As the COVID-19 pandemic advances into its third year with varied intensity, the world waits with bated breath to learn about the impact of the Omicron variant . Fortunately, South Africa has established an efficient rapid mortality surveillance system which would alert us with the earliest evidence of such impact, if any. It is also reassuring that lessons from previous pandemic waves in India have resulted in an early scrambling of critical response preparedness activities. Nevertheless, despite pronouncements of ICU bed availability and oxygen supply by several States combined with personal protection through vaccination, the threat of individual susceptibility to the new variant from immune escape and underlying morbidity remains. Hence, any proof of lethality (or lack thereof) of this variant from South Africa could only serve as background information, underscoring the need for similar early warning mortality surveillance activities in India.
No reliable estimate
Mortality data are essential for pandemic management from both clinical and public health perspectives to guide patient care, protection of vulnerable population groups, and resource mobilisation and allocation. In all countries, data from COVID-19 mortality surveillance systems are limited on account of inadequate access to ante-mortem testing and inconsistent cause attribution. Hence, pandemic impact is now evaluated through estimation of excess mortality (additional deaths observed during the pandemic period in comparison with those recorded in previous calendar years). For adequate response, such evaluation needs reliable baseline pre-pandemic mortality measures, near-complete death reporting during the pandemic, and real-time data compilation and release at weekly or monthly intervals. For South Africa, the National Population Register serves as a useful data source, with weekly updates of deaths by sex, age, date of death, and place of registration. Shortcomings in completeness and timeliness of reporting exist, but the mortality surveillance team has established procedures to correct for such biases to report excess mortality estimates at the national and provincial level in near to real-time.
For India, excess mortality estimates are based on epidemiological models, some of which include analyses of month-wise death registration data released by several States. However, these estimation exercises were hampered by uncertainty in baseline pre-pandemic mortality levels, potential under-registration in 2021, and the likelihood for the data to include some delayed registrations. All these likely data biases were statistically accounted for by different analysts through varying methods and assumptions, resulting in national excess mortality estimates ranging from 2.8 million to 5.2 million deaths between April 2020 and June 2021. On the other hand, the national COVID-19 surveillance programme death toll for this period was about 4,30,000, considered to be a gross under-count. This 10-fold variation in estimates of COVID-19 mortality in India is clearly not helpful, and the absence of a reliable estimate constitutes a major impediment to our understanding of the magnitude of pandemic mortality, both at the national and global level.
Encouraging trends
Nevertheless, several recent developments have created prospects for improved estimation of excess mortality in 2020/2021, as well as for ongoing mortality surveillance and measurement programmes in India. First, the Civil Registration System (CRS) Report for 2019 indicates high levels of registration completeness across India. These data have now been corrected for under-reporting to compute reliable pre-pandemic mortality estimates by sex, age and location, as a baseline for evaluating pandemic impact. Next, the prompt release of information on registered deaths by some States in 2021 (summarised in COVID-19 mortality data reports by The Hindu ) indicates that efficient mechanisms for data compilation are functional in these States. Indeed, the Registrar General of India (RGI) had issued a circular in November 2017 which requires District and State Registrars to submit summary monthly returns of births and deaths registered within their jurisdiction. The circular also declares that “the vital rates generated through the CRS are exact and real data certified by registering authority, and therefore legally admissible, and prevail over any other estimates that may be done for proxy purposes due to the lack of data”. It is likely that the prompt release of registration data following the second wave would have been facilitated by these instructions. Third, the pandemic has evoked considerable media and public attention to epidemiological data including mortality, which led to the flurry of data reports to meet this demand. There is now a general expectation of continued public attention to such health data in times to come.
The road ahead
While the quick data release in 2021 augurs well for future availability, CRS data for 2020 and provisional data reports for 2021 should be released at the earliest, at least to an internal team of analysts. Local statistical capacity must also be established at State registration offices for data quality evaluation, adjustments for data bias, and basic trend and forecast analysis. Data dissemination protocols need to be standardised using a template for compilations of deaths tabulated by sex, broad age groups, month of death occurrence, and district of usual residence. The CRS death reporting forms include these variables, hence enabling such detailed tabulations each month. A lag period in death reporting across registration units and districts is anticipated, but updated counts in subsequent months are widely accepted in regard to pandemic surveillance. Information on deaths recorded in the national Sample Registration System and other household surveys could be used to estimate completeness of CRS data using record linkage methods. It should be feasible for State Registrars to direct such operations, with resources and technical support from local public health/academic institutions. Of course, subsequent detailed analysis of annual data complying with the standard data validation protocols implemented by the RGI would still be necessary for the annual CRS report. But the availability of provisional near real-time surveillance reports would enhance monitoring of the mortality impact of this and future epidemics.
In principle, mortality estimates developed from adjustments to empirical data are by design locally relevant and therefore more acceptable than modelled estimates. Even in routine times, there has been an undue dependence on epidemiological models to estimate mortality for India, as observed from various recent disease burden estimations. Analysis of recent CRS data has exposed the brittle nature of the outputs from these models. Moreover, the identified gaps in CRS data by location, sex and age from detailed analyses can help guide interventions to improve data quality for the future. In the current environment, establishing reliable, real-time mortality surveillance is an essential element of pandemic preparedness, and urgent steps to build on recent developments in this aspect would advance the cause of evidence-based health policy to deal with the pandemic in India.
Chalapati Rao is Honorary Associate Professor, Research School of Population Health, Australian National University
