The Burden of All-Cause Mortality Following Influenza-Associated Hospitalizations: Influenza Hospitalization Surveillance Network, 2010–2019

Abstract

Background

While the estimated number of US influenza-associated deaths is reported annually, detailed data on the epidemiology of influenza-associated deaths, including the burden of in-hospital vs post-hospital discharge deaths, are limited.

Methods

Using data from the 2010–2011 through 2018–2019 seasons from the Influenza Hospitalization Surveillance Network, we linked cases to death certificates to identify patients who died from any cause during their influenza hospital stay or within 30 days post discharge. We described demographic and clinical characteristics of patients who died in the hospital vs post discharge and characterized locations and causes of death (CODs).

Results

Among 121 390 cases hospitalized with laboratory-confirmed influenza over 9 seasons, 5.5% died; 76% of deaths were in patients aged ≥65 years, 71% were non-Hispanic White, and 34% had 4 or more underlying medical conditions. Among all patients with an influenza-associated hospitalization who died, 48% of deaths occurred after hospital discharge; the median number of days from discharge to death was 9 (interquartile range, 3–19). Post-discharge deaths more often occurred in older patients and among those with underlying medical conditions. Only 37% of patients who died had “influenza” as a COD on their death certificate. Influenza was more frequently listed as a COD among persons who died in the hospital compared with cardiovascular disease among those who died after discharge.

Conclusions

All-cause mortality burden is substantial among patients hospitalized with influenza, with almost 50% of deaths occurring within 30 days after hospital discharge. Surveillance systems should consider capture of post-discharge outcomes to better characterize the impact of influenza on all-cause mortality.

Influenza is associated with an estimated 140 000–710 000 hospitalizations and 12 000–52 000 deaths in the United States annually [1]. Adults aged ≥65 years, children aged <5 years, pregnant people, and those with chronic medical conditions are at increased risk for influenza-related complications, which may result in hospitalization or death [2–8]. National surveillance systems, including the National Center for Health Statistics (NCHS), which collects death certificate data from state vital statistics offices [9], and the Influenza-Associated Pediatric Mortality Surveillance System, which collects information on all laboratory-confirmed deaths in persons aged <18 years [9], capture data annually on the number of influenza-associated deaths in the United States. However, detailed epidemiologic and clinical data on influenza-associated deaths, as well as the burden of deaths that occur after an influenza-associated hospitalization, are limited [9].

Although less is known about the impact of influenza on post-hospitalization outcomes, one study in Arizona found that use of electronic vital records increased detection of severe acute respiratory infection–associated mortality by 30% [10]. Another study of patients hospitalized with influenza in Tennessee found that 14% of patients were readmitted to the hospital within 30 days of hospital discharge after linking cases to the state discharge database [11]. Data on causes of death (CODs) among patients with influenza are also limited. A recent test-negative study from Ontario, Canada, found that only 23% of influenza-positive adults aged >65 years had “influenza” recorded as a COD on their death certificates [12]. Influenza and pneumonia were more frequently reported CODs among those who died within 0–7 days of discharge, while circulatory system diseases and cancer were reported more frequently ≥8 days post discharge. However, this study did not include persons aged ≤65 years [12]. An improved understanding of the burden of all-cause mortality occurring up to 30 days following an influenza-associated hospitalization and the causes of influenza-attributable deaths among patients of all ages is important to better assess annual influenza disease burden and the burden of illness averted through influenza vaccination and to guide influenza prevention and control policies and communications.

The Centers for Disease Control and Prevention (CDC) began collecting data on all-cause mortality up to 30 days post-discharge from laboratory-confirmed influenza-associated hospitalizations through the Influenza Hospitalization Surveillance Network (FluSurv-NET) during the 2010–2011 influenza season. The objectives of this analysis were to characterize the incidence and epidemiology of deaths among laboratory-confirmed influenza-associated hospitalizations, determine the proportion of deaths that occurred in the hospital vs post discharge, examine trends over time and by age, describe the underlying CODs from death certificates, and describe death locations among those who were discharged from their influenza-associated hospitalization.

