Meta-Analysis and Systematic Review of the Measures of Psychological Safety

Liu et al · Measurement & Method, Critique & Boundary, Origins · medRxiv (preprint) · 2024 · Open access

Open this neighbourhood on the live map →

The measurement audit the field needed, and its most uncomfortable finding lands on the scale nearly everyone uses. Liu and colleagues searched five databases and included 217 studies, covering 238 independent psychological safety samples and data from 504,136 people, rating each study for the rigour of its psychometric reporting. Internal consistency was adequate, with average Cronbach's alpha between .77 and .81, but heterogeneity was extreme, an I squared of 99.92, and outside internal consistency there was very little evidence of anything: test-retest reliability, factorial validity and criterion validity were all identified as major gaps. Convergent correlations with related domains averaged r = .39.

Edmondson's scale accounted for 130 of the studies, 54.6% of the total, and was rated the least rigorous of the five named measures, scoring below average on three of five categories and no better than average on the other two. Detert and Burris's measure scored highest, despite appearing in only five studies. Because Edmondson's scale supplies more than half the samples it also drives the overall averages, so the reassuring headline and the unflattering detail come from the same place. The authors add a conceptual argument to the psychometric one: the scale was built to capture team psychological safety as a unidimensional construct, and Edmondson and Lei have themselves since argued the construct is multidimensional and interpersonal, which the low convergent correlations are consistent with.

Two facts belong beside any use of this. It is a medRxiv preprint, posted 9 February 2024, explicitly not certified by peer review and carrying medRxiv's own notice that it should not be used to guide clinical practice; no published version has appeared. And its literature search was run on 26 October 2021, so it describes the field as it stood more than two years before posting and longer now. Within those bounds it is the strongest measurement evidence this corpus holds. Every one of the 217 studies used self-report and none observed behaviour, and under 5% sampled military or paramilitary populations.

Liu, J.J.W., Ein, N., Plouffe, R.A., Gervasio, J., St. Cyr, K., Nazarov, A. and Richardson, J.D. (2024) 'Meta-analysis and systematic review of the measures of psychological safety', medRxiv preprint, 2024.02.09.24302562. Not peer reviewed.

Save this paper to a bibliography, keep notes on it, and export citations. Create a free account →

Neighbours (14)

View the source paper →