Generic global regression models for growth prediction of Salmonella in ground pork and pork cuts

Tasja Buschhardt, Tina Beck Hansen, Martin Iain Bahl, Donald W. Schaffner, Søren Aabo

Research output: Chapter in Book/Report/Conference proceedingConference abstract in proceedingsResearchpeer-review

186 Downloads (Pure)

Abstract

Introduction and Objectives
Models for the prediction of bacterial growth in fresh pork are primarily developed using two-step regression (i.e. primary models followed by secondary models). These models are also generally based on experiments in liquids or ground meat and neglect surface growth. It has been shown that one-step global regressions can result in more accurate models and that bacterial growth on intact surfaces can substantially differ from growth in liquid culture.
Material and Methods
We used a global-regression approach to develop predictive models for the growth of Salmonella for three pork matrices: on the surface of shoulder (neck) and hind part (ham), and in ground pork. We conducted five experimental trials and inoculated essentially sterile pork pieces with a Salmonella cocktail (n = 192). Inoculated meat was aerobically incubated at 4 °C, 7 °C, 12 °C, and 16 °C for 96 h. One part of obtained logtransformed cell counts was used for model development and another for model validation. The Ratkowsky square root model and the relative lag time (RLT) model were integrated into the logistic model with delay. Fitted parameter estimates were compared to investigate the effect of meat structure on bacterial growth and goodness-of-fit was evaluated by root mean squared errors (RMSE). We used the Acceptable Simulation Zone (ASZ) approach and cross-validation with model-independent data to investigate if generic predictive models could accurately describe microbial growth across all studied pork products and compared our models to already existing generic models.
Results
Our results indicated that the growth of Salmonella was affected by product characteristics such as pH and structure, but storage temperature was shown to be the only variable needed to predict growth independent of pH and structural differences. RMSE of 0.54 suggested acceptable goodness-of-fit for the Salmonella generic growth model. Model evaluations of the generic growth model showed that described growth responses on pork neck and in ground pork were highly accurate with 86 and 98% of all model independent observations within the ASZ, respectively. Although growth descriptions showed less accuracy in the case of pork ham, a fail-safe model could still be developed. Model evaluation also showed that our model performed better than generic existing models.
Conclusions
We suggested that generic model with fewer variables might provide a more suitable approach to bacterial growth modeling in fresh pork if pH and the type of pork product are unknown. Our study provides a “readyto-use” global regression model relevant for a wide range of time and temperature combinations and various fresh pork products. The model should be a useful tool to control growth of Salmonella in meat and set critical limits for temperature during production and storage of fresh pork.
Original languageEnglish
Title of host publication10th International conference on predictive modelling in food
Number of pages1
Place of PublicationCordoba, Spain
Publication date2017
Article number020
Publication statusPublished - 2017
Event10th International Conference on Predictive Modelling in Food: ICPMF10 - Cordoba, Spain
Duration: 26 Sep 201729 Sep 2017
Conference number: 10
http://www.icpmf10.com/

Conference

Conference10th International Conference on Predictive Modelling in Food
Number10
Country/TerritorySpain
CityCordoba
Period26/09/201729/09/2017
Internet address

Fingerprint

Dive into the research topics of 'Generic global regression models for growth prediction of Salmonella in ground pork and pork cuts'. Together they form a unique fingerprint.

Cite this