Attribution of Pediatric Acute Gastroenteritis Episodes and Emergency Department Visits to Norovirus Genogroups I and II

Abstract

Background

Norovirus is a leading cause of acute gastroenteritis. With vaccines in development, population-based estimates of norovirus burden are needed to identify target populations, quantify potential benefits, and understand disease dynamics.

Methods

We estimated the attributable fraction (AF) for norovirus infections in children, defined as the proportion of children testing positive for norovirus whose gastroenteritis was attributable to norovirus. We calculated the standardized incidence and emergency department (ED) visit rates attributable to norovirus using provincial gastroenteritis visit administrative data.

Results

From 3731 gastroenteritis case patients and 2135 controls we determined that the AFs were 67.0% (95% confidence interval [CI], 31.5%–100%) and 91.6% (88.8%–94.4%) for norovirus genogroups I (GI) and II (GII), respectively. Norovirus GII AF varied by season but not age. We attributed 116 episodes (95% CI, 103–129) and 59 (51–67) ED visits per 10 000 child-years to norovirus GII across all ages, accounting for 20% and 18% of all medically attended gastroenteritis episodes and ED visits, respectively.

Conclusions

In children, a large proportion of norovirus GII detections reflect causation, demonstrating significant potential for norovirus GII vaccines. Seasonal variation in the norovirus GII AF may have implications for understanding the role asymptomatic carriage plays in disease dynamics.

This study estimates the attributable fraction for norovirus infections in children and the corresponding attributable incidence and emergency department visit rates. In children, a large proportion of norovirus genogroup II detections reflect causation.

Acute gastroenteritis remains a leading cause of disease and death worldwide [1], and norovirus has emerged as the leading cause of both community [2–4] and medically attended [3, 5, 6] gastroenteritis. With vaccines advancing into clinical trials [7], knowledge of the magnitude and distribution of disease caused by norovirus is needed to identify target populations, quantify the potential benefits of vaccine introduction, and explicate crucial disease dynamics.

Most studies have measured norovirus detection without accounting for asymptomatic carriage [8–10]. Consequently, detection may not reflect causal attribution, leaving a gap in our current knowledge of norovirus dynamics. Studies addressing attribution are limited to children <5 years old, low- and middle-income countries, and/or cases with a primary presentation of diarrhea [4, 11, 12]. Many studies have relied on diarrheal specimens submitted for routine enteric diagnostics [6, 13–15] and excluded cases with isolated vomiting (ie, without concomitant diarrhea) [6, 15]. These approaches are problematic because half of norovirus cases in children present with isolated vomiting [16], and children serve as a reservoir for gastrointestinal viruses [17].

To close this gap in our knowledge, we combined administrative data with data from a large case-control study of acute gastroenteritis that employed enhanced enteropathogen detection in children aged 0–18 years. Our objectives were to determine (1) the proportion of children testing positive for norovirus whose gastroenteritis was attributable to norovirus (ie, the attributable fraction [AF]), and (2) provincial standardized acute gastroenteritis incidence and emergency department (ED) visit rates attributable to norovirus in children.

METHODS

Case-Control Study

Participants were enrolled in the Alberta Provincial Pediatric EnTeric Infection TEam (APPETITE) study between 14 December 2014 and 13 August 2018 in Alberta, Canada [18]. Alberta has a single-payer, publicly funded healthcare system and a population of 4.3 million, including 960 000 children <18 years old [19].

Case patients were children <18 years old with acute gastroenteritis, defined as the occurrence of ≥3 vomiting and/or diarrhea episodes in a 24-hour period with <7 days of symptoms (Supplementary Methods) [18]. They were recruited in the EDs of the Alberta Children’s Hospital (Calgary, Alberta), and Stollery Children’s Hospital (Edmonton, Alberta) and via Health Link, Alberta’s nurse triage telephone resource, if caregivers were advised to provide ongoing supportive care at home (ie, no physician assessment required) (Supplementary Methods). Controls were children <18 years old without vomiting, diarrhea, rhinorrhea, cough, or fever, and they were recruited through the participating EDs if they sought care for a noninfectious health complaint (eg, injury) or from a public health clinic in Calgary if presenting for vaccination. Participants provided data regarding demographics and family characteristics, and for case patients, gastroenteritis symptoms and medical treatment sought.

