Background and Epidemiology
Among humans, annual epidemic influenza illness is caused by two types of influenza viruses, influenza A and influenza B (1). Influenza virus types A and B are further subclassified through serologic and genetic testing. For influenza A viruses, differences in two viral surfaces glycoproteins, hemagglutinin (HA) and neuraminidase (NA), permit categorization into different subtypes. In the nomenclature of influenza A viruses, these different HA and NA proteins are represented by the letters “H” and “N”, respectively. Since the late 1970s, influenza A(H1N1) and influenza A(H3N2) have been the common influenza A viral subtypes circulating among humans (2). Influenza B viruses are separated into two distinct genetic lineages (Yamagata and Victoria) on the basis of differences in the HA glycoprotein. Influenza B viruses from both lineages have co-circulated during most influenza seasons since the 1980s (3, 4).
Influenza viruses undergo constant genetic change, which has substantial impact on induced immunity and considerations for vaccine composition. Two main types of changes are recognized. Point mutations and recombination events occur in the viral genome, resulting in constant emergence of new virus variants. This phenomenon is termed “antigenic drift” (1, 2). While it occurs among both influenza A and B viruses, influenza A viruses undergo antigenic drift more rapidly than influenza B viruses (5). Frequent emergence of antigenic variants through antigenic drift is the virologic basis for seasonal influenza epidemics, and necessitates consideration of adjustment of vaccine viruses each season.
In addition to antigenic drift, larger genetic change events, termed “antigenic shift”, can occur among influenza A viruses. Antigenic shift occurs less frequently than antigenic drift, and generally arises though genetic reassortment among different viruses. These events can lead to new or substantially different influenza A viruses, for which there is little pre-existing immunity in the population. Such viruses can be associated with widespread pandemic influenza illness, if they exhibit efficient and sustained transmission among humans (1). In April 2009, a novel influenza A(H1N1) virus caused the most recent worldwide pandemic. This virus was antigenically distinct from human influenza A(H1N1) viruses in circulation from 1977 through spring 2009 (6, 7).
Each season, viruses belonging to influenza A subtypes A(H1N1) and A(H3N2) and to both B lineages co-circulate (1, 2). Natural infection and vaccination induce the production of antibodies to influenza HA and NA surface glycoproteins, which reduce likelihood of infection (8). However, antibodies produced against one influenza virus type or subtype confer limited or no protection against another type or subtype (9).
Currently available seasonal influenza vaccines are either trivalent or quadrivalent in composition. Trivalent vaccines contain HA derived from one A(H1N1) virus, one A(H3N2) virus, and one B virus (representing one influenza B lineage). Quadrivalent vaccines have the same HA composition as trivalent vaccines, with the addition of HA from a second influenza B virus (such that both influenza B lineages are represented).
The exact timing of the onset, peak, and end of influenza activity vary, and cannot be predicted precisely from one season to the next. In general, however, annual epidemics of influenza typically occur in the United States during the fall and winter. Influenza activity often begins to increase in the U.S. during October, and can extend as late as May. Peak activity most commonly occurs during the winter. During the 36 seasons from 1982-83 through 2017-18, peak activity occurred most commonly during February (in 15 [42%] seasons); however peak activity was observed during December (7 [19%]), January (6 [17%]), and March (6 [17%]) during other seasons (10).
Surveillance systems and research studies use different case definitions to characterize influenza activity and illness. Some outcomes (e.g., influenza illness confirmed by viral culture or polymerase chain reaction [PCR]) are more specific than others (e.g., influenza-like illness defined by a clinical case definition, without confirmatory diagnostic testing). Studies that report rates of clinically-defined outcomes without laboratory confirmation of influenza (e.g., respiratory illness requiring hospitalization during influenza season) can be difficult to interpret because of coincident circulation of other respiratory pathogens (e.g., respiratory syncytial virus) (11). More specific burden estimates are provided by surveillance studies based on laboratory-confirmed influenza (LCI). However, less specific outcomes are useful in national surveillance of influenza activity, and are used in some components of routine U.S. influenza surveillance. Increases in health care provider visits for acute febrile respiratory illness occur annually, coinciding with periods of increased influenza activity, making influenza-like illness (ILI) surveillance systems valuable in understanding and describing the seasonal and geographic occurrence of influenza each year (12).
