Clinical Factors Associated with Positive Active Bleeder CT Angiography in the Emergency Setting

Author

Lu Mao

Statistical analysis

Placeholder for description of statistical methods.

Characteristics by CT imaging result

Patient demographics and clinical variables are summarized by CT imaging result for bleeding status in Table 1 below. Quantitative variables are summarized by median (inter-quartile range, IQR) and categorical variables by n (%).

Table 1:

Patient characteristics by CT imaging for bleeding status.
Characteristic CT-Negative, N = 2701 CT-Positive, N = 581 p-value2
Age at time of exam 61 (48, 74) 66 (50, 73) 0.6
Gender

0.6
    Female 128 (47%) 25 (43%)
    Male 142 (53%) 33 (57%)
Race

0.11
    White 244 (90%) 54 (93%)
    Black/African American 13 (4.8%) 0 (0%)
    Asian/Pacific Islander 8 (3.0%) 4 (6.9%)
    Other/Unknown 5 (1.9%) 0 (0%)
BMI 27 (23, 32) 28 (24, 33) 0.4
Disposition

<0.001
    Admitted (incl. ICU) 184 (68%) 57 (98%)
    Discharged/Home 79 (29%) 1 (1.7%)
    Hospice/Assisted Living 3 (1.1%) 0 (0%)
    Deceased 2 (0.7%) 0 (0%)
    Inpatient 2 (0.7%) 0 (0%)
Endoscopy Positive 15 (5.6%) 10 (17%) 0.005
IR Positive 9 (3.3%) 18 (31%) <0.001
Surgery Positive 0 (0%) 1 (1.7%) 0.2
Systolic BP 130 (113, 147) 117 (101, 138) 0.016
Diastolic BP 74 (63, 82) 66 (56, 78) 0.008
Heart Rate 85 (74, 98) 84 (73, 97) 0.6
Respiratory Rate 18.0 (16.0, 20.0) 18.0 (16.0, 19.8) 0.3
Oxygen Saturation 97.00 (95.00, 99.00) 98.00 (96.00, 99.00) 0.019
Cancer 71 (26%) 15 (26%) >0.9
Hypertension 132 (49%) 32 (55%) 0.4
Diabetes 60 (22%) 12 (21%) 0.8
Coronary Artery Disease 35 (13%) 15 (26%) 0.013
Peptic Ulcer Disease 3 (1.1%) 3 (5.2%) 0.071
Alcohol Abuse 30 (11%) 5 (8.6%) 0.6
Chronic Livear Disease 25 (9.3%) 5 (8.6%) 0.9
Portal HTN (varices) 16 (5.9%) 4 (6.9%) 0.8
Diverticulosis 16 (5.9%) 8 (14%) 0.050
End Stage Liver Disease 8 (3.0%) 2 (3.4%) 0.7
H. Pylori 2 (0.7%) 1 (1.7%) 0.4
Abd surgery within 3 months 43 (16%) 15 (26%) 0.072
Hb (g/dL) (record all prior to CT) 11.00 (8.25, 13.40) 10.50 (7.60, 12.70) 0.2
Hct (g/dL) (record all prior to CT) 34 (26, 41) 32 (27, 38) 0.3
Platelets 258 (191, 331) 242 (192, 317) 0.6
BUN 18 (13, 27) 24 (14, 31) 0.093
Creatinine 0.91 (0.71, 1.13) 0.99 (0.82, 1.27) 0.037
Glucose 115 (100, 142) 129 (107, 163) 0.029
AST 21 (16, 34) 23 (16, 46) 0.3
ALT 20 (13, 32) 24 (15, 43) 0.2
aPTT 31 (26, 35) 29 (26, 37) 0.8
INR 1.20 (1.00, 1.40) 1.20 (1.00, 1.60) 0.3
Anticoagulation 59 (22%) 11 (19%) 0.6
Antiplatelets 46 (17%) 15 (26%) 0.12
SSRI 34 (13%) 11 (19%) 0.2
Steroids 22 (8.1%) 7 (12%) 0.3
Any Medication 130 (48%) 31 (53%) 0.5
1 Median (IQR); n (%)
2 Wilcoxon rank sum test; Pearson’s Chi-squared test; Fisher’s exact test

Predictive value of CT imaging against gold standard

Gold standard for active bleeding status is defined as either endoscopy, interventional radiology (IR), or surgery positive (see Table 1). A cross-tabulation between the imaging result and gold standard is shown in Table 2 below.