METHODS

Study Setting

We analyzed FluSurv-NET data from the 2010–2011 through 2018–2019 influenza seasons. FluSurv-NET conducts population-based surveillance for laboratory-confirmed influenza-associated hospitalizations among all ages [13]. FluSurv-NET captures influenza-associated hospitalizations among persons residing in 70 counties in 10 Emerging Infections Program states (California, Colorado, Connecticut, Georgia, Maryland, Minnesota, New Mexico, New York, Oregon, and Tennessee) and 3 Influenza Hospitalization Surveillance Project states (Michigan, Ohio, and Utah). The FluSurv-NET catchment area represents about 9% of the US population. Capture of influenza-associated hospitalizations among residents of the catchment area should be nearly 100% as cases are identified through review of multiple sources, including laboratory, clinical, and notifiable disease databases [13]. The demographic makeup of the FluSurv-NET catchment area is generally similar to that of the US population [14].

The CDC determined that this activity met the requirement for public health surveillance; therefore, CDC institutional review board (IRB) approval was not required. Sites that participate in FluSurv-NET obtained human subjects and ethics approvals from their respective state and local health departments and academic partner IRBs as needed.

Case Ascertainment and Data Collection Methods

A FluSurv-NET case was defined as hospitalization of a FluSurv-NET catchment area resident, with admission date during a designated influenza season (1 October –30 April) and a positive influenza test within 14 days before hospital admission or any time during hospitalization [13, 14]. Trained surveillance staff conducted medical record abstractions using a standardized case report form. Data elements included age, sex, race and ethnicity, laboratory testing results for influenza type and subtype, place of residence, underlying medical conditions (obtained from problem lists, history and physical exam notes, progress notes, and discharge summaries), hospital course including symptom onset date, length of stay, intensive care unit (ICU) admission, mechanical ventilation, and discharge disposition including death during hospitalization.

For each case, surveillance staff used either the NCHS Electronic Death Registration System or their state health or vital records department to link and abstract death certificate data for FluSurv-NET patients who died in the hospital or within 30 days post-hospital discharge. Sites used a variety of methods to match FluSurv-NET cases to death certificates, with 6 sites using a probabilistic approach (Link Plus, Link King), 4 sites using a deterministic approach (SAS), and 3 sites obtaining data directly from vital statistics programs. All sites used last name, first name, and date of birth as person-level identifiers in the match. Sites could also use other optional fields (county, middle name, social security number) if available. Information abstracted from the death certificates included date, location, and underlying CODs. Patients with missing death certificate data or death certificates that listed a date of death greater than 30 days after hospital discharge (n = 88) were excluded from the analysis.

The COD for each patient was recorded as either International Classification of Diseases, 9th or 10th Revisions (ICD-9 or ICD-10), codes or free text on death certificates. We categorized COD using NCHS 113-COD codes [15]. Causes of death were grouped into nonmutually exclusive categories including influenza, pneumonia and influenza (P&I) [1], respiratory and circulatory (R&C; influenza includes ICD10 codes J09–J11, P&I includes ICD10 codes J09–J18, and R&C includes ICD10 codes I00–I99 and J00–J99), cardiovascular disease, sepsis, immunosuppressive condition, chronic lung disease, chronic kidney disease, chronic metabolic disease, liver disease, neurological disease, blood disorders, and other conditions.

Statistical Analyses

Descriptive statistics were presented as frequencies, percentages, or medians. We calculated the proportion of deaths that occurred in the hospital vs within 30 days post discharge by season and age group. We compared demographic and clinical characteristics, hospital course, and CODs of patients who died in the hospital vs within 30 days post discharge. We analyzed post-discharge deaths to compare CODs among those who died 0–9 days vs 10–30 days post discharge and described the locations of deaths.

Data were analyzed using SAS (version 9.4, Cary, NC). Fisher exact or χ² tests were used to compare proportions, and the Wilcoxon rank sum test was used to evaluate distribution differences between continuous variables. All comparisons were 2-sided, and P values <.05 were considered statistically significant.

