Limits...
Spatial Dynamics and High Risk Transmission Pathways of Poliovirus in Nigeria 2001-2013

View Article: PubMed Central - PubMed

ABSTRACT

The polio eradication programme in Nigeria has been successful in reducing incidence to just six confirmed cases in 2014 and zero to date in 2015, but prediction and management of future outbreaks remains a concern. A Poisson mixed effects model was used to describe poliovirus spread between January 2001 and November 2013, incorporating the strength of connectivity between districts (local government areas, LGAs) as estimated by three models of human mobility: simple distance, gravity and radiation models. Potential explanatory variables associated with the case numbers in each LGA were investigated and the model fit was tested by simulation. Spatial connectivity, the number of non-immune children under five years old, and season were associated with the incidence of poliomyelitis in an LGA (all P < 0.001). The best-fitting spatial model was the radiation model, outperforming the simple distance and gravity models (likelihood ratio test P < 0.05), under which the number of people estimated to move from an infected LGA to an uninfected LGA was strongly associated with the incidence of poliomyelitis in that LGA. We inferred transmission networks between LGAs based on this model and found these to be highly local, largely restricted to neighbouring LGAs (e.g. 67.7% of secondary spread from Kano was expected to occur within 10 km). The remaining secondary spread occurred along routes of high population movement. Poliovirus transmission in Nigeria is predominantly localised, occurring between spatially contiguous areas. Outbreak response should be guided by knowledge of high-probability pathways to ensure vulnerable children are protected.

No MeSH data available.


Related in: MedlinePlus

The expected number of LGAs (of 774) reporting at least one case of poliomyelitis during each six month period.The shaded area represents 95% of the distribution of outcomes using 1000 simulations of the radiation model with the actual number of LGAs reporting a case overlaid in blue.
© Copyright Policy
Related In: Results  -  Collection

License
getmorefigures.php?uid=PMC5036822&req=5

pone.0163065.g003: The expected number of LGAs (of 774) reporting at least one case of poliomyelitis during each six month period.The shaded area represents 95% of the distribution of outcomes using 1000 simulations of the radiation model with the actual number of LGAs reporting a case overlaid in blue.

Mentions: We first simulated 1000 datasets using the best-fitting model and calculated the mean number of LGAs predicted to experience an outbreak for each simulation (Fig 3). A comparison of these simulations with the observed values indicated reasonable agreement (mean R2 value 0.62 [95% CI: 0.54, 0.70]). However, the model failed to capture the short-term dynamics and lagged behind the observed data, underestimating the number of infected LGAs at the start of a peak. Secondly we examined the geographic distribution of cases in 1000 simulated datasets at each time point, comparing the performance of the model in identifying which LGAs would become infected with the observed data. The geographic distributions of cases were well-estimated with an average AUC of 0.81 (SE 0.02) suggesting that it performs well in estimating the probability and location of outbreaks (S5 Fig).


Spatial Dynamics and High Risk Transmission Pathways of Poliovirus in Nigeria 2001-2013
The expected number of LGAs (of 774) reporting at least one case of poliomyelitis during each six month period.The shaded area represents 95% of the distribution of outcomes using 1000 simulations of the radiation model with the actual number of LGAs reporting a case overlaid in blue.
© Copyright Policy
Related In: Results  -  Collection

License
Show All Figures
getmorefigures.php?uid=PMC5036822&req=5

pone.0163065.g003: The expected number of LGAs (of 774) reporting at least one case of poliomyelitis during each six month period.The shaded area represents 95% of the distribution of outcomes using 1000 simulations of the radiation model with the actual number of LGAs reporting a case overlaid in blue.
Mentions: We first simulated 1000 datasets using the best-fitting model and calculated the mean number of LGAs predicted to experience an outbreak for each simulation (Fig 3). A comparison of these simulations with the observed values indicated reasonable agreement (mean R2 value 0.62 [95% CI: 0.54, 0.70]). However, the model failed to capture the short-term dynamics and lagged behind the observed data, underestimating the number of infected LGAs at the start of a peak. Secondly we examined the geographic distribution of cases in 1000 simulated datasets at each time point, comparing the performance of the model in identifying which LGAs would become infected with the observed data. The geographic distributions of cases were well-estimated with an average AUC of 0.81 (SE 0.02) suggesting that it performs well in estimating the probability and location of outbreaks (S5 Fig).

View Article: PubMed Central - PubMed

ABSTRACT

The polio eradication programme in Nigeria has been successful in reducing incidence to just six confirmed cases in 2014 and zero to date in 2015, but prediction and management of future outbreaks remains a concern. A Poisson mixed effects model was used to describe poliovirus spread between January 2001 and November 2013, incorporating the strength of connectivity between districts (local government areas, LGAs) as estimated by three models of human mobility: simple distance, gravity and radiation models. Potential explanatory variables associated with the case numbers in each LGA were investigated and the model fit was tested by simulation. Spatial connectivity, the number of non-immune children under five years old, and season were associated with the incidence of poliomyelitis in an LGA (all P < 0.001). The best-fitting spatial model was the radiation model, outperforming the simple distance and gravity models (likelihood ratio test P < 0.05), under which the number of people estimated to move from an infected LGA to an uninfected LGA was strongly associated with the incidence of poliomyelitis in that LGA. We inferred transmission networks between LGAs based on this model and found these to be highly local, largely restricted to neighbouring LGAs (e.g. 67.7% of secondary spread from Kano was expected to occur within 10 km). The remaining secondary spread occurred along routes of high population movement. Poliovirus transmission in Nigeria is predominantly localised, occurring between spatially contiguous areas. Outbreak response should be guided by knowledge of high-probability pathways to ensure vulnerable children are protected.

No MeSH data available.


Related in: MedlinePlus