Table 2.

Proportion of alerts (mean and standard error) given either a suggestion of no action (i.e. no concern indicated either directly following review of the data or after the simulated call) or action (concern indicated with either follow-up, medication adjustment or sending a community nurse) together with the percentage of raw agreement (i.e. the average proportion of rater-pairs agreeing of the total possible rater-pair agreements) with bootstrapped 95% confidence intervals. The Fleiss kappa statistic excluded cases in which not all clinicians provided a rating (total: arm A, 281 cases with 8 raters; arm P, 147 + 85 cases with 4 + 4 raters)

Indicated response		Arm A	Arm P
No action	Proportion of alerts	50.5 ± 6.2%	40.6 ± 7.1%
	Agreement across participants on response	66.8% (63.4–69.8)	57.5% (52.1–62.4)
Action (concern indicated)	Proportion of alerts	49.5 ± 6.2%	59.4 ± 7.1%
	Agreement across participants on response	66.4% (62.7–69.6)	72.8% (69.0–76.3)
	Fleiss’s Kappa statistic	.334 (.329–.340)	.305 (.291–.319)