Estimating the Per-Contact Probability of Infection by Highly Pathogenic Avian Influenza (H7N7) Virus during the 2003 Epidemic in The Netherlands by Amos Ssematimba, Armin R. W. Elbers, Thomas J. Hagenaars, Mart C. M. de Jong

Read on October 25^th, 2024 in Winston-Salem, NC, rated 5/5

Summary

SSematimba et al examine the per contact probability rates between poultry farms in the Netherlands in 2003 outbreaks of H7N7. H7N7 is a highly pathogenic strain for poultry. They used a series of probabilistic arguments backed by contact tracing data and some phylogenetic analysis to examine transmission routes. They found a little unsurprisingly that egg transport represents a very high probability of introducing infection to a new farm.

Main Points

![[Pasted image 20241025132811.png]]

Abstract

Estimates of the per-contact probability of transmission between farms of Highly Pathogenic Avian Influenza virus of H7N7 subtype during the 2003 epidemic in the Netherlands are important for the design of better control and biosecurity strategies. We used standardized data collected during the epidemic and a model to extract data for untraced contacts based on the daily number of infectious farms within a given distance of a susceptible farm. With these data, we used a maximum likelihood estimation approach to estimate the transmission probabilities by the individual contact types, both traced and untraced. The estimated conditional probabilities, conditional on the contact originating from an infectious farm, of virus transmission were: 0.000057 per infectious farm within 1 km per day, 0.000413 per infectious farm between 1 and 3 km per day, 0.0000895 per infectious farm between 3 and 10 km per day, 0.0011 per crisis organisation contact, 0.0414 per feed delivery contact, 0.308 per egg transport contact, 0.133 per other-professional contact and, 0.246 per rendering contact. We validate these outcomes against literature data on virus genetic sequences for outbreak farms. These estimates can be used to inform further studies on the role that improved biosecurity between contacts and/or contact frequency reduction can play in eliminating between-farm spread of the virus during future epidemics. The findings also highlight the need to; 1) understand the routes underlying the infections without traced contacts and, 2) to review whether the contact-tracing protocol is exhaustive in relation to all the farm’s day-to-day activities and practices.

Notes

They set out the following sets of equations and estimated using their contact rate data: $\text{Prob of farm infected} = \big ( 1 - \Pi_{all i}(1-p_i){\Sigma_dw_dC_{i,d}{Infected}}\big)$ And then the probability of a farm escaping infection is then $\big ( (1 - p_i){\Sigma (1 - w_d)C_{i,d}{Inf}}\big)$ And then the probability of a farm escaping infection during an outbreak is:

$\big( \Pi_{all \text{i}}(1-p+i){C_i{Escape}}\big)$

“The estimated conditional probabilities, conditional on the contact originating from an infectious farm, of virus transmission were: 0.000057 per infectious farm within 1 km per day, 0.000413 per infectious farm between 1 and 3 km per day, 0.0000895 per infectious farm between 3 and 10 km per day, 0.0011 per crisis organisation contact, 0.0414 per feed delivery contact, 0.308 per egg transport contact, 0.133 per other-professional contact and, 0.246 per rendering contact. We validate these outcomes against literature data on virus genetic sequences for outbreak farms.” (Ssematimba et al., 2012, p. 1)

“Following the detection of the first outbreak, a control programme, as stipulated by the European Union, was implemented. This programme consisted of stamping out of infected flocks, movement restrictions and establishment of protection and surveillance zones.” (Ssematimba et al., 2012, p. 1)

“However, in comparison with preventive culling, emergency vaccination would have the important disadvantage that its effect suffers from a 7 to 14 days protection delay 12]. This delay would prolong the time until epidemic control is obtained especially in the high density poultry areas (de Jong and Hagenaars [13] and the references therein).” ([Ssematimba et al., 2012, p. 1)

“Plausible mechanisms include movements of humans (professional and non-professional visitors, employees and farmers themselves), vehicular traffic (for example, delivery trucks), other fomites (such as tools, cell phones and shared farm equipment) and other vectors such as wind, rodents and insects 9,14–17]. These transmission events involve transportation of the virus either in contaminated litter, faeces or skin and feathers that can colloid on the fomites or the vectors’ body. Therefore, in order to better control neighbourhood transmission, we need to understand deeper the steps involved in the whole virus dissemination process; a quite complex task.” ([Ssematimba et al., 2012, p. 2)

“the probability of HPAI virus transmission may be contact-specific but will also depend on the contact patterns” (Ssematimba et al., 2012, p. 2)

“They found an increased risk of HPAI virus introduction in layer-finisher type poultry compared to other poultry types. Their analysis gave some clues on the risk factors for HPAI virus introduction such as poultry type and flock size.” (Ssematimba et al., 2012, p. 2)

“Our analysis aims to give quantitative insight into the role of the different between-farm contacts in the spread of the virus during an epidemic.” (Ssematimba et al., 2012, p. 2)

“This dataset captured information on a total of 614 visits originating from 203 infectious farms. Out of these visits, 381 were to infected farms. The total number of receiving farms was 325 of which 149 were ultimately infected.” (Ssematimba et al., 2012, p. 2)