Persons of all ages are susceptible to influenza. Influenza incidence is difficult to quantify precisely, as many or most of those infected may not seek medical attention and are therefore not diagnosed. An estimated incidence of approximately 8% (varying from 3% to 11%) was derived through statistical extrapolation of U.S. hospitalization data and a meta-analysis of published literature (13). In a systematic review of randomized controlled trials which examined LCI events in the control (unvaccinated or placebo) arms of the included studies, an estimated 1 in 5 unvaccinated children and 1 in 10 unvaccinated adults were infected, with approximately half of these cases being symptomatic (14).
Burden of severe outcomes associated with influenza illness, such as hospitalization and death, may be estimated in several ways, such as through assessing rates of these events during influenza seasons, though mathematical modelling methods, and through studies that examine LCI. In typical influenza seasons, increases in deaths and hospitalizations are observed during periods when influenza viruses are circulating. Although not all excess events occurring during these periods are attributable to influenza, surveillance of these events is useful for monitoring season-to-season trends in influenza-associated outcomes. Estimates based on hospitalizations or deaths associated with pneumonia and influenza (P&I) diagnoses likely underestimate the burden of severe illnesses that are at least partly attributable to influenza, because this category excludes deaths caused by exacerbations of underlying cardiac and pulmonary conditions that are associated with influenza infection. Thus, some authors have used the broader category of respiratory and circulatory excess events for influenza burden estimates. A modeling analysis of population-based surveillance data for seasons following the 2009 pandemic (2010–11 through 2012–13) estimated that influenza was associated with 114,018—633,001 hospitalizations, 18,476–96,667 intensive care unit (ICU) admissions, and 4,866–27,810 deaths per year (15). An estimated 54%–70% of hospitalizations and 71%–85% of deaths occurred among adults aged ≥65 years. This model used a multiplier method to correct for under-detection in hospitalizations attributable to cases for which influenza testing was not performed and for suboptimal test sensitivity. Using similar methodology, for seven seasons from 2010-11 through 2015-16 it has been estimated that between 9.2—35.6 million illnesses, 4.3—16.7 million outpatient medical visits, 139,000—708,000 hospitalizations were attributable to influenza each season (16) Excess deaths were estimated to be 4,000-20,000 per season for P&I deaths and 12,000-56,000 for respiratory and circulatory deaths.
Influenza is an important cause of outpatient medical visits and hospitalizations among young children. In a population-based retrospective cohort study, hospitalization rates for children aged <5 years with acute respiratory illness (ARI) or LCI-associated fever averaged 0.9 per 1000 (range 0.4-1.5) for seasons 2000-01 through 2003-04 and 0.58 per 1000 (range 0.36 to 0.97) for the seasons 2004-05 through 2008-09 (17, 18). In a retrospective cohort study of children aged <15 years over 19 seasons (1974–75 through 1992–93), an estimated average of 6–15 additional outpatient visits and 3–9 additional antibiotic courses per 100 children per season were attributed to influenza (19). In a study conducted in a single county in Tennessee during the 2000–01 through 2010–11 seasons, estimated rates of influenza-related hospitalizations among children aged 6 through 59 months varied from 1.9 to 16 per 10,000 children per year; estimated rates of influenza-related emergency department visits ranged from 89 to 620 per 10,000 children per year (20).
