Table 2.

Analytical approaches used in HoPeNET machine learning algorithm development

Product/goal: creation of data-trained predictive tool for examining barriers in future trials that will inform changes in recruitment, screening and enrollment of AA/Black participants
Analysis	Data/tools
Correlation analysis (Corrplot)	Self-reported self-efficacy/engagement between initial/terminal time points with CAB outcome assessments
t-distributed stochastic neighbour embedding	Initial and final participant self-efficacy, knowledge, engagement, barriers and attitudes
Natural language processing	Transcribed community and investigator focus group data to identify within and between group differences and similarities in attitudes, knowledge, perceptions of bias, etc.
Structural equation modelling (SEM-LAVAAN) and group-based modelling (GBM-CrimCV)	Phase 2 focus group attitudes to clinical trial participation among community and investigator group models.
Path analysis	Path analysis of phase 3 changes in perceptions on group-based model.

AA, African-American; CAB, community advisory board.