Variation in incidence and notification of Campylobacter and Salmonella by general practice in the Thames Valley area

Abstract

ObjectivesTo test whether there is unexplained variation in a) incidence of diagnosed bacterial food poisoning; and b) notification of bacterial food poisoning between general practices. Study designObservational study using routine surveillance data collected between 1 January 2008 and 31 December 2009. MethodsPoisson regression, and the pseudo-R2 statistic, was used to test for the unexplained (i.e. after adjustment for measured confounders) variation in incidence between practices. A generalized linear model, and the pseudo-R2 statistic, was used to test for variation in notifications between practices. Both models were adjusted for demographic factors and organisational factors (Primary Care Trust and Quality and Outcomes Framework score). ResultsA total of 5766 incident cases (811 Salmonella and 4955 Campylobacter) were included. The adjusted incidence of Salmonella and Campylobacter was 128.3 cases per 100,000 persons per year. The adjusted incidence by general practice ranged from 9.8 to 281 per 100,000 (IQR: 90.2–151) persons per year. The median practice notification rate for Salmonella was 25% (range: 0%–100%), and 14.3% (range: 0%–87.5%) for Campylobacter.The Poisson regression model had a pseudo-R2 of 0.080 for the total number of Salmonella and Campylobacter cases, after adjustment for Primary Care Trust and practice deprivation, suggesting substantial variation. The Generalized Linear regression model (predicting notification by general practice) had a pseudo-R2 of 0.040 for Salmonella and Campylobacter, after adjustment for Primary Care Trust and practice deprivation, suggesting substantial unexplained variation. ConclusionSubstantial variation in the diagnosed incidence and notification of Salmonella and Campylobacter by general practice in the Thames Valley area exists. Practice-level factors are likely to account for some of the difference in testing and under-notification. This is important for interpreting data from surveillance systems. Further research is needed to inform interventions designed to increase notifications or improve testing.