What Makes a Healthcare Worker Tick?

The Challenge

Uttar Pradesh, India’s most populous state, has some of the country’s worst maternal and child health outcomes, including some the highest rates of infant and maternal mortality. How can community healthcare workers supported by the Indian government improve health and serve their communities more effectively?

Our Approach

Community healthcare workers known as ASHAs serve as links between the formal health system and communities in India. ASHAs in Uttar Pradesh had been operating well below their potential – only performing some of the activities asked of them. We wanted to understand what these workers were doing, what drove their behavior, and how we could better support them to improve health outcomes in their communities.

Key Results

• We applied our CUBES framework to discover that one-size-fits-all solutions to improving performance won’t work: ASHAs with similar job performances may be motivated by very different drivers.
• We used machine learning to group ASHAs into five segments, based on their actions and their contextual and perceptual drivers of behavior.
• We found that different segments require different interventions to target these behavioral drivers. For example, the ‘undertrained but motivated’ ASHA who works hard but lacks clinical knowledge might need better training, while the ‘financially troubled and hands-off’ ASHA needs support claiming her incentive payments.

Interventions to improve the effectiveness of ASHAs (Accredited Social Health Activists) in Uttar Pradesh have been a major focus of programs and policy. However, to create lasting change, we needed to better understand why ASHAs behave as they do, and why they do not perform all their tasks with each woman they visit.

We used our behavior framework, CUBES, to map all the potential factors driving community healthcare workers’ behaviors, including contextual factors (such as available infrastructure or supervisory support) and factors related to their own perceptions (such as self-efficacy, or incorrect beliefs about antenatal care). Multiple narratives emerged, highlighting the unique characteristics and needs of different ASHAs. 

We then undertook one of the largest studies on front-line workers conducted to date – speaking to 1,500 ASHAs across the state of Uttar Pradesh, and 5,000 mothers served directly by them, to understand and quantify all of the factors influencing their behavior. 

Our data confirmed the heterogeneity we had seen in our qualitative work with CUBES: ASHAs differed in the number of hours they worked, what motivated them, and the types of support they received from their supervisors. Armed with this information, we used machine learning algorithms to divide the ASHAs into subgroups based on their job performance and the underlying factors driving their behaviors. We identified distinct segments of ASHAs – for example, some were independent, skilled and capable of high performance even without supervisory support. Others believed in the importance of their role but lacked up-to-date knowledge and planned their time poorly. In another segment, ASHAs were knowledgeable and well supported, but so worried by financial concerns that they struggled to put in the time needed for their work.

These findings open many possibilities for governments and large-scale health programs. For example, a simple questionnaire could help ASHAs’ supervisors identify which segment each ASHA belongs to. They can then provide the right kind of support and connect the workers with specialized training and skills-building sessions. Technology-enabled job aids could be created to provide tailored motivational nudges and support based on the individual ASHA’s profile. We are conducting a similar study in Madhya Pradesh to see how our findings compare across states.