Table 2:

Cross-tabulation of CT imaging result and gold standard.
Gold standard Total p-value1
Negative Positive
CT Result


<0.001
    CT-Negative 248 (76%) 22 (6.7%) 270 (82%)
    CT-Positive 30 (9.1%) 28 (8.5%) 58 (18%)
Total 278 (85%) 50 (15%) 328 (100%)
1 Pearson’s Chi-squared test

Diagnostic metrics, such as sensitivity, specificity, positive predictive value (PPV), negative predictive value, and diagnostic odds ratio (DOR), along with their 95% confidence intervals (CI) are shown below.

Table 3: Diagnostic metrics of CT imaging against gold standard for bleeding status.
Metric Estimate 95% CI
Sensitivity 0.56 0.423, 0.688
Specificity 0.892 0.85, 0.923
PPV 0.483 0.359, 0.608
NPV 0.919 0.88, 0.946
DOR 10.5 5.4, 20.7

Factors associated with CT positive

A set of factors are picked from Table 1 based on their strength of asssociation with CT positivity.

Univariate analysis

First, a univariate logistic regression model is built for each factor to assess its association with CT result. The univariate odds ratios (OR; for CT positivity) are summarized in Table 4 below.

Table 4:

Univariate logistic regression analysis of CT positivity against clinical factors.
Characteristic OR1 95% CI1 p-value
Discharged

<0.001
    No 1.00 —
    Yes 0.04 0.00, 0.20
Systolic BP 0.99 0.98, 1.00 0.019
Diastolic BP 0.97 0.95, 0.99 0.006
Oxygen Saturation 1.14 1.02, 1.30 0.023
Coronary Artery Disease

0.019
    No 1.00 —
    Yes 2.34 1.15, 4.60
Diverticulosis

0.054
    No 1.00 —
    Yes 2.54 0.98, 6.11
Abd surgery within 3 months

0.084
    No 1.00 —
    Yes 1.84 0.92, 3.55
BUN 1.01 0.99, 1.02 0.28
Creatinine 1.06 1.00, 1.24 0.070
Glucose 1.00 1.00, 1.01 0.33
1 OR = Odds Ratio, CI = Confidence Interval

Multiple logistic regression

A multiple logistic regression model is fit below to evaluate each factor’s independent contribution to predicting CT positivity in the presence of others. Oxygen saturation and diverticulosis are two strongest independent predictors of a positive outcome.

Table 5:

Multiple logistic regression analysis of CT positivity against clinical factors.
Characteristic OR1 95% CI1 p-value
Discharged


    No — —
    Yes 1,758 12.7, 55,616,693,338,273,576 0.3
Systolic BP 1.00 0.98, 1.02 >0.9
Diastolic BP 1.02 0.99, 1.06 0.2
Oxygen Saturation 0.84 0.73, 0.95 0.009
Coronary Artery Disease


    No — —
    Yes 0.53 0.22, 1.30 0.2
Diverticulosis


    No — —
    Yes 0.23 0.08, 0.68 0.007
Abd surgery within 3 months


    No — —
    Yes 0.49 0.23, 1.07 0.068
BUN 1.01 0.99, 1.03 0.4
Creatinine 0.85 0.59, 0.95 0.2
Glucose 1.00 0.99, 1.00 0.2
1 OR = Odds Ratio, CI = Confidence Interval

The receiver operating characteristic (ROC) curve (by 10-fold cross-validation) of the logistic model in Table 5 for predicting CT positivity is plotted below.