Fig 1.
NetLogo views of Facilities A (upper panel) and B (lower panel) illustrating positions of and connections among agents and presence of water at a point in time during production. Circles, triangles and pentagons represent equipment surfaces in Zones 1, 2 and 3, respectively; agent water level is denoted by shape color darkness and is independent from floor conditions; employees and forklifts are denoted by specific icons (people and cars respectively); arrows represent the direction of directed agent links and lines without arrows represent undirected links; blue shaded areas represent water presence on the floor (darker colors representing puddles and lighter colors representing damp areas); brown patches denote wall-floor-junctures; grey patches denote doors, with dark grey patches being points of Zone 4 introduction; empty space is denoted by white patches; inactive space not represented by the model is denoted by black.
Table 1.
Agent characteristics by zone of agent-based models (Facilities A and B) representing two modeled packinghouses.
Table 2.
Baseline model input parameters, description, equation and distribution, summary values and sources for Listeria spp. introduction, growth, transmission, and reduction.
Table 3.
Corrective action scenarios and their virtual implementation within the agent-based models.
Fig 2.
Graphical comparison of baseline Facilities A and B using historical data at midday against simulated sampling results.
Results for Facility A and depicted in panels A, C and E, while results for Facility B are in panels B, D and F. Validation groupings investigated included (i) all agents (panels A and B), (ii) Zone category (panels C and D), and (iii) presence of water in the area (panels E and F). Lack of differences between historical (black, covering the mean (denoted with x) and 95% confidence intervals for contamination prevalence) and simulated sampling (colored boxplots, describing the mean (denoted with x), median, interquartile range, 5th and 95th percentiles and any outliers for contamination prevalence) groups indicated the model’s behavior could be considered representative of its respective facility.
Fig 3.
Boxplots describing Listeria contamination prevalence and concentration on contaminated agents on Wednesday at midday for Facility A and B baseline conditions.
Prevalence of Listeria contamination within each area of Facility A (panel A) and Facility B (panel B). Both facilities show higher prevalence in wet areas (blue) than dry areas (yellow), except for the Reject area. Log10 concentrations (CFU/cm2) of Listeria on contaminated agents within each area of Facility A (panel C) and Facility B (panel D) with median concentrations listed showing low level of contamination.
Fig 4.
Sensitivity plots of significant model input parameters against prevalence of Listeria contaminated agents at midday Wednesday of the second week of simulation for each facility.
Significant partial rank correlation coefficient (PRCC) values were determined using Bonferroni correction according to the number of parameters evaluated in Facilities A and B, respectively. (NR: Concentration of Listeria spp. per gram of contaminated raw produce (CFU/g); Pij: Probability of contact from contaminated surface in Zone i to another surface in Zone j, where i = j = Zone1, Zone 2, Zone 3 or Employee agent type; Pr: Rate of random event occurrences that introduces Listeria spp. from outside the room per hour; Pz: Probability that Listeria spp. is introduced into the room via objects from Zone 4 per hour; Rd: Prevalence of Listeria spp. in produce on day d, for d = Monday, Tuesday, Wednesday, Thursday, Friday; α: Proportion of Listeria spp. transferred to a surface upon contact with a contaminated raw produce; τij: Probability of Listeria spp. transfer from i to j agent given contact, where i = j = Zone1, Zone 2, Zone 3, or Employee agent type.; Nz: Amount of Listeria spp. introduced to an agent or patch from Zone 4 (CFU) per occurrence).
Fig 5.
Comparison of corrective action efficacy against baseline conditions in Facility A.
