Interventions:
The interventions were described. The rationale for their selection was clear as the proposed screening was compared with the usual care in hospitals in the USA (no screening).
Effectiveness/benefits:
The clinical and effectiveness data were from published studies. For each parameter in the model, the authors reported the main value, the range of values for the sensitivity analysis, and the source, but they did not report a systematic review to identify these sources. As a result, it is unclear if all the relevant evidence was included. Most of the data were from one academic institution, which might not be representative of other health care centres, where the strains of Clostridium difficile might differ. The use of QALYs was appropriate, given the impact of infections on morbidity and mortality. The sources for the utility weights were reported, but the instruments used to elicit them were not.
Costs:
For the stated perspectives of the hospital and third-party payer, it appears that all the relevant major costs were included. The sources for the unit costs and resource use were reported. A breakdown of individual costs items was provided, and the resource use quantities were presented separately from the costs. This will help when replicating the analysis for other settings. The time horizon and currency were reported, but the price year was not, which will limit any future inflationary exercises.
Analysis and results:
A published decision-tree model was adapted to synthesise the cost and outcome information. The details of the model were reported, with a diagram. Uncertainty in the model results was tested in one-way and probabilistic sensitivity analyses, but the results of the probabilistic sensitivity analysis were not reported. As the main limitations to their study, the authors reported that their model was a simplification of real life and could not include the full clinical heterogeneity among patients being screened, and the inputs were from studies of varying quality. The generalisability of the results to other settings was discussed.
Concluding remarks:
The study methods were adequate, but the results of the probabilistic sensitivity analyses were not reported. Given the scope of the study, the authors’ conclusions appear to be valid.