What’s the right measure of the impact of Covid-19?
On May 3, the day bridging the second and third phases of the lockdown, where did the different states of India stand with regard to deaths resulting from Covid-19? How do we decide which states are doing better?
Usually, media reports mention the total number of deaths, but these are not adjusted for the size or the age structure of the population.
That makes comparison between states difficult. A different number which stands out even more prominently in the reports is the Case Fatality Ratio (CFR), which refers to the deaths as a percentage of the cases diagnosed.
Which is the number that provides a better comparison of the states?
CFR has a clearly identifiable numerator: deaths attributed to Covid-19. The denominator can be very variable, based on the number of persons who tested positive for the virus and thus classified as cases. That, in turn, depends on the number of tests performed.
All who are tested for the virus will not be positive, but more the number of tests performed, higher the total number of persons detected, even if the fraction of positive tests remains constant. Quite often, when large numbers are tested, more mild cases turn up among the positives and they are less likely to die. So, a high testing rate, with more cases detected and a greater proportion of milder cases, will lower the estimate of CFR. A lower testing rate, with tests usually reserved for the more severe cases, will yield a smaller denominator of cases and, therefore, a higher CFR. South Korea, which tested liberally, including several asymptomatic persons, had a low CFR because of the large number and lower composite risk profile of the cases included in the denominator.
On the other hand, total deaths in any state are a clearly defined number at any given time, assuming that deaths are being correctly classified as Covid or non-Covid. However, the absolute number of deaths will also vary between states according to the size of their populations. If we standardise that measure and estimate deaths per one million population, we can compare different states more meaningfully. This figure yields a better picture of a state’s overall performance against the virus, capturing the benefit from the population level containment strategy as well as the clinical case management impact. It is this broadband package of public health and clinical interventions that has a combined effect on lowering deaths in any state. On the other hand, CFR usually gives us information only about the clinical treatment benefit. Only when we get a measure of all cases infected in a population -- severe, mild and asymptomatic included -- do we get an Infection Fatality Ratio (IFR), which invariably is lower than CFR.
The two statistics (CFR and deaths per million) also give very different pictures about the states (Charts 1 and 2). If we look at CFR, Delhi (1.6 ) looks better than Punjab (2.6) or Jharkhand (2.6), which have lower testing rates. On the other hand, Punjab has far fewer deaths per million population (0.63) than Delhi (3.14). Jharkhand is even better (0.08 deaths per million). Where would you rather be today? Meghalaya looks terrible on CFR but is very well placed when the actual death rate in the population is considered. Odisha and Kerala do well in both, while Maharashtra and Gujarat stay high in both while minimally changing places .
Statistics can be appealing or appalling, depending on how they are presented. The purpose of each statistic differs from another and the choice, of which to use when, is driven by the nature of assessment we wish to make. If performance of hospitals is being compared, CFR matters. If the overall state of the epidemic is being assessed, deaths per million is the statistic that tells us whether a state is in a safer zone than another. It is important that all of us -- policymakers, media and the public -- understand these measures and their distinctions so that we can soon get the measure of Covid-19.