Construction and validation of tertiary models for predicting growth of Salmonella Infantis in chicken liver during a processing chain deviation

Salmonella Infantis is a top human clinical isolate that is found at low levels in chicken liver after primary processing. However, temperature abuse of chicken liver during secondary processing can lead to growth of Salmonella and higher risk of salmonellosis. Therefore, a three-phase linear, polynomial regression, tertiary model (TM PR ) and a multiple layer feedforward neural network with two nodes in the hidden layer, tertiary model (TM NN ) for growth of Salmonella Infantis in chicken liver as a function of dose (10 1 –10 6 ), time (0–8 h), and temperature (18–30°C) were constructed, validated, and compared using the criteria of the Acceptable Prediction Zones (APZ) method. When the proportion of residuals in the APZ or pAPZ was ≥0.7, predictions were considered acceptable. The pAPZ for the dependent data ( n = 360) was 0.979 for the TM PR and 0.976 for the TM NN , whereas the pAPZ for the independent data for interpolation ( n = 72) was 0.968 for the TM PR and 0.964 for the TM NN . Thus, both the TM PR and TM NN were validated for interpolation, had similar performance, and can be used with confidence to predict the growth of Salmonella Infantis in chicken liver during a secondary processing deviation of temperature abuse. However, construction of the TM PR involved three steps, whereas construction of the TM NN involved one step. Thus, the TM NN was easier to construct and validate. Nonetheless, the final TM included the TM PR and TM NN because the TM PR predicted lag time and growth rate, whereas the TM NN did not.