Objectives: To determine the diagnostic accuracy of tests for the rapid diagnosis of bacterial food poisoning in clinical and public health practice and to estimate the cost-effectiveness of these assays in a hypothetical population in order to inform policy on the use of these tests.
Data sources: Studies evaluating diagnostic accuracy of rapid tests were retrieved using electronic databases and handsearching reference lists and key journals. Hospital laboratories and test manufacturers were contacted for cost data, and clinicians involved in the care of patients with food poisoning were invited to discuss the conclusions of this review using the nominal group technique.
Review methods: A systematic review of the current medical literature on assays used for the rapid diagnosis of bacterial food poisoning was carried out. Specific organisms under review were Salmonella, Campylobacter, Escherichia coli O157, Staphylococcus aureus, Clostridium perfringens and Bacillus cereus. Data extraction was undertaken using standardised data extraction forms. Where a sufficient number of studies evaluating comparable tests were identified, meta-analysis was performed. A decision analytic model was developed, using effectiveness data from the review and cost data from hospitals and manufacturers, which contributed to an assessment of the cost-effectiveness of rapid tests in a hypothetical UK population. Finally, diagnostic accuracy and cost-effectiveness results were presented to a focus group of GPs, microbiologists and consultants in communicable disease control, to assess professional opinion on the use of rapid tests in the diagnosis of food poisoning.
Results: Good test performance levels were observed with rapid test methods, especially for polymerase chain reaction (PCR) assays. The estimated levels of diagnostic accuracy using the area under the curve of the summary receiver operating characteristic curve was very high. Indeed, although traditional culture is the natural reference test to use for comparative statistical analysis, on many occasions the rapid test outperforms culture, detecting additional 'truly' positive cases of food-borne illness. The significance of these additional positives requires further investigation. Economic modelling suggests that adoption of rapid tests in combination with routine culture is unlikely to be cost-effective, however, as the cost of rapid technologies decreases; total replacement with rapid technologies may be feasible.
Conclusions: Despite the relatively poor quality of reporting of studies evaluating rapid detection methods, the reviewed evidence shows that PCR for Campylobacter, Salmonella and E. coli O157 is potentially very successful in identifying pathogens, possibly detecting more than the number currently reported using culture. Less is known about the benefits of testing for B. cereus, C. perfringens and S. aureus. Further investigation is needed on how clinical outcomes may be altered if test results are available more quickly and at a greater precision than in the current practice of bacterial culture.