AI to Predict Preterm Births in India

The power of predictive AI models is increasingly being harnessed to make more effective healthcare solutions (Ananf Ibn Shahibul, Wikimedia Commons).

The Government of India is training AI to predict preterm births and improve clinical diagnostic tools to combat neonatal mortality. 

The Department of Biotechnology announced the Interdisciplinary Group for Advanced Research on Birth Outcomes, known as GARBH-INi, on March 25 with the goal of creating state-of-the-art biotherapeutic, diagnostic, and risk assessment tools, according to GK Today. Healthcare IT News explains that the government has collected millions of biomedical and ultrasound samples from over 12,000 pregnant women to train AI models to recognize local risk factors and predict preterm births (defined as births earlier than 37 weeks) in advance, alerting medical staff to perform emergency C-sections or prepare to treat infectious diseases. Although the government has pursued similar AI technologies for cancer and life sciences research, this initiative is its largest yet, according to GK Today.

Dr. Joel de Lara, an Assistant Teaching Professor at the Kennedy Institute of Ethics at Georgetown University, told The Caravel, “Training an AI system on data from 12,000 Indian women could, in principle, help ‘localize’ predictions by capturing population-specific patterns that might be missed in global datasets. Many existing clinical models are trained on data from Western populations, which can limit their accuracy elsewhere. The reason is that there are region-specific biological, environmental, and social factors to consider (e.g., nutrition, access to care, prevalence of certain conditions--like low birth weight), and such a system could potentially improve risk prediction and allow clinicians to intervene earlier or tailor care more appropriately.”

Of the 15 million preterm births a year, India accounts for over 24 percent of the global burden, according to a study published by Kursheed et al in the Indian Journal of Obstetrics and Gynecology Research. According to The Times of India, preterm babies have low birth weight, reduced lung and vital organ capacity, and a high susceptibility to infectious diseases. As a result, over one million die every year. If they survive, they are at risk for long-term feeding difficulties, malnutrition and stunting, diabetes, and disabilities, such as cerebral palsy, per Skolnik et al.

In India, the most common causes of preterm death are intrauterine hypoxia, a complication involving the loss of the ability to breathe while in the womb, and congenital infections, which are diseases transmitted from mother to child. A study in The Lancet wrote, “In contrast to high-resource settings, where almost all preterm newborn babies survive, prematurity remains a leading cause of neonatal death globally. Furthermore, in low-resource settings, complications of prematurity are often listed as the cause of death, with little understanding of the factors that could inform programmes to reduce these deaths.” Therefore, GARBH-INi is a critical opportunity for India to better understand the causes of preterm deaths and learn to prevent them. 

In addition to predicting preterm births, GARBH-INi also uploads all the samples to a national database for further clinical research, according to Healthcare IT News. Alongside the urgent need for biomedical innovation, there are ethical considerations to ensure these new tools are both effective and safe. Dr. De Lara reflected, “It is necessary to establish not just that [the participants] understand the prospective benefits/risks, etc., of such interventions but that they understand how their data is being used, how it may be used in future systems, and potential risks there, too. Even if data is de-identified, health data can sometimes be re-identified, particularly as datasets are linked or reused. In future clinical uses, privacy concerns may arise if AI tools are integrated into digital health systems that store, transmit, or share patient data across institutions.”

The GARBH-INi is a vital next step in catching preterm complications early and addressing high neonatal deaths in India. However, for the GARBH-INi to make a difference in neonate mortality, these AI tools must be both representative of and available to low-income, rural, and marginalized women in India. Dr. De Lara noted, “From my perspective, what is more important are questions about benefit-sharing--will the communities contributing data actually benefit from the resulting technology? To ensure that, it is necessary to ensure that the dataset is sufficiently representative of the diversity of women and birthing people in India. There is immense diversity in India (regional, socioeconomic, caste, etc.), and a dataset of 12,000 may still underrepresent the most vulnerable populations.” To make the most of powerful new technologies under GARBH-INi, targeting the most vulnerable and ensuring safe deliveries of their babies should be priority number one.