Table 3

GRADE assessment of included studies

Participants (studies)	Risk of bias	Inconsistency	Indirectness	Imprecision	Publication bias	Magnitude of effect	Effects of residual confounding	Overall certainty of evidence	Study event rates (%)		Relative effect (95% CI)	Anticipated absolute effects
Participants (studies)	Risk of bias	Inconsistency	Indirectness	Imprecision	Publication bias	Magnitude of effect	Effects of residual confounding	Overall certainty of evidence	Risk of death prior to oxygen systems strengthening	Risk of death after oxygen systems strengthening	Relative effect (95% CI)	Risk of death prior to oxygen systems strengthening	Risk difference after oxygen systems strengthening
34 485 (4 observational studies)	Not serious	Not serious	Not serious	Not serious	No	Not large	No	Low	860/19887 (4.3%)	379/14598 (2.6%)	OR 0.52 (0.39 to 0.70)	43 per 1000	20 fewer per 1000 (from 25 fewer to 14 fewer)