E-Paper
Daily deaths best way to assess Covid-19 pandemic | Opinion
It must be noted that the US, where the Covid-19 epidemic has been raging, now has a doubling rate four times higher than India’s, writes K Srinath Reddy
The release of antibody-based serological surveillance studies from Delhi and Pune have revealed high rates of viral dissemination in the urban populations. Delhi‘s recently completed repeat survey, which took place in July-August , reported that 28.3% of Delhi’s residents had tested positive for anti-Covid-19 antibodies. This indicated a rise from the previously reported figure of 23.4% a month earlier. A survey conducted during the same period in five highly affected wards of Pune reported a prevalence of 51.5%. A sewage sampling study, of viruses excreted in faeces, estimated that 6.6% of the population in Hyderabad were probably infected.
Follow latest updates on Covid-19 here
The exact number of persons infected, whether symptomatic or asymptomatic, is uncertain as the antibody tests can yield false positive tests due to cross-reactive antibodies produced by other coronaviruses. The false positivity rates get amplified when mass surveys are done, due to variable rates of pre-test probability between the hospital cases and the general population (Bayes Theorem). To some extent, the false positives may be balanced by false negative results in persons who were infected early in the epidemic, as the antibody levels decline in about three months.
Despite these uncertainties, we may assume that at a fifth of Delhi’s population may be already infected, if not a fourth. Even in other big cities, a far higher number of persons would have been infected than identified by viral testing methods (Reverse Transcriptase-Polymerase Chain Reaction and Rapid Antigen Tests). While debates continue on how short these numbers are from the herd immunity threshold, there are pertinent questions on how this information affects our reliance on some widely used statistical measures for tracking the course of the epidemic.
Also read| Covid-19: What you need to know today
The popular measure is for infectivity in the population R, expressed as Ro or Rt. The number provides an estimate of how many others are infected by an infected person. The number used in the initial part of the epidemic is Ro. This assumes that everyone in the population is equally susceptible and derives its estimates from the numbers of cases and deaths being reported. However, susceptibility varies over time as some people get infected and recover and are presumed to be no longer susceptible. So a different number (Rt) is used when the epidemic advances and estimates of numbers recovered are also available. Though this is called ‘effective R’ (Re) by some, the term Rt is more conventionally used.
Click here for complete coronavirus coverage
However, both these estimates of R are dependent on testing rates and other diagnostic criteria employed to identify ‘cases.’ Infected individuals who are asymptomatic and untested do not enter into these calculations. The antibody tests now tell us that these ‘missed’ numbers are far higher than conventionally identified cases. So, the total numbers of infected and recovered persons are very difficult to estimate and the reported statistics are at a far distance from the actual. The numbers of infected, untested and recovered persons will vary from location to location in the country, and even within a state or city. Therefore, we cannot even assume a uniform or constant ratio of diagnosed to undiagnosed persons who were infected and have recovered, from estimates provided by an antibody survey in a part of the population. A generally applicable ‘correction factor’ is not possible to use, given such variability from place to place and over time. So, how much reliance should we place on the estimated R (even when measured as Rt) when the epidemic is advancing?
Also read: Covid-19 vaccine deals in the works with candidates, producers says Harsh Vardhan
The value of R will vary with the freedom the virus has to move in a population. It is higher in a crowded location-- especially an urban indoor one. It will be lower in a rural area where people are more mobile and meet in open areas. To derive a single value for a state, with a collage of urban and rural populations, is untenable. The added uncertainty, of the numbers infected but untested, makes it even more imprecise to draw conclusions on the changing rates of infectivity in any state just based on estimated R values.
The same problem arises with the ‘Doubling Time’, or ‘’Doubling Rate’, which has been popular on the social media. It has also been referred to by our officials to report progress. However, it must be noted that the US, where the epidemic has been raging, now has a doubling rate four times higher than India’s. That demonstrates the effect of the high cumulative base count. When the base count rises to a high number, its doubling time takes longer than when the count was low. Even if we consider only the time taken to add each increment of 100, 000 new cases, do we know how the numbers of infected but asymptomatic and untested persons changed in between? If the numbers of such persons are very high, as the antibody surveys suggest, how do we estimate the rates of increase in the actual numbers of infected persons? We will be painting a portrait looking only at changing shadows.
What then can we measure to track deaths? A comparison of daily case counts is subject to variations in the numbers and types of tests, as well as the criteria employed for testing persons. The uncertainty is compounded by the false positives (‘dead viruses’) and the false negatives (modest sensitivity of RT-PCR and even lower sensitivity of Rapid Antigen Tests). In any case, there is likely to be a big gap between the measured daily case count and total number of persons infected on that day if we go by the surveillance estimates.
We are left to depend on the one variable which is least subject to time-dependant variations in measurement methods. That is the ‘daily death count’ in a defined population. Is that count consistently falling? Or, how is a 7- day moving average of deaths behaving? Admittedly, there will be some undercounting of deaths. However, that proportion is unlikely to change over time. Even if the absolute numbers are short of the mark, time trends will be clearly discernible. We should try to improve the measurement of both in-hospital and out-of-hospital deaths, with clearly defined criteria for certification of death supplemented by ‘verbal autopsy’ techniques when needed. The cause of death assignment should not depend only on test results but also take into account the clinical history and any other investigations such as chest X-Rays or CT-scans. Even as we try to improve our measurement of Covid19 related deaths, the changes in daily death counts over time will be the best indicator for us to gauge whether the epidemic is waning in a city, district or a state. With deaths, we will not be questioning “what R we measuring?”
The author, a cardiologist and epidemiologist, is president, Public Health Foundation of India. He is the author Make Health in India: Reaching a Billion Plus. Views are personal.
