The challenges in counting the dead

India must increase investment in innovations for improving data collection and reporting

Workers in protective suits prepare to cremate the body of a person who died of COVID-19 at Nigam Bodh Ghat in New Delhi on May 29, 2020. FIle

Compared to other diseases, COVID-19 is highly transmissible. The long-term sequelae of infection are unravelling only now. The only goal is to prevent as many deaths as possible. The proportion of deaths due to COVID-19 (about 2%) is lower than that of SARS (about 9%) or the Middle East Respiratory Syndrome (34.4%) and probably higher than that of influenza (0.1%). However, the task of counting deaths attributed to COVID-19 is full of complex challenges.

Assessing fatalities

Globally, the disruption of health services during a pandemic results in the underestimation of any health indicator, including the assessment of fatalities. For estimating death rates, the total number of deaths are usually in the numerator; it is choosing the denominator that is the difficult aspect. The case fatality proportion uses reported deaths in the numerator and reported cases in the denominator; this helps in assessing and comparing clinical severity and effectiveness of clinical outcomes. Delayed reporting of deaths from hospitals, and incomplete and inaccurate reporting of COVID-19-related deaths will affect the numerator. It is recognised that case fatality will not reflect the deaths from COVID-19 as it varies with test rates and strategies used.

Comment | COVID-19 deaths may be higher than reported

The infection fatality rate (IFR) due to COVID-19 uses the same numerator. The denominator comprises the total number of infections, derived from seroprevalence studies in the total population. IFR provides real estimates of COVID-19 death rates, including the presence of unreported infections in the denominator. However, differences in sampling strategy and methods of seroprevalence studies, demographics and healthcare resources will result in incomparable seroprevalence estimates. As we witnessed in India, testing, provision of healthcare and treatment improve progressively. This too will change both the numerator and denominator over time.

In addition, we can estimate the crude mortality rate by having the number of people in the area in the denominator. Generally expressed as deaths per million, this provides the probability that any individual in the population will die from the disease. This too is affected by incomplete and inaccurate reporting of COVID-19 deaths.

Finally, all-cause excess mortality can be estimated by calculating the number of deaths above expected baseline levels, regardless of the reported cause of death, and provides an overall impact of the COVID-19 pandemic.

Data | Is India undercounting its COVID-19 deaths?

The seroprevalence surveys are useful to obtain an unbiased estimate of deaths due to COVID-19. Global evidence indicates that the IFR is less than 1% of total infections. We can estimate this only when we have realistic data on total infections in the country, obtained from repeated nationwide seroprevalence surveys.

In the absence of IFR, deaths per million is an alternative estimate. Table-1 shows the deaths per million in the capital cities of the nations that have more than 1,00,000 confirmed cases. The deaths per million population roughly correlates with a higher proportion of persons above 65 years. The exceptions are Peru and Brazil, which are known to take delayed action. An honourable exception is Germany, which has the best preventive and curative services compared to the rest of the countries.

Registering deaths

It is important to get baseline estimates of deaths to estimate the excess mortality. Currently, although mandatory, only about 86% of the deaths get registered, which has increased from 66.9% in 2009, indicating that 14% do not get registered in India independent of the COVID-19 crisis. The annual crude death rate in India is 0.73 per 100 persons. The deaths attributed to COVID-19 in 2020 are 0.028 per 100 persons, which varies from 0.21 in Delhi followed by Maharashtra, Tamil Nadu, Gujarat, Puducherry and Karnataka (Table-2). This is mostly a function of better death reporting, with three of these States having 100% death registration and two States registering more than 90% of deaths. The registration system has further challenges with a high proportion of unclassifiable deaths, long delays and irregular publication of statistics. In addition, lack of training, resources and systematic screening results in making medical ascertainment of every death far from practical in India. Once efforts are made to register every death, the government should invest in establishing the cause of death, linking the data to electronic health records and coding the data.

The deaths attributed to COVID-19 in all the South Asian countries are comparable and lower than the developed countries. We will know the actual number of deaths when all the deaths are compiled by the Civil Registration System, perhaps more than a year from now. Once known, attributing the cause of death, although difficult, can provide delineation of causes and excess mortality due to COVID-19 in India. As per the ICMR-IHME-PHFI’s India disease burden study, non-communicable diseases-related deaths caused by cancer, diabetes, cirrhosis, etc. constitute 72.5% of deaths, while 27.5% of deaths are caused by HIV/AIDS, injury, suicide, etc (Table-2).

