American Journal of Epidemiology 163 (11): S1. i.e. June 1st 2006


ABSTRACTS OF THE 2ND NORTH AMERICAN CONGRESS OF EPIDEMIOLOGY June 21-24,
2006

THE DEVELOPMENT OF AN EPIDEMIOLOGICAL CASE DEFINITION FOR CHRONIC FATIGUE
SYNDROME (CFS).

*T Osoba, S Gray, J Duffield (University of the West of England)

Current case definitions for CFS are designed for clinical use and not
appropriate for health needs assessment. A robust epidemiological
casedefinition for CFS is crucial in order to achieve rational allocation 
of resources to improve service provision for people with CFS. 

To identify the
clinical features that distinguishes people with CFS from those with other
forms of chronic fatigue and subgroups. General practice data on symptoms,
comorbidities and demographics of patients with unexplained chronic
fatigue based on 4 CFS clinical case definitions were obtained. 

Cases were assigned to disease and non disease groups by an expert panel, 
and multivariate discriminant analysis used to identify those clinical 
features which assigned at least 95% of patients studied to the correct 
group. Sensitivity analysis assessed the impact of the clinical research 
definitions on ascertainment, classification and regression tree analysis, 
cluster analysis to identify subgroups on the basis of the clinical 
manifestations of their illness and a review of basis of diagnosis to 
assess how far GPs' diagnoses were consistent with current definitions 
and specialist opinion. 

Preliminary analyses using classification and regression tree analysis 
included a 10-fold cross-validation approach to prevent over fitting. 

Reliability by Cronbach's alpha was 0.644. Risk and classification tables 
showed an overall correct classification rate of 81.2%. 
The analyses demonstrated that the application of the combination of 
the 4 discriminating variables-the defacto epidemiological
case-definition and pre-defined comorbid conditions had the ability
to differentiate between disease and non-disease cases with CFS with
'postexertional malaise' being the strongest predictor.