Specimen Collection and Testing

Rectal swab and/or stool specimens were collected from ED and Health Link participants (Supplementary Methods). Controls provided a stool specimen but not a rectal swab specimen. Specimens were comprehensively tested using routine enteric culture, the Luminex xTAG Gastrointestinal Pathogen Panel (GPP; Luminex Molecular Diagnostics, Toronto, Canada), and a gastroenteritis virus panel (GVP) [20]. Both GPP and GVP distinguished norovirus genogroups I (GI) and II (GII), the most common genogroups to cause acute gastroenteritis in humans.

We classified a case patient as positive for norovirus GI or GII if any specimen yielded positive results on either assay. Cycle threshold (Ct) values were obtained from GVP, which was a real-time reverse transcriptase polymerase chain reaction assay. The percentage of positive agreement between GPP and GVP was 74% for norovirus GI and 90% for GII; only 2 (0.2%) of the 911 norovirus GII–positive specimens were GPP positive and GVP negative, meaning that hey had no Ct value (Supplementary Table 1). Full details of specimen collection, retrieval, transport, storage, and testing have been described elsewhere [21].

The APPETITE study was approved by the Research Ethics Boards of University of Calgary and University of Alberta. Informed consent was provided by caregivers; assent was obtained from children when appropriate.

Provincial Data

To standardize APPETITE ED visit rates to the population of Alberta (Figure 1), we obtained data from the Alberta provincial Ministry of Health for 3 groups: APPETITE case patients and controls deterministically linked using their personal health number; children with gastroenteritis-coded ED visits, inpatient hospitalizations, and outpatient clinic visits; and all children <18 years old in Alberta. ED visits were identified from the National Ambulatory Care Reporting System, inpatient hospitalizations from the Discharge Abstract Database, and outpatient clinic visits from the Supplemental Enhanced Service Event, a province-wide physician claims database.

Figure 1.

Schematic of how data sources were used and analytic steps. Case patients and controls from the APPETITE (Alberta Provincial Pediatric EnTeric Infection TEam) study were used to calculate the attributable fraction (AF) of norovirus. Gastroenteritis visits identified from provincial databases were used to define gastroenteritis episodes. AF estimates, APPETITE case patients, and provincial gastroenteritis episodes and visits, and population counts were then used to estimate the incidence and visit rates attributable to norovirus and corresponding populations AFs (PAFs). Abbreviations: Ct, cycle threshold; DAD, Discharge Abstract Database; ED, emergency department; ICD, International Classification of Diseases; ICD-9, International Classification of Diseases, Ninth Revision; ICD-10-CA, International Classification of Diseases, Tenth Revision; NACRS, National Ambulatory Care Reporting System; OR, odds ratio; SESE, Supplemental Enhanced Service Event.

Visits and hospitalizations ending between 14 December 2014 and 31 March 2018 were included. Visits were identified using gastroenteritis-defining sets of International Classification of Diseases (ICD) codes, based on the International Statistical Classification of Diseases and Related Health Problems, Tenth Revision, Canada, from the Discharge Abstract Database and the National Ambulatory Care Reporting System, or the International Classification of Diseases, Ninth Revision, from the Supplemental Enhanced Service Event (Supplementary Tables 2 and 3 and Supplementary Methods); codes were included irrespective of the order in which they were listed. We obtained data for standardization from the provincial databases (Supplementary Methods). Alberta Health provided aggregate population totals for each year of the study, jointly stratified by potential standardization variables.

To define gastroenteritis episodes, we grouped gastroenteritis-coded visits occurring within 14 days of each other, regardless of visit venue (ie, hospital admission, ED, clinic) (Supplementary Figure 1). We summed visits by type for each episode and calculated overall and age- and season-specific incidence and visit rates for medically attended gastroenteritis in children in Alberta.

