Accuracy of history, physical examination, cardiac biomarkers, and biochemical variables in identifying dogs with stage B2 degenerative mitral valve disease

FIGURE 1. A flow chart indicating how data were partitioned before their inclusion in analyses

FIGURE 2. Estimated values of the log likelihood of being in stage B2 at different levels of categorical variables present in the explanatory multivariable model. Estimated marginal means were calculated when other variables in the explanatory multivariable logistic regression were held at a constant (continuous variables = mean, categorical variables = reference group). Estimated marginal means were compared using Fischer's Least Significant Difference test and significant pairwise comparisons are displayed using asterisks (*<.05, **<.01, ***<.001). Dots are used to indicate the estimated marginal mean and error bars are used to indicate the standard error of the mean

FIGURE 3. A receiver operating characteristic curve for the explanatory multivariable logistic regression analysis of risk factors associated with having stage B2 disease. 95% confidence intervals are represented by the blue area around the receiver operating curve. Area under the curve: 0.84 (0.82‐0.87)

FIGURE 4. Discriminatory performance of the explanatory multivariable logistic regression model when applied to the clean, complete and confounded samples. The area under the receiver operating characteristic curves were as follows: for the clean sample 0.84 (0.82‐0.87); for the complete sample 0.81 (0.79‐0.83); for the confounded sample 0.76 (0.72‐0.80)

FIGURE 5. Discriminatory performance of the explanatory multivariable logistic regression model in comparison to using NT‐proBNP alone or the vertebral heart score. The area under the receiver operating characteristic curves were as follows: multivariable logistic regression model 0.84 (0.82‐0.87); univariable NT‐proBNP model 0.77 (0.74‐0.80); univariable vertebral heart score model 0.76 (0.69‐0.83). NT‐proBNP, N‐terminal propeptide of B‐type natriuretic peptide; VHS, vertebral heart score

FIGURE 6. Variable importance plots displaying the 5 most important variables for the following classifiers: A, ridge regression; B, support vector machine with a linear kernel; C, Random Forest; and D, Gradient Boosting Machine (XGBoost). Variable importance is presented relative to the most important predictor in each model. Scores for ridge regression and the support vector machine with a linear kernel are the coefficients for each variable. Scores for random forest and gradient boosting machine are the mean Shapely value for each variable. BCS, body condition score; BUN, blood urea nitrogen; NT‐proBNP, N‐terminal propeptide of B‐type natriuretic peptide