Enrolled participants were >14 days and ≤10 years old, had NVSN protocol-defined AGE and sought medical care at the ED of Children’s Mercy Kansas City, Missouri, from December 2011 to June 2016; ED patients who were subsequently admitted were included in the analysis. Surveillance was year-round (September through August) except the initial 2011/2012 season that started December 2011. Detailed methods and NVSN scope were previously described.^7,15,16 Briefly, AGE was defined as ≥3 diarrhea episodes within 24 hours or ≥1 vomiting episode within 24 hours, or both; symptoms had to be present for <10 days. Systematically enrolled HC had to attend a scheduled well-child visit in an outpatient clinical care site within ±14 days of an enrolled AGE participant of a similar age group (ie, <6 months, 6–11 months, 12–23 months, 24–59 months, and ≥60 months). Per protocol, HC had no history of AGE symptoms for 14 days before enrollment. AGE stools were collected within 10 days of symptom onset and HC stools within 5 days of enrollment. Families provided demographic, epidemiologic, and clinical data. Immunization histories were obtained through Kansas and Missouri state vaccine registries and/or through the participant’s medical provider.

Children were enrolled via written consent from parent/s or legal guardian. This study was reviewed and approved by the Institutional Review Board of Children’s Mercy Kansas City and the Centers for Disease Control and Prevention.

Stools were stored at −80 °C until total nucleic acids were extracted with an automated EasyMag extraction system (bioMerieux, Durham, NC).¹⁷ Samples were tested via Luminex GPP assay (Luminex Corp, Austin, TX) along with positive and negative controls per manufacturer’s instructions. Targets includes viruses (RV, NoV, and adenovirus 40/41), bacteria (Clostridium difficile, Shigella, Salmonella, Campylobacter, Escherichia coli O157, enterotoxigenic E coli [ETEC], shiga-like toxin producing E coli [STEC], Vibrio cholera, and Yersinia) and parasites (Cryptosporidium, Giardia, and Entamoeba histolytica).

Samples that were given a target call of invalid by the Luminex data analysis software on initial testing were repeated, per manufacturer, and if both results were invalid, the patient’s stool sample was excluded from the analysis (n = 35). Because Luminex Corp had released notes on false-positive detection of Salmonella, E histolytica, Giardia, and Cryptosporidium by the multiplex Luminex GPP assay and our relatively high Luminex detection rate of these 3 organisms in nonsymptomatic HC, each Luminex detection was confirmed by a second methodology. Using manufacturers’ instructions, Progastro SSCS assay was used to confirm Salmonella (Hologic Inc, Marlborough, MA) and RIDA Gene Parasitic Stool Panel to confirm Giardia lamblia, Cryptosporidium spp., Entamoeba histolytica (R-Biopharm, Creve Coeur, MO). If the Salmonella, E histolytica, Giardia, and Cryptosporidium detected by Luminex GPP assay was not confirmed by a repeat testing assay, the sample was categorized as negative for that organism. The American Academy of Pediatrics recommends not testing children <12 months of age for C difficile owing to the high rate of asymptomatic colonization.¹⁸ As a result, data on C difficile detections in children <12 months were collected but were categorized as negative owing to the uncertainty in whether the child was a C difficile carrier or not. For this analysis, samples with ≥2 detected organisms were considered codetections.

Per NVSN protocol, participant enrollment is performed with the intended goal to have AGE and HC be similar as possible with regards to visit date, age, and race/ethnicity. However, because significant differences in patient demographics were noted between AGE and HC, we created a matched analytic dataset using a subset of our AGE and HC participants (case control subset). This dataset was created using a propensity score model that accounted for sampling time frame (ie, calendar month and year), patient race, and patient ethnicity. HC and AGE who did not have a corresponding comparator (ie, AGE and HC, respectively) within 14 days of their screened data were excluded from the propensity model. To further ensure balance, the propensity score models were stratified by 5 patient age groups (<6 months, 6–11 months, 12–23 months, 24–59 months, and ≥60 months). Optimal pair matching, which included sampling without replacement, was implemented with an intended goal of a 2:1 ratio of AGE:HC. This case control subset was used for demographic, risk factor, and seasonality comparisons between AGE and HC.

RV vaccination status was determined from provider records for all participants to quantify RV vaccine adherence of children who had RV detected. A complete schedule was defined as 3 doses of any vaccine or combination of vaccines, or 2 doses of monovalent vaccine. Participant vaccination status was classified as either unvaccinated, partial, or complete. RV vaccination dates were used to identify which AGE/HC children received the vaccine within 7 days of stool specimen collection.

Patient demographics (age group, gender, race/ethnicity) among the AGE participants were compared based on whether a pathogen was detected (positive) or not (negative). Select clinical symptoms (eg, maximum temperature, any diarrhea, number of vomiting episodes) and seasonality effect were contrasted between pathogens using the entire AGE subset. The case control subset was used to compare the potential risk factors for AGE illness (eg, current daycare ≥4 hours/week, exposure to persons with AGE in the prior week, number of persons residing in household). Last, pathogen detections were compared between matched AGE and HC.

