Building acceptance for machine learning in study selection within a systematic review institution: Experiences from the Norwegian Institute of Public Health

Article type
Authors
Ames H1, Jardim PSJ1, Borge TC1, Hestevik CH1, Himmels JPW1, Meneses-Eschavez J1
1The Norwegian Institute of Public Health
Abstract
Background:
The Norwegian Institute of Public Health’s (NIPH) machine learning (ML) implementation team within the Cluster for Reviews and Health Technology Assessments sprung from the need to produce more health technology assessments and systematic reviews, faster, during the COVID-19 pandemic. We introduced and scaled up ML among nonspecialists through using off-the-shelf software, beginning with a 6-month team whose mandate was to explore potential benefits of ML. We then continued with a team to build acceptance and use of the tools through teaching and peer support.

Objectives:
To present our experiences with starting up and implementing an ML team for evidence synthesis. More specifically, we will present key elements of our team’s strategy, including:
- Embedding process and performance evaluations into existing commissioned reviews (i.e., how we “tested” ML functions within ongoing work)
- Creating tailored training materials and providing peer support
- Onboarding new team members and keeping up to date
- Managing workflow changes
- Fostering acceptance and implementation of ML

Conclusions:
We will conclude by discussing lessons learned and the role of a nonspecialist, implementation-focused ML network or working group in providing training to evidence synthesis organizations.