Predicting Death from Coronary Heart Disease Using a Questionnaire

The ten-year coronary heart disease (CHD) mortality is reported for 18,322 male civil servants aged 40 to 64 according to questionnaire responses at entry into the Whitehall study. In all 1714 died, 723 from CHD. The predictive power of the questionnaire was examined with a view to its use as a screening tool in population studies. In predicting death from coronary heart disease the greatest specificity (true negative rate) was achieved with men reporting both angina (A) and a history of severe chest pain (possible myocardial infarction, PMI). This strategy (A plus PMI) achieved a specificity of 99% but a sensitivity (true positive rate) of only 7%. In contrast, in men reporting angina and/or PMI, specificity was 90% and sensitivity 29%. If this 'and/or' algorithm was extended to include the report of dyspnoea, diabetes, and/or attending a primary care physician with heart disease or hypertension, then specificity was still 85%, but sensitivity increased to 44%. This combination (11 questions in all) is therefore recommended for screening purposes. Identifying and excluding those who favour positive answers ('yes-set' responders), using questions such as the effect of weather on breathing, led to small increases in specificity but relatively large falls in sensitivity. Among subjects reporting chest pain, those who also complained of non-specific symptoms experienced only half the mortality of those with none of these additional complaints.