Estimated rates of influenza-associated hospitalization generally are substantially higher among infants and children <5 years than among older children (17, 18, 21-26). During 1993–2008, estimated annual rates of influenza-associated hospitalizations were 151.0 per 100,000 among children aged <1 years and 38.8 per 100,000 among children aged 1–4 years, compared with 16.8 per 100,000 among persons aged 5 through 49 years (23). Estimated hospitalization rates for children with high-risk medical conditions are generally higher than for those without (26-28). In a study of children hospitalized with confirmed influenza infection, length of stay was longer for those with high-risk conditions than for healthy children of the same age (4.7 vs. 3.0 days for those aged 6 through 23 months; 5.8 vs. 3.6 days for those aged 2 through 17 years) (21). Thirty-seven percent had an ACIP-defined high-risk condition; the most common high-risk conditions were asthma (45%), followed by neurological (23%), cardiovascular (21%), metabolic and immunosuppressive disorders (7% each). In another study, asthma was associated with 23% of the influenza hospitalizations and 15% of the outpatient visits (27).
Estimates of influenza mortality rates for children based on pneumonia and influenza diagnoses, respiratory and circulatory diagnoses, or laboratory confirmed influenza have generally been low, <1 per 100,000 person-years (29, 30). However, the absolute number of pediatric deaths varies from season to season (31). It is important to note that these deaths often occur in children with no other risk factors for severe influenza illness. In one study of the 2003-04 season, nearly half occurred in previously healthy children (30). In the United States, death associated with LCI among children aged <18 years has been a nationally reportable condition since October 2004 (32). Since reporting began, the annual number of reported influenza-associated pediatric deaths during regular influenza seasons has ranged from 37 deaths in the 2011-12 season to a high of 187 in 2017-18 (31). A larger number of deaths were reported during the 2009 influenza A(H1N1)pdm09 pandemic, for which 358 pediatric deaths were reported to CDC from April 15, 2009 through October 2, 2010 (33). In a review of pediatric mortality surveillance data covering the 2010-11 through 2015-16 seasons, death rates were inversely associated with age, with the highest rates occurring among those aged < 6 months. Among children aged ≥6 months, only 31% had received at least one dose of influenza vaccine in the season that death occurred (34).
In typical seasons, influenza-associated hospitalization occurs less frequently among younger adults as compared with children aged <5 years and adults aged ≥65 years. However, some authors have reported that influenza is an important cause of outpatient medical visits and worker absenteeism among healthy adults. In one economic modeling analysis, the average annual burden of seasonal influenza among adults aged 18–49 years without medical conditions that confer a higher risk for influenza complications was estimated to include approximately 5.2 million illnesses, 2.4 million outpatient visits, 31,800 hospitalizations, and 684 deaths (35). Studies of worker vaccination have reported lower rates of ILI (36, 37), lost work time (36-39), and health care visits (37, 39) in association with vaccination as compared with no vaccine or placebo. Influenza may be associated with greater workplace productivity losses among working adults than acute respiratory infections caused by other pathogens (40).
During the 2009 influenza A(H1N1)pdm09 pandemic adults aged <65 years appeared to be at higher risk for influenza-related hospitalizations and deaths (41) as compared with typical influenza seasons. During the 2009 influenza A (H1N1) pandemic period (April 2009 through May 1, 2010), the cumulative crude rates of LCI-related hospitalization for the Emerging Infections Program sites were 3.0 per 10,000 persons aged 18–49 years, 3.8 per 10,000 persons aged 50–64 years, and 3.2 per 10,000 persons aged ≥65 years. During the previous three seasons, rates had ranged from 0.3–0.7 per 10,000 persons aged 18–49 years to 0.4–1.5 per 10.000 persons aged 50–64 years and 1.4–7.5 per 10,000 persons aged ≥65 years (42). Adults aged 50–64 years had the highest mortality rate during the 2009 pandemic. This group was again severely affected during the 2013–14 season when H1N1pdm09 was the predominant virus, sustaining higher hospitalization rates than in previous seasons since the pandemic (43).
