Epidemiology of Salmonellosis Among Infants in the United States: 1968–2015

The Centers for Disease Control and Prevention (CDC) conducts United States surveillance for culture-confirmed salmonellosis. Clinical diagnostic laboratories submit isolates from ill persons to public health laboratories, where, during the study period, they were confirmed and serotyped by the Kauffmann-White Scheme.¹²  The CDC Laboratory-based Enteric Disease Surveillance (LEDS) system conducts passive, national laboratory surveillance of salmonellosis in all states and the District of Columbia. Data include self-reported demographics, serotype, and specimen source. Since 1996, the Foodborne Diseases Active Surveillance Network (FoodNet) has conducted population-based salmonellosis surveillance; since 2004, the catchment area has been stable and included Connecticut, Georgia, Maryland, Minnesota, New Mexico, Oregon, Tennessee, and selected counties in California, Colorado, and New York (approximately 15% of the United States population). Investigators attempt to collect demographic, hospitalization, and outcome data on every salmonellosis case via patient/parent interviews, electronic laboratory reports, and other health records; methods of collecting data have not changed over this period. Each surveillance system was analyzed separately; both systems count only 1 case when specimens collected ≤30 days apart yielded the same pathogen. When isolates from multiple specimen sources are available for the same case, data from the source indicating the most invasive infection (cerebrospinal fluid > blood > stool) are analyzed. We excluded cases after 2015 because the increasing use of culture-independent diagnostic tests for Salmonella increases pathogen detection but decreases availability of isolates for serotyping.^13,14 

We defined infants as persons aged <1 year. We defined syndromes based on specimen source: gastroenteritis from stool, bacteremia from blood, and meningitis from cerebrospinal fluid. We excluded cases with isolates from other sources. We combined meningitis and bacteremia as invasive infections in some analyses because of sparse data.

We identified cases of Salmonella gastroenteritis, bacteremia, and meningitis reported to national surveillance (LEDS) during 1968–2015 and compared syndromes by incidence, age, sex, month of specimen collection, and serotype. Race was available for only 11% and ethnicity for 6% in LEDS, and they were not assessed. We report incidence trends using 4-year age intervals; we used 12-year intervals for serotypes because of greater year-to-year variability. Because LEDS did not collect birth data until 1995 and reporters might have rounded age to months (eg, 4 months and 20 days to 5 months), we minimized any resulting bias by reporting median age with first and third quartiles per time interval or used categories (<1–2, 3–5, 6–8, and 9–11 months). We assessed seasonality by comparing the percentage of cases in each month with the expected percentage (8.3%) if cases were evenly distributed.

We assessed incidence, age, sex, and patient outcomes, including hospitalization, length of hospitalization, and death, using FoodNet data reported during 1996–2015. Comparisons by race were limited to 2004–2015 because data were missing for >20% of cases for most earlier years; ethnicity was not assessed because data were missing for >20% of cases for most years. Hospitalization was defined as admission occurring within seven days of specimen collection. Patient outcome (alive, dead, or unknown) was determined by interview or medical chart review and recorded 7 days after specimen collection or at hospital discharge, whichever was later.

We used census population estimates for infants during 1990–2015 to estimate annual incidence by syndrome for national surveillance. During 1970–1989, census estimates of infants were unavailable. We estimated infant population size during 1970–1989 by applying the mean proportion of children aged 0–4 years who were infants during 1990–2015 to census estimates for children aged 0–4 years during 1970–1989. We estimated annual incidence by syndrome for FoodNet using census population estimates for FoodNet surveillance sites. We report incidence trends by race using four-year intervals; incidence estimates were limited to groups with ≥10 cases of invasive salmonellosis. We express all incidence estimates per 100 000 infants. We compared incidence rate ratios (IRRs) and 95% confidence intervals (95% CI) for boys and girls in national surveillance (1990–2015) and FoodNet (1996–2015). We compared IRRs for Black and Asian infants with white infants in FoodNet (2004–2015). We selected the referent group based on largest infant population size and lowest incidence rate for invasive salmonellosis. We compared continuous variables using 95% CIs and proportions using a 1-sample test for binomial proportion and Fisher exact test with the mid-P method. Analyses were conducted using SAS version 9.4 (SAS Institute, Inc., Cary, North Carolina), OpenEpi version 3.01 (Dean et al., Emory University, Atlanta, Georgia), and Microsoft Excel 2013.

We identified 190 627 culture-confirmed Salmonella infections among infants reported to national surveillance during 1968–2015, which composed 165 236 (86.7%) cases of gastroenteritis, 6767 (3.5%) bacteremia, 371 (0.2%) meningitis (Table 1), and 18 253 (9.7%) with other or unknown specimen sources. The median annual incidence (range) during 1970–2015 was 93 (27–162) gastroenteritis cases, 3.9 (1.4–6.7) bacteremia cases, and 0.21 (0.07–0.49) meningitis cases (Table 1). Incidence of each syndrome increased during the late 1970s or the 1980s, declined during the 1990s–early 2000s, and gradually has increased since the mid-2000s (Fig 1).