Determining the AF

Our first objective was to determine the proportion of gastroenteritis case patients in whom norovirus was detected whose gastroenteritis was, in fact, due to norovirus, that is, the AF. We use the AF as distinct from the population AF (PAF) [22], with the latter reflecting the proportion of cases attributed to norovirus among all gastroenteritis cases (Supplementary Figure 2).

R software [23] was used for all analyses. For norovirus GI and GII, we separately calculated the odds ratio (OR) for the marginal effect of norovirus Ct on gastroenteritis using inverse probability weighting, which accounted for confounding by demographics, illness season and year, and codetected pathogens (Supplementary Methods). We calculated the genogroup-specific AF from the OR for each case as AF = (OR − 1)/OR (Supplementary Methods), and we averaged individual AF estimates to determine the AF for the genogroup.

Next, we estimated stratum-specific norovirus GII AFs for age, year, season (January–March, April–June, July–September, or October–December), and case recruitment source (ED or Health Link) (Supplementary Methods). There were too few norovirus GI–positive cases for further stratification. For the norovirus GII stratified AFs, the AF was estimated using the same model as for norovirus GII, minus variables accounted for by stratification (eg, age was removed from the exposure model for age strata). If AFs differed by >5% across strata, without substantial overlap of 95% confidence intervals (CIs), we used the stratum-specific AF estimates to adjust standardized incidence and visit rate estimates.

Standardized Incidence and Visit Rates

Our second objective was to estimate the attributable, standardized incidence rate (IR) and ED visit rate of norovirus in children in Alberta, by genogroup (Figure 1). First, we matched APPETITE case patients to the gastroenteritis episodes defined by the provincial visit data if the participants enrolled in the study during the episode ±7 days (Supplementary Methods). We considered all visits during the episode as associated with the pathogen(s) detected by the APPETITE study. Second, for matched case patients with norovirus, we discounted gastroenteritis episodes and visits by the case patients’ norovirus AFs. In doing so, knowledge of which pathogen for a given case was responsible for disease was unnecessary; we instead used the exact AF that reflects the probability that a gastroenteritis case would not have occurred in the absence of norovirus. Third, we determined which variables to use for standardization (Supplementary Methods). We standardized on age, year, geographic region, and season. We summed total provincial, total APPETITE, and norovirus-attributed APPETITE episodes and visits within each stratum defined by these variables.

Fourth, to account for underestimation of gastroenteritis visits identified using ICD codes, we applied an underestimation factor to stratum-specific estimates of Alberta gastroenteritis episodes and visits (Supplementary Methods). Fifth, we estimated APPETITE’s underlying population at risk by multiplying the stratum-specific Alberta population by the proportion of Alberta episodes or visits enrolled in APPETITE (Supplementary Methods). Finally, we calculated stratum-specific norovirus incidence and visit rates and standardized the rates to the Alberta population to obtain overall rates for each genogroup, and age-, year-, and season-specific rates for norovirus GII. We expressed all rates per 10 000 child-years and calculated the PAF, or the fraction of all gastroenteritis episodes or visits attributable to norovirus.

In secondary analysis, we estimated the norovirus-attributable rates of hospitalizations and outpatient clinic visits. Finally, we analyzed how sensitive our incidence and visit rates were to the assumption that the provincial gastroenteritis visits identified using ICD codes underestimated the actual number of visits. In this analysis, we did not multiply the stratum-specific totals of Alberta episodes and visits by the underestimation factor.

Missing Data

To account for missing data and obtain CIs for our estimates, we performed multiple imputation by chained equations using the R software package mice [24]. Forty data sets were imputed to make the loss of efficiency <1, assuming that ≤40% of data would be missing for a given variable [25]. We assessed the success of imputation by examining imputed variable distributions and diagnostic plots (Supplementary Figure 3). From each imputed data set, 10 000 bootstrap samples were drawn with replacement. The full estimation process, including calculation of the AFs and standardized incidence and visit rates, was performed on each bootstrap sample.