Table-2: Projected deaths for 2020 and its causes including COVID-19

State	Deaths in 2018	population	Projected Deaths for 2020	No of deaths due to IDs	Number of Deaths due to NCDs	Deaths due to COVID-19	Proportion of Deaths due to COVID-19*
Andaman & Nicobar Islands	2237	399000	2913	801	2112	21	0.72
Andhra Pradesh	375777	52504000	383279	105402	277877	2203	0.57
Arunachala Pradesh	3860	1519000	11089	3049	8039	3	0.03
Assam	142605	34668000	253076	69596	183480	155	0.06
Bihar	213989	121302000	885505	243514	641991	413	0.05
Chandigarh	23330	1193000	8709	2395	6314	26	0.30
Chattisgarh	177549	29109000	212496	58436	154059	104	0.05
Dadar nagar haveli	2174	1018000	7431	2044	5388	2	0.03
Delhi	145533	20193000	147409	40537	106871	4139	2.81
Goa	13072	1549000	11308	3110	8198	86	0.76
Gujarat	433256	68862000	502693	138240	364452	2695	0.54
Haryana	185842	29077000	212262	58372	153890	500	0.24
Himachal Pradesh	41833	7347000	53633	14749	38884	18	0.03
Jammu and Kashmir	39410	13305000	97127	26710	70417	490	0.50
Jharkhand	102729	37937000	276940	76159	200782	192	0.07
Karnataka	483511	66322000	484151	133141	351009	3398	0.70
Kerala	258530	35307000	257741	70879	186862	120	0.05
ladhak		295000	2154	592	1561	9	0.42
Madhya Pradesh	424257	83374000	608630	167373	441257	1033	0.17
Maharashtra	667900	123295000	900054	247515	652539	18306	2.03
Manipur	4476	3134000	22878	6292	16587	12	0.05
Meghalaya	14779	3256000	23769	6536	17232	6	0.03
Mizoram	5525	1204000	8789	2417	6372		0.00
Nagaland	828	2171000	15848	4358	11490	8	0.05
Odissa	328799	43852000	320120	88033	232087	296	0.09
Puducherry	12839	1537000	11220	3086	8135	91	0.81
Punjab	213234	30099000	219723	60424	159299	636	0.29
Rajasthan	443173	78273000	571393	157133	414260	811	0.14
Sikkim	3386	670000	4891	1345	3546	1	0.02
Tamil nadu	574006	76049000	555158	152668	402489	5159	0.93
Telangana	136528	37473000	273553	75227	198326	654	0.24
Tripura	29080	4032000	29434	8094	21339	43	0.15
Uttar pradesh	906653	227943000	1663984	457596	1206388	2176	0.13
Uttarkhand	47894	11270000	82271	22625	59646	136	0.17
West Bengal	490530	97516000	711867	195763	516103	2149	0.30
Total	6950607	1347054000	9833494	2704211	7129283	46091	0.47

*Source: World Bank

Deaths cannot be hidden in the case of COVID-19; sooner or later, they get reported due to a vigilant media and an active civil society. Due to such efforts, at least five cities in India have reconciled the death numbers. To obtain unbiased, reliable estimates, nationally representative verbal autopsy surveys such as the Million Death Study can be helpful. In the meantime, India has to adopt a strategy of transparency in data flow and reporting, with increased investments in innovations for improving data collection and reporting.

Data |  Tracking death rates helps gauge India's COVID-19 response better

Due to these inherent challenges, India and other countries with low mortality are subject to increased, and welcome, scrutiny. However, commentators should not be tempted to use estimates beyond the range that can be inferred by evidence. The assumptions about undercounting deaths should be realistic. With no supporting evidence such as a surge in numbers from burial grounds or in the neighbourhood, it is counter-intuitive to assume that we have under-counted thousands of deaths. Alas, counting the deaths is a deadly challenge out there.

Giridhara R. Babu is a Professor of Epidemiology at the Indian Institute of Public Health, PHFI, Bengaluru. The author acknowledges Deepa R., Daisy Solomon of IIPH Bengaluru and Jeevan Raksha for help in data compilation and proof reading