20.08.2013 Views

PhD thesis - University of Hertfordshire Research Archive

PhD thesis - University of Hertfordshire Research Archive

PhD thesis - University of Hertfordshire Research Archive

SHOW MORE
SHOW LESS

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

have been increased, necessitating additional statistical power to identify true<br />

differences which might have existed. However, data were captured on<br />

almost 20,000 cases <strong>of</strong> Campylobacter infection and therefore statistical<br />

power will have been adequate.<br />

The choice <strong>of</strong> denominator data in disease determinant analysis is important<br />

in ensuring that incidence estimates are not skewed. This is best illustrated in<br />

paper 3 (Campylobacter Sentinel Surveillance Scheme Collaborators,<br />

2003b), where the risk <strong>of</strong> campylobacteriosis associated with travel to<br />

Cumbria was inflated potentially by increased access to this area <strong>of</strong> the<br />

country by a number <strong>of</strong> participating health authorities situated in north or<br />

north west England. This was unavoidable, as health authority-specific<br />

denominator data was unavailable. It is important to bring such caveats to the<br />

attention <strong>of</strong> the reader so as to keep the findings in context.<br />

In paper 5 (Campylobacter Sentinel Surveillance Scheme Collaborators,<br />

2003a) and paper 8 (Gillespie et al., 2008) we were unable to control for all<br />

the factors under investigation in a single analysis, increasing the likelihood<br />

<strong>of</strong> uncontrolled confounding. This was especially apparent in the latter, where<br />

age, gender, ethnicity and socio-economic status were all investigated. This<br />

potential drawback could have been overcome by applying, for example,<br />

multivariate log-linked Poisson regression techniques to the data, which<br />

would also have allowed for the highly seasonal pattern <strong>of</strong> infection to be<br />

examined. Such techniques require denominator data stratified by all factors<br />

under investigation, however, and these were unavailable.<br />

Bearing in mind its advantages and disadvantages, case-case methodology<br />

is applicable to any data where an outcome is readily identifiable and<br />

exposure data are available, from whence they follow case-control study<br />

methodology. That is not to say however, that the analytical methods<br />

employed in the publications described were not improved over the course <strong>of</strong><br />

the study. The changing analytical strategy with regard to the case-case<br />

comparisons, described previously, is a case in point. However, there is no<br />

suggestion that the findings <strong>of</strong> the earlier studies are necessarily<br />

59

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!