RESULTS

Prevalence and Characteristics of Influenza-Associated Deaths

During 2010–2019, we identified 121 390 hospitalized FluSurv-NET cases; of those, 6687 (5.5%) died during their hospitalization or within 30 days post-hospital discharge (Figure 1A). The proportion of cases who died ranged by season from 4.1% in 2011–2012 to 6.2% in 2014–2015; the proportion ranged by age group from 0.5% among persons aged 0–4 years to 12.2% among persons aged ≥85 years (Figure 1B).

Figure 1.

Mortality among laboratory-confirmed influenza-associated hospitalizations by season and age group, Influenza Hospitalization Surveillance Network (FluSurv-NET) 2010–2019. A, stratifies by season; B, stratifies by age group. Mortality includes those who died during hospitalization and those who died within 30 days after discharge among all FluSurv-NET cases. Denominators overall and by season: overall, n = 121 390; 2010–2011, n = 5663; 2011–2012, n = 2364; 2012–2013, n = 11 434; 2013–2014, n = 9658; 2014–2015, n = 17 743; 2015–2016, n = 8790; 2016–2017, n = 17 489; 2017–2018, n = 29 695; 2018–2019, n = 18 554.

Among all deaths, including those who died in the hospital or after discharge (n = 6687), 3472 (51.9%) occurred in the hospital and 3215 (48.1%) occurred during the 30 days after discharge. The proportion of deaths that occurred post-hospital discharge increased with increasing age, from 12.5% among those aged <5 years to 59.7% among those aged ≥85 years (Figure 2). The overall proportion of those who died post-hospital discharge also varied by influenza season, ranging from 35.9% to 54.5%. Notably, during influenza A(H3N2)-predominant seasons (2012–2013, 2014–2015, 2016–2017, 2017–2018, and 2018–2019), higher proportions of patients died after hospital discharge compared with the other seasons (Table 1).

Figure 2.

Among patients with influenza-associated hospitalizations who died, the proportion of deaths that occurred during hospitalization vs within 30 days after hospital discharge by age group, Influenza Hospitalization Surveillance Network, 2010–2019.

Table 1.

Among Patients With Laboratory-Confirmed Influenza-Associated Hospitalizations Who Died, the Proportion Who Died in the Hospital Vs Within 30 Days Post-Hospital Discharge, by Season, Age Group, and Race/Ethnicity: FluSurv-NET, 2010–2019

Characteristic	Overall (n = 6687); %	2010–2011 (n = 307); %	2011–2012 (n = 96); %	2012–2013 (n = 614); %	2013–2014 (n = 541); %	2014–2015 (n = 1106); %	2015–2016 (n = 386); %	2016–2017 (n = 989); %	2017–2018 (n = 1649); %	2018–2019 (n = 999); %
Overall
Died in the hospital	51.9	59.3	60.4	50.3	64.1	45.5	58.8	51.4	49.5	52.3
Died after discharge	48.1	40.7	39.6	49.7	35.9	54.5	41.2	48.6	50.5	47.7
Age group
0–49 y
Died in the hospital	77.9	85.2	80.0a	83.7	83.3	70.2	81.4	81.0	71.1	72.8
Died after discharge	22.1	14.8	20.0a	16.3	16.7	29.8	18.6	19.0	28.9	27.2
50–64 y
Died in the hospital	69.1	69.7	83.3a	68.9	75.1	63.0	68.5	64.7	68.1	69.7
Died after discharge	30.9	30.3	16.7a	31.1	24.9	37.0	31.5	35.3	31.9	30.3
≥65 y
Died in the hospital	45.8	48.1	50.7	44.2	51.8	41.7	51.6	47.9	45.2	45.2
Died after discharge	54.2	51.9	49.3	55.8	48.2	58.3	48.4	52.1	54.8	54.8
Race/Ethnicity
Non-Hispanic White
Died in the hospital	49.7	54.5	64.4	49.8	63.4	44.3	53.0	49.6	46.5	49.9
Died after discharge	50.3	45.5	35.6	50.2	36.6	55.7	47.0	50.4	53.5	50.1
Non-Hispanic Black
Died in the hospital	56.5	84.4	66.7a	46.2	62.5	51.3	69.8	50.8	58.5	53.4
Died after discharge	43.5	15.6	33.3a	53.8	37.5	48.7	30.2	49.2	41.5	46.6
Hispanic
Died in the hospital	61.8	76.5	50.0a	57.9a	62.1a	51.4	72.4a	71.8	55.4	63.2
Died after discharge	38.2	23.5a	50.0a	42.1a	37.9a	48.6	27.6a	28.2	44.6	36.8
Otherb
Died in the hospital	63.1	66.7a	50.0a	65.4a	68.4a	55.6	75.0a	61.3	60.4	69.2
Died after discharge	36.9	33.3a	50.0a	34.6a	31.6a	44.4	25.0a	38.7	39.6	30.8