RESULTS

During the study period, 2725 ED case patients, 1006 Health Link case patients, and 2135 controls were recruited. At least 1 specimen was tested for 2684 ED case patients (98.5%), 635 Health Link case patients (63.1%), and 1342 controls (62.9%) (Table 1 and Supplementary Results). We detected norovirus GI in 0.9% of ED case patients (25 of 2684) and Health Link case patients (6 of 635) case patients, and 0.6% of controls (8 of 1342). We detected norovirus GII in 24.1% (646 of 2684), 35.3% (224 of 635), and 3.1% (41 of 1342) of ED case patients, Health Link case patients, and controls, respectively (Table 1). The frequency of codetected pathogens was similar for norovirus GII case patients recruited through the ED and Health Link (Table 1), but ED case patients had more enteric bacterial and parasitic codetected pathogens and more instances of >1 codetected pathogen (Supplementary Table 4).

Table 1.

Demographics and Norovirus Detection in APPETITE Study Participants

	Gastroenteritis Case Patients, No. (%)a
Demographics and Norovirus Detection	ED (n = 2725)	Health Link (n = 1006)	Controls, No. (%)a (n = 2135)
Age group, mo
<12	817 (30)	286 (28)	825 (39)
12–23	689 (25)	359 (36)	461 (22)
24–35	341 (13)	143 (14)	216 (10)
36–59	414 (15)	118 (12)	273 (13)
60–215	463 (17)	100 (10)	360 (17)
Data missing	1 (0)	0	0
Sex
Female	1259 (46)	456 (45)	935 (44)
Male	1466 (54)	550 (55)	1063 (50)
Data missing	0	0	137 (6)
Zone
Calgary	1609 (59)	534 (53)	1453 (68)
Edmonton	814 (30)	203 (20)	500 (23)
North, Central, or South	82 (3)	202 (20)	43 (2)
Data missing	220 (8)	67 (7)	139 (7)
Period
December 2014 to June 2015	292 (11)	166 (17)	21 (1)
July 2015 to June 2016	817 (30)	246 (24)	419 (20)
July 2016 to June 2017	828 (30)	267 (27)	904 (42)
July 2017 to June 2018	721 (26)	316 (31)	748 (35)
July 2018 to August 2018	67 (2)	11 (1)	43 (2)
Season enrolled
January–March	661 (24)	295 (29)	392 (18)
April–June	862 (32)	290 (29)	516 (24)
July–September	591 (22)	150 (15)	676 (32)
October–December	611 (22)	271 (27)	551 (26)
Specimen testedb	2684 (98)	635 (63)	1342 (63)
Norovirus detected	669 (25)	229 (36)	49 (4)
Norovirus GI detected	25 (0.9)	6 (0.9)	8 (0.6)
Ct, median (IQR)	24.4 (17.6–31.4)	19.8 (16.1– 21.2)	29.6 (25.9–34.4)
Codetected pathogen	10 (40)	4 (67)	0
Norovirus GII detected	646 (24)	224 (35)	41 (3)
Ct, median (IQR)	18.8 (15.4–25.1)	17.7 (16.1–20.0)	26.3 (22.7–28.5)
Codetected pathogen	257 (40)	82 (37)	14 (34)
Matched to a medically attended gastroenteritis episode	2065 (76)	199 (20)	NA

Abbreviations: APPETITE, Alberta Provincial Pediatric EnTeric Infection TEam; Ct, cycle threshold; ED, emergency department; GI, genogroup I; GII, genogroup II; IQR, interquartile range; NA, not applicable.

^aData represent no. (%) of case patients or controls unless otherwise specified.

^bNot all enrolled participants provided a specimen. This is because some ED case patients and all Health Link case patients and control participants agreed to collect rectal swab and/or stool specimens at home, but not all participants returned their specimens. In addition, some specimens were inadequate for testing.

