- © 2007 Canadian Medical Association or its licensors
[Two of the authors respond:]
We conducted our CMAJ study1 using cases defined by prescription for oral vancomycin therapy in response to the criticism of the clinical diagnosis case definition that we used in our earlier paper published in JAMA,2 as well as in response to the issues raised by van Staa and colleagues concerning changes in recording.3 We wished to tighten the case definition by looking at patients who received a prescription for oral vancomycin for whom the only indication was treatment of Clostridium difficile-associated disease.
We found that for about 90% of patients for whom a C. difficile toxin assay was performed, the result was not entered in the laboratory results section of the database. We also found that there were 2 different coding systems for medical diagnoses: OXMIS and READ. In practices using the OXMIS coding system, there was no medical diagnostic code for C. difficile-associated disease, so these practices may have been missed in our original study. We believe that the laboratory results and clinical diagnosis may have been included in the free-text section of the database, to which we did not have access. In addition, if the test was done but no results were entered into the database, these patients were excluded from our original study if they did not have a clinical diagnosis,2 and thus we possibly missed cases.
We believe that by defining cases by prescription for oral vancomycin in the present study, we were able to find patients from practices using the OXMIS coding system and possibly patients whose toxin assay results or clinical diagnoses might have been entered only in the free-text section of the database. As we indicated in our CMAJ article, when cases were defined on the basis of a prescription, the results were less likely to be affected by changes in data entry and recording over time.3 Finally, as we demonstrated several times in the article, the cases in the 2 studies did not overlap.