infection will continue until we reach yellow zone when growth is expected to steady for while until it reaches in the green zone i.e. the graph is reaching plateau. This graph says that if the current rate continues, our infection will grow until Mid-Dec, then it be will be stable (grow but at lower rate) for several weeks until early April.

Bottom graph: The bottom shows the growth factors. We would like to be in the region where the growth factor is less than 1, which happened at the end of July. The R (effective reproduction rate--how many people get infected by an infected people on an average-- is 0.88 which is still high.

(Matlab Code credit: Professor Milan Batista, University of Ljubljana, Slovenia)

SUMMARY OF RESULTS

Outbreak
03-Apr-2020
Start of acceleration
19-May-2020
Turning point
21-Aug-2020
Start of steady growth
11-Dec-2020
Start of ending phase
24-Mar-2021
End of epidemic (5 cases) 22-Oct-2021
End of epidemic (1 case)
22-Jan-2022

Are we testing enough?

Positivity/ case detection rate in comparable countries

(July-Aug 23)

Underestimation of the actual infection due to low test numbers

Clearly our infections are stable (remains same) though government numbers is showing otherwise (declining)

The bar shows the Benefit Cost ratio of Free Testing. The benefit is in between 2.38-11.90 for each TK investment.

Methodology: lower tests mean lower official cases, which will increase the infection rate, and so does the deaths. These deaths could be averted with free tests. Productivity is adjusted 60+ will not earn.

This sections included relatively older stuffs and other research work not related to projection

Projection using Machine Learning Tool

(Facebook Analytics Tool: Last Update July 7)

(Assumption: the infection will growth at an exponential rates, as per this model infection may go up to 288,000 by the end of July . Hover around the line to see date specific forecast. If you are visiting from mobile devise and having trouble in viewing click here https://plotly.com/~shafiun.ihe/42/ , and underneath you can see the static version of the graph (if you visit from mobile device and face difficulty in viewing, tap in gray area and then scroll left or right.

Summary of TIME SERIES MODEL

--July 31: 288K

--Aug 31: 386K

PEAK: MID-3rd Week of JULY WITH 4.2-5 THOUSAND DAILY CASES

(Note: If we don't test enough official statistics will be lower than these numbers)

Note: the official numbers will likely to be lower if testing is lower

Projected Total Cases: Various Models

Projected Daily Cases: Various Models

Test-adjusted actual cases are used

Peak

SIR Model: Already in peak

Time Series (GOM): End of July with 4500 cases

Data Analytics: Mid-End of July with 4-5.5K cases

(if testing remains low then we would not see 5.5 cases officially)

Impact of complete economy opening

Update: June 16, 2020

Changing districts ranking over time

Update: June 16

Data: IEDCR (some anomaly may be observed)

Bangladesh has a higher number of Covid-19 cases than Russia (no 2 in terms of total cases as of May 21) has had in the first 70 days. France and the UK, the other two worst-affected countries, have had few more cases than Bangladesh. The effect of eid shopping and home-going sprawls yet to be seen!

Covid-19 Spread (Dhaka City, Nganj and district with low case numbers are excluded

Covid-19 spread across the country

For interactive map visit

public.tableau.com/profile/shafiun.shimul#!/

For more on projection visit public.tableau.com/profile/shafiun.shimul#!/

DISTRICTS' CONDITIONS

Last Update (June 10)

District-level graph (districts with minimum 500 cases, as reported, Dhaka city excluded).

SOME WORKS DONE PREVIOUSLY

Cases are growing exponentially in some districts. These are Sylhet, Noakhali, Feni, Cox's Bazar, Chattogram Faridpur, Dhaka, Cumilla. Update: June 2, 2020

Many districts are experiencing steep growth. Chattogram, Rangpur , Cumilla can be next hotspots

Relatively high case numbers but growth is less worrying

Curve is flattening in Narayangan but not in Dhaka City. Still it has showing steady growth which is the biggest concern

Chattogram will catch up Narayanganj soon

District level growth from May 12- May 20, 2020

See the growth of confirmed cases in the top 20 (in terms of number) districts --Dhaka has shown steady increase--Red flag for Chattogram, Cumilla

Growth Factor of the districts (as of May 20, districts with minimum 200 cases)

Growth greater than one indicates a sign of increase and less than one is for a sign of decay, and so one means every day you have the same number of new cases. It is very simple statistic (new cases today/new cases of yesterday) but it can give a good insight. If gf is below one and the curve is moving downward then it is a good sign. Whenever it comes down to daily growth of cases, Gazipur, Kishorganj, Narshingdi even Narayanganj is in relatively good shape, but most other districts are not. Most importantly, Chittagong, Cumilla, Rangpur, Cox’s Bazar has worrisome growth. Dhaka is still worryingly. The next graph shows th week growth factor. If growth factor is greater than 2, then cases are doubling in a week, and if it is greater than two then cases are doubling less than a week. Again Chattogram, Cox’s Bazar, Cumilla, Rangpur are showing some bad signs.

Weekly growth factor

Daily growth factor

Work in progress

Does lock-down really work to reduce infection? an empirical evidence from Covid-19 cases

The primary aim of this study is to understand whether lockdown or those types of measures can affect the growth of infection in developing countries. To understand the impact of lockdown and other measures on infection growth, the study uses panel regression-based difference in difference and GMM estimation method. This study shows that lockdown type measures are not very effective in developing countries even though these types of arrangements are highly effective in developed nations. However, with lockdown, staying at home, income support programs, and other social distancing measures are found to be effective for both developed and developing countries. In addition, the timing of the lockdown also matters. Enforcing nationwide lockdown too early-when cases are very low-may not be effective or optimal and enforcing lockdown too late is ineffective. Even though this study does not find strong evidence of the effectiveness of lockdown for developing countries, this finding does not necessarily indicate that lockdown should not be enforced in developing countries; rather it suggests that lockdown should be combined with a strong drive in contact tracing, extensive testing, income support for the poor, effective management of informal or migrant workers—to make the lockdown effective. Merely declaring lockdown, without complementing other must-have measures, will endanger the economy without contributing much to reducing the growth of infection.

Understanding the correlation of case-fatality rate and socio-demographic characteristics

Abstract: Using correlational analysis and traditional OLS and LASSO, we find that ageing and health system capacity are the significant predictors of case-fatality rate. The country with larger ageing population is found to have higher death rates, whereas the country with higher health-system capacity (hospital beds per capita) have lower death rate. Other controls: tropical region, income, pollution, diabetics rate etc. are not statistically significant. These are correlations, not necessarily causation.

This a joint work with Fariha Kadir and Mohammed Ihsal-ul-Kabir.

Page updated

Google Sites

Report abuse