When the hospital closes, the heart stops: What's behind India's 'excess' COVID deaths?

A landmark study, “The mortality burden from COVID in low-income settings: evidence from verbal autopsies in India”, using verbal autopsies of 20,000 deaths reveals that only one-third of India’s pandemic excess mortality was directly caused by SARS-CoV-2. The rest — a hidden toll running into millions — was the collateral damage of a healthcare system brought to its knees. 

When India’s government reported 533,847 COVID-19 deaths as its final pandemic toll, public health experts across the world knew the number told only a fraction of the story. What no one had been able to do, until now, was ask the harder question: if the official count was a vast underestimate, what was actually killing people — the virus, or the collapse of everything around it?

The study by Anup Malani, Lee and Brena Freeman Professor of Law at the University of Chicago, published in BMC Public Health, offers the most granular answer yet. Drawing on WHO-compliant verbal autopsy surveys of roughly 20,000 deaths across India between 2018 and 2021 — conversations with the families of the dead, asking what their loved ones died of — Malani’s findings reframe the pandemic’s death toll in ways that carry profound implications for public health policy, healthcare infrastructure, and pandemic preparedness.

The central finding is both striking and sobering: India’s overall death rate rose by 81 percent during the pandemic. But SARS-CoV-2 was directly responsible for only 33 percent of those excess deaths. Another 23 percent were attributable to cardiovascular disease. In other words, for every three people who died because the virus found them, nearly two more died because the health system could not reach them — or they could not reach it.

The Two Tolls of a Pandemic

The official figure of roughly 534,000 COVID deaths has long been contested. Estimates of India’s actual pandemic mortality have ranged from 3 to 6 million excess deaths, depending on the methodology. The WHO’s own 2022 estimate placed India’s toll at 4.7 million — nearly ten times the official count — a figure the Indian government rejected and sought to suppress.

What has been missing in this debate is not just the number of deaths, but their cause. Excess mortality statistics — calculated by comparing total deaths during the pandemic to what would have been expected in normal years — can tell us that more people died. They cannot tell us why.

Malani’s verbal autopsy approach fills that gap. Using the Consumer Pyramids Household Survey (CPHS) conducted by the Centre for Monitoring the Indian Economy (CMIE) as his sampling frame, and working with physician panels coordinated by the Centre for Global Health Research under Prabhat Jha, Malani obtained cause-of-death data encoded in ICD-10 codes for a nationally representative sample. The result is the only large-scale, population-representative dataset that can attribute cause to pandemic deaths in India — rather than merely counting them.

The findings show that SARS-CoV-2 deaths spiked twice: in June 2020, immediately after India’s first lockdown, when the virus accounted for 23.3 percent of all deaths; and again in May 2021, in the devastating second wave driven by the Delta variant, when it was responsible for 35.8 percent of all deaths. These spikes alone confirm that the official toll was a severe undercount.

But the more revelatory finding concerns what happened in between, and around, those spikes: a sustained and deadly surge in cardiovascular deaths that mirrors, almost perfectly, the collapse of healthcare access.

The Cardiovascular Shadow: When the System Stopped Working

India’s second wave, peaking in April–May 2021, has been seared into public memory: overflowing crematoriums, oxygen tankers being escorted by police, families pleading on social media for ICU beds, hospitals turning away patients gasping for breath. The Lancet described it as a healthcare system “overwhelmed.” Delhi’s Chief Minister called it “a state of distress.” It was, in effect, a collapse.

Malani’s data captures, in mortality terms, what that collapse cost. Cardiovascular disease accounted for 23 percent of excess deaths across the pandemic period — a figure that would not be visible in any excess mortality estimate based purely on registered deaths, since death certificates rarely capture the full causal chain. A person who died of a heart attack because they could not get to a hospital, or whose hypertension went unmanaged because their clinic was shuttered, or whose cardiac procedure was postponed because every hospital bed was occupied by COVID patients — that person does not appear in any official COVID death count.

The Malani study makes them visible. In doing so, it adds a new dimension to the concept of indirect pandemic mortality — deaths that would not have occurred had the pandemic not disrupted the provision of care. This matters enormously for policy. If two-thirds of India’s excess deaths were not directly from SARS-CoV-2, then the pandemic’s death toll was not primarily a story of viral pathology. It was, at least as much, a story of healthcare fragility.

