The institutional review boards of the Oregon State Public Health Division and the Centers for Disease Control and Prevention (CDC) approved the study. All residents and staff of LTCFs were eligible for inclusion, excluding those cognitively or decisionally impaired. Written informed consent was obtained from each participant.

Prior to the onset of each winter norovirus season (December–March), LTCFs in Oregon were recruited to participate in the study. A total of 43 LTCFs were enrolled from November 2009 to January 2013. Individual study participants were recruited both before the norovirus season and when outbreaks occurred. An outbreak was suspected to be caused by norovirus based on clinical and epidemiologic characteristics [20]. Enrolled participants were classified as cases and exposed and nonexposed controls based on symptomatology and contact with a sick person (Figure 1). Only outbreaks that were reported to local health departments within 3 days of the onset date of the first case, had at least 10 sick individuals, had 20 potentially exposed individuals, and were still ongoing at the time of notification were selected for inclusion in the study. See Supplementary data for detailed study design and methods.

Clinical data were obtained directly from participants using a standardized questionnaire. The severity of gastrointestinal symptoms was assessed using a score system based on the duration and frequency of diarrhea or vomiting, fever, and treatment (Supplementary Table).

Saliva samples were collected on the day of symptom onset or exposure (day 0) from cases and exposed controls, respectively, and also on days 1, 14, and 21 to study the immune responses, which will be described in another manuscript. Serum and stool specimens were collected from cases and exposed controls on days 0 and 21 and stored at −20°C and 4°C, respectively. Single saliva, serum, and stool samples were collected from nonexposed controls within 7 days after onset of the outbreak.

Norovirus RNA from 50 µL of 10% fecal suspension was detected using GI/GII TaqMan real-time quantitative polymerase chain reaction (RT-qPCR) [9]. The P2 domain was amplified by RT-PCR using oligonucleotide primer sets specific for GI.1 and GII.4 viruses detected in the study. Genotyping and phylogenetic analysis were performed using MEGA5 software [21] and norovirus reference sequences [6]. All stool samples were also tested for enteropathogens using xTAG Gastrointestinal Pathogen Panel (Luminex Molecular Diagnostic).

Secretor genotype, defined by a single nucleotide polymorphism at position 428G>A in the FUT 2 gene, was determined by pyrosequencing as described previously [22]. HBGA phenotypes were determined from saliva by enzyme immunoassay using monoclonal antibodies (1 µg/mL) against blood group antigens A or B (Ortho Clinical Diagnostics), H antigen, Lewis (Le) a, Lewis b, Lewis x (Santa Cruz Biotechnology), or Lewis y (Calbiochem) [23]. Three saliva samples from each participant were tested for HBGA phenotypes.

To determine whether norovirus strains could bind saliva from secretor and nonsecretor individuals, 100 µL of GII.4 Den Haag, GII.4 New Orleans, or GII.4 Sydney virus-like particles (VLPs) at 0.5, 1.0, and 5.0 µg/mL were added to saliva-coated wells. VLP and saliva were genotype matched by outbreak. VLP binding was detected with polyclonal antibody against the norovirus strain and 3,3′,5,5′ tetramethylbenzidine substrate. Norovirus GII.4 Den Haag and GII.4 Sydney VLPs were kindly provided by Dr Charles Arntzen (Arizona State University) and Dr Ralph Baric (University of North Carolina), respectively.

From November 2009 through January 2013, 39 norovirus outbreaks were reported at 30 (70%) of 43 LTCFs (Supplementary Figure). Ten (26%) of 39 outbreaks met the inclusion criteria of our study and resulted in 386 illnesses, 29 hospitalizations, and 5 associated deaths.

Sixty-two cases (65% aged ≥70 years), 34 exposed controls (9% aged ≥70 years), and 18 nonexposed controls (5% aged ≥70 years) were enrolled (Table 1). Most cases were LTCF residents (79%), whereas the exposed control group predominantly consisted of staff who provided direct patient care (68%), and 50% of the nonexposed controls were nonpatient care staff. Symptoms included diarrhea (84%), fatigue (81%), vomiting (76%), and nausea (74%). Presence of both vomiting and diarrhea (62%) was reported more frequently than either symptom alone (14% and 22%, respectively) (Table 2). Adults aged ≥70 years were more likely than younger patients to report both symptoms (73% vs 45%; P = .035). Two (3%) of 62 cases reported to have had a similar illness within the preceding year.

Illness duration was longer in cases aged ≥70 years (n = 29; median, 4; interquartile range [IQR], 3–4) than aged <70 years (P = .041), with 19 (60%) lasting >3 days and 4 (13%) lasting >5 days (Figure 2A). The median severity score was 5.0 (n = 62; range, 2–10). The median score was higher in those aged 50–69 and 70–89 years (median, 6) than in those aged 21–49 and ≥90 years (median, 5; Figure 2B). No correlation was observed among severity score, illness, or shedding duration for the group or after age stratification (n = 29; Figure 2C).

The 10 studied outbreaks occurred in 7 LTCFs, with more than 1 outbreak detected in 2 LTCFs (Table 3). Norovirus was detected in stool samples from 42 cases, 4 exposed controls, and 1 nonexposed control. No sequence variation among samples from the same outbreak was detected except for outbreaks A, F, and H. Phylogenetic analysis of P2 sequences (Figure 3) showed that 18 (95%) of the 19 cases in outbreak A were typed as GII.4 New Orleans and 1 case as GII.4 Den Haag. Positive samples from exposed and nonexposed controls (A_268_Exp_Cv and A_205_NonExc, respectively) as well as convalescent stool samples had 100% sequence homology with the outbreak strain.

