Clinical Features and Outcomes of Immunocompromised Children Hospitalized With Laboratory-Confirmed Influenza in the United States, 2011–2015

Abstract

Background

Existing data on the clinical features and outcomes of immunocompromised children with influenza are limited.

Methods

Data from the 2011–2012 through 2014–2015 influenza seasons were collected as part of the Centers for Disease Control and Prevention (CDC) Influenza Hospitalization Surveillance Network (FluSurv-NET). We compared clinical features and outcomes between immunocompromised and nonimmunocompromised children (<18 years old) hospitalized with laboratory-confirmed community-acquired influenza. Immunocompromised children were defined as those for whom ≥1 of the following applies: human immunodeficiency virus/acquired immunodeficiency syndrome, cancer, stem cell or solid organ transplantation, nonsteroidal immunosuppressive therapy, immunoglobulin deficiency, complement deficiency, asplenia, and/or another rare condition. The primary outcomes were intensive care admission, duration of hospitalization, and in-hospital death.

Results

Among 5262 hospitalized children, 242 (4.6%) were immunocompromised; receipt of nonsteroidal immunosuppressive therapy (60%), cancer (39%), and solid organ transplantation (14%) were most common. Immunocompromised children were older than the nonimmunocompromised children (median, 8.8 vs 2.8 years, respectively; P < .001), more likely to have another comorbidity (58% vs 49%, respectively; P = .007), and more likely to have received an influenza vaccination (58% vs 39%, respectively; P < .001) and early antiviral treatment (35% vs 27%, respectively; P = .013). In multivariable analyses, immunocompromised children were less likely to receive intensive care (adjusted odds ratio [95% confidence interval], 0.31 [0.20–0.49]) and had a slightly longer duration of hospitalization (adjusted hazard ratio of hospital discharge [95% confidence interval], 0.89 [0.80–0.99]). Death was uncommon in both groups.

Conclusions

Immunocompromised children hospitalized with influenza received intensive care less frequently but had a longer hospitalization duration than nonimmunocompromised children. Vaccination and early antiviral use could be improved substantially. Data are needed to determine whether immunocompromised children are more commonly admitted with milder influenza severity than are nonimmunocompromised children.

Influenza is a common cause of morbidity in children, and the highest hospitalization rates occur in infants aged less than 6 months [1, 2]. Data on immunocompromised children with influenza come from small descriptive studies that focused on specific immunocompromising conditions. Results of influenza studies in mostly adult patients have suggested that immunocompromised patients are less likely than those who are not immunocompromised to present with classic symptoms of fever, chills, myalgia, and cough [3, 4]. Outcomes among immunocompromised patients with influenza can include higher rates of hospitalization and intensive care unit (ICU) admission [4]. Limited data suggest that up to 15% of children with cancer and influenza experience respiratory failure [5] and that the rate of influenza-associated death in children who have undergone a solid organ transplant can range from 0% to >20% [6, 7]. However, studies of children with cancer and pediatric solid organ transplant recipients with influenza A(H1N1)pdm09 virus, specifically, have found more favorable outcomes [7–10]. In addition, recent data from transplant recipients suggested that influenza-related morbidity is lower in those vaccinated against influenza and for whom antiviral therapy was initiated early [11].

The Centers for Disease Control and Prevention (CDC) recommends annual influenza vaccination for all people aged 6 months or older who do not have a contraindication and early influenza antiviral treatment for all hospitalized patients with suspected or confirmed influenza [12, 13]. However, data on the use of influenza vaccination and antiviral treatment in immunocompromised children are limited. In this cross-sectional study, we compared clinical features and outcomes of immunocompromised and nonimmunocompromised children hospitalized with laboratory-confirmed influenza over 4 seasons.

METHODS

Study Design and Setting

We performed this cross-sectional study using data from CDC’s Influenza Hospitalization Surveillance Network (FluSurv-NET), which conducts active surveillance of acute care hospitals and laboratories in selected counties in California, Colorado, Connecticut, Georgia, Maryland, Michigan, Minnesota, Ohio, Oregon, New Mexico, New York, Tennessee, and Utah; the total catchment population is >27 million people (~9% of the US population) [14]. Cases of laboratory-confirmed influenza in catchment-area residents who were hospitalized during each influenza season (October 1–April 30) were reported to FluSurv-NET. Laboratory-confirmed influenza was defined as a positive result from at least 1 influenza test (rapid antigen test, reverse-transcription polymerase chain reaction, immunofluorescence antibody staining, or viral culture). Influenza testing was performed for clinical purposes at the discretion of the healthcare providers at each participating facility. Cases were identified via hospital admission, discharge, and infection control databases and by contacting hospital laboratories. For each case, demographic and clinical information was collected via medical chart abstraction and recorded into a centralized database [15]. We analyzed data from 4 influenza seasons, 2011–2012 through 2014–2015.

Participants

We included all children (<18 years of age) with laboratory-confirmed community-acquired influenza, defined as a positive influenza test result (according to any method) between 14 days before and 3 days after admission. We excluded cases that were missing outcome data for ICU admission, mechanical ventilation, and/or death.

