Pr. Hunter, Modelling the impact of prior immunity, case misclassification and bias oncase-control studies in the investigation of outbreaks of cryptosporidiosis, EPIDEM INFE, 125(3), 2000, pp. 713-718
Cryptosporidiosis is the most common cause of outbreaks of disease linked t
o mains water supply in the United Kingdom and the second commonest in the
United States. Recent evidence has suggested that prior population immunity
may have an impact on the epidemiology of waterborne outbreaks and in part
icular prior immunity may reduce the power of case-control studies for demo
nstrating association between disease and water consumption behaviour. Howe
ver, the degree of impact of prior immunity on the power of epidemiological
studies is not yet clear. This paper reports the results of some simple ma
thematical models of outbreaks of waterborne disease in populations with va
rying levels of immunity due to prior water and non-water exposure. The bas
ic outbreak model was run on a spreadsheet. To determine the impact of prio
r immunity on case-control studies, further analysis was done using a Monte
Carlo method to simulate sampling from cases and controls. It was found th
at moderate degrees of prior immunity due to water associated disease could
markedly reduce the relative risk of water consumption on illness in water
borne outbreaks. In turn this would reduce the power of case-control studie
s. In addition, this model was used to demonstrate the impact of case miscl
assification and recall bias on case-control studies. Again it was found th
at within the model, the results of case-control studies could be significa
ntly affected by both these sources of error. Anyone conducting epidemiolog
ical investigations of potentially waterborne outbreaks of disease should b
e aware of the epidemiological problems. Mistakes from case-control studies
will be minimized if the outbreak team pays considerable attention to the
descriptive phase of the investigation and if case-control studies are cond
ucted as soon as possible after an outbreak is detected.