A groundbreaking study led by UC Berkeley Psychology Postdoctoral Research Fellow Aakash Chowkase showed the official number of COVID-19 deaths in Pune, India may be significantly underreported.
Published in Nature's Scientific Reports on May 6th, 2024, the research estimates that from March 2020 to December 2021, Pune experienced approximately 14,770 excess deaths, while only 9,093 were officially recorded as COVID-19-related. This discrepancy suggests that the true impact of the pandemic in the region could be around 60% higher than official figures indicate.
"Our results confirm that, like most other places, officially reported COVID-19 mortality in the Pune region was an underestimate,” Dr. Chowkase said. “We estimated an undercount factor of 1.6, meaning that the actual number of COVID-19-related deaths could be 1.6 times higher than reported. Even some of the world’s best healthcare systems saw undercount factors around 1.5, so Pune's performance is relatively comparable."
The findings have significant implications for public health policy in Pune and beyond, Dr. Chowkase said. He emphasized the need for preparedness. "Based on our findings, decision-makers in Pune and similar regions should prepare for a worst-case scenario of up to 2.5 times the reported number of pandemic-caused deaths in future health crises. This kind of planning is crucial for ensuring that healthcare systems are not overwhelmed."
“The study not only contributes to a better understanding of COVID-19's true impact in India but also demonstrates the potential of innovative and cost-effective research methods in contributing to achieving health equity in data-scarce settings,” Dr. Chowkase said. “By utilizing both traditional statistical techniques and community-based approaches, the research offers a comprehensive view of the pandemic's toll in Pune, and it provides valuable insights for future public health planning and crisis response,” he added.
The research used several advanced statistical models to estimate excess deaths, including simple averaging, the Farrington surveillance algorithm, and the Poisson model. These models are considered the gold standard in overcounts of deaths because they account for factors such as seasonality, population growth, and contemporary mortality trends, Dr. Chowkase said. However, they rely heavily on high-quality data, which was often lacking from available data sets.
To overcome data limitations, the research team also used a novel approach called the "wisdom of crowds.” This method involves surveying a diverse sample of Pune residents to estimate the number of COVID-19-related deaths.
"The 'wisdom of crowds' method provides a novel and frugal way to estimate critical public health data, especially in settings where traditional data collection is challenging,” Dr. Chowkase said. “It has been used across various fields, including epidemiology, business, and public policy because of its versatility and effectiveness."
The study was a collaborative effort involving over 40 volunteers and researchers from five countries: India, the United States, the United Kingdom, Germany, and Canada. The team included experts in epidemiology, public health, machine learning, psychology, statistics, engineering, data science, and education. Local grassroots organizations in Pune, such as Jnana Prabodhini’s Nagari Vasti Abhyas Gat and 'Swa'-Roopwardhinee, played a crucial role in distributing the surveys among low-income groups and urban slum residents. Dr. Chowkase co-led this initiative with Abhishek Dedhe of Carnegie Mellon University and Dr. Pranav Pandit of UC Davis as a part of JPF Analytics, which is a non-profit computational social science platform.
Funding for the study came from various sources, including Carnegie Mellon University and Johns Hopkins University. However, the majority of the work was volunteer-driven and unfunded.
