Epidemiology of Nontyphoidal Salmonella Bloodstream Infections in Children

This study was approved by the Institutional Review Board of Kaiser Permanente Northern California. Informed consent was waived.

This was a retrospective cohort study of the electronic health records of all children <18 years old with NTS found in blood cultures collected from January 1, 1999 to December 31, 2018 in a fully integrated inpatient and outpatient system in Northern California. NTS was defined as Salmonella spp. that were not of serotype typhi or paratyphi. The serogroup and serotype of all Salmonella spp. were extracted from the microbiology records.

Additionally, from January 1, 2009 to December 31, 2018 all stool study results (ie, culture or stool enteric polymerase chain reaction [PCR]) positive for NTS were reviewed for blood culture acquisition and, as applicable, blood culture result. Children with a stool study result positive for NTS were defined as having NTS gastroenteritis. We identified all children from this time who had NTS gastroenteritis and at least 1 blood culture collected without evidence of BSI. Children with NTS gastroenteritis and BSI were matched 1:2 by age and year of specimen collection with children with NTS gastroenteritis and unknown BSI (NTS from stool and no blood culture obtained).

Using the International Classification of Diseases, Ninth and Tenth Revision codes, we determined if a child had an underlying immunocompromising condition and other chronic medical condition, as defined by Feudtner.^23,24  Immunocompromised was defined as the presence of malignancy, acquired or hereditary immunodeficiencies, or receipt of immunosuppressants (antineoplastic agents, high-dose glucocorticoids (2 mg/kg for >2 weeks), tumor necrosis factor inhibitors, calcineurin inhibitors, and mycophenolate mofetil) in the previous 6 months. These data were validated during a manual chart review.

Clinical data from initial and follow-up visits were manually abstracted, including maximum temperature, duration of fever, and presence of diarrhea, bloody stool, vomiting, and abdominal pain. The maximum temperature was either the highest value in a medical setting or parental measured home temperature. The duration of fever was defined as duration at time of blood culture or, for children without blood cultures, total duration of fever. Antibiotic treatment (intravenous and oral) and duration were captured. For those admitted to the hospital, the length of stay and reason for hospitalization (ie, appearance at initial visit, positive blood culture) were assessed. We categorized children as having a focal infection when they had one of the following: cellulitis, septic arthritis, osteomyelitis, pneumonia, endocarditis, myocarditis, meningitis, cholangitis, pancreatitis, or adenitis. Those without a focal infection were categorized as having an uncomplicated NTS BSI. Other adverse outcomes identified included the persistence of BSI (defined as positive blood culture on separate days) and death.

If obtained, white blood cell count, neutrophil count, hematocrit, and cerebrospinal fluid analysis and culture results within 3 days of blood culture acquisition were captured. Additionally, for children with NTS BSI, stool culture and stool enteric PCR results were collected. On September 17, 2014, the microbiology laboratory switched from stool bacterial cultures to stool enteric PCR. When a stool enteric PCR result was positive for Salmonella it reflexed to stool culture to identify the serogroup and serotype. If the stool enteric PCR result was positive for Salmonella and the blood grew NTS, the serotype of NTS was presumed the same.

The incidence rate of NTS BSI by age and year was calculated by using annual midpoint membership information, which is available aggregated by age in years. By using a test for trend, the trend in rates of NTS BSI was determined. Risk factors for complicated NTS BSI using clinical and laboratory data were calculated. Predictors for complicated NTS BSI and NTS gastroenteritis with, without, and unknown BSI were compared by using χ², Fisher’s exact test, or Kruskal Wallis. Statistical significance was defined as P ≤ .05. The statistical program used was SAS 9.4 (Cary, NC, USA).

