Improving the control of measles epidemics in Niger through a combination of anticipatory and real-time tools

Principal Investigators: Dr. Anton Camacho, Dr Ousmane Guindo

Research Snapshot: Measles in Niger: who will be sick tomorrow?

This study in Niger developed and validated an early warning alert system and a forecasting tool to better anticipate and interpret measles outbreaks, generating insights and positioning the tools for operational decision-making in future outbreaks.

[.cta_link]Read the Snapshot[.cta_link]

‍

What did the study set out to achieve?

Measles contributes heavily to childhood mortality and morbidity in humanitarian settings, and rapid outbreak containment is critical. Niger faces an annual measles season. While its timing is relatively consistent, the size and location of hotspots varies, making it difficult to plan health responses. Emergency campaigns are relied upon, but their impact is easily compromised by delays in outbreak detection and resource mobilization. While prediction methods for measles have been developed elsewhere, these techniques have yet to be validated in a low resource setting.

Three types of tools were developed and evaluated against retrospective data:

1. Maps of outbreak risk to inform proactive vaccine allocation;
2. A simple alert system to improve early detection of outbreaks;
3. Forecasting tools to interpret epidemic trends and inform response prioritization.

‍

What were the key findings?

• While risk maps were unreliable, likely due to data quality/availability issues, the alert system and forecasting tools proved operationally promising.
• The final warning system generates district-level alerts with Level 1 or 2 indicating how likely it is that a current trend in cases will turn into an full on outbreak (64% chance for Level 1 and 95% for Level 2); these alerts prove more reliable compared to existing systems, which are prone to false alerts.
• If followed by response within 4 weeks, Level 1 Alerts may help interventions to treat or offset 73% of cases (60% for Level 2).
• The final forecasting tools assess which alerted outbreaks are nearing their peak with 65% accuracy for Level 1 Alerts and 75% for Level 2; they also identify which outbreaks will have 100 or more additional cases (74% accurate for Level 1 and 84% for Level 2).
• Ultimately, simple models/approaches proved equally or more robust than complex ones when applied to a data frail humanitarian context.

‍

What does this mean for policymakers and practitioners?

‘Anticipate Measles’ built data driven tools adapted to a humanitarian context. Building such tools is challenging--complicated by issues of extreme data frailty and pragmatic constraints.

The alert and forecasting tools have been integrated into an online dashboard used by the Ministry of Health and Médecins sans Frontières for operational decision making. Local actors have been given multiple trainings on the dashboard, the warning system, and the forecasting tools so that they can be used confidently during future measles seasons.

Ultimately, the team found that the complex approaches and research questions frequently used in the Global North may be poorly adapted to the pragmatic realities and data limitations of humanitarian contexts. Finding solutions demanded flexibility and close collaboration with local actors to identify "sweet spot" questions that are both operationally useful and feasible to answer. For example, while it is unrealistic to answer "exactly when will this epidemic peak?", we can respond to "is the peak at least 4 weeks away"--which ends up being enough information to determine whether teams still have time to run a vaccination campaign before the outbreak starts to slow down on its own.

‍