Data include in-hospital deaths and those who died within 30 days after discharge. The numbers of post-discharge deaths that occurred among pediatric cases aged <18 years by season are as follows: 2010–2011, n = 3; 2011–2012, n = 0; 2012–2013, n = 20; 2013–2014, n = 3; 2014–2015, n = 17; 2015–2016, n = 6; 2016–2017, n = 6; 2017–2018, n = 15; 2018–2019, n = 15. The following seasons were predominated by influenza A(H3N2) viruses: 2012–2013, 2014–2015, 2016–2017, 2017–2018, and 2018–2019.

Abbreviation: FluSurv-NET, Influenza Hospitalization Surveillance Network.

^aSample size <30; estimates may not be reliable.

^bOther race/ethnicity includes the following groups: non-Hispanic Asian or Pacific Islander, non-Hispanic American Indian/Alaska Native, and multiracial.

Among the 6687 deaths, 76.4% occurred among adults aged ≥65 years, 52.2% were female, and 71.3% were non-Hispanic White (Table 2). Among the 114 703 patients who did not die, 49.7% were aged ≥65 years, 53.8% were female, and 53.0% were non-Hispanic White (data not shown). The majority of patients who died had 2 or more underlying conditions (81.5%), the most common being cardiovascular disease (64.1%), chronic metabolic disease (45.6%), and chronic lung disease, excluding asthma (36.2%). Among all patients who died, 42.2% had pneumonia, 54.1% received ICU care, 36.7% required mechanical ventilation, and 66.6% received antiviral treatment, and the median length of ICU stay was 4 days (interquartile range [IQR], 2–9; Table 2).

Table 2.

Demographic and Clinical Characteristics of Patients With Laboratory-Confirmed Influenza-Associated Hospitalizations Who Died, Stratified by Age Group and Whether They Died During Hospitalization or Within 30 Days Post-discharge: FluSurv-NET 2010–2019