The study was determined to be consistent with routine public health surveillance and exempt from human subjects regulations by the CDC Human Research Protection Office. The FluSurv-NET sites obtained human subjects and ethics approvals from their respective state health department and academic partner institutional review boards.

Variables

The primary exposure was the presence of at least 1 preexisting immunocompromising condition (ie, human immunodeficiency virus/acquired immunodeficiency syndrome, cancer, stem cell transplantation, solid organ transplantation, receipt of nonsteroidal immunosuppressive therapy, immunoglobulin deficiency, complement deficiency, asplenia, and/or another rare condition) (Supplementary Table 1). Patients were categorized as nonimmunocompromised if they had none of the aforementioned conditions. Because steroid use is common, and we did not have dosing data, patients with isolated receipt of steroids were classified in the primary analysis as nonimmunocompromised.

Symptoms and signs at hospital admission were first collected by FluSurv-NET in 2014–2015 and therefore are reported for only this season. Analysis of pain-based symptoms was restricted to children aged ≥4 years who could report such complaints reliably.

Primary outcomes were all-cause mortality during the hospitalization, ICU admission, and duration of hospitalization. Secondary outcomes included pneumonia, mechanical ventilation, and length of ICU admission (among patients for whom complete ICU data were available). Patients were classified as having had pneumonia if it was listed as a discharge diagnosis.

The following variables were considered potential confounders a priori: age, sex, race/ethnicity, the presence of at least 1 preexisting medical condition apart from an immunocompromising condition (abnormal upper airway, prematurity in children aged <2 years, chronic lung disease, cardiovascular disease, neuromuscular disorder, neurologic disorder, sickle cell disease without asplenia, renal disease, and/or liver disease), receipt of seasonal influenza vaccination, receipt of antiviral treatment, and influenza season. Vaccine verification was completed using standard methodology and a variety of sources, including medical charts, state registries, and provider and/or patient/proxy interviews [16, 17]. Antiviral treatment was defined as receipt of an influenza antiviral medication before or during the hospitalization. Antiviral treatment was not included as a covariate in the model for the ICU admission outcome because the majority of patients for whom data on antiviral timing were available received antiviral treatment concurrent with or subsequent to ICU admission (data not shown). We included influenza season in the model to control for seasonal variations in the predominant circulating influenza virus type and subtype.

Statistical Methods

All statistical analyses were performed using SAS 9.4 (SAS Institute, Cary, North Carolina). A P value of < .05 was considered statistically significant. Immunocompromised and nonimmunocompromised groups were compared using the χ² or Fisher exact test for categorical variables and the 2-sample t or Mann-Whitney U test for continuous variables. Modeling was used to control for potential confounding of the association between immunocompromised status and the outcomes (multivariable logistic regression for the dichotomous primary outcomes and Cox proportional hazards models for time to discharge [a hazard ratio (HR) of <1 represents an increased length of stay]). The proportional hazards assumption was assessed for each covariate using log-log likelihood curves. The Fine and Gray method was used to address death as a competing risk for hospital discharge [18].

Because vaccination was included as a covariate in all models, only children who were vaccine eligible on the basis of age (i.e., ≥6 months old) were included in the multivariable analysis. An indicator variable was used for the missing/unknown category in all models. For each model, a sensitivity analysis that included those who had received steroid therapy in the immunocompromised group was performed. We performed post hoc subgroup analyses using cancer, nonsteroidal immunosuppressive therapy, and solid organ transplantation as the primary exposure.

RESULTS

Immunocompromising Conditions

In FluSurv-NET in the 2011–2012 through 2014–2015 influenza seasons, 5415 children were hospitalized with influenza. We excluded 102 children who did not have an influenza test date within the specified range and 51 children for whom incomplete outcome data were available. Of the remaining 5262 children, 242 (4.6%) were immunocompromised (Figure 1). The most common immunocompromising conditions (nonexclusive) were having received nonsteroidal immunosuppressive therapy (n = 145 [59.9%]), cancer (n = 95 [39.3%]), being a solid organ transplant recipient (n = 34 [14.0%]), immunoglobulin deficiency (n = 24 [9.9%]), and asplenia (n = 24 [9.9%])(Figure 2; Supplementary Table 2). Of the 145 patients who received nonsteroidal immunosuppressive therapy, 53 (36.6%) had no other immunocompromising condition.

Figure 1.

Flowchart of included/excluded cases of children hospitalized with laboratory-confirmed influenza in the United States, 2011–2015.

Figure 2.

Frequency of immunocompromising conditions among children hospitalized with laboratory-confirmed influenza in the United States, 2011–2015. Conditions were not mutually exclusive. Nonsteroidal immunosuppressive therapy included chemotherapy for cancer (within 2 weeks of admission), antibody-based agents (alemtuzumab, basiliximab, daclizumab, trastuzumab, rituximab, infliximab, and/or OKT-3), immunosuppressants (cyclosporine, azathioprine, and/or leflunomide), and antirejection medications (tacrolimus, sirolimus, mycophenolate mofetil, and/or anti–thymocyte globulin).