Cases were highest during August–October for gastroenteritis (10.9% to 11.4%) and bacteremia (10.1% to 10.8%). The number of meningitis cases was much higher than expected (if evenly distributed by month) in October (17.8% vs 8.3%; 95% CI, 14.0–22.1). (Fig 2). This increase was not associated with a specific serotype, period, or age (data not shown).

Median (first–third quartile) ages were 4 (3–7) months for gastroenteritis, 4 (2–7) months for bacteremia, and 2 months (1–3) for meningitis (Table 1). Comparing national surveillance data for 1968–1971 with 2012–2015, median age increased from 4 (1–7) to 5 (3–8) months for gastroenteritis and from 3 (1–4) to 4 (<1–8) months for bacteremia but decreased from 1 (1–4) to <1 (<1–2) month for meningitis. The decrease in age with meningitis occurred primarily because of an increase in cases among infants aged <3 months during 2004–2015 (Fig 1). Incidence of gastroenteritis was slightly higher for boys (73) than girls (67) (IRR, 1.1; 95% CI, 1.1–1.1), as was incidence of bacteremia (4.0 vs 3.6; IRR, 1.1; 95% CI, 1.0–1.2) and meningitis (0.20 vs 0.16; IRR, 1.2; 95% CI, 0.9–1.6) (Table 1).

During 1968–2015, the most frequently reported serotypes for gastroenteritis were Typhimurium (21.5%), Newport (8.7%), and Heidelberg (7.4%). The most frequently reported serotypes for invasive salmonellosis were Heidelberg (28.3%), Typhimurium (12.6%), and Enteritidis (10.2%) (Table 2). Comparing serotypes during 1968–79 with 2004–2015, the percentage of infections decreased from 23.2% to 12.3% for Typhimurium, 8.8% to 2.8% for Heidelberg, and 6.0% to 2.1% for infantis. The percentage of infections increased from 7.3% to 12.8% for Newport and 2.2% to 6.8% for Javiana. However, the percentage of unknown serotype or partially serotyped isolates increased, from 1.0% to 10.7%. The percentage that was Enteritidis was stable at 5.4%. Serotype 4,[5],12:i:- emerged after the 1990s, accounting for 2.5% of cases during 2004–2015. During 2012–2015, Newport (13.9%), Typhimurium (9.9%), and Javiana (7.9%) were the most frequently identified serotypes, but Heidelberg (14.2%) and Enteritidis (8.8%) caused the most invasive infections.

During 1996–2015, FoodNet sites reported 13 151 cases of infant salmonellosis, including 12 241 (93%) cases of gastroenteritis, 624 (4.7%) bacteremia, 25 (0.2%) meningitis (Table 1), and 261 (2%) with other or unknown specimen sources. Median annual incidence (range) for gastroenteritis was 120 (101–149), bacteremia 6.2 (3.6–9.8), and meningitis 0.25 (0.00–0.84) (Table 1). Incidence of bacteremia decreased from 7.6 (95% CI, 6.2–9.1) during 1996–1999 to 5.1 (95% CI, 4.0–6.1) during 2012–2015; incidence of overall salmonellosis, gastroenteritis, and meningitis did not change significantly over time. Incidence of meningitis was 1.4 times higher for boys than girls (0.29 vs 0.20; 95% CI, 1.1–1.9), incidence of gastroenteritis was 1.1 times higher (126 vs 113; 95% CI, 1.1–1.1), and incidence of bacteremia was 1.1 times higher (6.2 vs 5.9; IRR, 1.1; 95% CI, 1.0–1.1).

During 2004–2015, race data were available for 413 (90%) infants with invasive salmonellosis and 8018 (88%) of infants with gastroenteritis. Compared with white infants, incidence of invasive salmonellosis was 2.7 times higher for Black infants (10.4 vs 3.9; 95% CI, 2.6– 2.8) and 1.8 times higher for Asian infants (6.8 vs 3.9; 95% CI, 1.7–1.8). Compared with white infants, the incidence of gastroenteritis was 1.2 times higher for Black infants (119 vs 99; 95% CI, 1.19–1.21) and 0.89 times lower for Asian infants (88 vs 99; IRR, 0.89; 95% CI, 0.88–0.90).

