Skip to main content
. 2021 Mar 1;35(2):755–770. doi: 10.1111/jvim.16083

TABLE 6.

The results of a predictive logistic regression model fitted on the training data with features selected by backward stepwise elimination

Variable β Odds ratio (95% CIs) P
Intercept −10.00 4.53 e−5 (4.26 e−6 ‐ 4.33 e−4) <.001
Appetite (decreased) 2.99 19.81 (3.73‐154.60) .001
BCS (≤3)
BCS (4) −1.83 0.16 (0.05‐0.52) .003
BCS (5) −1.39 0.25 (0.08‐0.77) .02
BCS (6) −1.11 0.33 (0.10‐1.04) .06
BCS (7) −1.58 0.21 (0.05‐0.74) .02
BCS (≥8) −2.01 0.13 (0.03‐0.58) .009
Creatinine −0.02 0.98 (0.97‐0.99) <.001
Murmur (soft)
Murmur (moderate) 0.73 2.07 (1.09‐4.18) .03
Murmur (loud) 1.59 4.89 (2.59‐9.85) <.001
Murmur (thrilling) 1.93 6.89 (3.19‐15.61) <.001
Log10(NT‐proBNP) 3.66 38.98 (19.38‐81.56) <.001

Notes: Creatinine underwent analysis in Systéme International (SI) units (μmol/L), so the coefficient reported reflects the change associated with a 1‐unit increase measured on this scale. BCS, body condition score; CI, confidence intervals; e, multiply by 10x; log10, logarithmic transformation to the base 10; NT‐proBNP, N‐terminal propeptide of B‐type natriuretic peptide; P, statistical significance; β, regression coefficient. A hyphen (‐) indicates that this category was used as a reference group for comparisons. The following levels of categorical variables were chosen as reference groups to provide a suitable point of comparison: appetite, not reduced; BCS, ≤3; murmur intensity, soft. For continuous variables that have been log transformed, the OR represents the change in odds associated with a 10‐fold increase in a variable.