SEARCH

SEARCH BY CITATION

Keywords:

  • asthma;
  • logistic regression;
  • questionnaire

Background Asthma is becoming increasingly prevalent and a number of research groups are investigating its genetic and environmental basis.

Objective To produce a brief screening tool suitable for determining phenotype in asthma research.

Methods The scores from eight questions on symptoms and history were obtained from 678 adults and 244 children from high asthma-incidence caucasian families. An independent physician diagnosis was also obtained with the use of a modified NHLBI-CSGA questionnaire and pulmonary function test. Stepwise logistic regression was applied to determine which of the eight questions had greatest predictive value for asthma, and the quality of the resultant models was evaluated using an independent set of 643 adults and 239 children.

Results For adults, the most parsimonious model used responses from three of the eight questions. It had sensitivity and specificity of 0.94 and 0.96, respectively. For children, responses to two questions gave a model with sensitivity and specificity of 0.97. For both age groups, negative predictive values were above 0.87. Positive predictive values were 0.58 and 0.78 for adults and children respectively. The latter emphasize the need for conformation, by physician, of ‘affected’ calls made by this initial screen.

Conclusion The brief questionnaires described are potentially useful in a research setting, as a preliminary screening mechanism of low cost. Their use will reduce the numbers of subjects that must undergo detailed phenotyping.