Active machine-learning: An iterative process whereby the accuracy of the predictions made by the algorithm is improved through interaction with reviewers as they screen additional records [1].

Artificial intelligence: Simulation of human intelligence in machines that are programmed to think like humans and mimic their actions [2].

Level 2 automation: Tools enable workflow prioritization, e.g., prioritization of relevant abstracts; however, this does not reduce the work time for reviewers on the task but does allow for compression of the calendar time of the entire process [3].

Level 4 automation: Tools perform tasks to eliminate the need for human participation in the task altogether, e.g., fully automated article screening decision about relevance made by the automated system [3].

Reviewer compatibility: A setting in systematic review software that allows you to restrict certain users from screening each other’s records. For example, if Reviewer A and Reviewer B are restricted, if Reviewer A screens a record, it will be removed from the list of records for Reviewer B. You may also assign a certain range of reference identification numbers to reviewers. These settings will ensure that two junior reviewers will not screen the same records.

Stakeholders: A person or group with a vested interest in a particular clinical decision and the evidence that supports that decision. For example, local government, or health insurance groups [4].

Training set: A set of records which contribute to the active machine-learning algorithm.