Inference for infectious diseases from multivariate serological survey data

The aim of this project is to develop new methods that will aid our understanding of how infectious diseases spread in a population, and hence improve our ability to control those diseases by vaccination. The methods to be developed will be used with serological data, that is, data obtained from blood tests. These tests determine whether a person has, or has not, been infected by one or more infections. Large studies based on such data are commonly used to plan vaccination programmes, and to monitor how successful existing vaccination programmes are at maintaining levels of immunity sufficient to prevent the occurrence of large epidemics. However, most of the methods that are currently employed to analyse serological data are used to look at each infection in isolation. The project is to develop new methods to analyse serological data on several infections at the same time. The reason why this might be fruitful is that the occurrences of different infections within the same individuals are likely to be correlated. For example, children who go to nurseries are more likely to get infected by all childhood infections that are transmitted by close contacts. The idea behind the project is to use the correlations between different infections (which can be measured using serological data) to tell us about contact patterns, how they vary between individuals, and how they vary with age. The way we propose to do this is by developing new statistical models for this type of data, based on relevant hypotheses about what might be causing the correlations to arise. The results from these models can then be used to improve our understanding of the spread of infections, for example by providing better estimates of the proportion of children that need to be vaccinated to prevent large epidemics, or by helping to identify how infections are transmitted if this is not known. The project involves a collaboration between the two applicants, who have long experience of statistical modelling of infectious diseases, and the Head of the Health Protection Agency?s Seroepidemiology Programme. So far, much data on different infections have been collected, and initial analyses have been undertaken to verify that the project?s rationale is well-founded. In this application, we seek funding for a researcher to work on the project full-time for three years under the supervision of the applicants.

This project is to develop new statistical methodology to analyse multivariate serological survey data, and thus derive better estimates of key infectious disease parameters relevant to vaccination policies. Such data are commonly available, since samples are typically banked and tested for several infections. However, the methods of analysis that are currently used do not fully exploit the multivariate features of the data. This is unfortunate, since serological data contribute importantly to defining vaccination and other control strategies, understanding the epidemiology of endemic infections, monitoring susceptibility levels and predicting outbreaks.

We propose to develop new analysis tools to describe, model and infer new, relevant knowledge from associations between different infections. Such associations may be generated by individual variations in susceptibility, contact rates, or test performance. The associations that have been observed exhibit complex patterns over age, which vary according to transmission routes. The rationale of the project is that new information about heterogeneities, contact rates and transmission routes may be obtained from such patterns.

We shall use modelling approaches to quantify the levels of individual heterogeneity in contact rates and how they vary with age, and thus obtain better estimates of fundamental quantities such as reproduction numbers and critical immunization thresholds. The methods may also provide new tools to derive information on route of transmission, if this is not already known. Thus, the project will help to improve the modelling of infectious diseases.

We have already collected extensive data on many infections, and have plans to obtain more. Some initial analysis has been undertaken to verify that the rationale of the project is well-founded. The statistical methods to be used include smoothing, dependence measures for current status data, frailty modelling, time-varying frailty models, copulas, mixture models, epidemic models. The five specific objectives of the project strike a careful balance between methodological development, application to data and popularization within the epidemiological community.

We seek funds for a named post-doctoral Research Fellow for three years. The applicants have considerable experience in modelling infectious diseases, especially serological survey data. The project will benefit from advice from the Health Protection Agency Seroepidemiology Programme, the Head of which is a named collaborator on the application.