	All Deaths	Deaths by Age Group			Deaths by Hospital Discharge Status
Characteristic	n = 6687	<65 y n = 1579	≥65 y n= 5108		In-Hospital n = 3472	Within 30 Days Post Discharge n = 3215
	n (%)	n (%)	n (%)	P Value*	n (%)	n (%)	P Value**
Age, y				<.001			<.001
0–4	40 (0.6%)	40 (2.5%)	...		35 (1.0%)	5 (0.2%)
5–17	45 (0.7%)	45 (2.5%)	...		39 (1.1%)	6 (0.2%)
18–49	404 (6.0%)	404 (25.6%)	...		307 (8.8%)	97 (3.4%)
50–64	1090 (16.3%)	1090 (69.0%)	...		753 (21.7%)	337 (10.5%)
≥65	5108 (76.4%)	...	5108 (100.0%)		2338 (67.3%)	2770 (86.2%)
65–74	1128 (16.9%)	...	1128 (22.1%)		609 (17.5%)	519 (16.1%)
75–84	1596 (23.9%)	...	1596 (31.2%)		769 (22.1%)	827 (25.7%)
≥85	2384 (35.7%)	...	2384 (46.7%)		960 (27.6%)	1424 (44.3%)
Sex				<.001			.041
Male	3194 (47.8%)	841 (53.3%)	2353 (46.1%)		1700 (49.0%)	1494 (46.6%)
Female	3493 (52.2%)	738 (46.7%)	2755 (53.9%)		1772 (51.0%)	1721 (53.5%)
Race/Ethnicitya				<.001			<.001
Non-Hispanic White	4766 (71.3%)	933 (59.1%)	3833 (75.0%)		2007 (68.3%)	2030 (74.5%)
Non-Hispanic Black	805 (12.0%)	336 (21.3%)	469 (9.2%)		392 (13.1%)	295 (10.9%)
Non-Hispanic American Indian/Alaska Native	24 (0.4%)	14 (0.9%)	10 (0.2%)		11 (0.5%)	8 (0.2%)
Non-Hispanic Asian/Pacific Islander	324 (4.8%)	57 (3.6%)	267 (5.2%)		174 (5.9%)	105 (3.7%)
Hispanic	314 (4.7%)	130 (8.2%)	184 (3.6%)		158 (5.6%)	99 (3.7%)
Influenza type/Subtypeb				.50			.25
Influenza A	5391 (80.6%)	1276 (80.8%)	4115 (80.6%)		2798 (80.6%)	2593 (80.7%)
2009 H1N1c	764 (29.3%)	403 (58.2%)	361 (18.9%)	<.001	494 (35.6%)	270 (22.2%)	<.001
H3N2c	1838 (70.6%)	290 (41.9%)	1548 (81.1%)		893 (64.4%)	945 (77.8%)
Influenza B	1243 (18.6%)	286 (18.1%)	957 (18.7%)		642 (18.5%)	601 (18.7%)
Influenza vaccination				<.001			<.001
Not vaccinated	2139 (32.0%)	723 (45.8%)	1416 (27.7%)		1205 (34.7%)	934 (29.1%)
Vaccinated	3128 (46.8%)	474 (30.0%)	2654 (52.0%)		1483 (42.7%)	1645 (51.2%)
Unknown	1420 (21.2%)	382 (24.2%)	1038 (20.3%)		784 (22.6%)	636 (19.8%)
Antiviral treatment	4453 (66.6%)	1119 (70.9%)	3334 (65.3%)	.006	2259 (65.1%)	2194 (68.2%)	.001
Number of underlying condition categoriesd				<.001			<.001
0	320 (4.8%)	169 (10.7%)	151 (3.0%)		237 (6.8%)	83 (2.6%)
1	910 (13.7%)	298 (18.9%)	612 (12.1%)		513 (14.8%)	397 (12.5%)
2	1523 (22.9%)	324 (20.5%)	1199 (23.6%)		805 (23.2%)	718 (22.6%)
3	1612 (24.2%)	322 (20.4%)	1290 (25.4%)		799 (23.0%)	813 (25.5%)
≥4	2286 (34.4%)	464 (29.4%)	1822 (35.9%)		1113 (32.1%)	1173 (36.8%)
Underlying Medical Condition
Asthma	708 (10.6%)	227 (14.4%)	481 (9.4%)	<.001	372 (10.7%)	336 (10.5%)	.73
Chronic lung disease, excluding asthma	2422 (36.2%)	503 (31.2%)	1919 (37.6%)	<.001	1210 (34.9%)	1212 (37.7%)	.015
Chronic metabolic disease	3050 (45.6%)	626 (39.6%)	2424 (47.5%)	<.001	1553 (44.7%)	1497 (46.6%)	.13
Cardiovascular disease, excluding hypertension	4288 (64.1%)	647 (41.0%)	3641 (71.3%)	<.001	2124 (61.2%)	2164 (67.3%)	<.001
Hemoglobinopathy/blood disorder	394 (5.9%)	111 (7.0%)	283 (5.5%)	.028	199 (5.7%)	195 (6.1%)	.56
Immunosuppressive condition	1498 (22.4%)	468 (29.6%)	1030 (20.2%)	<.001	751 (21.6%)	747 (23.2%)	.12
Neurologic conditione	2166 (33.9%)	310 (21.2%)	1856 (37.7%)	<.001	917 (27.9%)	1249 (40.4%)	<.001
Neuromuscular condition	573 (8.6%)	99 (6.3%)	474 (9.3%)	<.001	264 (7.6%)	309 (9.6%)	.003
Obesity	1621 (24.2%)	564 (35.7%)	1057 (20.7%)	<.001	944 (27.2%)	677 (21.1%)	<.001
Renal disease	1995 (29.8%)	320 (20.3%)	1675 (32.8%)	<.001	999 (28.8%)	996 (31.0%)	.049
ICU admissionf	3618 (54.1%)	1278 (80.9%)	2340 (45.8%)	<.001	2586 (74.5%)	1032 (32.1%)	<.001
ICU length of stay,g median (interquartile range), d	4.0 (2.0–9.0)	5.0 (2.0–11.0)	4.0 (2.0–8.0)	<.001	4.0 (1.0–9.0)	4.0 (2.0–8.0)	.67
Mechanical ventilationh	2454 (36.7%)	1079 (68.3%)	1375 (26.9%)	<.001	1994 (57.4%)	460 (14.3%)	<.001
Extracorporeal membrane oxygenationi	103 (1.5%)	91 (5.8%)	12 (0.2%)	<.001	97 (2.8%)	6 (0.2%)	<.001
Pneumoniaj	2824 (42.2%)	780 (49.4%)	2044 (40.0%)	<.001	1724 (49.7%)	1100 (34.2%)	<.001
Hospital length of stay, dk				<.001			<.001
0	112 (1.7%)	36 (2.3%)	76 (1.5%)		104 (3.0%)	8 (0.2%)
1–7	3613 (54.0%)	739 (46.8%)	2874 (56.3%)		1829 (52.7%)	1784 (55.5%)
8+	2960 (44.3%)	804 (50.9%)	2156 (42.2%)		1539 (44.3%)	1421 (44.2%)

