A Simulation Model for Predicting the Potential Growth of Salmonella as a Function of Time, Temperature and Type of Chicken

Thomas P. Oscar, Agricultural Research Service, USDA, 1124 Trigg Hall, UMES, Princess Anne, MD 21853

410-651-6062; 410-651-6568 (fax); toscar@mail.umes.edu

Abstract

The growth of Salmonella Typhimurium on the surface of autoclaved ground

chicken breast and thigh burgers incubated at constant temperatures from 8 to 48C

in 2C increments was investigated and modeled. Growth of S. Typhimurium on

breast and thigh meat was very similar. Consequently, secondary models were

developed with the combined dataset for breast and thigh meat. A hyperbola model

and a cardinal temperature model were used to model lag time and specific growth

rate, respectively, as a function of temperature. The lag time and specific growth

rate models were combined in a computer spreadsheet to create a simulation model

that predicted the potential growth (log10 increase) of S. Typhimurium on cooked

chicken as a function of time and temperature. The outputs of the simulation model

were integrated with a previously developed risk assessment model for Salmonella

to continue the process of developing an objective Process Risk Model for

assessing the microbiological safety of chicken.

Introduction

Mathematical models that predict the growth of Salmonella are limited in

their ability to predict food safety because they do not consider other pathogen

events (contamination, reduction and dose-response) that determine the exposure

and response of consumers to pathogens of food origin. In other words, growth

models only predict the potential growth of the pathogen and not the actual growth.

One way to overcome this limitation is to integrate growth models with risk

assessment models that predict the actual change in the pathogen load of a food as it

moves from farm-to-table. Recently, an approach for doing this was developed

(Oscar, 1999d) and made available on the Internet (www.arserrc.gov/mfs/) as

version 2.0 of the Poultry Food Assess Risk Model or Poultry FARM.

Objective

To develop a simulation model that provides the input settings for pathogen event 6

in the risk assessment model for Salmonella in Poultry FARM, the growth of

Salmonella on cooked chicken.

Poultry FARM is a Process Risk Model for assessing the microbiological safety of chicken. It contains simulation models for assessing the risk of salmonellosis and campylobacteriosis from chicken produced by different farm-to-table scenarios. The exposure section of the risk assessment model for Salmonella in Poultry FARM consists of six pathogen events or nodes: (1) contamination of raw chicken; (2) non-thermal inactivation during cold storage; (3) growth during distribution and meal preparation; (4) thermal inactivation during cooking; (5) recontamination of cooked chicken; and (6) growth on cooked chicken. Each pathogen event is modeled by linking two types of probability distributions. A discrete distribution is used to model the incidence of the event, whereas a pert distribution, defined by minimum, median and maximum values, is used to model the extent (log10 cycle change)

of the event.

USDA, ARSPoultry Food Assess Risk Model

Poultry FARM, Version 2.0www.arserrc.gov/mfs/

Methods

Kinetic data for development of the model were collected using a single strain of

Salmonella Typhimurium (ATCC 14028). Autoclaved ground chicken breast and thigh burgers

were inoculated on their surface with 106 cells of S. Typhimurium in a 1.2 cm2 inoculation well

and then incubated at constant temperatures from 8 to 48C in 2C increments for a total of 42

growth curves, 21 with breast meat and 21 with thigh meat. Viable cell counts were graphed as

a function of sampling time and then lag time (h) and specific growth rate (log10 CFU/h) were

determined by non-linear regression (Prism) using a two-phase linear model. Lag time was then

modeled as a function of temperature using a modified form of the hyperbola model that was

developed in this study. Specific growth rate was modeled as a function of temperature using a

cardinal temperature model (Rosso et al. 1993). The models for lag time and specific growth

rate were combined in a computer spreadsheet (Excel) to create a simulation model that

predicted the potential growth of S. Typhimurium on cooked chicken as a function of time and

temperature. Simulation was accomplished using a spreadsheet add-in program (@Risk).

