摘要
DIAGNOSTIC ACCURACY AND CASE FINDING Accurate diagnostic and case ascertainment methods are a prerequisite for epidemiological research. In epilepsy, a common problem is diagnostic accuracy as it can be diagnosed only by taking a history of the index event or by chance observation of a seizure.' The diagnosis is fundamentally a discretionary judgement which depends on the skill and experience of the physician and the quality of witness information available.2 Common sources of confusion are syncope or psychogenic attacks.'-6 As many as 10%-20% of cases referred to specialised units with seemingly intractable seizures do not have epilepsy.' 2 4-6 Up to 30% of patients developing will eventually be classified as having epilepsy.7-8 The inclusion of patients with non-epileptic attacks in the chronic epilepsy group may artificially inflate the proportion of cases.8 By contrast, many patients with have the condition for some time before the correct diagnosis is achieved.' 9 10 Most studies lack clear and reproducible diagnostic definitions, a recent exception being one in Ecuador. 12 Case ascertainment also poses problems in epilepsy. Some patients with seizures never seek medical attention either through concealment, denial, or ignorance. 13-14 It is likely, therefore, that field studies miss patients unless sensitive screening techniques for all epileptic phenomena are included in the case ascertainment. This has not yet been achieved. It could be argued that patients not presenting to a medical agency should not be considered a problem. This may hold sway in clinical practice but in epidemiology it is important that all cases are included.'5 The commonest method of case ascertainment is a retrospective review of medical notes, for seizures, antiepileptic drugs (AEDs), EEG, or a diagnostic coding.' There are major sources of inaccuracy and underreporting is common. The extent of this was shown by a study in Warsaw that found a prevalence rate of 51/1000 based on a survey of medical records alone which rose to 104/1000 in a sample of 0-5% of the community.'3 Similarly, in Guam, incidence rates based on field surveys were twice as high as those based on medical records only.'6 Studies employing record reviews have covered total populations,'7-23 a random sample,13-24 or selected groups such as sick funds policy holders,25 army draftees,26 hospital attenders,27-28 school children,29-3' government employees,32 those with learning disabilities,33 or general practitioners' lists.34-37 A second approach has been the use of a register of cases but unless precautions are taken, these may present the same diagnostic problems as a review of existing records.93839 An advantage of a register set up for research purposes is that the methodology may be planned in advance.' An epidemiological approach that does not rely on prior diagnosis is to carry out a community survey with a sensitive and specific screening questionnaire.40 This strategy works for tonic clonic seizures and other seizures with florid clinical symptomatology. For other seizure types, however, this may not be accurate as a pragmatic screening instrument for these seizures has not yet been designed. An attempt to design such a questionnaire had to be abandoned due to the low specificity of questions relating to absence and myoclonic seizures.40 Community surveys using a screening questionnaire have been carried out in various locations. 12 13 41-66 Entire populations, random samples, or subgroups have been selected. Such surveys depend crucially on the adequacy of the screening methods and these are difficult to design. It is important to reach a balance between sensitivity and specificity, which has not yet been satisfactorily achieved.40 A screening questionnaire and its validation are not easily transferable between different populations due to cultural and social influences and must in all cases be piloted and validated for each population.