Efficacy was calculated using Eq 2 for each area (i.e., “wet” and “dry”) within a model for each applicable corrective action and displayed by median efficacy (red line marker) and interquartile range (black dot crossed by line marker). Positive efficacy indicated a lower Listeria prevalence in the model with a corrective action compared to the baseline model and thus effectiveness of the corrective action, while zero or negative efficacy indicated that the corrective action is predicted to not be able to reduce the agent contamination prevalence. Panel A: Facility A Wet area. Panel B: Facility A Dry area. PR_01-PR_03: Random Event Occurrence Reduction (125%; 150%; 175% event delay from baseline respectively). LR_01-LR_03: Random Load Reduction (1–3 Log10, respectively). PZ_01-PZ_04: Z4 Event Occurrence Reduction (25%; 50%; 75%; 100% reduction from baseline respectively). LZ_01-LZ_03: Z4 Load Reduction (1–3 Log10, respectively). EC_01-EC_04: Reduction of Listeria Prevalence in incoming produce (25%; 50%; 75%; 100% reduction from baseline, respectively). MI_01: Cleaning Effectiveness Improvement (increased Listeria removed during reduction events increased by 3 log10). MI_02: Weekend Deep Clean (Removal of Listeria from all agents regardless of cleanability status every Sunday). AI_01: Enhanced Flume Water Treatment (2 log10 removal of Listeria in flume agent per hour of production). AI_02/AI_02C1/AI_02C2: Broad Model-based Master Sanitation Restructuring (Agent cleaning and sanitation schedules were fully reassigned according to a mean contamination probability; Facility A was given a daily schedule for both “Cleaning Only” and “Cleaning & Sanitation” respectively). AI_03/AI_03C1/AI_03C2: Directed Model-based Master Sanitation Schedule Restructuring (Sanitization of agents with a mean contamination probability ≥66% was set to a daily frequency; Facility A was given a daily schedule for both “Cleaning Only” and “Cleaning & Sanitation” respectively). AI_04: Transmission Pathways Modification Corrective Action (Drain compartmentalization). CI_01: EC_02 and AI_02C2 were applied simultaneously. CI_02: EC_02 and AI_04 were applied simultaneously. CI_03: AI_02C2 and AI_04 were applied simultaneously.
Fig 6.
Comparison of corrective action efficacy against baseline conditions in Facility B.
Efficacy was calculated using Eq 2 for each area (i.e., “wet” and “dry”) within a model for each applicable corrective action and displayed by median efficacy and interquartile range. Positive efficacy indicated a lower Listeria prevalence in the model with a corrective action compared to the baseline model and thus effectiveness of the corrective action, while zero or negative efficacy indicated that the corrective action is predicted to not be able to reduce the agent contamination prevalence. Panel A: Facility B Wet area. Panel B: Facility B Dry area. PR_01-PR_03: Random Event Occurrence Reduction (125%; 150%; 175% event delay from baseline respectively). LR_01-LR_03: Random Load Reduction (1–3 Log10, respectively). PZ_01-PZ_04: Z4 Event Occurrence Reduction (25%; 50%; 75%; 100% reduction from baseline respectively). LZ_01-LZ_03: Z4 Load Reduction (1–3 Log10, respectively). EC_01-EC_04: Reduction of Listeria Prevalence in incoming produce (25%; 50%; 75%; 100% reduction from baseline, respectively). MI_01: Cleaning Effectiveness Improvement (increased Listeria removed during reduction events increased by 3 log10). MI_02: Weekend Deep Clean (Removal of Listeria from all agents regardless of cleanability status every Sunday). AI_01: Enhanced Flume Water Treatment (2 log10 removal of Listeria in flume agent per hour of production). AI_02/AI_02C1/AI_02C2: Broad Model-based Master Sanitation Restructuring (Agent cleaning and sanitation schedules were fully reassigned according to a mean contamination probability). AI_03/AI_03C1/AI_03C2: Directed Model-based Master Sanitation Schedule Restructuring (Sanitization of agents with a mean contamination probability ≥66% was set to a daily frequency). AI_04: Transmission Pathways Modification Corrective Action (Separation of forklift area assignment). CI_01: EC_02 and AI_03 were applied simultaneously. CI_02: EC_02 and AI_04 were applied simultaneously. CI_03: AI_03 and AI_04 were applied simultaneously.