Covid-19: What you need to know today
This column is about my favourite topic, testing, but from a different perspective — positivity rates.
India’s positivity rate (the number of people testing positive for Covid-19 as a proportion of those tested) has been rising for the past four weeks.
The table below captures the average weekly positivity rates for states that report testing data regularly, and also at the national level.
India’s positivity rate has increased from 5.37% in the week ended May 24 to 7.74% in the week ended June 21.
The World Health Organization says a positivity rate of between 3% and 12% is adequate — or indicates that a country/state/city is testing enough. The view among scientists and doctors at the Harvard Global Health Institute is that anything over 10% indicates insufficient testing.
So, why should we worry about a 7.74% positivity rate? Before we answer that, it’s important to understand why progressive positivity rates — or the trend line of positivity rates — matter. Any disease, Covid-19 included, can be said to be on the wane if we find fewer cases as we test more people from the population. Conversely, it can be said to be on the rise if we find more.
India’s positivity rate has increased by 2.37 percentage points over the past four weeks. This is a period when overall testing has increased. In the week ended May 31, India conducted 803,616 tests a week, an average of 114,802 a day. This number was 1.176 million, or an average of 168,051 a day for the week ended June 21. Thus, at a time when testing increased, so did the positivity rate — or, as we tested more, we found more (not fewer) positive cases. Statistically, a measure such as the positivity rate will increase with more testing up to a point, and then start decreasing. The positivity rate in many countries JUNE 7 JUNE 14 JUNE 21 may not have followed 4.80 1.26 1.44 2.09
this pattern (especially 3.45 9.77 3.14 1.65 because of rules governing 14.45 5.63 5.96 3.75 who can be or needs to be 7.75 5.01 6.12 5.76 tested) but it’s easy to see why 1.53 2.48 5.69 2.62 this should be so — not just for 13.17 23.61 29.37 24.02 Covid-19 but for just about
anything. The number of tests 0.73 1.96 1.99 1.88
at which the positivity rate 9.71 8.29 9.43 10.54
starts decreasing is a perfect 3.54 7.49 8.42 11.05
numerical measure of adequacy 1.34 1.21 1.38 1.52 (which means it 1.90 2.76 2.94 2.37 answers a question often posed 1.10 2.58 2.77 3.02 in this column: how do we 3.41 8.92 5.04 2.12 know we are testing enough?). 5.34 3.37 3.25 2.53 India’s aggregate data on 18.19 20.22 20.43 20.62 positivity rates is muddied by
the fact that most states mention 0.00 2.59 9.10 0.83 the number of samples 3.71 2.48 2.85 1.77
tested, not the number of individuals. 1.97 4.02 3.97 4.81 Many states require 0.94 0.89 1.01 1.59 infected (at least seriously 2.09 3.41 2.25 2.18 infected) people who have been 6.47 13.78 9.60 8.00 hospitalised or isolated in an 30.15 29.52 25.73 14.94 institutional facility to be 2.92 3.84 3.22 5.65 tested before discharge. Given 5.51 8.92 7.33 5.04 this, the positivity rates of
many states will actually be 2.57 3.87 4.75 4.20
higher. Tamil Nadu is one of 6.66 6.99 7.59 7.74
the few states that mentions ALL FIGURES IN % both the number of samples and individuals tested.
So, how should the table accompanying this piece be interpreted? In general, any state where the number is below 10%, and where it has seen an increase and then a plateau or a decrease, can be said to be testing adequately (both conditions have to be met).
States where the number is trending upward, despite being below 10%, are clearly not testing enough. And states where the number is above 10% and is continuing to trend upward have a real problem on their hands.
Delhi, Gujarat, Haryana (to some extent), and Maharashtra all seem to belong to the third category of states. Telangana and Tamil Nadu (especially the latter) appear to have things under control.
But many of the states are seeing an increase in positivity rates (even though the number itself is below 10%).
Four weeks may be too narrow a time-frame to base conclusions on, but over a longer period (say eight weeks; and it’s important to pick the right weeks; March and April saw very few cases in India, so, in hindsight, it does not make sense to pick weeks from those months), a trend analysis of positivity rates may be able to tell us when we are testing enough to stay ahead of the disease.
Right now, we aren’t.