Table 4

Top 3-performing Combination of Terms with the Best Sensitivity (keeping Specificity ≥50%), Specificity (keeping Sensitivity ≥50%), and Best Optimization of Sensitivity and Specificity (based on abs [sensitivity-specificity]<1%) for Detecting Studies of Causation in EMBASE in 2000

Search Strategy OVID search*

Sensitivity (%)

(n = 215)

Specificity (%)

(n = 27,554)

Precision (%)†

Accuracy (%)

(n = 27,769)


Top 3-performing combination of terms with best Sensitivity


risk:.mp.

OR exp methodology

OR exp epidemiology

91.6 (87.9 to 95.3)

60.9 (60.3 to 61.4)

1.8 (1.6 to 2.0)

61.1 (60.5 to 61.7)

risk:.tw.

OR exp methodology

OR exp epidemiology

91.2 (87.4 to 95.0)

63.0 (62.4 to 63.6)

1.9 (1.6 to 2.2)

63.2 (62.6 to 63.8)

risk:.mp.

OR exp methodology

OR exp mortality

90.7 (86.8 to 94.6)

65.1 (64.5 to 65.7)

2.0 (1.7 to 2.3)

65.3 (64.7 to 65.8)


Top 3-performing combination of terms with best Specificity


cohort.tw.

OR relative risk:.tw.

53.0 (46.4 to 59.7)

94.6 (94.4 to 94.9)

7.1 (5.9 to 8.4)

94.3 (94.0 to 94.6)

confidence interval.tw.

OR relative risk:.tw.

50.7 (44.0 to 57.4)

94.5 (94.2 to 94.7)

6.7 (5.4 to 7.9)

94.1 (93.8 to 94.4)

OR relative risk:.tw.

OR cohort:.tw.

53.5 (46.8 to 60.2)

94.4 (94.1 to 94.6)

6.9 (5.7 to 8.1)

94.1 (93.8 to 94.3)


Top 3-performing combination of terms with best optimization of Sensitivity and Specificity


risk.tw.

OR mortalit:.tw,

OR cohort.tw.

81.9 (76.7 to 87.0)

81.4 (80.9 to 81.8)

3.3 (2.8 to 3.8)

81.4 (80.9 to 81.8)

risk:.tw.

OR cohort:.mp.

OR confidence interval:.tw.

81.9 (76.7 to 87.0)

81.2 (80.8 to 81.7)

3.3 (2.8 to 3.8)

81.3 (80.8 to 81.7)

risk.tw.

OR mortalit:.tw.

OR cohort:.tw

81.9 (76.7 to 87.0)

81.2 (80.8 to 81.7)

3.3 (2.8 to 3.8)

81.2 (80.8 to 81.7)


*Search strategies are reported using Ovid's search engine syntax for EMBASE.

†Denominator varies by row.

: = truncation; mp = multiple posting – term appears in title, abstract, or subject heading; exp = explode, a search term that automatically includes closely related indexing terms; tw = textword (word or phrase appears in title or abstract).

Haynes et al. BMC Medical Informatics and Decision Making 2005 5:8   doi:10.1186/1472-6947-5-8

Open Data