Hospitalization rates during typical influenza seasons are generally highest for adults aged ≥65 years (12). Risk of hospitalization may be greater among older adults with high-risk underlying medical conditions than for those without such conditions. One retrospective analysis of data from three managed-care organizations collected during 1996–97 through 1999–2000 estimated that the risk during influenza season among persons aged ≥65 years with high-risk underlying medical conditions was 55.6 for pneumonia and influenza-associated hospitalizations per 10,000 persons, compared with 18.7 per 10,000 among lower risk persons in this age group. Persons aged 50–64 years who had underlying medical conditions also were at substantially increased risk for hospitalization during influenza season compared with healthy adults aged 50–64 years (12.3 versus 1.8 per 10,000 person-periods) (44).
Deaths associated with influenza are most frequent among older adults. From the 2010-11 through 2015-16 seasons, an estimated 9,000-43,000 influenza-related deaths occurred among adults aged ≥65 years, corresponding to >75% of estimated annual average deaths across all age groups (16). In comparison, the average annual mortality was estimated to be 200-1,300 deaths among persons aged <18 years and 2,700-11,000 deaths among persons aged 18–64 years.
Some studies have examined potential associations between acute respiratory illnesses and acute vascular events such as myocardial infarction and ischemic stroke (45-48). A self-controlled case series analysis of 364 hospitalizations for myocardial infarction found an increased risk for myocardial infarction within 7 days of laboratory detection of influenza (incidence ratio [IRR]=6.05, 95%CI 3.86—9.50 for all influenza; IRR=10.11, 95%CI 4.37—23.38 for influenza B, and IRR=5.17, 95%CI 3.02—8.84 for influenza A) (45). A self-controlled case series analysis of the United Kingdom Myocardial Ischaemia National Audit Project and the General Practice Research Database found that the risk of acute myocardial infarction was significantly higher 1-3 days after the onset of an acute respiratory infection (IRR=4.19, 95%CI 3.18—5.53). This effect was greatest among those aged ≥80 years (46). In an analysis of hospitalization data, admissions for myocardial infarction and stroke among persons aged ≥75 years were correlated with circulation of influenza (47). A retrospective analysis of medical record data reported that persons hospitalized for an ischemic stroke had increased odds of having had a recent previous hospitalization for ILI. In this study, the association between ILI and stroke was greatest for younger adults (those under 43 years of age), and decreased with increasing age (though it remained statistically significant) (48).
Pregnant women are vulnerable to severe symptoms and illness attributable to influenza. Physiologic changes associated with pregnancy, such as altered cardiopulmonary mechanics and changes in cell-mediated immunity, might contribute to enhanced susceptibility (49). A retrospective cohort study of pregnant women conducted in Nova Scotia during 1990–2002 compared medical record data for 134,188 pregnant women to data from the same women during the year before pregnancy. The rate ratio for hospital admissions was significantly increased during all trimesters of pregnancy, and increased with successive stages (for the third trimester, relative risk [RR] 7.9 [95%CI 5.0–12.5] among women with comorbidities and 5.1 [95%CI 3.6–7.3] among those without comorbidities) (50). A cohort study conducted in New Zealand found increased risk of LCI-associated hospitalizations among pregnant women compared with women who were not pregnant. Risk was increased for all trimesters and increased with successive stages of pregnancy (RR=2.5, 95%CI 1.2—5.4 for first trimester; RR=3.9, 95%CI 2.4—6.3 for second trimester; and RR=4.8, 95%CI 3.0—7.7 for third trimester) (51). A systematic review and meta-analysis of observational studies concluded that influenza infection during pregnancy was associated with an increased risk for hospitalization relative to infection in non-pregnant individuals (OR=2.44, 95%CI 1.22—4.87), but not for death (52).
