Skip to main content

Table 4 Abstrackr’s mean performance across three trials for each of the screening projects

From: Technology-assisted title and abstract screening for systematic reviews: a retrospective evaluation of the Abstrackr machine learning tool

Performance metric Topic
Antipsychotics
N = 12,763 records
Bronchiolitis
N = 5893 records
Child Health SRs
N = 5243 records
Diabetes
N = 47,385 recordsa
Precision, % (SD) 15.1 (2.6) 38.1 (2.6) 64.7 (2.0) 14.8 (2.6)
False negative rate, % (SD) 21.2 (8.3) 7.3 (2.2) 3.5 (1.4) 17.9 (2.3)
Proportion missed, % (SD) 0.1 (0.1) 0.1 (0.1) 6.4 (1.7) 0.1 (0.01)
Workload savings, % (SD) 64.5 (9.8) 70.0 (3.7) 9.5 (3.5) 88.4 (2.7)
  1. Standard deviations for proportions relate to the range of values observed across trials and not the mean variance across trials
  2. SD standard deviation
  3. aIncluded some duplicates, as three EndNote libraries were combined to create the dataset