Skip to main content
. 2022 Jun 23;5(6):e2218172. doi: 10.1001/jamanetworkopen.2022.18172

Table 4. Full Model and Final Determinants of Misdiagnosed Site of Infection .

Variables Full model OR (95% CI) Reduced model ORa,b
Age (per decade) 1.2 (1.1-1.3) 1.1
Sex (male) 0.9 (0.7-1.3) Not selected
Use of immunosuppressants 1.1 (0.6-1.8) Not selected
Use of antibiotics 0.9 (0.6-1.3) Not selected
Altered mental status 1.0 (0.7-1.7) Not selected
SBP 1.0 (1.0-1.0) Not selected
CRP 1.0 (1.0-1.0) Not selected
BF 1.0 (1.0-1.0) Not selected
Occurrence of Loeb criteria 1.3 (0.8-2.2) Not selected
Dementia 1.9 (0.9-3.7) 1.4
Positive urine sediment test result without Loeb criteriac,d 1.4 (0.9-2.1) 1.1
Site of infection at ED
Abdominal 2.2 (1.3-3.7) 1.4
Bone and joint 6.1 (2.8-13.4) 3.4
CNS 11.0 (4.0-29.7) 5.1
Intravascular 12.0 (5.3-27.2) 6.7
Respiratory 1 [Reference] 1 [Reference]
Urogenital 1.6 (1.0-2.7) 1.2
Skin and soft tissue 1.1 (0.5-2.2) 1.0

Abbreviations: BF, breathing frequency; CNS, central nervous system; CRP, C-reactive protein; ED, emergency department; OR, odds ratio; SBP, systolic blood pressure.

a

Model selection is achieved by shrinkage, so the ORs with CIs cannot be determined.

b

The tuning parameter (λ) is chosen by cross-validation.25 The value of λ that gives the simplest model but also lies within 1 SE of the optimal value of λ was selected.

c

The Loeb criteria are dysuria, suprapubic pain or tenderness, frequency, or urgency. A positive urine sediment test result indicates nitrite positive or more than 5 white blood cells per high-power field.

d

Occurred in 36 of 163 misdiagnosed patients (22.1%) and 183 of 1267 nonmisdiagnosed patients (14.4%).