AF Results

The AFs for norovirus GI and GII were 67.0% (95% CI, 31.5%–100%) and 91.6% (88.8%–94.4%), respectively. We observed negligible differences in the AF for norovirus GII across age groups and years of enrollment (Supplementary Figure 4). The AF for norovirus GII infections in October–December was 95.0% (95% CI, 91.8%–98.2%), dropping to 86.8% (80.6%–93.0%) in January–March. In the former period, half of controls with norovirus GII had another virus codetected (Supplementary Table 5). Restricting case patients to those recruited through the ED, the AF for norovirus GII was 89.7% (95% CI, 86.3%–93.0%), compared with 95.0% (92.9%–97.1%) when restricting case patients to those recruited through Health Link. Accordingly, Health Link had overall lower Ct values for norovirus GII (Supplementary Table 5). Because of these differences, we calculated season-specific and enrollment site–specific AF estimates for norovirus GII and used these values to estimate incidence and visit rates (Figure 2).

Figure 2.

Attributable fraction (AF) for norovirus genogroup II (GII; n = 870 case patients) by enrollment group and season. Summary AFs are averages of individual cycle threshold–based AFs for each child who tested positive for norovirus GII.

Estimating IRs

During the study period, there were 187 549 medically attended childhood gastroenteritis episodes in Alberta, corresponding to an IR of 566 per 10 000 child-years (Figure 3 and Supplementary Results). We estimated the IRs of medically attended norovirus in Alberta to be 5 (95% CI, .5–9) and 116 (103–129) per 10 000 child-years for norovirus GI and GII, respectively (Supplementary Table 6). Norovirus GII caused 20% (116 of 566) of the gastroenteritis episodes in Alberta during the study period. The incidence was highest in January–March, with 173 (95% CI, 141–205) norovirus GII episodes per 10 000 child-years. The overall IR of gastroenteritis in this period was 669 per 10 000 child-years, of which norovirus GII accounted for 26% (173 of 669).

Figure 3.

Incidence rates (IRs) of all gastroenteritis and gastroenteritis attributed to norovirus genogroup II (GII). Light turquoise bars represent the total IR of medically attended gastroenteritis in Alberta, Canada, overall (A), by season (B), and by age (C). Within these, the dark turquoise bars represent IRs of medically attended norovirus GII in Alberta. Percentages indicate the population attributable fraction, or the proportion of all gastroenteritis episodes that can be attributed to norovirus GII. Data tables under each graph provide IRs for norovirus GII gastroenteritis and all gastroenteritis. IRs for all gastroenteritis reflect episodes of gastroenteritis formed by linking International Classification of Diseases–coded gastroenteritis visits in Alberta, from December 2014 to March 2018. Norovirus GII gastroenteritis IRs were generated by standardizing IRs in the APPETITE (Alberta Provincial Pediatric EnTeric Infection TEam) study population to the full Alberta population.

Norovirus GII incidence varied by age, with the highest incidence in children 12–23 months of age (IR, 479 per 10 000 child-years; 95% CI, 389–569) (Supplementary Table 6), and lowest among those aged 5–17 years. Norovirus GII caused 58 (95% CI, 46–70) episodes of medically attended gastroenteritis per 10 000 child-years in this age group, representing 19% of the 308 per 10 000 child-years gastroenteritis episodes (Figure 3).

Visit Rates

We identified 108 682 ED visits in Alberta coded as gastroenteritis related during the study period, equivalent to 328 gastroenteritis ED visits per 10 000 child-years. In 53% of medically attended gastroenteritis episodes in Alberta (98 694 of 187 549), children had ≥1 ED visit (range, 1–10 visits). Norovirus GI caused only 3 (95% CI, −1 to 7) ED visits per 10 000 child-years. We attributed 59 (95% CI, 51–67) ED visits per 10 000 child-years to norovirus GII, representing 18% of gastroenteritis ED visits (59 of 328) in Alberta during the study period (Figure 4).

Figure 4.

Emergency department (ED) visit rates of all gastroenteritis and gastroenteritis attributed to norovirus genogroup II (GII). Light yellow bars represent the total gastroenteritis ED visit rate in Alberta, Canada, overall (A), by season (B), and by age (C). Within these, the dark yellow bars represent the norovirus GII ED visit rate in Alberta. Percentages indicate the population attributable fraction (PAF), or the proportion of all gastroenteritis ED visits that can be attributed to norovirus GII. Data tables under each graph provide ED visit rates for norovirus GII gastroenteritis and all gastroenteritis. Visit rates for all gastroenteritis reflect International Classification of Diseases–coded gastroenteritis ED visits in Alberta, from December 2014 to March 2018. Norovirus GII gastroenteritis visit rates were generated by standardizing visit rates in the APPETITE (Alberta Provincial Pediatric EnTeric Infection TEam) study population to the overall Alberta population <18 years old.

