Lady Hardinge Medical College: A leader in StillBirth research

UNICEF estimates that ~1.9 million StillBirth(SB) occurred in 2023, with the highest incidence in sub-Saharan Africa and South Asia [1]. While SB rates have declined in India over the last two decades, it remains a significant public health challenge and the government has set a goal of achieving single-digit stillbirth rates [2, 3].

Large-scale data on causes of death in SB, understanding underlying risk factors, and designing preventive strategies play a key role in advancing progress toward this goal. In this context, Lady Hardinge Medical College (LHMC) has made significant contributions by collecting extensive data on the epidemiological profile of SB cases, assessing probable causes, and establishing baseline assessments [4, 5]. 

Verbal Autopsy: Role in SB Research and Challenges

When a direct diagnosis is not available, Verbal Autopsy (VA) is a commonly used technique to determine the cause of SB. Frontline staff and physicians conduct VA interviews, maintain handwritten notes and collect detailed information about the circumstances surrounding the SB from affected family members. They then analyze the data and identify the Cause of Death (COD). VA studies are widely used by researchers to identify risk factors and establish action plans to reduce stillbirth rates [6, 7, 8].

However, due to the manual processes, conducting VA studies at scale remains challenging. VA interviews are long and time-intensive. Given the low physician-patient ratios in India, it is not feasible for physicians to solely conduct VAs at scale. Since the interviews are subjective, the inferences are thus vulnerable to variability in interpretation by frontline staff and are difficult to standardize. Additionally, since the workflow relies on memory and handwritten notes, VA forms are prone to being lost or incomplete [9]. 

These challenges result in low completeness, interpretation inconsistencies and reduced data quality—ultimately, limiting the ability of running large scale VA studies to generate generalizable insights.

AI-powered VA Registry to accelerate SB Research

To address the challenges, clinicians from LHMC are collaborating with us to develop AI systems to automate VA processes that are currently manual. Oxalis Technologies will build a mobile application that frontline staff can use to record VA interviews and log images of any additional handwritten notes directly from the field. Foster’s AI system will process the ambient conversations, transcribe and translate them to English [10], extract relevant information from images and automatically generate answers for all the questions in the structured VA forms from the source files [11]. We will conduct a study to validate the tool and measure the performance gains. Since the source files are securely stored, any newly defined clinical variables can be retrospectively populated automatically by querying the database—eliminating the need for repeat interviews or manual data processing.

This novel initiative, led by LHMC, applies AI in VA studies to identify COD and is expected to enable large-scale data collection, empower researchers to rapidly test hypotheses and generate meaningful insights into SB causes, and accelerate the identification of risk factors and development of intervention strategies to reduce SB rates.