Increased severity of influenza among pregnant women was reported during the pandemics of 1918–19, 1957–58, and 2009–10 (53-58). During the 2009(H1N1) pandemic, severe infections among postpartum (delivered within previous 2 weeks) women also were observed (54, 57, 59). In a case series conducted during the 2009(H1N1) pandemic, 56 (20%) deaths were reported among 280 pregnant women admitted to intensive care units. Among U.S. deaths due to pandemic influenza reported to CDC, five percent of all US deaths from pandemic influenza involved pregnant women, even though they represented <1% of the population (60, 61). Among the deaths, 36 (64%) occurred in the third trimester. Pregnant women who were treated with neuraminidase inhibitor antivirals >4 days after symptom onset were more likely to be admitted to an intensive care unit (57% versus 9%; relative risk [RR]=6.0, 95%CI 3.5–10.6) than those treated within 2 days after symptom onset (61).
Some studies of pregnancy outcomes have suggested increased risk for pregnancy complications attributable to maternal influenza illness; others have not. A review of data from the National Inpatient Sample (a publicly available hospital discharge database; www.hcupexternal icon–us.ahrq.gov/nisoverview.jspexternal icon)external icon covering the 1998–99 through the 2001–02 seasons and including over 6.2 million hospitalizations of pregnant women, reported increased risk for fetal distress, preterm labor, and cesarean delivery among those women with respiratory illness during influenza seasons, compared with women without respiratory illness (62). A study of 117,347 pregnancies in Norway during the 2009–10 pandemic noted an increased risk for fetal death among pregnant women with a clinical diagnosis of influenza (adjusted hazard ratio [aHR]=1.91; 95%CI 1.07–3.41) (63). A cohort study conducted among 221 hospitals in the United Kingdom observed an increased risk for perinatal death, stillbirth, and preterm birth among women admitted with confirmed 2009(H1N1) infection (64). In a retrospective cohort study of 86,779 pregnancies in which 192 cases of LCI were identified during the 2012-14 and 2013-14 seasons, women infected during the first trimester had a significantly lower mean length of gestation than uninfected women (38 vs. 39 weeks). The infants of those infected with influenza B had a 4% lower mean percent of optimal weight (65). However, other studies of infants born to women with LCI during pregnancy have not shown higher rates of prematurity, preterm labor, low birth weight, or lower Apgar scores compared with infants born to uninfected women (66-68).
Influenza symptoms often include fever, which during pregnancy might be associated with neural tube defects and other adverse outcomes (69). A meta-analysis of 22 observational studies of congenital anomalies following influenza exposure during the first trimester of pregnancy noted associations with several types of congenital anomalies, including neural tube defects, hydrocephaly, heart and aortic valve defects, digestive system defects, cleft lip, and limb reduction defects (70). However, many of the included studies were conducted during the 1950s through 1970s, and a nonspecific definition of influenza was used (any reported influenza, ILI, or fever with influenza, with or without serological or clinical confirmation). A 2005 meta-analysis of fifteen observational studies noted an association between maternal fever and neural tube defects (71). Associations between maternal fever and congenital heart defects (72) and orofacial cleft (73) have been reported in some studies; in one study of congenital anomalies such as orofacial clefts, congenital heart defects, and omphalocele, the association with maternal fever was ameliorated among those mothers who had taken multivitamins (74).
Persons with “chronic debilitating diseases” (specifically those with cardiovascular disease, pulmonary disease, diabetes and Addison’s disease), were included (along with persons aged ≥65 years and pregnant women) among those recommended to receive vaccine as early as 1960. These groups had been noted to have contributed most to the excess deaths observed during the 1957 influenza pandemic (75). Since that time, the list of conditions potentially conferring increased risk of severe illness attributable to influenza has expanded to include persons with chronic medical conditions of most body systems, immunocompromised persons, those who are extremely obese, and American Indian/Alaska Native populations (76). In general, risk groups have been added as a result of increased observed risk of hospitalization or other severe outcomes during some influenza seasons. For any given group, the degree of increased risk may not be the same in all seasons. However, the potential for increased risk of severe illness cannot be predicted in any given season. Moreover, most risk categories comprise a heterogeneous group of conditions, the severity of which varies among different individuals, as well as with the presence or absence of medical treatment for the underlying condition. It is possible that these factors may alter illness risk for any given individual.
