The interventions were clearly described and appear to have been relevant to the study setting. It is possible that they were relevant to other settings.
The reporting was generally good. The clinical evidence was from a cohort study, which was retrospective and open to bias, such as selection bias, which was apparent in the differences in symptom count between groups. It was unclear whether the groups contained sufficient numbers of patients to capture significant differences between them, and there was no indication that adjustments were made for potential confounding factors. All the evidence came from one institution, which might not be representative of other health care centres. For example, the diagnostic coding for preseptal cellulitis might have differed.
The authors did not explicitly report the perspective, but the costs appear to have been relevant to the UK NHS. The sources for the resource use and costs were well reported. The costs were presented at a category level. No discounting was reported, which was appropriate for the short time horizon. The price year was not reported, making reflation exercises difficult. All the economic data were from one hospital, and the impact of variations in the cost estimates was not assessed.
Analysis and results:
The reporting of the effectiveness outcomes was limited and there was no synthesis of costs and benefits; only the net cost saving was reported. Reasonable steps could have been taken to assess the uncertainty in the results. The authors discussed the retrospective design as a limitation of their analysis, and they compared the incidence of preseptal cellulitis in their centre with that in another published study, with similar findings. The differences between the groups were adequately tested, using appropriate statistical methods, but it was unclear if adjustments were made for differences in the patients’ baseline characteristics, some of which were not reported, such as the patients' gender.
The methods were reasonably well reported, but had some drawbacks, which should be considered when assessing the validity of the authors’ conclusions.