Mapping Maternal & Child Health Data Sources for Local Decision-Making in Tribal PHCs

Background

Despite heavy investment in digital health information systems, frontline facilities in resource-constrained settings remain paradoxically data-rich yet information-poor. Health information systems built with an upward gaze for accountability reporting often fail to support the decisions that medical officers, data entry operators, and ANMs need to make locally. This work, part of Arun Mitra’s PhD thesis on participatory data science in ITDA-Rampachodavaram, Andhra Pradesh, asks how the maternal and child health (MCH) data ecosystem at primary health centre (PHC) level is actually structured, used, and constrained. It is the foundational landscape paper of the broader participatory data science for MCH programme.

Approach

A qualitative descriptive study was conducted across three tribal PHCs (Boduluru, Gangavaram, Vadapalli) in Alluri Sitarama Raju District, Andhra Pradesh. The methods combined participatory data discovery, document review, key informant interviews, and observation of routine data practices. Data sources were classified using a dual framework: the HEALTHY classification for content domains and the Keller typology for data origin, while the ODI data ecosystem mapping approach guided analysis of actors, flows, and value exchanges across routine HMIS, community-generated, and programme data.

What we found

• Twenty-eight distinct MCH-relevant data sources were identified, roughly 66% digital and 34% paper-based.
• Cross-tabulation showed heavy concentration in the Healthcare × Administrative cell, with most wider social determinants of health largely absent.
• Data architecture exhibited a pronounced upward gaze: information flowed from community to district with minimal feedback returning to frontline workers.
• Health workers maintained extensive paper-based parallel systems, WhatsApp networks, and personal tracking spreadsheets to compensate for portal complexity, connectivity failures, and authentication barriers, recovering the real-time, individual-level, locally actionable access that formal systems lacked.

Outputs & impact

The study reframes the data-rich, information-poor paradox as an architectural mismatch rather than data scarcity, and the emergence of compensatory information practices as evidence of system design failure rather than worker deficiency. It argues that context-blind digitalization can actively harm health system functioning, and proposes a shift from accountability infrastructure toward bidirectional decision-support ecosystems that integrate informal-system learning. The work is available as a medRxiv preprint (Mitra, Jayaraman, Ondopu, Malisetty, Niranjan, Shaik, Soman, Gaitonde, Bhatnagar, Niehaus, K.S, Roy; posted March 2026) and includes a reusable data source summary table and classification framework for MCH data integration at PHC level.