Data extraction is the process of sourcing and recording relevant results and details from the primary research studies included in the systematic review. Standardized data extraction tools facilitate the extraction of the same types of data across all of the included studies and are required for JBI systematic reviews. Reviewers should practice using the data extraction tool so they are consistently applied. The protocol should detail what data the reviewers will extract from the included studies and the data extraction tool should be attached in the appendices. Among the most important detail to extract is the decision threshold used.
As well as recording the final results of the study it is important to extract the details that inform generalizability and context of the primary studies. The STARD (Standards for Reporting of Diagnostic Accuracy) checklist and flow diagram provides detailed guidance on what studies of DTA to report and the majority of items are incorporated into the standard data extraction template that is appended to this chapter (Appendix II) (Gatsonis. 2003). You can download the STARD checklist and STARD flow diagram: http://www.stard-statement.org/
To reduce errors in data extraction it is recommended that two independent reviewers extract data and use the standardized instrument.
Studies of diagnostic test accuracy that comply with the STARD statement should include a 2×2 table that classifies patient test results and disease status as shown below (Table 9.4):
Table 9.4: Condition status (reference test results)
Index test
Outcome
Condition positive
Condition negative
Total
Index test positive
True positives (a)
False positives (b)
Test positives (a + b)
Index test negative
False negatives (c)
True negatives (d)
Test negatives (c + d)
Total
Disease/condition positives (a + c)
Disease/condition negatives (b + d)
N (a + b + c + d)
This should essentially include all quantitative data that is needed for the extraction.