Abstract
<jats:p>Background: Cholera outbreaks can spread rapidly, which means that the optimal decision-making window for a large, coordinated response is very narrow. Alerts for triggering interventions need to balance tradeoffs between wasting resources on false positives and delaying decisions until they lose effectiveness. A systematic evaluation of such tradeoffs across settings is needed to understand which alerts may have the greatest public health utility and where. Methods: Using weekly suspected cholera surveillance across 4,081 subnational administrative units across 34 countries in Africa from 2010-2023, we evaluated 24 alert definitions (4 alert types with different numeric thresholds) over a 1-year post-alert period on five utility dimensions - potential health impact, potential intervention efficiency, positive predictive value (PPV) for large outbreaks, proportion of missed outbreaks, and timeliness of alert trigger. The dimensions were combined into a utility score, which was used to identify the best alert across the continent and by country. For top-performing alerts, we estimated the reduction in potential health impact for additional delays in response using Bayesian hierarchical models. Results: Fifty suspected cases for three consecutive weeks was the definition with the highest utility score across most contexts. In administrative units with 50,000 to 500,000 people, the year following such an alert experienced a mean of 376 suspected cases (standard deviation: 618.3) and 2 cases per 1000 population (SD: 4.1). Forty percent of such alerts (N alerts: 265) were followed by a 1-year period with over 300 cases, yet the definition missed 40% of outbreaks with over 300 cases (N outbreaks: 278) and was triggered 5.3 weeks (SD: 5) after outbreak start. Each week of delay was estimated to result in an additional 20% reduction (95% CrI: -23 to -17) of potential health impact in the outbreak response. One hundred cases over a three-week period was another definition that had high utility, particularly in administrative units with smaller populations and country-specific evaluations. Conclusion: We present a decision analytic framework that can be deployed in a short decision-making window using case-based surveillance to trigger large-scale cholera response activities with moderately high utility across most African transmission contexts. Future work should consider adaptations based on local data availability and priorities and examine the generalizability of early case-based signals outside Africa.</jats:p>