# Identifying Gaps in Mobile Data Collection by Frontline Health Workers in Bangladesh

## TL;DR

This study examines data gaps in mobile health surveys by frontline workers in Bangladesh, finding issues like missing IDs and technical problems affecting data quality.

## Contribution

The study identifies specific data gaps and their root causes in mHealth data collection by frontline workers in a low-income setting.

## Key findings

- National ID and birth ID data were missing in 84% of cases, while phone numbers were missing in 77%.
- Operational barriers like poor connectivity and syncing errors contributed to incomplete data submission.
- Social hesitancy and recall bias were key factors in missing socio-demographic and economic data.

## Abstract

Background: Mobile health (mHealth) tools are replacing paper-based surveys for frontline health workers, promising speed and cost-effectiveness. Yet, in low- and middle-income settings, there is scope for research on the accuracy of the information captured.

Objectives: To assess the quality of socio-demographic and economic data collected through the mHealth platform by BRAC (the largest non-profit in Bangladesh) Shasthya Kormis or SKs (frontline health workers), and to identify reasons for data gaps.

Methods: A mixed-methods study (2021) analyzed secondary mHealth records for 388 households drawn via two-stage cluster sampling from the catchment areas of 30 randomly selected SKs working across 61 districts. Descriptive statistics in R quantified missing values and irregular entries in household registration, visits, and member forms. Complementary insights were obtained from 24 in-depth interviews with SKs; transcripts were thematically coded using an iteratively refined codebook.

Findings: Core demographic variables were largely complete, but considerable gaps persisted: national ID/birth ID (84% missing), phone numbers (77%), household assets (39–70%), and land-size data. Several explanations were deduced: reluctance of community members to share sensitive information, sometimes to secure social benefits; recall or estimation difficulties for ages and land measurements; and operational barriers: poor connectivity, offline notetaking, and syncing errors that deterred submission in a timely manner.

Conclusions: While mHealth simplifies nationwide community data collection in this case, data quality is affected by social hesitancy, recall bias, and technical issues.