Norovirus GII ED visit rates were highest in January–March (visit rate, 88 visits per 10 000 child-years; 95% CI, 67–108 visits) followed by April–June (66 visits per 10 000 child-years; 48–85 visits) (Supplementary Table 6). The proportion of all gastroenteritis ED visits and incidence attributable to norovirus GII followed a similar seasonality (Figure 4).

The rate of ED visits attributed to norovirus GII was highest in children 12–23 months old (visit rate, 282 visits per 10 000 child-years; 95% CI, 199–365 visits) (Supplementary Table 6). The share of all gastroenteritis ED visits attributable to norovirus GII was similar to the proportion of episodes in all age groups except 5–17-year-olds, in whom only 14% of ED visits (24 of 164) were attributable to norovirus GII (Figure 4).

We identified 101 191 outpatient clinic visits and 6512 inpatient hospitalizations in Alberta coded as gastroenteritis related during the study period, equivalent to 305 clinic visits and 20 hospitalizations per 10 000 child-years (Supplementary Table 6). In 50% of medically attended gastroenteritis episodes in Alberta (93 888 of 187 549), children had ≥1 clinic visit (range, 1–14 visits), and 3% (6129 of 187 549) had ≥1 hospitalization (range, 1–8).

Sensitivity Analysis

Without adjustment for underestimation, norovirus GII caused 110 gastroenteritis episodes (95% CI, 98–123 episodes) and 56 ED visits per 10 000 child-years (49–64 visits), comparable to the primary analysis.

DISCUSSION

By accounting for noncausal detections and standardizing our study rates to the population, we offer a more accurate population-based picture of the role of norovirus in childhood gastroenteritis than has been available from previous studies. We determined that when norovirus GII is detected in children with acute gastroenteritis, most such illnesses can be attributed to this pathogen. There is meaningful variation by season but not by age. Norovirus GII accounted for 20% of medically attended gastroenteritis episodes and 18% of gastroenteritis ED visits in children in Alberta, demonstrating a substantial burden. In contrast, the AF for norovirus GI was 67%, causing only 5 gastroenteritis episodes and 3 ED visits per 10 000 child-years.

The US New Vaccine Surveillance Network found that 21% of children <5 years old seeking care for gastroenteritis tested positive for norovirus [5]. They estimated 141 norovirus-associated ED visits per 10 000 child-years in this age group [5]. Our estimate for norovirus GII in children <5 years old is remarkably close, at 148 per 10 000 child-years. Had the New Vaccine Surveillance Network study included rate estimates by genogroup and accounted for noncausal detection, the difference in our results would likely have been somewhat larger. However, our results differ from those reported in Israel /,where the estimated norovirus-associated ED visit rate was only 9.9 per 10 000 child-years among children <5 years old [26]. The substantial difference in visit rate between this study and our own may stem from their exclusion of hospitalized patients from ED rates or highlight the importance of region-specific data to assess enteropathogen burden.

We are unaware of previous studies estimating the AF of norovirus, although multiple studies have estimated the PAF, sometimes referring to it as the AF. In a cohort of Peruvian children, Saito et al [11] estimated an unadjusted norovirus PAF of 7.8% for children <12 months and 23.1% for children 12–23 months old. Using quantitative molecular data and adjusting for other pathogens, the Global Enteric Multicenter Study estimated norovirus GII PAFs from 0.4% to 4.4%, depending on country and age group [12]. The Malnutrition and Enteric Disease Study estimated that norovirus GII caused 5.2% of diarrhea in children <12 months and 5.4% in children 12–24 months old; in the latter age group, the PAF reached 19.2% at 1 site [4]. Because the PAF is affected by the incidence of norovirus and the prevalence of other pathogens, our higher Canadian PAF estimates indicate that norovirus plays a larger role in this setting relative to other pathogens than in the low- and middle-income countries studied previously.

