Patients Hospitalized With Laboratory-Confirmed Influenza During the 2010–2011 Influenza Season: Exploring Disease Severity by Virus Type and Subtype

Abstract

Background. The 2010–2011 influenza season was dominated by influenza A(H3N2) virus, but influenza A(H1N1) pdm09 (pH1N1) and B viruses cocirculated. This provided an opportunity to explore within-season predictors of severity among hospitalized patients, avoiding biases associated with season-to-season differences in strain virulence, population immunity, and healthcare seeking.

Methods. Population-based, laboratory-confirmed influenza hospitalization surveillance data were used to examine the association between virus type/subtype and outcomes in children and adults. Multivariable analysis explored virus type/subtype, prompt antiviral treatment, medical conditions, and age as predictors for severity (intensive care unit admission or death).

Results. In children, pH1N1 (adjusted odds ratio [aOR], 2.19; 95% confidence interval [CI], 1.11–4.3), chronic metabolic disease (aOR, 5.23; 95% CI, 1.74–15.69), and neuromuscular disorder (aOR, 4.84; 95% CI, 2.02–11.58) were independently associated with severity. In adults, independent predictors were pH1N1 (aOR, 2.21; 95% CI, 1.66–2.94), chronic lung disease (aOR, 1.46, 95% CI, 1.12–1.89), and neuromuscular disorder (aOR, 1.68; 95% CI, 1.11–2.52).Antiviral treatment reduced the odds of severity among adults (aOR, 0.47; 95% CI, .33–.68).

Conclusions. During the 2010–2011 season, pH1N1 caused more severe disease than H3N2 or B in hospitalized patients. Underlying medical conditions increased severity despite virus strain. Antiviral treatment reduced severity among adults. Our findings underscore the importance of influenza prevention.

influenza, influenza A(H1N1) pdm09, severe influenza, influenza hospitalizations, influenza surveillance

In April 2009, the influenza A(H1N1) pdm09 (pH1N1) virus emerged in Mexico and the United States and rapidly spread through the globe [1]. Children and young adults were disproportionally affected [2, 3], whereas older persons were relatively spared because of cross-protective immunity from prior exposures to related influenza viruses [4]. From April 2009 to April 2010, the Centers for Disease Control and Prevention (CDC) estimated that approximately 60.8 million cases, 274 304 hospitalizations, and 12 469 deaths occurred in the United States due to pH1N1 [5]. Because of the high levels of immunity in the elderly, who usually have the highest rates of mortality from influenza, the number of deaths during the pandemic was somewhat lower than that reported during a severe influenza season. Viboud et al used “years of life lost” to estimate the impact of the 2009 pandemic-associated mortality, accounting for the young age distribution of deaths during the 2009 pandemic [6]. The authors concluded that pH1N1 resulted in more than a million years of life lost in the United States, which is within the range of deaths seen during the 1968 pandemic. Despite the substantial morbidity and mortality documented, some studies suggested that pH1N1 did not cause more severe clinical disease than that caused by seasonal influenza A and B viruses in previous seasons [7–9]. Comparison of 2009 pandemic data with data from past influenza seasons is challenging because of season-to-season differences in virulence of circulating strains, levels of population immunity, healthcare-seeking behavior, and effectiveness of the vaccine distributed during a particular season. Additionally, there was an unprecedented increase in use of influenza antivirals during the 2009 pandemic that would have to be taken into account during such analysis.

The first US influenza season after the 2009 pandemic was dominated by seasonal influenza A (H3N2) virus, but pH1N1 and B viruses co-circulated [10]. Because there was no significant antigenic change in the pH1N1 virus circulating during the 2010–2011 influenza season, the season provided a unique opportunity to directly compare the pH1N1-associated clinical severity with that caused by H3N2 and B viruses and explore independent predictors of severity among hospitalized persons. We used data from the Influenza Hospitalization Surveillance Network (FluSurv-NET), a population-based surveillance system for laboratory-confirmed influenza hospitalizations, to compare demographic and clinical characteristics of cases by type and subtype of influenza viruses.

METHODS

Study Setting

We used FluSurv-NET data from October 2010 through April 2011. This network conducted population-based influenza-associated hospitalization surveillance in 81 selected counties in California, Colorado, Connecticut, Georgia, Idaho, Maryland, Michigan, Minnesota, New York, Oklahoma, Ohio, Oregon, Rhode Island, Tennessee, and Utah. The surveillance area encompassed a total of 276 reporting hospitals serving >29 million children and adults and represented 9% of the US population.

Data collection through FluSurv-NET was determined by the CDC to be for routine public health surveillance and was not subject to institutional review board approval for human research protections.

Definitions

A case of influenza-associated hospitalization was defined as a person residing in the surveillance area who was admitted to 1 of the surveillance hospitals ≤14 days after or ≤3 days before a positive influenza test. Laboratory testing for influenza was ordered at the discretion of clinicians providing clinical care. Laboratory confirmation was defined as a positive result from viral culture, direct or indirect fluorescent antibody staining, rapid antigen test, or reverse transcription polymerase chain reaction. Information on demographic characteristics, medical history, influenza vaccination status, clinical course during hospitalization, and treatment with influenza antiviral medications was collected through review of medical charts using a standardized questionnaire. Obesity was defined as a body mass index ≥30 kg/m² for adults and as a body mass index ≥95th percentile for children (excluding those aged <2 years). In adults, a body mass index ≥40 kg/m² was considered morbid obesity.

A patient was considered immunosuppressed if diagnosed with immunoglobulin deficiency, cancer, or human immunodeficiency virus/AIDS or treated with oral or injectable steroids for a minimum of 2 weeks before admission. Preexistent underlying medical conditions were obtained during medical chart review. When date of patient's influenza vaccination was not available in the medical record, a vaccination registry or the patient's primary care provider was consulted or the patient/caregiver was interviewed to obtain vaccination history.

Statistical Analysis

Age and type/subtype for influenza-associated hospitalization rates were calculated using the National Center for Health Statistics' vintage 2010 bridged-race postcensal population estimates for the counties included in the surveillance catchment area. To calculate rates using all cases, we adjusted the data for cases of influenza A virus without subtype information, using the age distribution of hospitalized cases with virus subtype information available. Ninety-five percent confidence intervals (CIs) for rates were derived using the delta method [11]. Only data from patients with information on influenza virus type and subtype were used for further analysis. We excluded hospital-acquired influenza cases (defined as those with onset >3 days after hospitalization) because reasons for hospitalization could be attributed to causes other than influenza. Pregnant women with influenza were also excluded because of the potential for complicating factors associated with disease severity in this group that would not be comparable with factors present in the rest of the case population.

Descriptive analyses were used to summarize patients' demographic and clinical characteristics, such as age, sex, race, presence of underlying chronic medical conditions, influenza virus type and subtype, antiviral use, length of hospital stay, intensive care unit (ICU) admission, assisted ventilation, and others. A patient was classified as having severe influenza if admitted to an ICU or if the patient died during hospitalization. We used 2-sided χ² test (or Fisher exact test when appropriate) to compare virus type/subtype with demographic, clinical outcomes, and severity for children (aged <18 years) and adults (aged ≥18 years) separately. Unconditional logistic regression models for children and adults were fit separately to determine whether virus type/subtype was an independent predictor of influenza severity and to explore other potential risk factors. Variables known as potential confounders or that were associated with disease severity in the univariable analysis (P < .10) were included in the multivariable regression models; variables that changed odds ratio for severe influenza by ≥10% when removed from the analysis were also kept in final models [12]. To avoid treatment bias introduced by physicians' inclination to treat more severe cases with antivirals [13], we only included patients treated with antivirals in our models and controlled for the potential effect of prompt treatment (antiviral initiated <3 days of admission) vs late treatment (antiviral initiated ≥3 days after admission). (Among 749 children with illness onset information, median number of days from illness onset to hospitalization was 2 [interquartile range [IQR], 1–4); among 2498 adults with illness onset information, the median was 2 days [IQR, 1-3]. Assuming that treatment initiated up to 5 days after illness onset could be beneficial, we decided to use a cutoff of <3 days after admission as an indication of prompt antiviral treatment.) For the model, we used individual underlying chronic medical conditions that had shown a potential association with disease severity. For all predictors and associations between variables, differences were considered significant at P < .05. Data analysis was performed using SAS version 9.3 software.

RESULTS

During October 2010 through April 2011, there were 6284 laboratory-confirmed influenza hospitalizations identified in the network. Overall, hospitalization rates associated with H3N2 were 10.8 per 100 000 population, higher than those associated with influenza B and pH1N1 (7.2 and 5.6 per 100 000 population, respectively). Rates varied by age and were highest among those aged <5 years and those aged ≥65 years, by type and subtype. Influenza B– and pH1N1-associated hospitalization rates were comparable among young children (12.3 and 11.1 per 100 000 population, respectively) and adults aged ≥65 years (6.9 and 5.8 per 100 000 population, respectively) but lower than rates associated with H3N2 (20.4 and 49.6 per 100 000 for young children and adults aged ≥65 years, respectively). Rates associated with pH1N1 and H3N2 among those aged 50–64 years were comparable and much higher than the rate associated with influenza B (Figure 1).

Figure 1.

Laboratory-confirmed influenza hospitalization rates by virus type and subtype and age group, United States, 2010–2011 influenza season.

After excluding 2169 cases with incomplete information on virus subtype and 494 cases that were either hospital-acquired cases or cases that occurred in pregnant women, 3621 (57.6%) hospitalizations were further analyzed (830 children and 2791 adults). Among children, 351 (42.3%) had at least 1 underlying medical condition identified, the most common being asthma/reactive airway disease (19.6%), developmental delay (9%), and chronic lung disease (6%). The majority (85.5%) of adults had underlying medical conditions identified; the most common were chronic cardiovascular disease (37.6%), chronic metabolic disease (36%), chronic lung disease (24%), and asthma (20%).

Demographics by Virus Type and Subtype

In general, demographic characteristics in hospitalized children were similar by virus type and subtype, although the age distribution of children with H3N2 was slightly shifted toward children aged <5 years (P = .02) (Table 1).

Table 1.

Demographic and Clinical Characteristics of Hospitalized Children and Adults With Laboratory-Confirmed Influenza, by Virus Type and Subtype, United States, 2010–2011 Influenza Season

Characteristics	Children (n = 830)			P Value	Adults (n = 2791)			P Value
	pH1N1 (n = 163)	H3N2 (n = 252)	B (n = 415)		pH1N1 (n = 761)	H3N2 (n = 1497)	B (n = 533)
	No. (%)	No. (%)	No. (%)		No. (%)	No. (%)	No. (%)
Age group, y				.02				<.0001
<2	78 (47.8)	129 (51.2)	182 (43.8)		…	…	…
2–4	22 (13.5)	53 (21.0)	71 (17.1)		…	…	…
5–17	63 (38.6)	70 (27.8)	162 (39.0)		…	…	…
18–49	…	…	…		348 (45.7)	286 (19.1)	195 (36.6)
50–64	…	…	…		305 (40.1)	301 (20.1)	112 (21.0)
≥65	…	…	…		108 (14.2)	910 (60.8)	226 (42.4)
Sex				.19				.24
Male	98 (60.2)	129 (51.2)	221 (53.2)		371 (48.7)	682 (45.5)	237 (44.5)
Female	65 (39.8)	123 (48.8)	194 (46.7)		390 (51.2)	815 (54.5)	296 (55.5)
Underlying medical conditionsa	66 (40.5)	98 (38.9)	187 (45.1)	.26	629 (82.6)	1321 (88.2)	436 (81.8)	<.0001
BMIb				.23				<.0001
Underweight	10 (16.7)	10 (10.2)	28 (16.3)		23 (3.7)	64 (5.1)	23 (5.2)
Normal	41 (68.3)	62 (63.3)	97 (56.4)		156 (24.9)	396 (31.8)	153 (34.4)
Overweight	5 (8.3)	11 (11.2)	14 (8.1)		162 (25.8)	376 (30.2)	117 (26.3)
Obese	4 (6.7)	15 (15.3)	33 (19.2)		171 (27.3)	295 (23.7)	109 (24.5)
Morbid obese					114 (18.2)	112 (9.0)	43 (9.6)
Vaccination ≥14 d prior to hospitalizationc	5 (22.7)	11 (42.3)	16 (32.0)	.35	222 (33.4)	761 (55.9)	203 (41.9)	<.0001
Antiviral therapy	79 (48.5)	146 (57.9)	206 (49.6)	.07	563 (74.1)	1085 (72.5)	365 (68.5)	.08
Antiviral started <3 d from admissiond	72 (93.5)	139 (96.5)	186 (92.5)	.29	511 (91.7)	987 (92.6)	334(92.5)	.82
Length of stay ≥5 d	34 (20.9)	27 (10.7)	96 (23.1)	.0003	286 (37.7)	614 (41.0)	193 (36.2)	.09
ICU admission	36 (22.1)	32 (12.7)	62 (14.9)	.03	192 (25.3)	202 (13.5)	85 (15.9)	<.0001
Mechanical ventilation	17 (10.5)	4 (1.6)	27 (6.5)	.0005	105 (13.8)	86 (5.7)	37 (6.9)	<.0001
ECMO	2 (1.2)	1 (0.4)	0	.08	6 (0.8)	3 (0.3)	1 (0.2)	.07
Bacterial coinfection	6 (3.7)	7 (2.8)	12 (2.9)	.85	51 (6.7)	82 (5.5)	36 (6.7)	.38
Discharge diagnosese
Pneumonia	47 (28.8)	39 (15.5)	85 (20.5)	.004	294 (38.6)	496 (33.1)	171 (32.1)	.02
Bronchiolitis	18 (11.0)	25 (9.9)	47 (11.3)	.85	…	…	…	…
Seizure	6 (3.7)	4 (1.6)	26 (6.3)	.01	…	…	…	…
Encephalitis	1 (0.6)	1 (0.4)	3 (0.7)	.87	8 (1.0)	17 (1.1)	5 (0.9)	.93
Stroke	…	…	…	…	2 (0.2)	4 (0.3)	0	.491
ARDS	5 (3.1)	3 (1.2)	7 (1.7)	.362	43 (5.6)	36 (2.4)	11 (2.0)	<.0001
Death	0	1 (0.4)	1 (0.2)	.725	28 (3.7)	59 (3.9)	22 (4.1)	.92

Bold numbers indicate statistically significance using χ² test (or Fisher exact when appropriate). Abbreviations: ARDS, Acute Respiratory Distress Syndrome; BMI, body mass index; ECMO, Extracorporeal Membrane Oxygenation.

^a Underlying medical conditions include asthma, chronic lung disease, chronic cardiovascular disease, renal disease, hemoglobinopathy, neuromuscular disorder, immunosuppressive condition, developmental delay, or cognitive dysfunction.

^b Excludes those aged <2 years.

^c Excludes children aged <6 months.

^d Among 2013 adults and 431 children who received treatment with influenza antiviral medications.

^e Discharge diagnoses were collected as predefined categories; stroke was not collected for children, and bronchiolitis and seizures were not collected for adults.

Among adults, distribution of cases by virus type and subtype varied significantly by age (P < .0001). Among those with H3N2, 60.8% of hospitalizations were in those aged ≥65 years, compared with 42.4% and 14.2% among influenza B and pH1N1 respectively. Hospitalizations associated with pH1N1 were highest among those aged 18–49 years (45.7%) and 50–64 years (40.1%). Morbid obesity was more common among patients hospitalized with pH1N1 (18.2%) than among those with H3N2 (9%) or influenza B (9.6%; P < .0001). Other differences in demographic characteristics by influenza virus type and subtype are presented in Table 1.

Clinical Characteristics by Virus Type and Subtype

Children hospitalized with pH1N1 and influenza B were more frequently hospitalized for ≥5 days than those with H3N2 (approximately 20% vs 10.7%, respectively; P = .0003). ICU admission was also more frequent among children hospitalized with pH1N1 (22.1%) than among those with H3N2 (12.7%) and influenza B (14.9%; P = .031). A higher percentage of children with pH1N1 (10.5%) and influenza B (6.5%) required mechanical ventilation compared with those with H3N2 (1.6%; P = .0005). Discharge diagnoses of pneumonia were also more common in children with pH1N1 (28.8%) and influenza B (20.5%) than in those with H3N2 (15.5%; P = .004). Discharge diagnosis of seizure was more common in children hospitalized with influenza B (6.3%) than in those with H3N2 (1.6%) or pH1N1 (3.7%; P = .01). Two children died (1 with H3N2 and 1 with influenza B) (Table 1).

Among adults, no difference was observed in length of hospital stay by virus type and subtype. However, 25.3% of adults with pH1N1were admitted to the ICU compared with 13.5% of those with H3N2 (and 15.9% of those with influenza B–associated hospitalizations; P < .0001). Twice as many adults with pH1N1 (13.8%) required mechanical ventilation than did those with H3N2 (5.7%) and influenza B (6.9%; P < .0001). Similarly, acute respiratory distress syndrome was 2 times more frequent among those with pH1N1 (5.6%) than among those with H3N2 (2.4%) and B (2%; P < .0001) (Table 1). ICU admission, mechanical ventilation, and acute respiratory distress syndrome were statistically significantly more common in those with pH1N1 than in those with H3N2 and influenza B viruses even after further age stratification (Table 2).

Table 2.

Selected Characteristics of Adults Hospitalized With Laboratory-Confirmed Influenza by Age Group and Virus Type/Subtype, Unied States, 2010–2011 Influenza season

Selected Characteristics	≥65 Years (n = 1244)			P Value	50–64 y (n = 713)			P Value	18–49 y (n = 829)			P Value
	pH1N1 (n = 108)	H3N2 (n = 910)	B (n = 226)		pH1N1 (n = 302)	H3N2 (n = 299)	B (n = 112)		pH1N1 (n = 348)	H3N2 (n = 286)	B (n = 195)
	No. (%)	No. (%)	No. (%)		No. (%)	No. (%)	No. (%)		No. (%)	No. (%)	No. (%)
ICU admission	31 (28.7)	117 (12.8)	226 (15.9)	<.0001	85 (28.2)	50 (16.7)	20 (17.8)	.001	76 (21.8)	35 (12.2)	29 (14.8)	.004
Mechanical ventilation	13 (12.1)	46 (5.0)	15 (6.6)	.01	55 (18.2)	27 (9.0)	8 (7.1)	.0005	37 (10.6)	13 (4.5)	14 (7.2)	.02
ARDS	2 (1.8)	20 (2.2)	7 (3.1)	.68	26 (8.5)	11 (3.6)	2 (1.8)	.005	14 (4.3)	5 (1.7)	2 (1.0)	.04
Death	3 (2.8)	46 (5.0)	13 (5.7)	.50	16 (5.3)	8 (2.7)	5 (4.5)	.26	9 (2.6)	5 (1.7)	4 (2.0)	.76
Underlying medical conditions	100 (92.6)	842 (92.5)	200 (88.5)	.13	269 (88.2)	261 (86.7)	97 (86.6)	.83	260 (74.7)	218 (76.2)	139 (71.3)	.50

Bold numbers indicate statistically significance using χ² test (or Fisher exact when appropriate). Abbreviations: ARDS, Acute Respiratory Distress Syndrome; ICU, intensive care unit.

Factors Associated With Severity

Overall, there were 141 (17%) children with severe influenza. Of those, 38 (26.9%) were associated with pH1N1 virus infection, compared with 34 (24.1%) and 69 (48.9%) children with H3N2 and influenza B, respectively (P = .03). Other characteristics associated with severe influenza in children included presence of underlying medical conditions; 51.1% of those with ≥1 underlying medical condition had severe disease compared with 40.5% of those without any (P = .02). Among selected underlying medical conditions, severity was more common in those with diagnosis of chronic lung disease (13.5% vs 4.5%; P < .0001), chronic metabolic disease (7.8% vs 2.5%; P = .001), or neuromuscular disorder (11.3% vs 3.2%; P < .0001) than those without these diagnoses (Table 3).

Table 3.

Characteristics Associated With Influenza Disease Severity for Children and Adults, United States, 2010–2011 Influenza Season

Characteristics	Children			Adults
	(n = 689)	(n = 141)	P Value	(n = 2255)	(n = 536)	P Value
	Nonsevere	Severe		Nonsevere	Severe
	No. (%)	No. (%)		No. (%)	No. (%)
Virus type/subtype			.03			<.0001
pH1N1	125 (18.1)	38 (26.9)		556 (24.7)	205 (38.2)
H3N2	218 (31.6)	34 (24.1)		1255 (55.6)	242 (45.1)
B	346 (50.2)	69 (48.9)		444 (19.7)	89 (16.7)
Age, y			.75			.005
<2	327 (47.5)	62 (43.9)		…	…
2–4	120 (17.4)	26 (18.4)		…	…
5–17	242 (35.1)	53 (37.6)		…	…
18–49	…	…		684 (30.4)	145 (27.0)
50–64	…	…		551 (24.4)	167 (31.2)
≥65	…	…		1020 (45.2)	224 (41.8)
Sex			.10			.75
Male	363 (52.7)	85 (60.3)		1039 (46.1)	251 (46.8)
Female	326 (47.3)	56 (39.7)		1216 (53.9)	285 (53.2)
Underlying medical conditionsa			.02			.29
Yes	279 (40.5)	72 (51.1)		1920 (85.1)	466 (86.9)
No	410 (59.5)	69 (48.9)		335 (14.9)	70 (13.1)
Selected conditions
Asthma
Yes	127 (18.4)	36 (25.5)	.053	463 (20.5)	105 (19.6)	.63
No	562 (81.6)	105 (74.5)		1792 (79.5)	431 (80.4)
Chronic lung disease
Yes	31 (4.5)	19 (13.5)	<.0001	508 (22.5)	165 (30.9)	<.0001
No	658 (95.5)	122 (86.5)		1747 (77.5)	371 (69.2)
Chronic vascular diseaseb
Yes	29 (4.2)	10 (7.1)	.14	835 (37.0)	216 (40.3)	.16
No	660 (95.8)	131 (92.9)		1420 (63.0)	312 (59.7)
Chronic metabolic disease
Yes	17 (2.5)	11 (7.8)	.001	798 (35.4)	207 (38.6)	.16
No	672 (97.5)	130 (92.2)		1457 (64.6)	329 (38.6)
Neuromuscular disorders
Yes	22 (3.2)	16 (11.3)	<.0001	158 (7.0)	53 (9.9)	.02
No	667 (96.8)	125 (88.6)		2097 (93.0)	483 (90.1)
Immunosuppressive status
Yes	31 (4.5)	4 (2.8)	.37	298 (13.2)	87 (16.2)	.07
No	658 (95.5)	137 (97.2)		1957 (86.8)	449 (83.8)
Developmental delayc
Yes	39 (5.8)	10 (8.0)	.36	…	…
No	628 (94.2)	115 (92.0)		…	…
Renal disease
Yes	12 (1.7)	1 (0.7)	.71	344 (15.2)	91 (16.9)	.32
No	677 (98.3)	140 (99.3)		1911 (84.8)	445 (83.1)
Morbid obese
Yes	…	…		219 (9.7)	68 (12.7))	.04
No	…	…		2036 (90.3)	468 (87.3)
Antiviral started <3 d from admissiond			.14			<.0001
Yes	305 (95.0)	92 (91.1)		1452 (93.6)	380 (87.7)
No	16 (5.0)	9 (8.9)		99 (6.4)	53 (12.3)
Vaccination ≥14 d prior to hospitalizatione			.35			.80
Yes	209 (33.1)	47 (37.3)		973 (47.6)	213 (46.9)
No	424 (66.9)	79 (62.7)		1072 (52.4)	241 (53.1)

Bold numbers indicate statistically significance using χ² test (or Fisher exact when appropriate).

^a At least 1 chronic medical condition stated in medical chart, including asthma, chronic lung disease, cardiovascular disease, chronic metabolic disease, renal disease, hemoglobinopathy, neuromuscular disorder, immunosuppressive condition, developmental delay, recent history of cancer, and cognitive dysfunction.

^b Excludes hypertension alone.

^c Excludes children with developmental delay who also had neuromuscular disorder.

^d Among those treated with influenza antiviral medications.

^e There were 17 (3.8%) and 807 (29%) missing dates of antiviral treatment for children and adults respectively.

In adults, there were 536 severe influenza cases; of those, 205 (38.2%) were associated with pH1N1, compared with 242 (45.1%) with H3N2 and 89 (16.7%) with influenza B (P < .0001). Disease severity was also associated with age; there was a higher frequency of adults aged 50–64 years among severe cases than among nonsevere cases (31.2% vs 24.4%; P = .005). There was no statistically significant association between having ≥1 underlying chronic medical condition and severe influenza in adults (P = .29). When evaluating the contribution of individually selected underlying medical conditions, influenza severity was significantly higher among those with chronic lung disease (30.9% vs 22.5%; P < .0001), neuromuscular disorder (9.9% vs 7.0%; P = .02), and morbid obesity (12.7% vs 9.7%; P = .04)(Table 3).

Independent Predictor of Disease Severity

Among children treated with antiviral agents, we fitted a model comparing severity by virus type/subtype while controlling for age group, presence of selected underlying medical conditions, and prompt treatment with antivirals. The final model included 402 children, of which 95 (23.6%) had severe cases. Severity was independently associated with pH1N1 virus infection (adjusted odds ratio (aOR), 2.19; 95% CI, 1.11–4.33), presence of chronic metabolic disease (aOR, 5.23; 95% CI, 1.74–15.69; P = .003) and neuromuscular disorder (aOR, 4.84; 95% CI, 2.02–11.58; P = .0004) (Table 4).

Table 4.

Unconditional Logistic Regression Analysis of Factors Associated With Severe Influenza Among Children and Adults, United States, 2010–2011 Influenza Season

	Adjusted Odds Ratioa	95% CI	P Value
Among Children
Virus type/subtype
pH1N1	2.19	1.11–4.33	.02
B	0.99	.55–1.77	.99
H3N2	Reference
Age, y
<2	1.17	.61–2.22	.62
2–4	1.25	.64–2.44	.50
5–17	Reference
Selected underlying medical conditions
Asthma	1.26	.66–2.39	.48
Chronic lung disease	2.11	.97–4.57	.06
Chronic vascular  disease	1.15	.35–3.74	.82
Chronic metabolic  disease	5.23	1.74–15.69	.003
Neuromuscular  disorder	4.84	2.02–11.56	.0004
Antiviral started <3 d from admission
Yes	0.50	.20–1.25	.14
Among Adults
Virus type/subtype
pH1N1	2.21	1.66–2.94	<.0001
H3N2	1.10	.79–1.53	.57
B	Reference
Age, y
18–49	Reference
50–64	1.20	.88–1.64	.24
≥65	0.94	.67–1.32	.723
Underlying medical conditions
Chronic lung disease	1.46	1.12–1.89	.005
Chronic vascular  disease	1.19	.92–1.54	.19
Chronic metabolic  disease	1.11	.87–1.42	.38
Neuromuscular  disorder	1.68	1.11–2.52	.01
Morbid obesity	1.17	.82–1.68	.38
Antiviral started <3 d from admission
Yes	0.47	.33–0.68	<.0001

Vaccination status was included in the model but did not affect odds ratio, therefore it was excluded to improve final model fit. Similarly, immunosuppressive condition was not kept in the final model for adults. Abbreviation: CI, confidence interval.

^a Adjusted for all the variables listed in the table.

There were 1826 adults treated with antiviral drugs, of which 399 (22%) were classified as having severe cases. Vaccination status and presence of immunosuppressive conditions were included in the initial model and later removed because they were not found to be confounders and made the model fit poorly. Adults hospitalized with pH1N1 were twice as likely to have severe influenza as those with H3N2 (aOR, 2.21; 95% CI, 1.66–2.94; P < .0001). After controlling for influenza virus type and subtype, age group, selected medical conditions, and antivirals, independent predictors of severity in adults included chronic lung disease (aOR, 1.46; 95% CI, 1.12–1.89) and neuromuscular disorder (aOR, 1.68; 95% CI, 1.11–2.52). The aOR of severe disease among patients treated with antivirals within <3 days from admission was 0.47 (95% CI, .33–.68; P < .0001), indicating that prompt treatment with antivirals independently reduced the odds of severity. (Table 4).

DISCUSSION

Our data suggest that during the 2010–2011 season, the first US season after the 2009 pandemic, persons of all ages hospitalized with pH1N1 were twice as likely to have severe influenza as patients hospitalized with H3N2 or influenza B virus infections; this was true even after controlling for age, presence of certain comorbidities, and prompt treatment with antiviral drugs. Certain underlying chronic medical conditions were also independently associated with increased severity among children and adults. Even among those treated with antiviral drugs, the odds of severe disease in children with chronic metabolic disease or neuromuscular disorder were about 5 times those without. We also found that adults with chronic lung disease or neuromuscular disorder were at increased risk of disease severity, despite virus strain and appropriate antiviral treatment.

Certain underlying medical conditions were also independently associated with disease severity. Similar to our study, Kawai et al analyzed data from the 2010–2011 influenza season and reported that cases associated with pH1N1 were slightly more severe than cases ofH3N2 and influenza B. However, this study looked only at outpatients and used duration of fever as their main outcome of interest [14]. To our knowledge, our study is the first to assess the independent effect of pH1N1 on disease severity among influenza-associated hospitalizations accounting for the contribution of influenza H3N2 and B viruses circulating within the same season. As such, we were able to avoid potential biases related to varying medical care–seeking behavior, host immune adaptation, degree of matching between circulating and vaccine-selected viruses and likelihood of being prescribed antiviral drugs, which might change from season to season.

Increased disease severity associated with pH1N1 strains was also described in ferrets when clinical characteristics associated with pH1N1 were compared with seasonal H1N1, H3N2, and influenza B virus infections [15]. The study reported that ferrets infected with pH1N1 virus displayed more severe outcomes and required a longer recovery period than those infected with other virus strains. Although all influenza viruses infect the respiratory epithelium from the nasal passages to bronchioles, pH1N1 tends to infect pneumocytes and intraalveolar macrophages, causing extensive areas of inflammation in the alveoli, which could partially explain the increased severity [15–18].

During the 2010–2011 season, the percentage of patients hospitalized with pH1N1 who were admitted to the ICU or needed mechanical ventilation was similar to that reported during the 2009 pandemic in the United States [2, 3], which suggested that the clinical severity described among pH1N1 hospitalizations during the 2010–2011 season was not dissimilar to that seen in the pandemic. In fact, virologic surveillance detected little, if any, antigenic changes in the pH1N1 viruses circulating during the 2010–2011 influenza season compared with strains circulating during the 2009 pandemic [10]. However, several studies done during and after the 2009 pandemic have not found differences in clinical severity attributable to pH1N1 when comparing clinical outcomes with that of seasonal influenza A (H3N2 and H1N1) and influenza B [7–10]. Those studies used data from pre-2009 pandemic influenza seasons as comparison, when antiviral treatment was not used as widely. Many of these studies also lacked large enough samples of hospitalized patients for evaluating ICU admissions as the analysis' endpoint or for controlling for other factors that might impact influenza disease severity, such as age and presence of underlying medical conditions [7–9]. Furthermore, valid comparisons across seasons need to take into account season-to-season differences in strain virulence, population immunity, and healthcare-seeking behavior.

Prepandemic studies comparing disease severity in children by viral type/subtype have often suggested that influenza B viruses cause milder influenza illness than influenza A viruses [19–23]. However, in our data, influenza B was associated with approximately 15% of ICU admissions and approximately 7% of mechanical ventilation among both children and adults, which was similar to that described for patients hospitalized with H3N2, although less than that associated with pH1N1 (>20% and >10%, respectively). Esposito et al described the clinical impact of the 2007–2008 and 2008–2009 influenza seasons in children; after controlling for age and sex, children with influenza B had significantly higher hospitalization rates than those with seasonal H1N1 [20]. A recent paper describing the 2010–2011 season in Spain, reported that influenza B virus infections caused as much severe disease among those hospitalized as that caused by pH1N1, but they reported on a small sample (total of 130 patients aged ≥14 years) and were not able to proceed with multivariable analysis controlling for age, comorbidities, and other potential confounders [24].

Influenza can be severe in both children and adults [25, 26], particularly in persons with underlying medical conditions. Regardless of virus type and subtype, we observed an independent contribution of selected underlying medical conditions to influenza severity, even when prompt antiviral therapy was initiated after admission. This was true for both children and adults and underscores the importance of annual influenza vaccination, especially for those with chronic medical conditions at increased risk of influenza-related complications and death [27]. Annual influenza vaccination has been recommended in the United States for all persons aged ≥6 months since 2010 and is the most effective way to prevent influenza disease morbidity and mortality. Nonetheless, in the Unietd States, the national average of influenza vaccination coverage for those aged ≥6 months was 43% during the 2010–2011 season [28].

Prompt empiric antiviral treatment should be initiated when influenza is suspected, even before laboratory results are known [29]. Our data showed that adults treated with antiviral drugs within <3 days of admission were half as likely to be severe compared with those for whom treatment was initiated later. We did not find a similar protective effect of prompt antiviral use among children, but this could be because of our small sample size, with fewer ICU/deaths in this age group compared with adults. Alternatively, our cutoff point of <3 days of admission indicating prompt antiviral treatment may not be adequate for children. The only study that demonstrated a benefit of oseltamivir in severely ill children hospitalized with influenza compared length of hospital stay between those treated within <24 hours of admission with those treated later (>24 hours) [30]. However, improved clinical outcomes have been described among hospitalized adults treated with antiviral drugs promptly after hospitalization or up to 5 days from illness onset [31–33]. Despite documented benefits, the use of influenza antiviral medications has decreased significantly among hospitalized patients since the 2009 pandemic [34]; this is worrisome considering that the use of antiviral agents could reduce the duration of illness and prevent complications.

Our study has a number of limitations. First, influenza testing was ordered at physicians' discretion, and therefore not all persons with influenza were tested or testing could be biased toward more severe cases. However, this bias would not affect our main findings because influenza testing was ordered independently of the knowledge of virus type or subtype. Cases missed because of no testing or less-sensitive influenza diagnostic tools likely lead to underestimation of our hospitalization rates. Second, there may have been underreporting of underlying diseases in the patient's medical chart, and therefore these were not accounted for in our analyses. As stated above, this bias would not affect the association between severity and influenza strain because underlying disease data would have been collected independent of the knowledge of virus type/subtype. Finally, we had a large number of influenza A–associated hospitalizations that did not contribute data to our predictors of severity analysis because of missing subtype information. However, demographic characteristics and clinical outcomes did not differ between cases excluded and those included in our analysis (data not shown).

In summary, in the 2010–2011 season, pH1N1 caused more severe disease than H3N2 or influenza B in hospitalized patients. Insight into strain-specific pathogenicity can be useful, and annual assessment of associated severity might be important to guide messaging about influenza vaccination and treatment recommendations. All persons aged ≥6 months should receive annual influenza vaccination. Persons with comorbidities should be specifically targeted with vaccine and early antiviral treatment to reduce the risk of influenza-related complications.

Notes

Financial support. The Influenza Hospitalizations Surveillance Network (FluSurv-NET) is a collaboration of state health departments, academic institutions, and local partners and is funded by the Centers for Disease Control and Prevention (CDC). This publication was supported in part by Cooperative Agreement number CDC-RFA-CK12–1202 and 5U38HM000414 from the CDC. Its contents are solely the responsibility of the authors and do not necessarily represent the official views of CDC.

Potential conflicts of interest. W. S. is a member of data safety monitoring boards for Merck and Sanofi-Pasteur and occasionally consults for Pfizer, GlasoSmithKline, and Dynavax.All other authors report no potential conflicts.

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.

Author notes

a

Present affiliation: University of Oklahoma Health Sciences Center, Oklahoma City.