A foodborne disease outbreak is defined as two or more similar illnesses resulting from consumption of a common food (Council of State and Territorial Epidemiologists, 2011). CDC’s PulseNet has used pulsed-field gel electro-phoresis (PFGE) to identify outbreaks where Salmonella serotype PFGE patterns are isolated more frequently than expected (CDC, 2018b). Once investigated, federal, state, and local agencies submit data on all foodborne disease outbreaks to CDC’s Foodborne Disease Outbreak Surveillance System (FDOSS) (CDC, 2018a). Outbreaks reported to FDOSS are attributed to foods based on the results of an epidemiological investigation or one or more of the following additional pieces of evidence: microbiological evidence of the pathogen in the suspect food, traceback investigation that identifies a common source of contamination, findings from an environmental assessment, or other data. FDOSS then categorized outbreak vehicles according to the Interagency Food Safety Analytics Collaboration Project (IFSAC) guidelines (Richardson et al., 2017). The major food category was used to identify produce outbreaks and specific food subtype was used to assign each produce item to a fruit or vegetable category.

During multistate outbreak investigations, PulseNet is used to identify illnesses that are part of an outbreak. In these investigations, officials in state and local health departments have provided additional epidemiologic information on ill people included in the investigation to CDC’s Outbreak Response and Prevention Branch (ORPB). These data are maintained in line lists of ill people associated with an investigation.

We included Salmonella outbreaks in the analysis that were identified by CDC’s PulseNet, investigated during the period 2009–2015, and were multistate outbreaks (contaminated food item caused illnesses in two or more states). These years were selected because complete data on outbreak line lists were available for analysis. We excluded outbreaks that were linked to shelf-stable produce-derived foods or products, including ground pepper, moringa leaf powder, and sprout seeds (all other outbreaks linked to sprouts were included). We also excluded outbreaks where the contaminated food item was sold frozen or otherwise shelf stable.

We requested data from FDOSS and ORPB for Salmonella outbreaks meeting the inclusion criteria, including line lists of cases, case demographic characteristics, case-patient specimen Salmonella isolation dates, case-patient states of residence, investigation start dates, and whether the outbreak-associated produce item was recalled. Outbreaks were analyzed together and also stratified by fruit versus vegetable source. To analyze differences between fruit and vegetable outbreaks, we used Wilcoxon rank-sum tests. We assessed age and gender for outbreak-associated case-patients, duration and geographic spread for each outbreak, and whether a product recall occurred. To identify aggregate median case-patient age, we calculated median ages for each outbreak, creating a distribution of medians. We then calculated a median of median ages. Age data were also categorized into groups (<5, 5–17, 18–59, and >59 years). We then compared age group percent distributions for fruit and vegetable outbreaks with the U.S. Census Bureau on U.S. population age distribution for 2012 (U.S. Census Bureau, 2018b). We chose 2012 because it is the midpoint year of our analysis. To assess geographic spread, we assigned a case location based on the reported state of patient residence and summarized the number of states involved per outbreak. In-depth geospatial mapping of multistate foodborne outbreaks is described elsewhere and was not conducted in this analysis (Chen et al., 2017).

We calculated median values for velocity metrics for Salmonella outbreaks with 25 or more illnesses, including outbreak duration (time from first isolation date to last isolation date); time to 25 illnesses; time to nth percentile of illnesses per outbreak (25th, 50th, and 75th percentiles); average illness onsets per day; outbreak peak number of illnesses; time to peak illnesses defined as the day with the highest number of illnesses identified (limited to outbreaks that had a day with four or more illnesses); and median time from the day of peak illnesses to the last outbreak illness. We used outbreaks of 25 or more illnesses to adequately ascertain characteristics of outbreak epidemic curves. A minimum of four illnesses was chosen for calculation of outbreak peak illnesses because lower thresholds resulted in multiple peaks. Outbreaks with peaks lower than four illnesses per day were not included in the analysis of time to the day of peak illnesses or time from the day of peak illnesses to the day of last outbreak illness.

Of note, we based all dates on reported case Salmonella isolation dates, which represent when a culture specimen was collected. This differs from the illness onset date, which is considered to be an average of 3 d before the isolation date. However, the actual period can vary in length among cases because of other factors such as when individuals visit a health care provider.

We plotted epidemic curves for Salmonella outbreaks of 25 or more illnesses included in this analysis. Outbreaks were standardized to the first isolation date as day zero. Each subsequent illness was plotted as the number of days following the isolation date for the first illness in the outbreak. We then grouped outbreaks by category (fruit or vegetable) and displayed them by outbreak vehicle. Data were obtained from FDOSS and from ORPB on February 8, 2017. We performed all analyses using RStudio, version 3.4 (RStudio).