Comparing 2004–2007 with 2012–2015, incidence (95% CI) of gastroenteritis increased from 87 (83–92) to 102 (97–107) for white infants, decreased from 127 (117–138) to 105 (96–115) for Black infants, and increased from 68 (54–84) to 107 (90–126) for Asian infants. During that time, incidence of invasive salmonellosis increased from 2.8 (2.1–3.6) to 4.3 (3.4–5.4) for white infants, decreased from 15.8 (12.4–19.9) to 6.7 (4.6–9.5) for Black infants, and increased from 4.3 (1.6–9.5) to 7.7 (3.9–13.8) for Asian infants (Fig 3).

Among infants with gastroenteritis with information on hospitalization and death, 2790 (26%) were hospitalized and 7 (0.1%) died. Among infants with bacteremia, 407 (70%) were hospitalized and 9 (1.6%) died. Among infants with meningitis, 22 (96%) were hospitalized and 1 died. Among infants who were hospitalized, the median (interquartile range) length of hospital stay was 2 (2–3) days for gastroenteritis, 4 (3–7) days for bacteremia, and 20 (10–33) days for meningitis (Table 1).

These analyses characterize the epidemiology of infant salmonellosis by syndrome using 48 years of passive national surveillance data and 20 years of active sentinel surveillance data. Both systems show that infants continue to experience substantial morbidity from salmonellosis, including gastrointestinal and invasive illness. However, the incidence of bacteremia has decreased since the 1990s. Infants with meningitis were younger than those with bacteremia or gastroenteritis, and serotype Heidelberg caused the highest percentage of invasive infections, even in the most recent 4 years when the overall incidence of Heidelberg infections markedly decreased.¹¹ 

The annual incidence varied over time for all syndromes. Incidence estimates based on primarily passive (national) surveillance suggest moderate increases for each syndrome during the past 20 years. Incidence estimates based on active (FoodNet) sentinel site surveillance are 25% to 50% higher than estimates from passive surveillance and suggest steady rates for meningitis and gastroenteritis and a decreasing rate for bacteremia. Different patterns over time might be due to better reporting from FoodNet active surveillance sites or to different population structure, patterns of diagnosis, or reporting compared with the entire country.

These data do not provide information to assess sources of Salmonella or risk factors for invasive disease; demographic characteristics are described as risk indicators. Reported risk factors for infant salmonellosis include exposure to animals, exposure to a household member or attending day care with a child with recent diarrhea, international travel, riding in a shopping cart with raw meat or poultry, and other indoor and outdoor environmental exposures.^15–18  Use of infant formula and lack of breastfeeding are correlated risk markers.^15,19 Salmonella has only rarely been identified in commercial formula.^20,21  Breastfeeding might be protective because of immunity (eg, maternal antibodies from breastmilk) or to lack of exposure to bottles contaminated by household members.

Infants are at increased risk of invasive infection.^22,23  Reported risk factors for extraintestinal salmonellosis include age, serotype, other infections such as HIV and rotavirus, and underlying health conditions such as sickle cell anemia and gastrointestinal conditions.^22–25  In children, invasive infections are often secondary to gastroenteritis; decreased gastric acidity or gut immaturity may increase vulnerability to invasive infections.^25–28 

The median age of infants differed by syndrome and over time; infants with meningitis are younger than infants with gastroenteritis and bacteremia, and the average age for meningitis decreased over time, primarily from an increase in meningitis among infants <3 months of age over the past 2 decades. A similar increase in bacterial meningitis from all causes was reported for infants <2 months old from 1998–1999 to 2006–2007, whereas incidence of bacterial meningitis decreased over the same period for all other age groups.²⁹  Increased survival of premature infants,³⁰  changes in exposure patterns, or changes in testing or care-seeking practices might contribute to the increase in meningitis cases among the youngest infants.

Serotypes Heidelberg and Enteritidis accounted for a larger proportion of invasive salmonellosis than of gastroenteritis, and Typhimurium accounted for a smaller proportion. However, Typhimurium still accounts for a substantial burden of infant invasive infections. Our findings are consistent with other reports that Heidelberg and Enteritidis are more likely to cause invasive infections than Typhimurium and Newport.^31,32  Among all persons, both Heidelberg and Typhimurium incidence have decreased dramatically in the past 20 years, most likely because of vaccination of poultry.^14,33,34 

During 2012–2015, Typhimurium and Newport were the most common serotypes causing infant salmonellosis. The number and percentage of cases caused by Heidelberg decreased since the 1980s for gastroenteritis, but Heidelberg and Enteritidis still cause 23% of invasive infections in infants. Heidelberg, Typhimurium, and Enteritidis infections among adults are often associated with egg and poultry products, and Newport infections are often associated with meat products. The extent to which these sources contribute to infant salmonellosis is unknown. However, declines in serotypes causing infant salmonellosis paralleling declines in overall salmonellosis suggest that food is the ultimate source. It seems likely that infants are exposed from contamination of their food or environment directly from food (eg, poultry, meat, eggs), or from the fecal matter of infected caretakers. Avenues for further exploration include whether infant formula could be a more common source than is recognized and whether the presence of certain food products in the home or the food consumption habits of caretakers are risk factors for infant salmonellosis.