Results

Growth of S. Typhimurium on cooked chicken breast and thigh burgers was very similar.

Consequently, the data for breast and thigh meat were combined and one lag time and one

specific growth rate model were developed. The lag time and specific growth rate models

developed fit the data well (Fig. 1) and produced predictions that had low bias (the median

relative error of the predictions was close to zero) and high accuracy (the mean absolute relative

error of predictions was close to zero) (Fig. 2). In the simulation model (Fig. 3), probability

distributions (pert distributions), which were defined by minimum, median and maximum

values, were used to model the lag time and specific growth rate model parameters (not shown)

and the times and temperatures of abuse. A temperature abuse scenario with the settings shown

in Fig. 3 was simulated for 10,000 iterations to demonstrate how the model could be used to

generate input settings for the previously developed risk assessment model for Salmonella.

Results of the simulation indicated that under the specified conditions of temperature abuse,

Salmonella had the potential to grow on 17.7% of the 10,000 servings of chicken simulated and

that the extent of this potential growth ranged from 1.6 x 10-4 to 1.03 log10 with a median log10

increase of 0.146.

References

Oscar, T.P., 1999. USDA, ARS Poultry Food Assess Risk Model (Poultry FARM). In: Satterfield B. (Ed.), Proceedings of the 34th National Meeting on Poultry Health & Processing, Delmarva Poultry Industry, Inc., Georgetown, 96-106.

Rosso, L., Lobry, J.R., Flandrois, J.P., 1993. An unexpected correlation between cardinal temperatures of microbial growth highlighted by a new model. J. Theor. Biol. 162, 447-463.

Fig. 1A. Hyperbola Model forLag Time (LT)

6 10 14 18 22 26 30 34 38 42 46 500

10

20

30

40

50

LT = [40.7/(T-5.2)]1.4

R2 = 0.994

Temperature (C)

Lag

tim

e (h

)

Fig. 1B. Cardinal Temperature Model forSpecific Growth Rate (SGR)

6 10 14 18 22 26 30 34 38 42 46 500.00

0.25

0.50

0.75

1.00Tmin = 5.7 CTopt = 40 CTmax = 49.3 Copt = 0.73 log10/h

Temperature (C)

SGR

(lo

g 10/

h)

Fig. 2A. Residual Plot for the Lag Time Model

10 14 18 22 26 30 34 38 42 46 50

-100

-50

0

50

100

Prediction Bias = -3.9%Prediction Accuracy = 10.1%

Temperature ( C)Rel

ativ

e E

rror

(%)

Fig. 2B. Residual Plot for theSpecific Growth Rate Model

10 14 18 22 26 30 34 38 42 46 50

-100

-50

0

50

100

Prediction Bias = 0.9%Prediction Accuracy = 8.6%

Temperature ( C)

Rel

ativ

e E

rror

(%)

Conclusions

The simulation model developed is by no means a perfect model for predicting

the growth of Salmonella on cooked chicken. Some of the important factors that

were not considered in the development of this model are: (1) strain variation,

(2) physiological state of the pathogen, (3) pathogen density, (4) competing

microorganisms, (5) fluctuating temperature and (6) cookery method. Clearly,

more work is needed to improve this model. Nonetheless, the important

advances made were the discovery of the modified hyperbola model for lag

time, the use of probability distributions and simulation in predictive modeling

and the development of a predictive model that integrates with a risk assessment

model to continue the process of creating an objective Process Risk Model for

chicken.

Abuse Scenario Pert Minimum Median MaximumTime (h) 2.3 0 2 6

Temperature (8-48oC) 22.2 20 22 25

Growth ParametersLag time (h) 3.5

Specific growth rate (log10/h) 0.261

Potential growth (log10) 0.000

Iterations 10,000

Incidence(%) Minimum Median Maximum17.7 1.64E-04 0.146 1.03

Fig. 3. Simulation Model for Growth of Salmonella on Cooked Chicken

Extent (log10 increase)Risk Model Inputs