What characteristics classify 'experience' with data abstraction?




Poster session 1


Monday 24 October 2016 - 10:30 to 11:00


All authors in correct order:

Saldanha I1, Wen J2, Schmid C3, Li T1
1 Cochrane United States, and Cochrane Eyes and Vision US Satellite, USA
2 Johns Hopkins Bloomberg School of Public Health, USA
3 Brown University School of Public Health, USA
Presenting author and contact person

Presenting author:

Ian Saldanha

Contact person:

Abstract text
Background: Cochrane recommends that data abstraction should be done independently by at least two individuals. In practice, individuals with complementary levels of data abstraction experience are often paired for data abstraction. However, what data abstraction experience really means is unclear.

Objectives: To identify characteristics that best classify an individual’s level of experience in performing data abstraction for systematic reviews.

Methods: We surveyed faculty, staff, and students at two schools of public health, two evidence-based practice centers (EPCs), and one Cochrane Center who had abstracted data from at least one study for a systematic review. We asked questions on respondent’s current status (faculty, staff, student, other), number of articles abstracted, number of systematic reviews published, and self-rated level of experience with data abstraction. Masked to their responses, we categorized each respondent as either a more experienced or less experienced data abstractor based on our subjective assessment of their quality of work. We then calculated the sensitivity and specificity of using 15 predefined items (or combination of items) and cut-offs in classifying data abstractor experience. We considered the items/combination of items with the highest total of sensitivity and specificity as having the best accuracy.

Results: We included 45 participants; 23 were classified as less experienced and 22 as more experienced data abstractors. The item on having published three or more vs two or fewer systematic reviews had the best accuracy (sensitivity = 0.73 and specificity = 0.74) (Table).

Conclusions: Among the items/combination of items, having authored three or more published systematic reviews was the most predictive of being a more experienced data abstractor, and may help other systematic review teams form pairs for data abstractors.