Eligible clinical cases (identified by either search method) were pooled and verified, duplicate entries excluded. Only the first hospitalization of any given patient was counted. Only cases providing written documentation of a definite or suspected diagnosis were considered eligible for this study and were included in a final listing of 255 clinical cases. Eligible cases were sorted by “CD+” for “Clinical diagnosis present”, and “CD−” for “clinical diagnosis absent” in each selleck chemical diagnostic category: “meningitis”,
“encephalitis” (ENC), “myelitis” (MYE), “ADEM” (ADEM). Cases with a discharge diagnosis of “meningitis” were further classified as “aseptic meningitis” (ASM), “bacterial meningitis” (BM) or “unspecified meningitis” (UM). In 7 cases “meningitis” was coded as one of the discharge diagnoses, but the letter indicated that the diagnosis had, in fact, been excluded during hospitalization. These cases were Fluorouracil tagged with “ND” for “no diagnosis”. An independent investigator (BR), who
had not previously been involved in the care of the patients, reviewed the medical records in a blinded fashion using the structured clinical report form (CRF). The extracted data in the CRF were confined to the variables required almost for the Levels 1–3 of the respective BC case definitions. [7] and [8]. The following labels were applied to all cases in each category (MEN, MYE, ENC, ADEM): “BC+” for “Brighton Collaboration Definition fulfilled”, “BC−” for “Brighton Collaboration
Definition not fulfilled”. The clinical tags were then unblinded and compared to the respective diagnostic categories according to the BC algorithm. In the absence of a gold standard for the diagnoses of encephalitis, meningitis, myelitis and ADEM, sensitivities and specificities cannot be calculated. The new test (i.e. the BC algorithm) was therefore tested against an imperfect, previously available reference test (i.e. the clinician’s diagnosis in the discharge summary). As a result, we determined overall rates of agreement (ORA), positive percent agreement (PPA) and negative percent agreement (NPA), respectively, including the 95% confidence intervals for a total sample size of 255 cases (See Appendices A1 and A2) [33] and [34]. Kappa scores were calculated (Stata Version 9.0se; College Station, TX) in order to find the probability of exceeding agreement expected by chance alone, when comparing the BC definition to the clinical assessment. Cases with discordant results between the physician’s diagnosis and BC category were reviewed individually.