Table 2:

Efficiency and comprehensiveness of physician-generated searches with and without search filters

Clinical queries therapy filter		Nephrology filter		Search performance		Percentage-point change from nonfiltered search (99% CI)	p value^*

Broad	Narrow	Broad	Narrow	Efficiency^†; Comprehensiveness^‡	Mean, %
No filter (physician-provided searches)				Efficiency	6.0

				Comprehensiveness	45.9

x				Efficiency	6.1	0.1 (−1.8 to 2.0)	0.86

				Comprehensiveness	45.2	−0.7 (−3.2 to 1.8)	0.38

	x			Efficiency	22.9	16.9 (9.9 to 23.8)	< 0.003

				Comprehensiveness	37.9	−8.0 (−13.8 to −2.2)	< 0.003

		x		Efficiency	4.9	−1.1 (−2.4 to 0.1)	0.006

				Comprehensiveness	54.5	8.6 (1.9 to 15.3)	< 0.003

			x	Efficiency	6.1	0.1 (−1.9 to 2.0)	0.93

				Comprehensiveness	52.4	6.5 (0.1 to 12.9)	< 0.003

x		x		Efficiency	4.9	−1.1 (−3.1 to 0.9)	0.09

				Comprehensiveness	53.9	8.0 (0.5 to 15.5)	0.002

x			x	Efficiency	6.2	0.2 (−2.4 to 2.7)	0.84

				Comprehensiveness	51.8	5.9 (−1.2 to 13.1)	0.01

	x	x		Efficiency	19.5	13.5 (7.3 to 19.6)	< 0.003

				Comprehensiveness	45.8	−0.1 (−8.9 to 8.8)	0.98

	x		x	Efficiency	21.6	15.6 (8.9 to 22.3)	< 0.003

				Comprehensiveness	44.2	−1.7 (−10.2 to 6.8)	0.55

Note: CI = confidence interval.
↵* Bonferroni-adjusted α = 0.003.
↵† Efficiency is the proportion of articles retrieved by the search query that are relevant to the search question (search precision, positive predictive value).
↵‡ Comprehensiveness is the proportion of all relevant articles that are retrieved by the search query (search recall, sensitivity).