In a study of 4,756 adults hospitalized with influenza from October 2005 through April 2008, characteristics significantly associated with pneumonia (a potential severe complication of influenza) included underlying chronic lung disease and immunosuppression (77). Among patients with pneumonia, patients with a poor outcome (defined as ICU admission, need for mechanical ventilation, or death) were more likely to be affected by chronic lung disease, cardiovascular disease, renal disease, or immunosuppression. In a case control study of children 6 through 59 months of age during the 2005-06 through 2008-09 seasons, hospitalization for influenza was significantly associated with pulmonary, hematologic/oncologic, and neurologic conditions (78). In a systematic review of studies of risk factors for influenza-related complications among children, hospitalization was most strongly associated with neurologic disorders, prematurity, sickle-cell disease, immunosuppression, and diabetes (79).
Some early studies suggested increased severity of influenza among HIV-infected persons (80, 81). Studies conducted since the widespread availability of effective HIV antiretroviral therapy indicate that there may be no increased risk of severe disease among persons for whom HIV is well-controlled (82-87). However, influenza may be associated with secondary bacterial pneumonia, particularly Streptococcus pneumoniae, Haemophilus influenzae, and Staphylococcus aureus, which may be more severe among HIV-infected persons (88).
Prior to the 2009 pandemic, obesity had not been recognized as a risk factor for severe influenza illness. However, several studies during the 2009 pandemic noted a high prevalence of obesity among persons with severe illness attributable to A(H1N1)pdm09 (89-91). In a case-cohort study, among persons aged ≥20 years, hospitalization with illness attributable to laboratory-confirmed influenza A(H1N1)pdm09 was associated with extreme obesity (body mass index [BMI] ≥40) even in the absence of other risk factors for severe illness (odds ratio [OR]=4.7; 95%CI 1.3–17.2) (92). Death was associated with both obesity, defined as BMI ≥30 (OR=3.1; 95%CI 1.5–6.6) and extreme obesity (OR=7.6; 95%CI 2.1–27.9). A Canadian cohort study covering 12 seasons (1996–97 through 2007–08) found that persons with a BMI of 30.0–34.9 and those with a BMI ≥35 were more likely than normal-weight persons to have a respiratory hospitalization during influenza seasons (OR=1.45; 95%CI 1.03–2.05 for BMI 30–34.9 and OR=2.12; 95%CI 1.45– 3.10 for BMI ≥35) (93). A retrospective cohort study of Australian national health insurance data between 2006 and 2015 found that compared to adults with a healthy BMI, those with a BMI of 30 to <40 had a higher risk of influenza-associated hospital admission (aHR=1.57, 95%CI 1.22–2.01); those with a BMI of ≥40 had an even higher risk (aHR=4.81, 95%CI 3.23—7.17) (94). Conversely, a two-season prospective cohort study (2007–09) in the United States found no association between obesity and medically attended LCI, including both seasonal and pandemic virus circulation (95). In a four-season (2010-11 through 2013-14) cohort study conducted in Mexico of children and adults with viral ILI, influenza infection was significantly associated with increased risk of hospitalization among adults who were either underweight or morbidly obese. Among children in this study, influenza was not analyzed separately; obesity was associated with increased risk for hospitalization for ILI due to any viral pathogen (96).
Also during the 2009 pandemic, racial and ethnic disparities in the risk for influenza-related complications among adults were noted, including higher rates of severe influenza illness among blacks and among American Indians/Alaska Natives and indigenous populations in other countries (97-102). These disparities might be attributable in part to the higher prevalence of underlying medical conditions or disparities in medical care among these racial/ethnic groups (101, 103). A more recent case-control study of risk factors for death from 2009 pandemic influenza that adjusted for factors such as pre-existing medical conditions, barriers to health care access, and delayed receipt of antivirals found that American Indian/Alaska Native status was not independently associated with death (104).