Optimal database combinations for literature searches in systematic reviews: a prospective exploratory study

Table 4 Performance of several databases and database combinations in terms of sensitivity and precision

	# results	# includes (N = 1746)	Overall recall^a	Median recall^b	Minimum recall^c	Percentage 100% recall^d	Precision^e	Number needed to read^f
Embase (EM)	85,521	1500	85.9%	87.3%	45.8%	13.8%	1.8%	57
MEDLINE (ML)	56,340	1375	78.8%	82.9%	50.0%	8.6%	2.4%	41
Web of Science (WoS)	48,561	1189	68.1%	72.5%	13.2%	6.9%	2.4%	41
Google Scholar (GS)	10,342	601	34.4%	38.0%	8.3%	5.2%	5.8%	17
EM-ML	100,444	1621	92.8%	94.6%	66.7%	24.1%	1.6%	62
EM-WoS	104,444	1585	90.8%	93.8%	57.9%	27.6%	1.5%	66
EM-GS	91,411	1570	89.9%	93.3%	65.8%	25.9%	1.7%	58
ML-WoS	75,263	1481	84.8%	87.1%	60.0%	15.5%	2.0%	51
ML-GS	62,230	1459	83.6%	89.8%	63.2%	15.5%	2.3%	43
WoS-GS	54,451	1320	75.6%	85.7%	23.7%	13.8%	2.4%	41
EM-ML-GS	106,334	1674	95.9%	97.4%	78.9%	41.4%	1.6%	64
EM-ML-WoS	119,367	1674	95.9%	97.1%	71.1%	37.9%	1.4%	70
EM-WoS-GS	110,334	1638	93.8%	98.1%	65.8%	44.8%	1.5%	67
ML-WoS-GS	81,153	1528	87.5%	92.6%	70.0%	29.3%	1.9%	53
EM-ML-GS-WoS	125,257	1716	98.3%	100.0%	78.9%	74.1%	1.4%	73

^aOverall recall: The total number of included references retrieved by the databases divided by the total number of included references retrieved by all databases
^bMedian recall: The median value of recall per review
^cMinimum recall: The lowest value of recall per review
^dPercentage 100% recall: The percentage of reviews for which the database combination retrieved all included references
^ePrecision: The number of included references divided by the total number of references retrieved
^fNumber Needed to Read: The total number of references retrieved divided by the number of included references

ISSN: 2046-4053