Table 1.
Test record metadata used as features in the model, along with their datatype.
All features can be considered time-variant, with each being a property of the herd at the time of the test.
Fig 1.
(A) Receiver operating characteristic (ROC) curve for the diagnostic model.
Performance is consistently better than SICCT testing alone for all decision thresholds. (B) The decision threshold choice, such that the herd-level specificity (HSp) is maintained at the level of the SICCT test and the herd-level sensitivity (HSe) is maximised.
Fig 2.
(A) Proportion (%) of herds by area that had a negative SICCT test result, but were correctly predicted by the diagnostic model to have a confirmed breakdown, over the year 2020.
(B) Proportion of herd tests by area that were misclassified by the model in the year 2020.
Table 2.
Confusion matrices for the diagnostic model (A), compared to the SICCT test alone at herd level (B), in the year 2020.
Ground truth here refers to whether the herd had a subsequent breakdown confirmed within 90 days of testing (as opposed to true test outcomes or individual disease status).
Table 3.
Results of simulating the transmission of bTB within and between herds in two areas (Derbyshire and Devon) using an individual-based simulation model, for the existing testing regime (SICCT only) and as augmented by the diagnostic model (with model).
Confidence intervals shown in square brackets.
Fig 3.
The relative importance of model features (risk factors), as tested by SHAP importance testing, with a random control variable.
Only feature whose absolute SHAP values are significantly greater than the random feature (Mann-Whitney U test, p < 0.01) are shown. Features marked * refer to previous tests or breakdowns and will be left-censored where the previous test or breakdown is before the first date in the dataset.