Using Causal AI to Help More Women Deliver their Babies in Hospitals

The Challenge

In rural Uttar Pradesh, India, a state with more than 230 million people, 10 times more mothers and babies die during or after birth than in the US. Around 80% of women there deliver their babies in hospitals. We wanted to find out why the remaining 20% were still delivering at home, and how we might be able to change that.

Our Approach

Using causal artificial intelligence, we were able to narrow down potential interventions to address the problem with far greater accuracy than we would have with a more traditional approach.

Key Results

• Instead of 16 predictors, we found two causal pathways to mothers delivering their babies at home:
  1. They held a very specific belief that delivering at home is safer for the mother and the baby
  2. They lacked a delivery plan – such as arranging transport to get to the hospital
• To save lives, policymakers and community leaders need to expand transportation access, shift beliefs about facility safety, and help women plan for their deliveries.

The state of Uttar Pradesh in India has one of the highest child and mother mortality rates in the country. To save more lives, we need to get more women to give birth in facilities. Yet despite immense and expensive efforts by the government and partners, 20% of women in Uttar Pradesh still deliver their babies at home.

To save these mothers and infants, we turned to causal artificial intelligence to deliver the insights about why that last one-fifth of mothers are still delivering at home. 

What we found altered many of our previous notions. While causal AI confirmed some things we already knew -- such as that education was critical in leading women to deliver in facilities -- it also provided new answers to key questions around where to invest. For example, the government in Uttar Pradesh was asking a multimillion-dollar question: Should they focus on building more facilities or expanding access to transportation, such as ambulances to deliver mothers to the hospital? 

Contrary to previous convictions and to results from predictive AI models, our causal model revealed that distance from hospitals played no role in where someone delivered their baby. What was absolutely critical, however, was the mother’s ability to find transport and get to a hospital -- no matter how close they lived to the clinic.

Furthermore, many believed that those who gave birth at home were doing so largely because it was tradition. Our model uprooted this belief and showed two key reasons women delivered at home:

1. They held a very specific belief that delivering at home is safer for the mother and the baby

2. They lacked a delivery plan – such as arranging transport to get to the hospital.

Our results have huge implications for the government in Uttar Pradesh – from resource allocation to smarter messaging by community health workers. To save lives, we need to expand transportation access, shift beliefs about facility safety, and help women plan for their deliveries.