From 1999 to 2018, there were 211 cases of NTS BSI. The mean number of cases per year was 10.6 (range 3–25). There were 14 952 802 evaluable children over the 20 years, with an average of 747 640 per year. The incidence rate of NTS BSI was 1.4 per 100 000 children per year. The trend of increasing cases over time was not statistically significant (Fig 1). The incidence rate by age in years <1, 1 to 4, 5 to 9, and 10 to 17 was 7.47, 2.31, 0.97, and 0.76, respectively. The incidence rate in females and males was 1.32 and 1.49, respectively.

In the 211 children with NTS BSI, 55% (116) were <5 years old with 40% (46) of those <1 year (Table 1). The distribution of NTS cases in the first year was comparable with 6, 15, and 25 cases occurring in infants 0 to 2, 3 to 5, and 6 to 11 months old, respectively. The mean maximum temperature was 100.7°F, and the predominant gastrointestinal symptom was diarrhea. Ten percent of children had underlying comorbidities. Forty-eight (22.8%) children had multiple days of positive NTS blood cultures. Multiple days of positive blood cultures did not predict focal infection. There were no deaths.

Seven (3%) of 211 children had focal infections, including 2 with osteomyelitis, 1 with pneumonia, 1 with meningitis, 1 with pancreatitis, and 2 with myocarditis. There was a single focal infection in the first year of life, a 7-week-old with NTS meningitis. Two of the children with focal infections had an underlying comorbidity (28.6%), which was greater than those without complications (P = .05). Compared with children without focal infections, children with focal infections were less likely to have diarrhea and had a higher neutrophil count on complete blood count (Table 1). Compared with children without focal infection, children with focal infection had a statistically different distribution of NTS serotype and none had S. heidelberg.

Forty-eight percent of Salmonella isolates were serogrouped but were not further serotyped (Table 2). The most common serogroup causing BSI was Group B, occurring in 113 (54%), and of Group B, the most common serotype was Salmonella enterica serovar heidelberg (henceforth, S. heidelberg), occurring in 45 (21.3%) of overall cases. Of those 45 children with S. heidelberg, 18 (40%) occurred in the first year of life. Further serotyping was not performed on 41 cases of Group B. Less commonly identified serotypes included S. typhimurium (Group B) and S. enteriditis (Group D), occurring in only 2 and 10 cases, respectively. Reviewing only those serotyped, the distribution was wide, with the median number of BSI cases per serotype 2, mode 1 (appearing 15 times), and mean 4.2 (range 1–45; Table 2). Fig 2 (online only) reveals the yearly distribution of NTS by serotype or serogroup (if not serotyped) when at least 5 cases occurred during the study period. Yearly cases of S. heidelberg remained elevated from 2006 to 2014.

From 2009 to 2018, 58 children had NTS gastroenteritis with BSI, including 11 who were <1 year and 1 <3 months of age. Among those <1 year old, their ages ranged from 2 to 11 months, and the mean age was 7.6 months. During the same time, there were 198 children with NTS gastroenteritis with negative blood cultures. Forty-six were <1 year of age, with a mean age of 5 months, and included 4 neonates. On univariate analysis, children with NTS gastroenteritis with BSI compared with those with negative blood cultures had a longer fever duration (4.4 vs 2.5 days), were more likely to have S. heidelberg, and had less bloody diarrhea but did not have a statistically significant difference in laboratory studies (eg, white blood cell count, neutrophils, or hematocrit; Table 3). Children with NTS BSI were more likely to be admitted and receive both IV and oral antibiotics. On multivariable analysis, prolonged fever duration, S. eidelberg, and lack of bloody stool remained statistically significant predictors (Fig 3). Age, S. enteritidis, and S. typhimurium were not predictors of NTS BSI.

The 58 children with NTS gastroenteritis and BSI were matched 1:2 by age and year of specimen collection with children with NTS gastroenteritis without blood cultures obtained. These 116 children with NTS gastroenteritis without blood cultures obtained were compared with the 198 children with negative blood cultures. The former was more likely to have stool collected/ordered in a clinic, be afebrile, and have a shorter duration of fever when febrile (Table 3). More children with negative blood cultures were admitted to the hospital and administered antibiotics. A similar proportion of Salmonella were unknown serogroup and serotype. NTS serotypes Enteritidis, Heidelberg, and Typhimurium were not statistically different between children with negative versus no blood culture.

