The legends in the map show the classes in the order- Handling well so far > Need to manage better > Still retrievable by social distancing > Untraceable outbreak > Worst case scenario, where ‘Handling well so far’ is the best possible situation and ‘Worst case scenario’ is the worst one. The indexing of the zones is done from criteria stemming from plots generated by COVID-19 confirmed case data. The plots show Bangladeshs’ district wise trajectories in confirmed COVID-19 cases from March 7 to June 3. These show daily new cases of COVID-19 vs time, where the cases per day are averaged over the 3 separate averages i.e. 3-day rolling average. 5-day rolling average and 7-day rolling average of the number cases respectively. We opted for averaging 3 averages instead of only 1 because, different districts had different lag time i.e. 3 days, 5 days or 7 days regarding delivery of test report. Thus, finding the rolling average of a district for 3-days which otherwise has a lag of 7-days will plot for false cases and vice-versa.
The averaging method was not used in case of Criteria 3 (details in the description)
For Criteria 1 & 2 the ‘case’ axis (y-axis) is exclusive to each curve based on their highest confirmed case. For Criteria 3 it depends on the total confirmed cases within the given timeline.
Criteria of classification:
We have selected the zones following mainly 3 criteria from the plots, they are as follows:
||· Criteria 1:
Highest number of cases in that district in a day within the given timeline
|· Criteria 2:
Trend of the curves (focusing on terminal point)
|· Criteria 3:
Exponential Growth Curve
||Downward or Flat Curve within 15%
||Curve within >15% – 25%
||Curve within >25% – <50%
||Curve at >=50%
||<20% Cases from the Peak
The criteria mainly depend on the plots given below. They can be described as follows:
Criteria 1: Highest number of cases in that district in a day within the given timeline
Population movement and pattern are easily understandable from change in number of cases per day. Initially, we plotted all the curves in one chart to see the relative visualization and found that, the trends were heavily dependent on the highest number of cases at each point as they are for each day. Thus, we determined this criteria classes from that relative plot. Also, we couldn’t use case per population data for there was no change on the plots (details in the FAQs). This criterion serves as a good evidence for how districts are managing in the lockdown situation.
Criteria 2: Trend of the curves (focusing on terminal point)
The main reason for plotting the curves was to understand the trend of the data. And, with time for the change in the case-trend, mainly in the most recent time, it was important to understand whether number of cases in each district were increasing, decreasing, or remaining the same, regardless of the total number of cases. So, the trend of the curve at its end-point represents the most recent situation.
Criteria 3: Exponential Growth Curve
Any analysis with population data is better of with any kind of population projection in order to understand how the data may change in the future. And, finding the exponential growth is a very effective way to understand that. The points are plotted with the cumulative of the cases and the range was determined by the averaging the exponential growth factors (formula: new case/previous case) for each district.
Going back to the table, each district would have 3 indices, one for each criterion. We then weighted those 3 criteria by the indices to get 1 value for zoning. Finally, we named our categories or, Risk Zones as per the list:
|Index value after weighting Criteria 1, 2 & 3 (Map 1)
||Index value after weighting Criteria 1 & 2 (Map 2)
||Handling well so far
||Need to manage better
||Still retrievable by social distancing
||Worst case scenario
We also chose to go for 5-indices categories instead of the conventional 3-indices categories for zoning (red, yellow, green) because, the data were highly intersecting or overlapping in many cases. Some districts were not completely in danger but had more than average number of cases, and thus they were in the middle of being in yellow or red. Moreover, some districts had an upward trend but had very few cases to even be in yellow zone. Thus, we categorized them in 5 indices which in turn gave us 5 zones. The naming of the 5 zones have been completely substantial. The ones in the Worst case scenario needs no description. The ones in Untraceable outbreak have shown symptoms of community transmission and the lockdown has been very ineffective in those regions. The ones in Still retrievable by social distancing are the textbook-average ones. We need more time to estimate their trend, but social distancing can still work for alleviating the situation. The ones in Need to manage better are very close to having normal situation and they have the resources for it. Given every single factor for COVID-19, the ones in Handling well so far are the ones which are able to manage the situation in the best possible way.
Weighting the aforementioned Criteria 1 and Criteria 2, we have sorted out the indices as in the table above. The map showing the Risk Zones is as follows: