For weeks now, I have been asking for the Tamil Nadu Health Department, which releases the Daily Media Bulletin, to publish the number of [COVID-19] tests done in each district — every day.
Let me explain why this is vital, for two reasons, with concrete examples from official data.
Also read: Coronavirus | Tamil Nadu caps testing fee in private sector at ₹3,000
Currently, the government’s publishes the number of samples tested in the entire State each day, and the number of positive cases in each district. While this gives some indication of the prevalence of the infection (and, therefore, risk to people in that district) on average across the State, it is not enough to understand the risk in each district with enough confidence.
For example, the number of positive cases in Madurai has gone up sharply from 268 (100 active) on June 1, that is almost three months after the first case in Tamil Nadu, to 493 (164 active) on June 17. This is a jump of 84% (64 % increase in active cases) in 17 days. Should this cause great alarm in Madurai?
The simple answer is that we cannot tell for sure without knowing how many people were tested in Madurai in those 17 days.
Also read: Tamil Nadu making up for lost momentum by testing at higher rate
For example, if 10,000 of the roughly 2,81,000 samples tested in Tamil Nadu during these 17 days were done in Madurai, then the positivity rate is only 2.25% (and the spread is low), so the average resident should not be alarmed about increased risk. But if only 1,000 tests had been done, then the positivity rate is 22.5% (which implies massive spread), and residents should exercise extreme caution — perhaps to the extent of not coming out of their houses despite the Government relaxing conditions.
(Cumulative Tests Done per Million (10 lakhs) of Population from inception, published June 7, 2020)
Therefore, it is vitally important, for every district in Tamil Nadu, to know the number of tests done each day. The one day the Government released this kind of information (on June 7, after weeks of demands from many sources), the data raised serious questions and concerns (blue bars in graph 1 above).
Also read: Ramping up testing is the way forward in TN, say experts
It shows that sample testing varies greatly by district, after adjusting for the population (tests per 10 lakh population). Why should this be the case? From a modelling priority, one could argue for testing more where the known prevalence is higher. However, analysing the rate of positive tests (also provided by the government) we can see that does not fit either.
(Rate of positive tests (# of positives/# of total tests) (red line) with tests per million (blue bars) by District (7/6/20))
In other words, the Government has neither done testing based on population in each district, nor based on the prevalence of COVID-19 infections. The Health Department must explain what other logic or rational drove their decisions on the allocation of tests to districts.
| Date | Samples tested | #of positive results | Rate of +ve Tests |
| June 10 | 17,675 | 1,927 | 10.90% |
| June 11 | 16,829 | 1,875 | 11.14% |
| June 12 | 18,231 | 1,982 | 10.87% |
| June 13 | 17,911 | 1,989 | 11.10% |
| June 14 | 18,782 | 1,974 | 10.51% |
| June 15 | 18,403 | 1,843 | 10.01% |
| June 16 | 19,242 | 1,515 | 7.87% |
There is yet another problem that arises when the testing numbers by district are not revealed daily. Consider the period between June 10 and June 16, where the results (in the adjacent table) were reported by the government. On the surface, it appears that the rate of infection might be slowing, as both the number of positives, as well as the rate of positives (since the volume of tests is relatively stable) show a significant improvement.
Also read: In new lockdown, a second chance for Tamil Nadu
Of course, there can be many explanations for such changes in numbers. The simplest would be that — with all else remaining the same (including proportion of tests allocated to each district) — the prevalence of infections was actually falling.
But there is at least one other scenario where the drop could occur without the actual statewide prevalence falling at all.
The graph above shows that Chennai has by far the highest rate of infections (and many of the other 37 districts in Tamil Nadu have far lower positive rates), the drop of a few hundred daily cases could just as easily result from the volume of testing being reduced in Chennai, and increased in low prevalence areas (such as Vellore or Coimbatore).
In simple terms, reduced testing in the high infection areas around Chennai, and increased testing in other low infection areas, would yield this overall reduction in positive rate, even while maintaining the same testing volume. But, if such a shifting of testing volumes away from Chennai to other districts had happened, it would most likely have led to the proportion of Chennai cases as a percentage of total Tamil Nadu cases declining also. Strangely, the data shows exactly this trend (direct matching of declining rates and proportion)
I am not implying that the government disproportionately directed the tests to be done outside of Chennai between June 10 and 16 to reduce the statewide daily number of positive cases. I am simply proving that a strategy of such shifting of test volumes could explain the reduction in daily new case numbers & rates.
(Rate of Positive Tests in Tamil Nadu (Blue Line) are highly correlated to the proportion of cases in Chennai (Red Line))
This is the second reason that the Government must release the daily testing numbers for each district in Tamil Nadu (just as they state the number of positives by districts each day), and not just the aggregate number for the whole State. We, the citizens, need both the numerator (# of positives in each district) as well as the denominator (number of samples tested in that district) in order to understand the risk (and model our behaviour commensurate to the risk), as well as to understand the true nature of statewide prevalence without doubting the process/methodology.
Before I finish, I want to show one more graph to show how trends have changed even over the past couple of weeks — which should be helpful for those who don’t quite understand how “unusual” or “unlikely” the corresponding curves in the graph above are — and why they, therefore, cause concern about possible “test-shifting” from high-prevalence to low-prevalence districts.
(No visible pattern of Rate of Positive Tests in Tamil Nadu (Blue Line) and the proportion of cases in Chennai (Red Line))
With increasing testing, the daily cases and rates have started rising again, from the recent low of 1,515 (7.87%) on June 16 to 2,174 (8.54%) on June 17, and 2,141 (8.01%) on the next day.
Until the Tamil Nadu Health Department starts releasing the testing statistics by district (daily and retrospectively back to March), we simply cannot tell how much risk we are exposed to by district, and whether the situation is improving or worsening in each of them, as well as across the whole State.
The longer they continue their irresponsible “shrouding” and non-transparent actions, the greater the charge of potential scenario/data manipulation they will face, and the consequences they will have to bear — sooner or later.
(The author is a DMK MLA who represents the Madurai Central constituency)