Demographic Characteristics

Immunocompromised children were older than the nonimmunocompromised children (median [interquartile range (IQR)], 8.8 [5.4–13.3] vs 2.8 [0.7–7.3] years, respectively; P < .001) and more likely to be non-Hispanic white (43.4% vs 34.7%, respectively; P = .049) (Table 1).

Table 1.

Characteristics of Immunocompromised and Nonimmunocompromised Children Hospitalized With Laboratory-Confirmed Influenza

Characteristic	Nonimmunocompromised Children (N = 5020)	Immunocompromised Children (N = 242)	Pa
Demographic characteristics
Age (median [IQR]) (years)	2.8 (0.7–7.3)	8.8 (5.4–13.3)	<.001
Age group (n [%])
<5 years	3183 (63.4)	51 (21.1)	<.001
5–17 years	1837 (36.6)	191 (78.9)
Sex (n [%])
Male	2821 (56.2)	145 (59.9)	.25
Female	2199 (43.8)	97 (40.1)
Race/ethnicity (n [%])
Non-Hispanic white	1743 (34.7)	105 (43.4)	.049
Non-Hispanic black	1266 (25.2)	50 (20.7)
Hispanic	988 (19.7)	49 (20.2)
Other	439 (8.7)	15 (6.2)
Unknown/missing	584 (11.6)	23 (9.5)
Preexisting medical conditions (n [%])
≥1 of the conditions below	2457 (48.9)	140 (57.9)	.007
Abnormal upper airway	174 (3.5)	11 (4.5)	.37
History of prematurityb	358 (16.7)	3 (23.1)	.54c
Chronic lung disease	1382 (27.5)	65 (26.9)	.82
Cardiovascular disease	343 (6.8)	31 (12.8)	<.001
Chronic metabolic disease	187 (3.7)	22 (9.1)	<.001
Neuromuscular disorder	129 (2.6)	8 (3.3)	.48
Neurologic disorder	718 (14.3)	43 (17.8)	.13
Sickle cell disease (without asplenia)	216 (4.3)	4 (1.7)	.044
Renal disease	81 (1.6)	33 (13.6)	<.001
Liver disease	12 (0.2)	7 (2.9)	<.001c
Influenza vaccine (n [%])d
Yes	1580 (39.2)	139 (57.7)	<.001
No	2186 (54.2)	87 (36.1)
Unknown/missing	268 (6.6)	15 (6.2)
Presenting sign/symptom (n [%])e
Fever	1257 (77.6)	82 (87.2)	.028
Cough	1175 (72.5)	67 (71.3)	.79
Dyspnea/respiratory distress	635 (39.2)	12 (12.8)	<.001
Wheeze	308 (19.0)	11 (11.7)	.077
Nasal congestion	895 (55.2)	46 (48.9)	.23
GI symptomsf	593 (36.6)	28 (29.8)	.18
Chest paing	78 (10.3)	9 (11.4)	.77
Myalgiag	119 (15.8)	16 (20.3)	.30
Sore throatg	169 (22.4)	21 (26.6)	.40
Headacheg	147 (19.5)	21 (26.6)	.14
Influenza diagnostics and treatment
Testing method (n [%])h
Rapid only	853 (17.0)	12 (5.0)	<.001
Molecular	3917 (78.0)	221 (91.3)	<.001
Culture	226 (4.5)	15 (6.2)	.22
IFA	408 (8.1)	22 (9.1)	.59
Unknown	73 (1.5)	1 (0.4)	.18
Time to first influenza test from symptom onset (mean [SD])i	3.36 (4.31)	2.70 (3.81)	.023
First influenza test sent on the day of admission (n [%])	3771 (75.1)	175 (72.3)	.33
Influenza type (n [%])
A	3683 (73.4)	173 (71.5)	.72
B	1265 (25.2)	67 (27.7)
A and B	29 (0.6)	1 (0.4)
Unknown	43 (0.9)	1 (0.4)
Season (n [%])
2011–2012	453 (9.0)	14 (5.8)	.079
2012–2013	1771 (35.3)	85 (35.1)
2013–2014	1176 (23.4)	49 (20.2)
2014–2015	1620 (32.3)	94 (38.8)
Antiviral treatment (n [%])
Yes	3772 (75.1)	205 (84.7)	.002
No	1236 (24.6)	36 (14.9)
Unknown/missing	12 (0.2)	1 (0.4)
First antiviral medication (n [%])j
Oseltamivir	3737 (99.8)	204 (99.5)	.35c
Other	7 (0.2)	1 (0.5)
Antivirals <2 days after symptom onset (n [%])k
Yes	991 (27.2)	68 (35.4)	.013
No	2651 (72.8)	124 (64.6)
Antivirals <2 days after admission (n [%])l
Yes	3393 (90.6)	170 (82.9)	<.001
No	351 (9.4)	35 (17.1)

Abbreviations: GI, gastrointestinal; IFA, immunofluorescence assay; IQR, interquartile range; SD, standard deviation.

^aMedians between nonimmunocompromised and immunocompromised children were compared by using the Mann-Whitney U test. Proportions of immunocompromised and nonimmunocompromised patients were compared for each variable using the χ² test, except as noted.

^bAmong nonimmunocompromised (n = 2140) and immunocompromised (n = 13) children and those aged <2 years.

^cFisher exact test.

^dAmong nonimmunocompromised (n = 4034) and immunocompromised (n = 241) children aged ≥6 months.

^eAmong nonimmunocompromised (n = 1714) and immunocompromised (n = 94) children hospitalized during the 2014–2015 season.

^fNausea/vomiting or diarrhea.

^gAmong nonimmunocompromised (n = 754) and immunocompromised (n = 79) children aged ≥4 years.

^hTests are not mutually exclusive. Each case could include up to 4 positive results.

ⁱAmong nonimmunocompromised (n = 4850) and immunocompromised (n = 225) children for whom complete date data were available.

^jAmong nonimmunocompromised (n = 3744) and immunocompromised (n = 205) children for whom antiviral name data were available.

^kAmong nonimmunocompromised (n = 3642) and immunocompromised (n = 192) children for whom antiviral timing data were available.

^lAmong nonimmunocompromised (n = 3744) and immunocompromised (n = 205) children for whom complete date data were available. Includes children who received antiviral medication before admission.

Preexisting Medical Conditions and Influenza Vaccination Status

Immunocompromised children were more likely than nonimmunocompromised children to have at least 1 preexisting medical condition apart from an immunocompromising condition (Table 1). Characteristics of immunocompromised children are compared with those of nonimmunocompromised children with or without an underlying condition(s) in Table 2.

Table 2.

Characteristics of Children Hospitalized With Laboratory-Confirmed Influenza According to Immunocompromised Status and the Presence of an Underlying Medical Condition(s)

Characteristic	Nonimmunocompromised Childrena		Immunocompromised Children (N = 242)	Pb
	Without an Underlying Condition (N = 2563)	With an Underlying Condition (N = 2457)
Demographic characteristics
Age (median [IQR]) (years)	1.5 (0.4–4.8)	4.9 (1.6–9.3)	8.8 (5.4–13.3)	<.001
Sex (n [%])
Male	1409 (55.0)	1412 (57.5)	145 (59.9)	.11
Female	1154 (45.0)	1045 (42.5)	97 (40.1)
Race/ethnicity (n [%])
Non-Hispanic white	970 (37.9)	773 (31.5)	105 (43.4)	<.001
Non-Hispanic black	459 (17.9)	807 (32.8)	50 (20.7)
Hispanic	550 (21.5)	438 (17.8)	49 (20.2)
Other	240 (9.4)	199 (8.1)	15 (6.2)
Unknown/missing	344 (13.4)	240 (9.8)	23 (9.5)
Influenza vaccine (n [%])c
Yes	586 (32.4)	994 (44.7)	139 (57.7)	<.001
No	1098 (60.7)	1088 (48.9)	87 (36.1)
Unknown/missing	126 (7.0)	142 (6.4)	15 (6.2)
Influenza treatment (n [%])
Antiviral treatment
Yes	1843 (71.9)	1929 (78.5)	205 (84.7)	<.001
No	714 (27.9)	522 (21.3)	36 (14.9)
Unknown/missing	6 (0.2)	6 (0.2)	1 (0.4)
Antivirals <2 days after symptom onsetd
Yes	493 (27.8)	498 (26.7)	68 (35.4)	.036
No	1283 (72.2)	1368 (73.3)	124 (64.6)
Antivirals <2 days after admissione
Yes	1670 (91.4)	1723 (89.9)	170 (82.9)	<.001
No	158 (8.6)	193 (10.1)	35 (17.1)
Outcomes
All-cause mortality (n [%])	19 (0.7)	16 (0.7)	0 (0)	.40
ICU admission (n [%])	400 (15.6)	609 (24.8)	22 (9.1)	<.001f
Duration of hospitalization (mean [SD]) (days)	3.0 (5.2)	3.9 (6.4)	4.3 (6.2)	<.001g
Pneumonia (n [%])	465 (18.1)	525 (21.4)	26 (10.7)	<.001
Mechanical ventilation (n [%])	114 (4.4)	162 (6.6)	7 (2.9)	<.001

Abbreviations: ICU, intensive care unit; IQR, interquartile range; SD, standard deviation.

^aUnderlying conditions were defined as an abnormal upper airway, history of prematurity (for children aged <2 years), chronic lung disease, cardiovascular disease, chronic metabolic disease, neuromuscular disorder, neurologic disorder, sickle cell disease without asplenia, renal disease, and liver disease.

^bMedians were compared using the Kruskal-Wallis test, means were compared using analysis of variance, and proportions were compared using the χ² test.

^cAmong children aged ≥6 months (1810 nonimmunocompromised children without an underlying condition, 2224 nonimmunocompromised children with an underlying condition[s], and 241 immunocompromised children).

^dAmong children for whom complete date data were available (1776 nonimmunocompromised children without an underlying condition, 1866 nonimmunocompromised children with an underlying condition[s], and 192 immunocompromised children).

^eAmong children for whom complete date data were available (1828 nonimmunocompromised children without an underlying condition, 1916 nonimmunocompromised children with an underlying condition[s], and 205 immunocompromised children).

^fPairwise comparisons: nonimmunocompromised children without an underlying condition versus immunocompromised children (P = .007) and nonimmunocompromised children with an underlying condition(s) versus immunocompromised children (P < .001).

^gPairwise comparisons: nonimmunocompromised children without an underlying condition versus immunocompromised children (P = .002) and nonimmunocompromised children with an underlying condition(s) versus immunocompromised children (P = .34).

Among children ≥6 months of age, immunocompromised children were more likely than nonimmunocompromised children to have received seasonal influenza vaccination at least 2 weeks before their hospitalization (57.7% vs 39.2%, respectively; P < .001).

Presenting Symptoms/Signs During the 2014–2015 Season

Among the children hospitalized during the 2014–2015 influenza season (94 immunocompromised and 1620 nonimmunocompromised children), immunocompromised children were more likely to present with fever (87.2% vs 77.6%, respectively; P = .028) and less likely to present with respiratory distress (12.8% vs 39.2%, respectively; P < .001). The frequency of cough (>70% in both groups) and other symptoms did not differ significantly between the groups (Table 1).

Influenza Diagnostics and Treatment

Immunocompromised children were more likely to have undergone molecular testing (91.3% vs 78.0%, respectively; P < .001) and less likely to have undergone only rapid antigen testing (5.0% vs 17.0%, respectively; P < .001) than nonimmunocompromised children (test results were not mutually exclusive). The mean duration between symptom onset and the first influenza test was shorter for immunocompromised patients (mean ± SD, 2.70 ± 3.81 vs 3.36 ± 4.31 days, respectively; P = .023). Approximately 75% of the children in both groups had their first influenza test performed on the day of admission. Influenza A was the most common virus type, and the distribution of influenza seasons did not differ significantly between groups (Table 1).

Treatment with influenza antiviral medications was more common in immunocompromised children (84.7% vs 75.1%, respectively; P = .002). Oseltamivir was the most commonly administered influenza antiviral medication (>99% in both groups). Among children for whom antiviral timing data were available, immunocompromised children received antiviral treatment more commonly within 2 days of symptom onset (35.4% vs 27.2%, respectively; P = .013) (Table 1) and less commonly within 2 days of hospital admission (82.9% vs 90.6%, respectively; P ≤ .001) (Table 1). The timing of antiviral treatment relative to the first influenza test did not differ between the 2 groups (mean ± SD, 0.66 ± 1.60 vs 0.52 ± 1.32 days among immunocompromised vs nonimmunocompromised children, respectively; P = .2).

Outcomes

All-Cause Mortality

All-cause mortality was not included as an outcome in the multivariable analysis because of its infrequency (0 [0%] immunocompromised children vs 35 [0.7%] nonimmunocompromised children; P = .41).

ICU Admission

In the unadjusted analysis, immunocompromised children were less likely than nonimmunocompromised children to require ICU admission (odds ratio [OR] [95% confidence interval (CI)], 0.40 [0.26–0.62]) (Table 3). This result was observed regardless of whether nonimmunocompromised children had another underlying medical condition(s) (Table 2). Other univariable associations with ICU admission are summarized in Table 3.

Table 3.

Association Between Patient Characteristics and ICU Admission Among Children Hospitalized With Laboratory-Confirmed Influenza

Characteristic	No ICU Admission (N = 4231)	ICU Admission (N = 1031)	ICU Admission (OR [95% CI])
Immunocompromised
No	4011 (79.9)	1009 (20.1)	Reference
Yes	220 (90.9)	22 (9.1)	0.40 (0.26–0.62)
Demographic characteristics (n [%])
Age group
<5 years	2662 (82.3)	572 (17.7)	Reference
5–17 years	1569 (77.4)	459 (22.6)	1.36 (1.19–1.56)
Sex
Male	2390 (80.6)	576 (19.4)	Reference
Female	1841 (80.2)	455 (19.8)	1.03 (0.89–1.18)
Race/ethnicity
Non-Hispanic white	1465 (79.3)	383 (20.7)	Reference
Non-Hispanic black	1024 (77.8)	292 (22.2)	1.09 (0.92–1.30)
Hispanic	854 (82.4)	183 (17.6)	0.82 (0.67–1.00)
Other	382 (84.1)	72 (15.9)	0.72 (0.55–0.95)
Unknown/missing	506 (83.4)	101 (16.6)	0.76 (0.60–0.97)
Preexisting medical conditions (n [%])
≥1 of the conditions below	1969 (75.8)	628 (24.2)	1.79 (1.56–2.06)
Abnormal upper airway	123 (66.5)	62 (33.5)	2.14 (1.56–2.92)
History of prematuritya	271 (15.2)	90 (24.1)	1.77 (1.35–2.32)
Chronic lung disease	1072 (74.1)	375 (25.9)	1.69 (1.46–1.95)
Cardiovascular disease	281 (75.1)	93 (24.9)	1.39 (1.09–1.78)
Chronic metabolic disease	149 (71.3)	60 (28.7)	1.69 (1.24–2.30)
Neuromuscular disorder	84 (61.3)	53 (38.7)	2.68 (1.88–3.80)
Neurologic disorder	523 (68.7)	238 (31.3)	2.13 (1.79–2.53)
Sickle cell disease (without asplenia)	206 (93.6)	14 (6.4)	0.27 (0.16–0.46)
Renal disease	95 (83.3)	19 (16.7)	0.82 (0.50–1.34)
Liver disease	15 (78.9)	4 (21.1)	1.10 (0.36–3.31)
Influenza vaccine (n [%])b
Yes	1364 (40.3)	355 (39.9)	Reference
No	1804 (53.3)	469 (52.8)	1.00 (0.86–1.17)
Unknown/missing	218 (6.4)	65 (7.3)	1.15 (0.85–1.55)
Influenza diagnostics and treatment (n [%])
Influenza type
A	3097 (80.3)	759 (19.7)	Reference
B	1076 (80.8)	256 (19.2)	0.97 (0.83–1.14)
A and B	22 (73.3)	8 (26.7)	1.49 (0.66–3.35)
Unknown	36 (81.8)	8 (18.2)	0.91 (0.42–1.96)
Season
2011–2012	377 (80.7)	90 (19.3)	Reference
2012–2013	1532 (82.5)	324 (17.5)	0.89 (0.68–1.15)
2013–2014	976 (79.7)	249 (20.3)	1.07 (0.82–1.40)
2014–2015	1346 (78.5)	368 (21.5)	1.15 (0.89–1.48)
Antiviral treatment
Yes	3131 (78.7)	846 (21.3)	Reference
No	1093 (85.9)	179 (14.1)	0.61 (0.51–0.72)
Unknown/missing	7 (53.8)	6 (46.2)	3.17 (1.06–9.46)

Abbreviations: CI, confidence interval; ICU, intensive care unit; OR, odds ratio.

^aAmong children aged <2 years not admitted to an ICU (n = 1780) and those admitted to an ICU (n = 373).

^bAmong children aged ≥6 months not admitted to an ICU (n = 3386) and those admitted to an ICU (n = 889).

When we controlled for age, sex, race/ethnicity, the presence of ≥1 preexisting medical condition apart from immunocompromising conditions, vaccination, and influenza season, immunocompromised children remained less likely to receive intensive care than nonimmunocompromised children (adjusted OR [95% CI], 0.31 [0.20–0.49]) (Table 4). In a sensitivity analysis that included having received steroid therapy (n = 152) as an immunocompromising condition, the effect was attenuated but the adjusted odds of ICU admission remained lower in the immunocompromised group (adjusted OR [95% CI], 0.55 [0.41–0.74]). Children with cancer, children who received nonsteroidal immunosuppressive therapy, and pediatric solid organ transplant recipients also had lower adjusted odds of ICU admission than the nonimmunocompromised children (Table 4). Demographic and clinical characteristics of the subgroups are detailed in Supplementary Table 3.

Table 4.

ICU Admission in Immunocompromised vs Nonimmunocompromised Children Hospitalized With Laboratory-Confirmed Influenza

Analysis Type	ICU Admission (aOR [95% CI])a
Primaryb	0.31 (0.20–0.49)
Sensitivity, steroidsb,c	0.55 (0.41–0.74)
Subgroup
Cancerd	0.16 (0.06–0.43)
Nonsteroidal immunosuppressive therapye	0.30 (0.17–0.55)
Solid organ transplantationf	0.28 (0.09–0.93)

Abbreviations: aOR, adjusted odds ratio; CI, confidence interval; ICU, intensive care unit.

^aControlling for age, sex, race/ethnicity, at least 1 preexisting medical condition apart from an immunocompromising condition, influenza vaccination, and influenza season.

^bAmong 4275 children aged ≥6 months.

^cIncluding receipt of steroid therapy as an immunocompromising condition.

^dAmong 4128 children aged ≥6 months (excluding immunocompromising conditions apart from cancer).

^eAmong 4178 children aged ≥6 months (excluding immunocompromising conditions apart from having received nonsteroidal immunosuppressive therapy).

^fAmong 4068 children aged ≥6 months (excluding immunocompromising conditions apart from being a solid organ transplantation recipient).

