Research map / Large language models / AI and society
Deploying ADVISER: Impact and Lessons from Using Artificial Intelligence for Child Vaccination Uptake in Nigeria
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's full text. Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Describes the field deployment of ADVISER, an ILP-based AI framework that allocates heterogeneous vaccination interventions (phone calls, pickup, vouchers, drives) to mothers in Oyo, Nigeria, with cost-reducing approximations and lessons learned.
Problem
Low childhood vaccination uptake in Nigeria drives many child deaths; limited resources (health workers, vehicles, budget) make optimal allocation of heterogeneous vaccination interventions hard to deploy in practice under cost and data constraints.
Approach
Deploys the ADVISER ILP framework (learned per-intervention success probabilities, guided local search for vehicle routes, greedy pruning, branch-and-bound) on cloud infrastructure, using a 1km grid travel-time matrix approximation to cut API costs, and reports real-world uptake gains and post-deployment survey feedback.
Key results
- ADVISER achieved 73.9% cumulative vaccination success (Jan-June 2023) versus 43.6% for the business-as-usual phone-call baseline (Aug-Dec 2022).
- In post-deployment surveys, 93.63% of mothers agreed vaccine drives were critical, 83.92% agreed travel vouchers were critical, and 85.71% agreed pickup service was critical.
- Grid-based travel-time approximation avoided over 7 million distance queries that would have cost more than USD 3000 in Google Maps API charges during testing.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| Use ADVISER's optimized intervention allocation instead of business-as-usual phone-call reminders | Vaccination uptake in Oyo · Oyo deployment, Jan--June 2023 vs Aug--Dec 2022 | Cumulative Success: 43.6% -> 73.9% | Helped |
| Add targeted additional phone calls | Vaccination interventions · Post-deployment survey, Additional Phone Calls | Mothers agreeing intervention was helpful: 72.9% | Helped |
Surprising findings
- Phone-call reminders were less often judged critical (72.9%) than other interventions, and the deployment's real costs of vaccine drives, pickup service and phone calls were substantially higher than the model assumed.
Limitations
- No randomized controlled trial; baseline compared on a different time period (Aug-Dec 2022 vs Jan-June 2023).
- Cold-start problem for newly designed interventions lacking historical data; success probabilities estimated from a survey.
- Intervention costs rose substantially during deployment (labor, fuel, stipends), straining non-profit resources.
Experimental setup
Datasets: vaccination tracker database (over 90,000 registered mothers) community survey of 1,020 mothers post-deployment survey
Benchmarks: field deployment in Oyo, Nigeria (13,783 mothers, 99 health centers)
Metrics: cumulative vaccination success rate post-deployment survey agreement rate on intervention critic
Baselines: business-as-usual phone call reminders
Builds on: ADVISER
Closest papers by meaning
- Promoting Fair Vaccination Strategies through Influence Maximization: A Case Study on COVID-19 Spread (AAAI 2024 · similarity 0.79)
- Statistically Valid Post-Deployment Monitoring Should Be Standard for AI-Based Digital Health (NeurIPS 2025 · similarity 0.78)
- Evaluating Index-based Treatment Allocation in Underresourced Communities (AAAI 2025 · similarity 0.78)
- A Decision-Language Model (DLM) for Dynamic Restless Multi-Armed Bandit Tasks in Public Health (NeurIPS 2024 · similarity 0.78)
- Bridging the AI Gap: Evaluating the Impact of an AI Education Program for Caregivers on Parental Leave (AAAI 2025 · similarity 0.78)
- Unlocking the Potential of Global Human Expertise (NeurIPS 2024 · similarity 0.77)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.