The CI for the AF of norovirus GI is too wide for definitive conclusions. The low point estimate reflects the preponderance of codetected pathogens among norovirus GI case patients and their absence among controls, as well as only slightly higher detection of norovirus GI among case patients (0.9%) than controls (0.6%). More work is needed to understand how much of the disease associated with norovirus GI is truly causal. Similar proportions of norovirus GII case patients and controls had codetected pathogens, and detection of norovirus GII among case patients was >8 times that among controls, producing a high AF.

The AF can provide insight into norovirus disease dynamics. Noncausal detection, a possibility when the AF for a case is <100%, could reflect asymptomatic excretion or postsymptomatic shedding [10, 11, 27]. The AF was similar across age groups, suggesting that asymptomatic carriage and/or postsymptomatic shedding are equally likely among detected cases regardless of age. Because older children have likely been exposed to norovirus previously, the constancy of the AF may have implications for how previous exposures affect asymptomatic carriage. Intra-annual variation in norovirus GII AF was out of phase with established norovirus GII incidence seasonality, which we also observed. The AF was highest at the beginning of the norovirus season in October–December, potentially reflecting the introduction of new strains and lower host immunity [28, 29]. Although norovirus incidence peaked in January–March, the AF was 8% lower during this period relative to the start of the norovirus season. This may be due to the preponderance of other circulating pathogens, such as rotavirus and sapovirus, or prolonged carriage of norovirus particles from earlier infections [10, 11, 27].

The number of gastroenteritis episodes and ED visits attributable to norovirus GII indicates that effective childhood vaccines targeting this genogroup would meaningfully reduce morbidity and healthcare use from norovirus, and by extension, gastroenteritis. That more than half of children seeking care for gastroenteritis had ≥1 ED visit demonstrates the importance of considering ED care in the impact of gastroenteritis. The high frequency of norovirus among our Health Link case patients, the majority of whom did not seek care, suggests that a norovirus GII vaccine would also considerably reduce costs to society, and potentially school absenteeism, as was accomplished by the rotavirus vaccine [30].

Few of our enrolled case patients visited outpatient clinics or were hospitalized. As such, our estimates for clinic visit and hospitalization rates attributable to norovirus lacked precision. In addition, we relied on ICD coding to identify gastroenteritis visits among the general population, which likely underestimated the number of gastroenteritis visits. This hypothesis was confirmed through investigation of ICD codes used for APPETITE participants whom we knew had visited an ED with gastroenteritis. Approximately 25% of these cases were coded generically (eg, unspecified viral infection, fever). To avoid overestimating the number of gastroenteritis visits in the province, we did not use the generic codes when identifying visits in the general population. Instead, we adjusted for underascertainment based on the proportion of APPETITE participants with known ED visits who were identified by ICD codes for gastroenteritis. Our sensitivity analysis suggested that this factor had minimal impact on the estimates.

One-third of controls and Health Link case patients did not have a specimen tested, the majority of whom did not return a specimen. Among case patients, data from follow-up interviews suggested this was, in part, driven by rapid resolution of symptoms. Given the bias attendant in a complete case approach, we used multiple imputation by chained equations [24] to incorporate into the analysis case patients without a specimen tested. Diagnostics indicated that the imputed values were consistent with the observed data.

Although we were able to adjust for viruses according to their Ct level, we lacked quantitative information on bacterial load. Thus, detections were treated the same in our models irrespective of bacterial load, which could have resulted in residual confounding. We also acknowledge the possibility that yet-to-be discovered agents might have caused some of the diarrhea we attributed to norovirus, though our microbiologic evaluation was intensive. Conversely, we might have under-ascertained norovirus in patients who provided only a swab specimen, especially among those whose presentation was vomiting without diarrhea.

In conclusion, 92% of childhood acute gastroenteritis episodes in which norovirus GII was detected could be confidently attributed to the virus. Seasonal variation and the lack of difference by age in the AF may illuminate the dynamics of symptomatic disease vs. asymptomatic carriage. Norovirus GII burden was substantial, causing 20% of medically attended gastroenteritis episodes and 18% of gastroenteritis-related ED visits each year. An effective immunoprophylaxis strategy against norovirus GII could save families and healthcare systems from the morbidity and costs of acute gastroenteritis in children in Canada and similar countries.