“From this dataset we selected visits to a farm that occurred up to seven days prior to and excluding its day of suspicion. For these contacts, we only considered same-day visits i.e., those that occurred on the same day that the person had visited an infectious farm.” (Ssematimba et al., 2012, p. 2)

“A farm was deemed exposed if the visit occurred during the period when the virus was likely to have been introduced onto the receiving farm, here referred to as the potential virus-introduction period.” (Ssematimba et al., 2012, p. 2)

“We quantified the contribution of the different contacts to the epidemic in terms of the number of new infections that they may have caused. This was obtained by multiplying their estimated per-contact probability with their frequency.” (Ssematimba et al., 2012, p. 3)

“For the base model, we used a uniform distribution to obtain wd ~ 1 7. In other words, we assumed that each of the seven days of the probable period of virus introduction was equally likely to be the actual day of virus introduction.” (Ssematimba et al., 2012, p. 3)

“We assessed two other distributions in which the estimated weighting factors wd were adjusted to sum to one over the 7-day period, namely; 1) a distribution in which the probability is decreasing exponentially over the 7-day period at a rate determined by the survival of HPAI virus in manure (in this case 14 days 23]) and, 2) a unimodal distribution with the most likely day being 4 days prior to the day of clinical suspicion.” ([Ssematimba et al., 2012, p. 3)

“In this way, we used the genetic data to validate the estimated probabilities per contact: too few or too many genetic matches would cast doubt on the estimated probabilities.” (Ssematimba et al., 2012, p. 4)

“For those pairs (i.e., with complete genetic information), we compared their genetic sequences to ascertain which ones were sufficiently ‘‘matching’’ for transmission between A and B not to be unlikely. The number of genetically matching pairs, minus an estimate of the expected number of ‘‘bychance’’ genetic matches, was then compared to the predicted number of pairs (amongst those with complete genetic information) in which virus transmission occurred (‘‘transmission pairs’’) Npredicted” (Ssematimba et al., 2012, p. 4)

“part from the unknown and crisis organisation contacts, feed deliveries had the lowest per-contact probability of virus transmission of 0.0414 and potentially caused 2.63% of the new case farms while the egg transports had the highest per-contact probability of 0.308 and may have potentially caused 2.04% of the new case farms.” (Ssematimba et al., 2012, p. 4)

“The estimates were very similar for most of the exposure types. The only differences found, but these were small, were in the per-contact probabilities for the crisis organisation contacts for both alternative distributions and the other-professional contacts for only the unimodal distribution (see Table 2). For both alternative distributions, the probabilities per crisis organisation contact were within the 95% CI of the default distribution whereas for the unimodal distribution, the per otherprofessional contact probability reduced from 13.3% to 0.0%.” (Ssematimba et al., 2012, p. 5)

“With respect to the effect of the potential difference in tracing efforts on case and non-case farms – hence a possibility of underrepresentation of the contacts to non-case farms, we found that, with the worst tracing efforts, the contacts to case farms would be twice as likely to be traced as those to non-case farms.” (Ssematimba et al., 2012, p. 5)

“This implies that, at worst, the estimated probabilities could be double their ‘unbiased’ counterparts.” (Ssematimba et al., 2012, p. 5)

“In terms of per-contact risk, the estimates reveal that egg transports have the highest risk with approximately 31% chance of transmission followed by the rendering visits with a chance of transmission of 25%. The unknown contacts in the distance band of 0–1 km have the lowest risk per contact although, as is clear from the 95% confidence bounds, its estimated percontact probability is not significantly different from those of the other unknown contact categories.” (Ssematimba et al., 2012, p. 5)

“We expect that the implementation of preventive culling within 1 km of an infectious farm during the epidemic 5] has had a (strong) censoring effect on the detection of infected farms with 1 km of an infectious farm, thus producing a downward bias on the transmission probability per unknown contact within 1 km.” ([Ssematimba et al., 2012, p. 5)

“Our results suggest that, apart from the unknown contacts, egg delivery contacts are interesting targets for improvements in biosecurity due to their high per-contact probability (31%) in infecting the receiving farms. They further suggest that the biosecurity applied to the crisis organisation contacts seems to be adequate at least for preventing the persons themselves from becoming important fomites between registered visits. Overall, these findings provide a scientific basis to conduct further studies, epidemiological or otherwise, to evaluate the impact of improved biosecurity and minimized contact-frequency in controlling the between-farm spread of HPAI virus during epidemics.” (Ssematimba et al., 2012, p. 6)

“Virus by Professionals During Outbreak Control Activities” (Ssematimba et al., 2012, p. 7) .

Questions

How can we leverage these data for different contexts?
Can we apply these to different contexts outside of poultry farms?
Could we use this type of survey to inform surveys for other livestock related outbreaks

Code/ Supplemental Data

In the article.

Implications for Infectious Diseases

Ssematimba A, Elbers ARW, Hagenaars TJ, Jong MCM de. Estimating the Per-Contact Probability of Infection by Highly Pathogenic Avian Influenza (H7N7) Virus during the 2003 Epidemic in The Netherlands. PLOS ONE 2012;7:e40929. https://doi.org/10.1371/journal.pone.0040929.

Imported: 2024-10-25