FIGURE 2
Example of CART (Classification and Regression Tree) Output for Food Safety Inspection Data
Y es
N = 573 n = 398 + p = . 69
Y es
No
N = 89 n = 36 + p = . 40
N = 159 n = 86 + p = . 54
N = 1, 000 n = 528 p = . 53
Y es
1. SD of scores > 1. 95?
N = 151 n = 41 p = . 27
N = 427 n = 130 p = . 30
4. 2 or 3 ex tra inspections in previous year?
2. ≥ 1 ex tra inspection in previous year?
N = 62 n = 5 p = . 08
N = 268 n = 44 p = . 16
3. Average interval between inspections ≥ 241 days?
No
N = 117 n = 3 p = . 03
No ( 4 or more ex tra inspections)
Y es
No
Source: Hartman, 1992.
• All public health and clinical laboratories report each week to CDC the total number of respiratory specimens tested for influenza. • All influenza-positive surveillance sam- ples received at CDC undergo next-gener- ation sequencing. • A total of 85 million patient visits for influenza-like illnesses (ILI) for any respi- ratory pathogen were reported by >3,000 outpatient care providers in the 2020– 2021 flu season. • For patient visits to healthcare profession- als, CDC compares visits for ILI in “non- influenza weeks” and “influenza weeks.” They add 2 standard deviations to the aver- age percentage of patient visits to healthcare professionals in “non-influenza weeks.” • Each year, approximately 15,000 nursing homes report positive ILI test results for patients and sta. • All death certificates in the U.S. are tabu- lated by the National Center for Health Statistics. Pneumonia, influenza, and/ or COVID-19 (PIC) deaths currently are combined into a new PIC category, and an increase of 1.645 standard deviations above the seasonal baseline of pneumonia and influenza deaths is considered the epi- demic threshold. One big dierence between these other types of surveillance and FoodNet is the latter con-
ducts a population survey by U.S. Postal Ser- vice mail and random digit dialing—whereas for ILI, surveillance is not done for cases that do not result in a visit to a healthcare practi- tioner. As such, CDC states that their surveil- lance “does not directly provide the number of influenza illnesses” (CDC, 2023). A distinction can be made, according to Wiki- pedia, between risk factors and risk markers, both of which can have predictive power: A risk marker is a variable that is quanti- tatively associated with a disease or other outcome, but direct alteration of the risk marker does not necessarily alter the risk of the outcome. For example, driving-while- intoxicated (DWI) history is a risk marker for pilots as epidemiologic studies indicate that pilots with a DWI history are signifi- cantly more likely than their counterparts without a DWI history to be involved in aviation crashes (Risk Factor, 2022). Risk Characterization and Predictive Modeling
Top 5 Risk Factors That Contribute to Foodborne Illness
• Food from unsafe sources • Inadequate cooking • Improper hot/cold holding temperatures • Contaminated equipment • Poor personal hygiene
Source: Florida Department of Agriculture and Consumer Services (https://www. fdacs.gov/Consumer-Resources/Health- and-Safety/Foodborne-Illnesses -FAQ/ What-are-the-risk-factors-that-contribute- to-foodborne-illness).
This situation has persisted for many years mostly unchanged (Bryan, 1978). None of the risk factors requires much elaboration here because all levels of government in the U.S. and around the world have programs to address foodborne illness and are striving to improve the processes to control it. These controls, however, have not always been in place. The novel The Jungle , written by Upton Sinclair and published in 1906, was intended to expose the dangerous labor prac- tices and cruel treatment of animals in the Chicago stockyards and U.S. slaughterhouses. According to Lohnes (2022), when The Jungle
Risk Characterization and Predictive Modeling for Foodborne Disease
Risk Characterization for Foodborne Disease The risk factors for foodborne illness are well-known (see Sidebar, page 23, at right).
23
September 2025 • Journal of Environmental Health
Powered by FlippingBook