Supplementary Data

Supplementary materials are available at The Journal of Infectious Diseases online. Consisting of data provided by the authors to benefit the reader, the posted materials are not copyedited and are the sole responsibility of the authors, so questions or comments should be addressed to the corresponding author.

Notes

Acknowledgments. We thank Ken Morrison and Michael Cooley at Alberta Health for their work coordinating and compiling the provincial data for this study; Brent Hagel for early discussions regarding the methods for this study; and Larry Svenson, James A. Dickinson, and Raymond Tellier for comments on an early version of this manuscript. We thank the patients and their families for their participation and the full APPETITE team for their contributions. We also thank the Department of Laboratory Medicine and Pathology, University of Alberta, DynaLIFE Dx Diagnostic Laboratory Services, community laboratories, and the Alberta Provincial Laboratory for Public Health, especially the bacteriology staff for their assistance with receiving, handling, and processing specimens; the emergency department research nurses and Pediatric Emergency Medicine Research Associate Program at the Alberta Children’s Hospital for recruiting study participants; the emergency department bedside nurses for assisting with obtaining rectal swab specimens; and the research assistants, research nurses, and the Little Bit of Help research volunteer program for their assistance with participant recruitment at the Stollery Children’s Hospital.

Author contributions. G. A. M. T. had full access to all of the data and had final responsibility for the decision to submit for publication. Conceptualization: G. A. M. T. and S. B. F.; methodology: G. A. M. T., X. L. P., B. E. L., L. C., S. A., O. G. V., S. M., and S. B. F.; data collection: B. E. L., S. A., K. K., and S. B. F.; specimen analysis: X. L. P., R. Z., and L. C.; software and data analysis: G. A. M. T.; data interpretation: G. A. M. T., X. L. P., R. Z., B. E. L., L. C., S. A., O. G. V., C. M. I., P. I.T., S. M., G. C., J. M., K. K., and S. B. F.; writing—original draft preparation: G. A. M. T.; writing—review and editing: G. A. M. T., X. L. P., R. Z., B. E. L., L. C., S. A., O. G. V., C. M. I., P. I. T., S. M., G. C., J. M., K. K., and S. B. F.; funding acquisition: G. A. M. T. and S. B. F.

Disclaimer. None of the funding sources had any role in the conduct of the study, writing of the manuscript, or the decision to submit for publication.

Financial support. This work was supported by the Alberta Health Services Maternal, Newborn, Child, Youth Strategic Clinical Network; Alberta Innovates (Team Collaborative Research Innovation Opportunity grant funding the APPETITE study); the Canadian Institutes of Health Research (Banting postdoctoral fellowship to G. A. M. T.), Alberta Innovates Health Programs (postgraduate fellowship to G. A. M. T.); the University of Calgary (Eyes High postdoctoral fellowship to G. A. M. T. ); and the Alberta Children’s Hospital Foundation (professorship in child health and wellness to S. B. F.).

Potential conflicts of interest. O. G. V. reports personal fees from Pfizer and GlaxoSmithKline (GSK), grants from Pfizer, and other support from GSK and Merck Canada, outside the current work. Unrelated to the current work, P. I. T. is a consultant to Takeda Pharmaceuticals on childhood digestive disorders and the Bill & Melinda Gates Foundation on neonatal infections; and a consultant to, scientific advisory board member of, and holder of equity in MediBeacon, for work unrelated to this study. S. B. F. is a consultant to Takeda Pharmaceuticals regarding vaccine development, outside the current work. All other authors report no potential conflict. All authors have submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest. Conflicts that the editors consider relevant to the content of the manuscript have been disclosed.

Presented in part: 2019 meeting of the Pediatric Academic Societies, Baltimore, Maryland, April 24-May 1, 2019; Society for Epidemiologic Research meeting, Minneapolis, Minnesota, June 18-21, 2019.