One (H_552) of 3 samples from outbreak H (GII.4 New Orleans) had 1 amino acid substitution in the P2 sequence. Sequences from the same participant at days 0 and 21 were identical.

Three of the 4 samples from outbreak F genotyped as GI.1, while the 4th sample (F_446_Ac) was typed as GII.4 New Orleans. However, since the convalescent stool sample from F_446 was also typed as GI.1, the GII.4 New Orleans virus at day 1 was considered unrelated to the outbreak.

Among 179 stool samples, 104 (55 acute, 49 convalescent, 47 acute/convalescent pairs) were collected from 62 cases (Table 3). Forty-two (76%) of 55 acute and 20 (42%) of 49 convalescent samples tested positive for norovirus. Six acute GII.4 New Orleans–positive stool samples also tested positive for Clostridium difficile toxin A/B (n = 5) and Giardia (n = 1).

A total of 59 stool samples (29 acute, 30 convalescent, 26 acute/convalescent pairs) were collected from 34 exposed controls. Of these, 3 (10%) of 29 acute and 1 (3%) of 30 convalescent tested positive for norovirus. One (6%) of 16 stool samples from nonexposed controls tested positive for norovirus.

The median collection date after onset of symptoms for acute positive stool was 3 days (n = 42; range, 0–8), with no significant difference among genotypes (P = .732). No significant differences were found between the mean acute viral loads in samples from cases in outbreaks with the same genotype (Table 3). In contrast, compared with outbreaks caused by GII.4 New Orleans (P = .029), GII.4 Den Haag (P = .009), and GI.1 Norwalk (P = .0281), the mean viral load in GII.4 Sydney outbreaks (2.33 × 10¹³ copies/g stool; range, 0.17–4.49 × 10¹³ copies/g stool) was significantly higher (Figure 4). Similarly, the mean viral load in GII.4 New Orleans outbreaks (7.81 × 10¹⁰ copies/g stool; range, 4.33–18.4 × 10¹⁰ copies/g stool) was significantly higher than in GII.4 Den Haag (P = .008) and GI.1 Norwalk (P = .0065) outbreaks. No difference was found between mean viral load from GI.1 Norwalk and GII.4 Den Haag outbreaks (P = .4695). Mean acute viral shedding was comparable in cases (n = 40; 1.3 × 10¹² copies/g stool) and asymptomatic exposed controls (n = 3; 1.22 × 10¹² copies/g stool; P = .4611). No correlation was observed between viral load and age (r = 0.1891; P = .2031).

Prolonged shedding (≥21 days) was detected in 16 (47%) of the 35 cases with positive acute stool. The mean acute viral load (3.02 × 10¹² copies/g stool) among cases with prolonged shedding was not significantly different from those that had a negative convalescent stool (5.91 × 10¹⁰ copies/g stool; P = .1012; Figure 5A). Nevertheless, cases with acute titers ≥10¹⁰ copies/g stool (n = 13) were significantly more likely to remain positive compared with those shedding <10¹⁰ copies/g stool (P < .001), with 9 (70%) lasting for at least 4 weeks (Figure 5B).

Of 114 individuals, 103 (90.4%) were secretors, of which 42 (40.4%) were homozygous and 61 (59.6%) were heterozygous. To investigate potential associations between HBGAs and norovirus infection or disease, the ABO, Lewis, and secretor status distributions among cases and exposed controls (n = 96) were stratified by infection (infected or noninfected) and disease (symptomatic or nonsymptomatic) and compared for each genotype or variant. Eighty-nine (93%) of 96 participants were secretor (Le^a−b+). All ABO types were detected among cases and exposed controls. Among 7 nonsecretors, 5 were Le^a+b− (Lewis positive) and 2 were Le^a−b− (Lewis negative). The 18 nonexposed controls included both secretors (n = 14; ALe^a−b+ n = 7; ALe^a−b− n = 1 and OLe^a−b+, n = 6) and nonsecretors (n = 4, Le^a+b−).

The blood type A/secretors (n = 4) of the GI.1 outbreak were symptomatic. Fifty-five (94.8%) of 58 symptomatic individuals in GII.4 outbreaks were secretors. However, secretors were as likely as nonsecretors to be infected. In GII.4 Den Haag outbreaks, 15 (60%) of 25 secretors were infected compared with 1 (50%) of 2 nonsecretors (P = 1.0). Similarly, 39 (80%) of 49 secretors and 2 (66%) of 3 nonsecretors were infected with GII.4 New Orleans viruses (P = .520), whereas GII.4 Sydney infected 5 (55%) of 9 secretors, but none of 2 nonsecretors (P = .455). In GII.4 outbreaks, we found that individuals regardless of blood type were infected, regardless of the GII.4 variant. In summary, no association between secretor status or blood type and symptomatic infection was found for GII.4 viruses.

Three of 7 nonsecretors were infected by either GII.4 Den Haag (n = 1; Le^a+b−) or GII.4 New Orleans (n = 2; Le^a+b− and Le^a−b−). To determine if nonsecretor individuals could be infected, saliva samples from all participants were incubated with different GII.4 VLPs (Figure 6). All VLPs bound to saliva samples from secretors. GII.4 Den Haag VLPs also bound to saliva samples from nonsecretors (n = 3, median, 0.432, range, 0.250–0.533; Figure 6A). In contrast, GII.4 New Orleans and Sydney VLPs were able to bind saliva from nonsecretors Le^a+b−, but not from Le^a−b− (Lewis negative), when the VLP concentration was increased to 5.0 µg/mL. Five of 6 Le^a+b− saliva samples were from controls, suggesting that lack of infection in these individuals could be the result of exposure to small amounts of virus. VLPs did not bind to saliva from either an exposed control (I_561) or a case (A_309) who were nonsecretor Le negative (Le^a−b−).