Covid case-count charade

Colossal innumeracy in the most consequential problem we face

If I anchored my investment thesis around a number that was wrong by 98%, I’d be summarily fired. So would you, whatever be your line of work. Since this isn’t an error but an absurdity, serious psychiatric help might be in order. Now that government sero-survey and private diagnostic labs’ antibody stats are out, India’s covid discourse seems to be centred around a number that’s off by 98% or thereabouts. Without getting into punishment or therapy, can we at least limit using detected case-counts, especially because there are serious policy consequences?

First, the facts. Delhi sero-survey revealed that 1 in 4 Delhiites have been infected. Since the data is a month old, Delhi government is on record saying that current prevalence is closer to 1 in 3. Detected cases are 2% of actual cases. 98% error. Private lab antibody test stats show positivity rates in excess of 1 in 5 across major cities, with a few other metros ending up close to where Delhi is at. As these discretionary tests underweight slums with higher prevalence, true numbers might be even higher. Detected case counts massively understate true case counts across cities, and likely across India too.  

Next, the good news. Our worst state, on covid deaths per million people, has a true disease fatality rate of under 0.1%. While I cannot calculate this metric precisely for all of India, it’s likely to be better than Delhi’s. Recovery rate from the disease isn’t 60-odd % but well over 99%. Fewer than 1 out of every 100 infected people show any symptoms at all. We have a far stronger basis to calm the f#*k down (shameless re-plug of my prior essay: https://www.linkedin.com/pulse/calm-fk-down-anand-sridharan/)

The absurdity. Record-high cases state after state. 50,000 cases a day. India unlocks amidst upward spiral.Two million cases by August. Every headline is utter nonsense. Imagine fretting over two million after we’ve already crossed hundred million. If this was limited to professional commentators seeking TRPs, I’d calm the f#*k down. However, government authorities, both centre and state, make this meaningless number central to their discourse. Availability bias is quite potent in buggy humans. Any available statistic sticks in our minds, especially with frequent repetition and highlighting. This distorts everything: emphasis, implied virulence, priorities, strategy, trade-offs, panic. In light of 98% error (I cringe beneath my face-mask, just mentioning this number), it’s worth stating categorically: detected case-count is reflective of testing, NOT disease. Every metric derived using detected case-count is meaningless: case doubling time, growth rate, active cases, recovery rate, fatality rate, cases per million. This is colossal innumeracy in the most consequential problem we face.

Now what? For starters, stop emphasizing nonsense. Ignore or downplay reported case count and its derivatives. Stop feeding bogus data into eye-catching graphs and pseudo sophisticated statistics that get blindly shared. If you ban nonsense and still have to say something, you’ll be forced to look for sense. I’m no expert, but it seems like falling sick is bad and dying is terrible. If our objective is to minimize such outcomes, we might be better off talking in those terms. Shift focus from meaningless reported case-counts to hospitalizations (not institutional quarantine), critical patients and deaths. As an illustration, sensible metrics are trending right while nonsensical ones are trending wrong in my city. Big difference.

What about testing? At 98% error, testing is evidently not an indicator of prevalence. So, where are we going with it? If we’re trying to minimize tragic outcomes, goal of testing seems to be to ensure timely and appropriate clinical care to those who need it most. Some measure of whether testing is effective at early detection of vulnerable people feels more crucial to highlight than a simplistic numerator that’s anyway missing 98% of true cases. Who and when seem as critical as how many. I’m not suggesting curtailing testing. Just being clear about what it’s real utility is.

Elephant in the room: lockdown. This is way above my pay grade. This also has the potential to degenerate into a shallow kill-the-old vs kill-the-poor binary, especially on the internet. I’ll leave policy prescriptions to smart-sounding folks on TV and just say the following: without the distraction and distortion of a nonsensical metric, we’ll have a less muddled path to better decisions. We’ll have clearer objectives, better framing, right questions, relevant evidence and real cost-benefit. Qualified people in positions of responsibility will have the elements of sound judgment in place. General public will be better informed to accept and act in line with this judgment.

Invert, always invert. This maxim has been central to doing justice to my day-job. If I seek to find good businesses, I have to get really good at weeding out bad ones. If we’re seeking a sensible course of action on a complex issue, a good starting point is to weed out nonsensical metrics.

[Views are entirely personal]