Deep Learning Models to Predict Fatal Pneumonia Using Chest X-Ray Images

Table 2

Univariate association of radiographic characteristics and mortality in the original dataset (n = 1274).

Radiographic characteristics	Frequency, no, (%)	Mortality rate (%)	OR (95% CI)	value
Cardiomegaly
No	711 (55.8)	15.5		^…a
Yes	563 (44.2)	23.6	1.69 (1.26–2.26)	<0.0005
Number of lobes involved with infiltrates
1	509 (40.0)	2.8		^…a
≥2	765 (60.0)	29.9	15.08 (8.66–28.41)	<0.0001
Location of infiltrates
Unilateral	666 (52.3)	8.0		^…a
Bilataral	608 (47.7)	31.2	5.25 (3.75–7.45)	<0.0001
Pleural effusion (location)
None	773 (60.7)	12.5		^…a
Unilateral	280 (22.0)	20.4	1.78 (1.22–2.59)	<0.005
Bilateral	221 (17.3)	40.3	4.69 (3.28–6.71)	<0.0001
Cavitation
No	1237 (97.1)	18.8		^…a
Yes	37 (2.9)	27.0	1.60 (0.68–3.46)	0.21

Univariate analysis was performed by Fisher’s exact test. values of <0.05 are shown in bold. ^aellipses indicate that this variable was used as the baseline variable in the univariate analysis.