A Nation Divided: The Interstate Picture

Perhaps the most politically explosive dimension of India’s pandemic mortality is the vast gulf between states in how honestly, and how accurately, deaths were reported. The 2025 release of India’s Civil Registration System (CRS) data for 2021 — buried, notably, beneath the noise of a military confrontation with Pakistan and barely covered by national media — revealed staggering interstate disparities.

Gujarat stands as the most glaring case. The state reported just 5,812 official COVID-19 deaths in 2021. The CRS recorded a total of 1.95 lakh (195,000) registered deaths that year — an increase of approximately 1,95,406 over the previous year. Research published in the Journal of Global Health calculated the state’s excess-to-official-COVID-death ratio at over 40:1. Gujarat’s reported COVID mortality, in other words, was not an undercount. It was, by any measure, a fabrication.

Madhya Pradesh and West Bengal follow a similar pattern. Madhya Pradesh reported 6,927 COVID deaths in 2021; CRS data indicates approximately 1,26,774 excess deaths. West Bengal officially counted 10,052 deaths; CRS points to over 1,52,000 additional deaths beyond baseline. Bihar, Rajasthan, Jharkhand, and Andhra Pradesh have all been identified as states with severe underreporting.

Kerala offers a striking contrast. The state reported 44,721 official COVID deaths in 2021. CRS data suggests approximately 66,655 excess deaths during the same period — a ratio of roughly 1.5:1. While that still represents undercounting, it is of a categorically different order. Researchers and public health experts have repeatedly pointed to Kerala’s functional civil registration system, higher rates of medical certification of deaths, and a more robust public health infrastructure as the reasons for its comparatively honest accounting.

The Salvatore et al. study published in 2025, using the CRS data as its foundation, estimated that India experienced approximately 2.4 million excess deaths in 2021 alone — a 7.2-fold undercount compared to the official COVID toll for that year. This figure sits within the range that Malani’s work would predict: if roughly a third of excess deaths were directly SARS-CoV-2 attributable, the true virus-caused toll would still be two to three times the official number, with the remainder attributable to cardiovascular disease and other conditions exacerbated or left untreated during the healthcare crisis.

Why This Study Matters: Method as Message

Malani’s study is as important for its methodology as for its findings. The verbal autopsy — a structured interview with the next of kin of a deceased person, using a standardised WHO instrument — is not a new tool. But it has rarely been deployed at this scale, with this degree of representativeness, and in direct response to a public health crisis of this magnitude.

India’s Sample Registration System (SRS) conducts verbal autopsies as part of its regular work. But as Malani notes, the SRS data for 2020–2021 had not been released at the time of this study — a gap that speaks volumes about the pace, and perhaps the willingness, of official data collection. In its absence, Malani’s work using the CMIE’s Consumer Pyramids survey fills a function that the state itself should have performed.

The implications extend beyond India. Low- and middle-income countries around the world face the same structural problem: death registries that are incomplete, cause-of-death certification that is rare or unreliable, and official COVID counts that reflect testing capacity and political will as much as epidemiological reality. Malani explicitly frames his study as a proof-of-concept for “ex post verbal autopsies as a method of measuring the burden of the pandemic in lower-income countries.” The message to global health surveillance systems is clear: we need better tools, and we need them deployed during — not four years after — the next crisis.

The Policy Reckoning India Still Owes Itself

The numbers in Malani’s study do not merely correct a historical record. They constitute an indictment of chronic underinvestment in India’s health infrastructure — and a warning about what the next pandemic will cost if nothing changes.

When nearly a quarter of excess pandemic deaths are cardiovascular in origin, it tells us that India’s primary health centres were unable to manage chronic disease during the crisis, that the consolidation of all healthcare resources toward COVID care left people with other conditions without treatment, and that the digital and physical distance between patients and providers — especially in rural areas, where over 70 percent of India’s annual deaths occur without medical attention — became, in a pandemic, a death sentence.

It also raises a question that the interstate data makes inescapable: why were some states able to count their dead honestly, while others could not — or would not? The gulf between Kerala and Gujarat is not simply a matter of data infrastructure. It is a matter of political accountability, the willingness of state governments to be measured, and the consequences they face when the numbers are bad.

Malani’s work cannot answer that question. But it sharpens it. By establishing, with the most rigorous methodology yet applied to this problem in India, that the pandemic killed not just through the virus but through the system’s failure to function, it places the burden of response squarely on those responsible for that system. The dead, it turns out, have been waiting four years for someone to ask how they died.