From 2009 to 2015, 38 Salmonella outbreaks related to produce were reported to CDC (Table 1). Fourteen were linked to fruits and 24 to vegetables. Fruit outbreaks were reported in all years except 2015, with the highest number of outbreaks (n = 5) occurring in 2012. Seven (50%) fruit outbreaks were related to cantaloupes, two each (14%) to man-goes and papayas, and one each (7%) to grapes, blueberries, and melons. Vegetable outbreaks were reported in all years, with the highest number of outbreaks occurring in both 2014 and 2015 (four each year). One vegetable outbreak started in 2008, but the outbreak was identified in 2009, so it was counted as a 2009 outbreak. Eight (33%) vegetable outbreaks were related to sprouts, six (25%) to tomatoes, five (21%) to cucumbers, three (13%) to lettuce, and one each (4%) to minipeppers and green onions.

Of the 38 outbreaks, 12 (32%) investigations resulted in a product recall. Of these, seven were vegetable products (six—sprouts; and one—cucumber) and five fruit products (three—cantaloupe, one—mango, and one—papaya). A median of 74% (range 22–100%) of outbreak cases occurred before the investigation start date. Among outbreaks with a recalled product, a median of 70% (range 23–100%) of cases occurred before the investigation start date.

Table 2 provides selected demographic characteristics for ill people in the 38 Salmonella outbreaks included in this analysis. Among all outbreaks, the median number of illnesses was 30 (range 3–907). The median percentage of ill people who were female was 61.9% (range 30–100%), and the median of outbreak median ages was 34 years. In 2012, the midpoint year of our analysis, the median U.S. population age was 37 (U.S. Census Bureau, 2018a). By outbreak age group, a median of 9.3% of ill people were aged <5 years, 11.1% 5–17 years, 53.2% 18–59 years, and 19.4% 60 years or older. U.S. census data from 2012 provide population estimates in these age cohorts (<5: 6.4%; 5–17: 17.1%; 18–59: 57.1%; and ≥60, 19.4%) (U.S. Census Bureau, 2018b). Illnesses were reported from a median of 11 states (range 2–40 states) per outbreak.

Overall, characteristics of Salmonella outbreaks linked to contaminated fruits and vegetables were similar. Although not statistically different, fruit outbreaks tended to have fewer illnesses than vegetable outbreaks (median 26 vs. 45, p = 0.4). The median proportion of females was similar when comparing fruit and vegetable outbreaks (median % 59.6 vs. 65.5, p = 0.21). The median age of ill people was not significantly different between fruit outbreaks and vegetable outbreaks (48 vs. 31, p = 0.20). However, fruit outbreaks had a higher median percentage of ill people in the <5 (15.7% vs. 7.3%, p = 0.01) and >59 age groups (33% vs. 16.3%, p < 0.01). The median percentage of ill people in the 18–59 age group was lower for fruit outbreaks than vegetable outbreaks (34.8% vs. 59.1%, p < 0.01).

Table 3 shows measures of outbreak velocity among Salmonella outbreaks associated with contaminated fruits and vegetables with 25 or more illnesses. Median overall outbreak duration was 77 d (range 35–242 d) or 11 weeks. Median time to the first 25 illnesses was 35 d (range 5–96 d). Additionally, median time to the 25th, 50th, and 75th percentiles of illnesses per outbreak was 25, 33, and 45 d, respectively. Using time to the measured percentiles, median quartile times were as follows: start to 25th = 25 d; 25th–50th = 7.5 d; 50th–75th = 12.5 d; and 75th to end = 22 d. Overall, median average illnesses per day was 0.82 (range 0.33–3.83), median peak daily number of illnesses was 9 (range 4–30), and median days to the peak number of illnesses was 38 (range 7–137 d). Finally, the median time from the day of peak outbreak illnesses to the last outbreak illness was 56 d (range 17–199 d).

When comparing fruit and vegetable outbreaks, we did not identify significant differences in outbreak duration, time to 25 illnesses, or time to 25th, 50th, or 75th percentiles of illnesses. Additionally, there were no significant differences when comparing median average illnesses per day (0.7 illnesses/day vs. 0.97 illnesses/day, p = 0.29), median peak illnesses (4 illnesses vs. 10 illnesses, p = 0.18), median time to peak illnesses (44 vs. 35.5 d, p = 0.96), or median time from day of peak illnesses to the last outbreak illness (56 vs. 57 d, p = 0.61).

All Salmonella outbreaks with 25 or more illnesses are represented as epidemic curves in Figure 1a (fruit outbreaks) and Figure 1b (vegetable outbreaks). The highest number of illnesses on a single day for fruit outbreaks was 16 illnesses for a cantaloupe outbreak and 11 illnesses for a mango outbreak. Seven of eight (88%) fruit outbreaks were over within 20 weeks (first illness date to last illness date). The highest number of illnesses on a single day for vegetable outbreaks was 30 illnesses for a cucumber outbreak and 19 illnesses for a sprout outbreak. Twelve of 16 (75%) vegetable outbreaks were over within 20 weeks of the first illness.