Duration of Hospitalization

In the unadjusted analysis, immunocompromised children had a longer length of hospitalization (mean ± SD, 4.3 ± 6.2 days) than nonimmunocompromised children (mean ± SD, 3.4 ± 5.8 days; P = .032). In the time-to-event models that considered time to hospital discharge, in which we controlled for the same covariates as for ICU admission, immunocompromised children had a longer duration of hospitalization (adjusted HR [aHR] of hospital discharge [95% CI], 0.89 [0.80–0.99]) (Table 5). In a sensitivity analysis that included having received steroid therapy as an immunocompromising condition, the effect was attenuated and was no longer significant (aHR of hospital discharge [95% CI], 0.93 [0.85–1.01]). In subgroups, only children with cancer had a longer duration of hospitalization than the nonimmunocompromised children (aHR of hospital discharge [95% CI], 0.77 [0.66–0.91]) (Table 5).

Table 5.

aHRs of Hospital Discharge in Immunocompromised vs Nonimmunocompromised Children Hospitalized With Laboratory-Confirmed Influenza

Analysis Type	Hospital Discharge (aHR [95% CI])a
Primaryb	0.89 (0.80–0.99)
Sensitivity, steroidsb,c	0.93 (0.85–1.01)
Subgroup
Cancerd	0.77 (0.66–0.91)
Nonsteroidal immunosuppressive therapye	0.91 (0.80–1.04)
Solid organ transplantationf	1.23 (1.00–1.52)

Abbreviations: aHR, adjusted hazard ratio; CI, confidence interval.

^aControlling for age, sex, race/ethnicity, at least 1 preexisting medical condition apart from an immunocompromising condition, influenza vaccination, and influenza season.

^bAmong 4272 children aged ≥6 months for whom length of hospitalization data were available.

^cIncluding receipt of steroid therapy as an immunocompromising condition.

^dAmong 4125 children aged ≥6 months for whom length of hospitalization data were available (excluding immunocompromising conditions apart from cancer).

^eAmong 4175 children aged >6 months for whom length of hospitalization data were available (excluding immunocompromising conditions apart from having received nonsteroidal immunosuppressive therapy).

^fAmong 4065 children aged >6 months for whom length of hospitalization data were available (excluding immunocompromising conditions apart from being a solid organ transplantation recipient).

Secondary Outcomes

Immunocompromised children were less likely than nonimmunocompromised children to be diagnosed with pneumonia (10.7% vs 19.7%, respectively; P = .005). Mechanical ventilation was similar among immunocompromised and nonimmunocompromised children (2.9% vs 5.5%, respectively; P = .079). ICU admission duration did not differ among immunocompromised (mean ± SD, 5.5 ± 8.4 days) and nonimmunocompromised (mean ± SD, 4.4 ± 8.6 days; P = .55) children for whom data were available.

DISCUSSION

In this large population-based study of children hospitalized with laboratory-confirmed influenza during 4 recent influenza seasons, we found that immunocompromised children were older and more likely to have another preexisting medical condition(s), to have received influenza vaccine, and to receive antiviral treatment than children who were not immunocompromised. Immunocompromised children were more likely to present with fever and about half as likely to present with respiratory distress and be diagnosed with pneumonia. Immunocompromised children were less likely than nonimmunocompromised children to receive intensive care, even when the latter group was stratified according to the presence of an underlying condition(s). After we controlled for this and other potential confounders, immunocompromised children remained less likely to be admitted to the ICU and had a slightly longer duration of hospitalization. Death was uncommon in both groups.

Although our finding that immunocompromised children hospitalized with influenza have milder disease is somewhat counterintuitive, previous reports have found mild courses of influenza, particularly with influenza A(H1N1)pdm09 virus, in immunocompromised children [8–10]. However, some data on immunocompromised children and adults with influenza have suggested that they have high rates of influenza-associated complications and worse outcomes than their nonimmunocompromised counterparts [4–6, 19, 20]. The population of immunocompromised children in our study did not have such severe outcomes. One possible explanation for this difference is that our group of immunocompromised patients was more heterogeneous. We used a broad inclusive definition of immunocompromised status, whereas much of the published literature on influenza in immunocompromised patients is restricted to immunocompromised patients with a particular condition, such as cancer, acquired immunodeficiency syndrome, or having undergone solid organ transplantation. Outcomes can be worse for children with certain immunocompromising conditions, although our findings related to ICU admission were similar in subgroup analyses of children with cancer, those who had received nonsteroidal immunosuppressive therapy, and solid organ transplant recipients. The surveillance data collected by FluSurv-NET do not allow for a more detailed classification of the patients’ degree of immunosuppression. For example, the degree of immunosuppression in children after stem cell transplantation varies over time, and FluSurv-NET does not collect data on transplantation timing. Additional prospective investigations in well-defined immunocompromised populations are needed to explore these issues.

The threshold for admitting immunocompromised children with influenza might be lower than that for nonimmunocompromised children. For example, immunocompromised children with relatively mild influenza signs and symptoms, such as isolated fever, might be admitted for monitoring because of their immunocompromised status, whereas nonimmunocompromised children might be admitted only with more severe illness. Our finding that immunocompromised children were less likely to present with respiratory distress in the 2014–2015 season supports this hypothesis. FluSurv-NET did not collect data on factors that would enable us to assess and control for disease severity at the time of admission, such as vital signs, level of responsiveness, and laboratory values. In addition, we were unable to compare hospitalization rates between immunocompromised and nonimmunocompromised children with influenza during the study period. Immunocompromised children with influenza have previously been reported to have high admission rates. A 2010 cohort study of pandemic H1N1 influenza found that 71% of solid organ transplant recipients with medically attenuated influenza were admitted [7]. Further exploration of disease severity on admission and admission rates might help clarify whether a bias toward admitting immunocompromised children with mild influenza exists. This phenomenon has been hypothesized to occur in other groups with influenza, such as patients with asthma and pregnant women, in whom reported outcomes are not as severe as anticipated [21, 22].

