India’s flawed ‘death rate’
The TLDR: In part one of this series, we explained Delhi’s rising number of deaths. Today, we look at India’s death toll or Case Fatality Rate—often touted as the lowest in the world. But a closer look shows that the numbers and methodology are greatly flawed. No, this doesn’t mean there are vast numbers of deaths that have been covered up. But it reveals serious problems with our data—and suggests that we may have far less reason to be smug about ourselves.
The Case Fatality
This is the most widely quoted statistic. And it basically shows what percentage of recorded cases led to death. Today, the total number of cases = 9,222,216. The total number of deaths = 134,699. So we divide the total deaths by total cases, and express that result as a percentage. Today, our CFR is 1.46%. In comparison, the United States’ CFR today is 2%.
But numerous experts point out that both the numerator and denominator are deeply flawed in this calculation—and hence, so is the result.
The denominator: total no. of cases
CFR only looks at recorded cases. But scientists say that the more accurate number is Infection Fatality Rate—where the denominator is the actual number of infections. Now, actual infections are difficult to detect across the world. But the extent to which they are undercounted in India is revealed in two different ways.
One: Serological surveys randomly test a certain population for the presence of antibodies—which in turn suggest that the person was exposed to the virus in the past two weeks. The last such survey in Delhi found 25% of the population had been infected—which is 30.2 million but Delhi’s recorded case count is only 540,541. More importantly, the survey also found that 43.5% of those who had been diagnosed with Covid in the past tested negative in this survey. So it showed sero-surveys are also an underestimate of the actual numbers.
Two: Unlike the gold standard PCR tests, antigen tests do not look for the virus’ genetic material but instead for its proteins. As a result, they are more likely to throw up false negatives. At the start of the pandemic, PCR accounted for 100% of the tests being conducted in India. Today, that number is 60%—the rest being antigen tests. As a result, a new NDTV analysis shows that the recorded case number may not include 3.4 million cases! (Watch a data scientist explain this here).
The numerator: total no. of deaths
The good news: is that the number of people dying of Covid is falling across the world. One big reason: Doctors just know more about the disease, and which treatments work best. For example, 60% of the daily deaths in Delhi occurred within 72 hours of hospitalisation in June/July. That has come down to 25% today. But those gains can be just as easily lost if hospitals—their staff and resources—become overwhelmed.
But that doesn’t mean we have a good handle on the actual number of deaths.
One: Medical certificates listing cause of death is mandatory if the person dies on the premises of a hospital. But even so, doctors only issue them in 21.1% of deaths. This means we do not know the medical reason for the other 80%.
Two: As we explained in part one, the official death tally also doesn’t include “suspected” Covid deaths. These include patients who have symptoms but aren’t tested, test negative or show an inconclusive result. Point to note: Covid tests throw up a false negative at least 30% of the time—and often need a second test. If the person dies before that happens, then they are labeled as ‘suspected’ cases by the hospital.
Three: Many states are also not reporting the deaths of confirmed Covid cases. Here’s how it happens:
- The medical certificate lists the chain of events that led to a death: The immediate cause of death, the antecedent cause that triggered the immediate cause, and finally the underlying cause of death.
- Now, this should ideally be attributed to Covid. But in many cases, the death is instead attributed to a preexisting underlying condition. For example: cancer, heart disease, diabetes etc.
- Example: In Vadodara, authorities were attributing 75-80% of deaths to underlying conditions. In Telangana in July, 53.8% of its recorded deaths were of this variety.
How bad is it? In Palakkad, Kerala, a team of volunteers cross-checked news reports and obituaries that mention Covid against the state’s official list. They found 45% of them were missing. Making it worse: the political pressure to undercount:
“Many states are to a larger or smaller extent attempting to under-report data so as not to attract negative media and political attention... Assam’s mortality from COVID-19 is lower than Bihar’s, but reporting from the state has shown that its death audit committees reclassified 60% of deaths of COVID-19 positive people as being ‘deaths from other causes.’”
Alternative ways to count
It isn’t easy to fix these data gaps and lags. But some experts have suggested other ways to get at our real death rate.
Alternative #1: Do not count current cases. The official total number of cases includes every documented Covid case: recoveries + under treatment + deaths. Since we do not know if those being treated will recover or die, one suggestion is that we exclude them entirely. So death rate would be (total number of deaths) divided by (total deaths plus recoveries). Since that shrinks the denominator, the CFR will inevitably be higher.
Alternative #2: A government advisory firm Sapio Analytics takes a different approach. To arrive at the current death rate, it only looks at the last 21 days—the average interval starting with initial infection and ending in either death or recovery. Here’s one example offered by its CEO:
“In the past 21 days, from May 7 to May 27, total Covid-19 cases were 78,956. Now divide this by 21 to get an average number of corona positive cases. So 78,956 divided by 21 gives 3,760. Now on May 27, the total deaths were 190. So the calculation is X % of 3760 = 190. So, X = 5.0%.”
Alternative #3: is way too nerdy for us to summarise. It basically adjusts India’s numbers for age—to remove the ‘youth’ advantage. And it compares ‘lagged CFR’ across countries, i.e. uses a case count from the past rather than current case totals (since again, we don’t know the outcome of these infections). It’s described at length here.
Point to note: As per each of these methods, our death rate is higher, or at the very least, less cause for comfort.
The bottomline: Data is our greatest weapon against infectious disease. We need to stop treating it instead as a badge of pride or shame.
Reading list
The Wire has the best explanation of why our death numbers are messed up. NDTV has data and charts that lay out the errors introduced by rapid testing. IndiaSpend via Business Standard has the big picture on the gaps and anomalies revealed by comparing different states. Outlook’s interview with Sapio’s CEO lays out their approach. Scroll has the age-adjusted approach to comparing global CFRs.