Table 4.

Bronchoalveolar lavage input for diagnosis and decision making

	Standard oxygen therapy N = 56a	High-flow nasal cannula oxygen therapy or non-invasive ventilation N = 49b	Invasive mechanical ventilation N = 378	p valuec
BAL fluid of good qualityd as declared by the pathologist	42 (75)	36 (73.5)	249 (65.9)	0.26
Highest class of BAL input reached
Class 1: brings no useful information	19 (33.9)	17 (34.7)	149 (39.4)	0.63
Class 2: in line with a diagnosis already mentioned	3 (5.4)	6 (12.2)	48 (12.7)	0.28
Class 3: suggests a diagnosis not previously envisaged	2 (3.6)	1 (2)	11 (2.9)	0.90
Class 4: allows discontinuing one or several treatments	11 (19.6)	6 (12.2)	20 (5.3)	< 0.01
Class 5: brings definitive diagnosis and/or allows new therapy initiation	21 (37.5)	19 (38.8)	150 (39.7)	0.95
Class 4 or 5 reachede	32 (57.1)	25 (51)	170 (45)	0.20

Numbers indicate counts and (%)

BAL bronchoalveolar lavage

^aIncluding 45 (80.4%) patients under oxygen therapy via standard nasal cannula or non-rebreathing mask, and 11 patients (19.6%) under oxygen therapy via rebreathing mask

^bIncluding only four patients (8.2%) under non-invasive ventilation

^cGroups were compared using χ² test

^dSee text for definition of “good quality” for BAL fluid

^eClass 4 or 5 BAL contribution to diagnosis and decision making was used as the outcome measure in logistic regression; see text for details