Concordance of historical questions used in risk-stratifying patients with headache

We sought to determine whether the manner and order in which historical questions used to risk stratify for subarachnoid hemorrhage are asked significantly alter the response/concordance.

Adult patients complaining of headache in the emergency department were presented with 1 of 2 questionnaires each containing 2 variations of the pertinent question and differing only in their order. Data were primarily analyzed using the kappa statistic to determine whether rates of concordance are greater than would be expected by chance alone. And, as a secondary outcome, a sample of 120 was predetermined to be adequate to achieve 80% power in detecting a difference of 20% to 25% between questionnaires comparing the influence of order on concordance.

The agreement corrected for chance for version 2, kappa = 0.51, is higher than the agreement corrected for chance for version 1, kappa = 0.28, a difference of 0.23 with a 95% confidence interval (-0.03 to 0.49; P = .08; SE, 0.13). The percentages of patients who answered the questions concordantly were 60% and 75%, respectively, for versions 1 and 2. The difference is 15% with a 95% confidence interval of (-2% to 32%, P = .08).

Although not statistically significant, our study indicates that up 38% answer these 2 very similar questions discordantly. Also, there appears to be a higher degree of concordance (15%) when patients are first asked, "When was the last time you had a headache this bad?"