Despite improvements in food safety, the rates of NTS BSI have not decreased over the last 2 decades. The incidence rate of NTS BSI was 1.4 per 100 000 children per year, with 55% occurring in children aged <5 years. Young children had the highest incidence rate of NTS BSI, with infants <12 months comprising 22% of all cases. Neonates and infants <2 months comprised a small proportion of our NTS BSI cases. In our 20-year cohort, there was only 1 case of NTS BSI in a neonate and only 1 case in a child <1 year of age with NTS BSI with a focal infection, a 7-week-old with NTS meningitis. Incidence rates of NTS BSI and gastroenteritis were highest in those <1 year;⁷  however, in children with NTS gastroenteritis in whom blood cultures were drawn, age was not a predictor of BSI. We found comparable rates of children developing NTS BSI as a complication of gastroenteritis.^9,25 

The Emerging Infections Program FoodNet Working Group estimates that there are 9.8 cases of Salmonella infection for each culture-confirmed case involving bloody diarrhea and 67.7 cases for each culture-confirmed case involving nonbloody diarrhea⁵  and so the percent developing BSI as a complication of gastroenteritis was an overestimation. Using incidence rates of NTS gastroenteritis published on FoodNet,²  6.7% of children aged 0 to 9 years and 8% of children aged 10 to 17 years with NTS gastroenteritis developed BSI. We suspect that more young children were tested for and subsequently found to have NTS gastroenteritis, which explained their lower rate of BSI.

Our findings supported previous work that children with NTS BSI typically do not have underlying conditions and recovered without issue.^15,18  Compared with other published pediatric case series of NTS BSI,^8,15  we found a lower proportion; only 10% of children had an underlying condition including only 2% with underlying immunosuppressing comorbidities. No child with uncomplicated NTS BSI was found to have an underlying immunodeficiency after the BSI. These findings suggest that in previously well children with an uncomplicated NTS BSI, there is no need for them to undergo additional immunologic evaluation.

Despite 20 years of NTS BSI, only 7 (3.3%) had complications, which was comparable to previous published reports.¹⁵  In a resource-rich setting, age was not a risk factor for complicated NTS BSI or case fatality. There were no deaths in our cohort. Complications were higher in children with underlying conditions because 9.5% with underlying chronic conditions and NTS BSI had a complication; however, no children with complications were immunocompromised. This contrasts with Zaida et al, who found 36% of children with underlying conditions, particularly immunocompromising conditions, and NTS BSI developed focal infections.¹⁵  We suspect the difference in our lower rates of complications reflected the changing landscape of immunodeficiency, including vigilance for BSI and earlier antibiotic treatments. In addition, immunosuppressive medications have changed over the last 20 years, with more steroid-sparing regimens, alternative chemotherapeutics, and the use of prophylaxis in the highest-risk children with malignancy. These factors may have decreased the rates of complications associated with NTS BSI. In a child with an underlying condition and NTS BSI, it is prudent to remain vigilant for complications.

Children with NTS BSI did not have markedly elevated temperatures, with a mean maximum temperature of 100.7°F. Like other publications,^15,26  only 80% of those children diagnosed with NTS BSI had diarrhea at presentation. Children with NTS gastroenteritis with BSI were not more likely to have a chronic condition or be immunocompromised compared with those with a negative blood culture. Comparing those with positive and negative blood cultures, children with BSI had longer fever duration and less bloody diarrhea. We also found no statistically significant difference in blood studies (eg, white blood cell, neutrophils, or hematocrit) or other symptoms. Less bloody diarrhea was more likely due to an increased diagnosis of Salmonella gastroenteritis⁵  and less likely a protective feature. Because there was no reliable way to differentiate between positive and negative blood cultures, we agree with current recommendations in the American Academy of Pediatrics²⁷  to obtain blood culture if starting antibiotics for Salmonella gastroenteritis. In addition, clinicians should be aware of the possibility of NTS BSI in children with prolonged fever, regardless of the height of temperature, with a history of diarrhea and/or known Salmonella gastroenteritis.