Pearson’s χ² test was used to compare categorical distributions and the Kruskal-Wallis test was used for continuous outcomes. Unadjusted logistic and quantile regression models compared risk factors and clinical presentations for each organism compared with the negative group. Reported P values for contrasting organisms were adjusted for multiple comparisons using the Sidak method. Unadjusted logistic models were used to compare risk factors between AGE and HC. The MatchIt package in R (R Core Team; Vienna, Austria) was used for creating the case control subset. All analyses were completed using Stata (StataCorp, College Station, TX).

When AGE participants having ≥1 detected pathogen were compared with AGE participants having no pathogens detected, those 12–23 months old were significantly over-represented (29.4% vs 20.2%), and those ≥60 months old were under-represented (18.4% vs 25.5%) (Table I). The distributions of sex, ethnicity, and race were similar between AGE test-positive and AGE test-negative groups. The AGE test-positive group had a slightly higher but nonsignificant proportion of Hispanic/Latino children compared with the AGE test-negative group (32.3% vs 28.9%, respectively). A comparison of clinical presentation by pathogen detection is provided in Table II.

NoV was the most frequently detected pathogen among children with AGE (n = 568 [22.7%]), followed by RV (n = 196 [7.8%]) (Table III). Among the children with AGE and RV detected, only 2 had received a RV vaccine within the last week, that is, might have been shedding vaccine virus. Among bacterial pathogens, Salmonella was the most common (n = 159 [6.4%]) followed by C difficile (participants ≥1 years of age; n = 132 [5.3%]) and Shigella (n = 113 [4.5%]).

Codetections occurred in 206 of 2503 AGE participants (8.2%) (Figure 1). In general, codetection proportions were higher for a bacterial pathogen (eg, Salmonella, 53.5%; C difficile, 54.5%) compared with a viral pathogen (eg, NoV, 20.8%; RV, 27.6%). Among codetections, nearly three-quarters (72.8% [n = 150]) of these 206 children had both a viral and a bacterial pathogen detected, followed by 11.2% (n = 23) who had multiple bacterial pathogens detected, 10.2% (n = 21) with multiple viral pathogens, and 5.8% (n = 12) who had a parasitic and bacterial/viral codetection. Most codetections in AGE participants (75%) included NoV, Salmonella, C difficile, and/or RV.

In the case control analysis, NoV was more commonly detected among matched AGE participants (n = 240 [22.0%]) compared with HC (n = 39 [6.8%]) (Table III). C difficile was the most commonly detected pathogen among HC (n = 40); the detection rate was slightly higher in HC than those with AGE (7.0% vs 5.3%, respectively). RV was detected in 15 HC; however, 10 of 15 of these HC had received a RV vaccination at their prior well-check visit within 1 week of stool specimen submission, suggesting most of the RV detections in HC may represent RV vaccine strain shedding. Only a single child with AGE had received RV vaccine within 7 days of stool collection. The detection rate of C difficile and E coli (E coli O157, ETEC, or STEC) did not differ significantly between matched AGE and HC.

Nearly one-quarter (23.8%) of AGE participants <24 months old tested positive for NoV, with a corresponding 7.5% NoV positivity among HC <24 months (Figure 2). Positivity rates for viral pathogens decreased with increasing age of enrolled participants, whereas for some bacterial organisms (eg, Shigella in children with AGE and C difficile in HC) the opposite effect was observed. Codetection occurred in 2.3% of all HC (13/579) and was predominately detection of a viral and bacterial organism (n = 8 [61.5%]). NoV and C difficile were the 2 most common organisms (both 30%) among HC participants with codetections.

Children with AGE and HC used in the case control analyses were similar in terms of age, sex, and race/ethnicity (Table I). As shown in Table IV, AGE participants were significantly more likely to report a sick contact compared with HC, both for contacts outside the home (15.6% vs 1.4%; OR, 13.01; 95% CI, 6.35–26.64) and inside the home (18.6% vs 2.1%; OR, 10.66; 95% CI, 5.90–19.27). The proportion with household income ≤$25,000 was higher among children with AGE (53.7%) than among HC (37.2%; OR, 1.96; 95% CI, 1.59–2.49). No significant differences in median household size were noted between matched AGE and HC participants. Children with AGE were also more likely to report daycare attendance compared with HC (41.4% vs 29.5%).

When considering all organism detections (ie, single-infection + codetections), year-to-year patterns for some organisms were similar between AGE and HC participants. Seasonal fluctuations in NoV among AGE have corresponding seasonal fluctuations among HC (Figure 3). Intermittent increases in Shigella detections were observed for both AGE and HC and correspond with the timing of a known concurrent local community outbreak (Figure 4). Pathogens such as NoV, RV, and Salmonella (among AGE) demonstrated consistent expected seasonal and annual patterns. Conversely, other organisms appear to have had a constant presence (eg, adenovirus 40/41, C difficile) or exhibited sporadic outbreaks (eg, Shigella).