# Disease outbreak data to inform decision-making: the role of disease spread models in the Lao P.D.R. African Swine Fever epidemic, 2019

## TL;DR

This study uses disease spread models to analyze the 2019 African Swine Fever outbreak in Laos, highlighting the importance of quality surveillance data for effective disease control.

## Contribution

The study demonstrates how surveillance data quality affects the accuracy of disease spread models in low-resource settings.

## Key findings

- Six clusters of ASF outbreaks were identified with radii ranging from 16 to 153 km.
- The basic reproduction number (R0) ranged from 13 to 32 between villages.
- Model accuracy was compromised when field data lacked temporal specificity.

## Abstract

Understanding the spread of African Swine Fever (ASF) between villages in the southeast-Asian, low − middle income country context is critical if this high impact disease is to be controlled by good policy and effective field activities in these resource-poor settings. Using governmental reporting data from the 2019 outbreak of ASF in Lao People’s Democratic Republic, spatial clustering techniques were used to identify clusters of outbreak villages. Then Approximate Bayesian Computation with Sequential Monte Carlo was used to estimate the transmission parameters of ASF virus between the villages within these clusters. We used a simple disease spread model to understand the impact of parameter estimation on predicted disease spread and thus decision-making. Six clusters of radius 16 to 153km were identified over the 7 month outbreak period. Within these clusters, the basic reproduction number (R0) ranged from 13 to 32 between-villages and whole-village infectious periods ranged from 62 to 68 days. The final model outputs were compared to the original field report data. We found that the ability of the estimated parameters to match field data was heavily reliant on how the original field surveillance data was reported. Specifically, in situations in which cases in a cluster appeared to have been reported as batches (lack of temporal specificity) our modelling approach failed to produce satisfactory outputs in terms of model fit and precision of estimates. This study demonstrates that surveillance for transboundary diseases not only has immediate benefit for disease response, but that good quality surveillance data is valuable for informing future planning for disease response via appropriately parameterised disease spread models. There is a need for ongoing quality control of surveillance and support for field veterinary services to ensure quality data that can be used to drive policy and decision-making.