S. heidelberg gastroenteritis cases have steadily decreased since the mid-1980s^28,29  and do not make up a large proportion of Salmonella gastroenteritis.^2,4  The incidence rate was 0.32 per 100 000 persons and ranked sixth overall from 2010 to 2014;^1,4,5,28  however, in our series, S. heidelberg represented the largest single serotype causing BSI in children. Although not seen in appreciable numbers before 2006, we suspect the high rates of Group B not serotyped were a reflection of S. heidelberg cases. Fifteen of our S. heidelberg BSI cases occurred from 2013 to 2014 during an outbreak associated with contaminated chicken³⁰  that centered in California and infected 634 total individuals and 490 in California; however, there were 23 additional S. heidelberg BSI cases from 2006 to 2012, when no outbreaks were reported. It is not clear if this trend will persist and if we will continue to see high rates of S. heidelberg BSI.

In our Northern California cohort, S. heidelberg was the strongest predictor of BSI in children with NTS gastroenteritis. The invasive nature of S. heidelberg was consistent with other publications.^9,13,20  This included a recent publication that reviewed the epidemiology of salmonellosis in infants from 1968 to 2015 and found S. heidelberg to be the most invasive serotype.²⁹  Our findings were not consistent with studies from lower-middle-income countries^8,11,22  that report S. typhimurium and S. enteritidis as the most common serotypes associated with NTS BSIs in children. Although the leading causes of Salmonella gastroenteritis, S. typhimurium, and enteritidis were not the leading causes of and did not predict NTS BSI in our cohort. Others have found Typhimurium,²²  Panama, Newport, Infantis, Virchow, Dublin,^16,25  and Choleraesus¹⁷  as the most common serotypes to cause invasive disease; however, we found S. heidelberg was 3 times more common than all 7 serotypes combined.

Because serotyping is done at state health departments, results are rarely back in time to affect patient management; however, in an ideal world, the rapid identification of stool would be possible to identify S. heidelberg. Rapid identification could prompt earlier evaluations and possibly antibiotics.

As with other retrospective studies, our study had limitations. Despite a large sample size of ∼750 000 children/year within a fully integrated inpatient and outpatient system, the extremely low rates of NTS BSI limit any generalizations that could be drawn about the clinical presentation of individual patients. The demographics of our cohort and distribution of serogroups/serotypes may not represent all of California or the United States. The acquisition of a blood culture was at the clinician’s discretion. Some bacteremic children may have been treated with antibiotics without a blood culture or had transient bacteremia, so an episode of BSI may have been missed. Not all Salmonella were subtyped, including 41 with Group B and 10 with Group D. It is possible some of the Salmonella not serotyped were misclassified as NTS. It is also possible that the proportion of Salmonella serotypes was different from that reported. Even with these limitations, we feel confident that our study accurately reported incidence rates of BSI and characteristics of children with NTS BSI.

NTS BSI rates are not decreasing, so providers should remain vigilant for NTS BSI in children. S. heidelberg is the most common cause of NTS BSI in children and a strong predictor of BSI in children with NTS gastroenteritis. Food safety, particularly with chicken and turkey (the source of Heidelberg), is necessary to prevent S. heidelberg gastroenteritis and BSI.

This study was supported by a grant from the Kaiser Permanente Northern California Community Benefit Program. We are grateful to David R. Vinson, MD, The Permanente Medical Group and the KP CREST Network, and Peter B. Cooch, MD, The Permanente Medical Group for their editorial assistance. They did not receive compensation for their contribution. This work was presented in part at the Infectious Diseases Week Conference 2019, Washington DC.

BSI

blood stream infections

NTS

non-typhoidal Salmonella

PCR

polymerase chain reaction