1. Which single design limitation most threatens valid estimates of sensitivity and representativeness? How would you address it within six weeks?
The biggest problem is that the AEFI system depends mostly on passive facility reporting, which misses many cases in the community and private clinics. This affects sensitivity and representativeness.
How to address it in six weeks:
Train community health workers (CHWs) and private clinics to report AEFIs using WhatsApp, SMS, or a simple online form. Collect reports from at least two sources (e.g., clinics and CHWs) and compare them using a simple capture–recapture method to estimate missing cases. Combine all data in one Excel or Google Sheet and analyze weekly.
2. Using the CDC surveillance attributes, propose one low-cost intervention to increase sensitivity. State the expected trade-offs, and list 2–3 indicators to detect impact from the intervention.
Intervention: Let CHWs send reports of any health problems after vaccination through SMS or WhatsApp.
Trade-offs: We will get more reports, but many might not be true AEFIs (more work to verify).
Indicators to measure impact:
Number of AEFIs reported per 10,000 doses (should increase).
Median time from event to report (should decrease).
Percentage of reports verified within 48 hours (shows improved response).
3. For a newly introduced vaccine, should the AEFI case definition be temporarily broadened to maximize early signal detection?
– If yes, what trigger would you use to revert to the prior definition?
– If no, why should this change not be implemented?
Yes, the AEFI case definition can be temporarily broadened when a new vaccine is introduced. This helps catch more possible reactions early.
When to return to the old definition: After 8–12 weeks, or when confirmed serious AEFIs stay stable for a month, and the verification workload becomes manageable.
If not broadened: The system may miss early safety signals. However, if the team cannot handle many reports, it’s better to strengthen capacity first before broadening.