Our findings suggest substantial room for improvement in the use of influenza vaccination and antiviral agents in children. Only 58% of the immunocompromised children and 39% of the nonimmunocompromised children in this study had received seasonal influenza vaccine. These frequencies correspond to children who developed influenza despite vaccination. The CDC estimated influenza vaccination coverage of 51.5% to 58.9% among US children aged 6 months to 17 years between 2011 and 2015 [23]. The higher vaccine coverage rate observed among immunocompromised children than among nonimmunocompromised children might reflect increased opportunity for vaccination because of frequent healthcare encounters, a greater likelihood of providers recommending vaccination, or decreased effectiveness of influenza vaccination in preventing hospitalization in this immunocompromised population. Findings from a large study of community-acquired pneumonia also suggested that influenza vaccination might not be as effective in this population as in others [24] and supported vaccinating close contacts of immunocompromised patients. It is possible also that vaccination protected against adverse outcomes in the immunocompromised group. Seasonal influenza vaccination attenuated adverse outcomes among adults hospitalized with influenza during the 2013–2014 influenza season [25]. Both seasonal influenza vaccination and early antiviral therapy have been associated with lower influenza-related morbidity in solid organ and stem cell transplant recipients [11].

Although antiviral use was higher than 75% in both groups, and higher in immunocompromised children than in nonimmunocompromised children, it could be improved. Antiviral treatment <2 days after symptom onset was poor in both groups but higher in immunocompromised children, perhaps because of earlier care seeking in this group. Antiviral administration <2 days after hospital admission was much higher than early antiviral use and, by this time point, more nonimmunocompromised patients had been treated. Influenza antiviral agents are most effective when given early, and they have been associated with a shorter duration of hospitalization and higher survival rates among critically ill children hospitalized with influenza [26, 27]. Earlier care seeking, particularly by children at high risk, and earlier initiation of antiviral therapy in the outpatient setting might have prevented some children from being hospitalized or from having an adverse outcome.

This study had limitations, including possible selection biases in testing and admission practices; individual clinicians decide whether to perform influenza testing and admit a patient, and these decisions might have been affected by a patient’s immunocompromised status.

In addition, the suboptimal sensitivity of rapid antigen testing might have resulted in some misclassification of influenza cases. Misclassification of immunocompromised status also could have occurred in the process of retrospective reviewing the charts. In addition, our models did not fully account for seasonal variation in vaccine effectiveness, which was of particular concern in the 2014–2015 season, in which circulating influenza A(H3N2) viruses had drifted from the vaccine virus. Unmeasured confounders or factors unrelated to influenza infection could have contributed to the outcomes. Last, our findings are limited to children with influenza-related hospitalization.

CONCLUSIONS

Influenza vaccination and antiviral use, especially early antiviral medication initiation, could be improved substantially in immunocompromised and nonimmunocompromised children. Our results suggest that, among children hospitalized with influenza, immunocompromised children are admitted for intensive care less frequently. Further investigation is needed to determine how and why influenza outcomes might differ between immunocompromised and nonimmunocompromised children, including whether a bias toward admitting immunocompromised children with relatively mild influenza exists.

Supplementary Data

Supplementary materials are available at Journal of the Pediatric Infectious Diseases Society online.

Notes

Acknowledgments. We thank all those involved in influenza hospitalization surveillance at each of the FluSurv‐NET sites, including Wendy Baughman, Stepy Thomas, Suzanne Segler, Krista Lung, Katie Dyer, Karen Lieb, Maria Gaitan, Christina Felsen, Ashley Coates, Maria Rosales, Craig Morin, Melissa McMahon, Sara Vetter, Dave Boxrud, Anna Strain, Patricia Ryan, Brian Bachaus, Emily Blake, Molly Hyde, Elisabeth Vaeth, Kathy Angeles, Lisa Butler, Emily Hancock, Sarah Shrum, Sarah Khanlian, Robert Mansmann, and Elizabeth Dufort.

Disclaimer. The findings and conclusions in this paper are those of the authors and do not necessarily represent the official position of CDC.

Financial support. This work was supported, in part, by grants from the National Institutes of Health/National Center for Advancing Translational Sciences (NACTS) (grant UL1 TR002378) (Georgia Clinical and Translational Science Alliance).

Potential conflicts of interest. E. J. A. has received research funding unrelated to this paper from Novavax, Pfizer, Regeneron, and Medimmune and has also consulted for AbbVie unrelated to this paper. All other authors: No reported conflicts of interest. 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.