Although infant ages differed, serotype distributions for meningitis and bacteremia were similar, suggesting that infants who develop meningitis and bacteremia are exposed to a similar distribution of sources, and that age-related factors influence clinical presentation. In contrast, infants with bacteremia have a similar age distribution but a different serotype distribution from infants with gastroenteritis, suggesting that differences in factors not related to age (eg, exposure to particular sources, susceptibility to particular serotypes) play a greater role in determining clinical presentation. Further study is needed to determine whether serotype differences by syndrome can be attributed to different exposure sources or routes, host factors, or invasiveness of serotypes.

Previous analyses using FoodNet data identified racial differences in incidence of Salmonella gastroenteritis and invasive salmonellosis among infants and showed that differences by race decreased between 1996 and 2008.^22,23  Our findings indicate that differences by race continued to decrease after 2008. The rate of invasive salmonellosis declined among Black infants but increased among Asian infants. Social determinates of health, including income inequality, discrimination, and differences in health care access and use, are drivers of poor health and might contribute to differences by race.^35–38  More information about social determinants of health is needed to better understand the reasons for differences by race and improve prevention measures. Differences in exposures, including differences in breastfeeding initiation and duration, and underlying health conditions might also contribute to differences by race.^23,25,39  Focusing on decreasing exposures that lead to illness and identifying subpopulations of infants with host factors that put them at increased risk for illness could help to decrease these illnesses. It is also possible that changes by race reflect changing access to and use of health care or diagnostic practices, rather than a true change in incidence. Salmonellosis incidence was slightly higher among boys than girls across syndromes and surveillance systems, consistent with previous reports.^22,23 

Infant salmonellosis peaked in the summer and fall for each syndrome, but the pattern differed for meningitis compared with gastroenteritis and bacteremia. Previous reports have found that Salmonella gastroenteritis and bacteremia are seasonal, occurring more frequently during warm and rainy seasons, suggesting that environmental factors play a role.^8,26–28  However, the distinct spike in meningitis cases in October is a new finding, and its implications are unknown.

Limitations of these analyses include that salmonellosis is underdiagnosed and underreported² ; the degree likely differs by syndrome because persons with more severe illness are more likely to receive medical care.^2,40  For bacteremia and meningitis, underestimation may occur because of treatment with antibiotics occurring before collecting specimens. The data do not provide information about factors that may contribute to the observed differences by race. Changes over time could be partly because of unrecognized changes in diagnosis and reporting of illness and race. The completeness and representativeness of passive national surveillance data, including age, varies over time and by reporting site. Extrapolated infant population size for 1970–1989 might bias incidence estimates if variability in birth rates or child mortality changed pediatric population distribution over time; however, birth rates were relatively stable throughout most of the 1970s–1990s, and child mortality improved for all ages.^41,42 

All syndromes of salmonellosis are associated with substantial morbidity and mortality among infants. We observed a higher rate of invasive illness among infants identified as Black and Asian than white. Although infants with meningitis had the highest case fatality and hospitalization rates, the largest mortality burden occurred among those with bacteremia, and the total number of hospitalization days was highest for infants with gastroenteritis. In addition to acute morbidity, risk of death, and stress to families, invasive salmonellosis among infants can be associated with moderate to severe sequelae, including fine motor, cognitive, behavioral, neuropsychologic, and auditory dysfunctions over a lifetime.^43–45 

To inform prevention efforts for infant salmonellosis, further research is needed into modifiable risk factors. Many infants are probably exposed to Salmonella in the home; they might become infected from the same exposures that could infect other family members, including consumption of foods and contact with contaminated surfaces, or they might acquire illness from family members.⁴⁶  Food is the major source of salmonellosis for the general population, so the most effective control measure might be to decrease Salmonella contamination of food.^14,47 

The authors thank Beau B. Bruce, Robert M. Hoekstra, and Robert V. Tauxe for providing useful feedback on the analysis plans or earlier drafts of this report. Many thanks to Yuri Springer for insightful feedback on analyses including race. The authors also thank the local, state, and federal public health professionals who made these analyses possible, as well as the Centers for Disease Control and Prevention’s Laboratory-based Enteric Disease Surveillance and Foodborne Diseases Active Surveillance Network teams that provided access to these data. The findings and conclusions in this report are those of the authors and do not necessarily represent the views of the Centers for Disease Control and Prevention.

95% CI

95% confidence interval

CDC

Centers for Disease Control and Prevention

FoodNet

Foodborne Diseases Active Surveillance Network

IRR

incidence rate ratio

LEDS

Laboratory-based Enteric Disease Surveillance