Abbreviations: FluSurv-NET, Influenza Hospitalization Surveillance Network; ICU, intensive care unit.

^* P value comparing aged <65 years with aged ≥65 years based on χ² test.

^** P value comparing in-hospital deaths with 30  days post-discharge deaths based on χ² test.

^aDenominator for race and ethnicity includes multiracial (n = 12, 0%) and unknown race (n = 442, 7%).

^bDenominator includes 37 (1%) cases with influenza A and B and 16 (0%) whose type could not be distinguished between A/B.

^cDenominator includes anyone with known influenza A subtype (n = 2602); those with Other (n = 16, 0%) and Unknown (n = 2810, 52%) subtype were excluded from this table.

^dUnderlying conditions used to classify this variable include the following: cardiovascular disease, chronic metabolic disease, chronic lung disease, blood disorder, immunocompromised conditions, neurologic conditions, neuromuscular conditions, obesity, and renal disease. Number of underlying conditions was missing among n = 36 (0%) cases.

^eData were not collected during the 2010–2011 influenza season (n = 307).

^fDenominator includes 57 (1%) cases with unknown ICU status.

^gThe 2010–2011 season was not included and missing n = 273 among those admitted to the ICU from 2011–2012 through 2018–2019 seasons.

^hDenominator includes 76 (1%) cases with unknown mechanical ventilation status.

ⁱDenominator includes 83 (1%) cases with unknown extracorporeal membrane oxygenation status.

^jPneumonia was defined using discharge diagnoses as well as abnormal chest X-ray findings within 3 days of hospital admission.

^kHospital length of stay missing for n = 2 deaths.

When comparing characteristics and clinical course for patients aged <65 years vs ≥65 years who died, those aged <65 years were significantly more likely to test positive for influenza A (H1N1)pdm09 (58.2% vs 18.9%), be unvaccinated (45.8% vs 27.7%), have a hemoglobinopathy or blood disorder (7.0% vs 5.5%), or have no underlying medical conditions (10.7% vs 3.0%). While decedents aged <65 years were less likely to have underlying conditions overall, conditions that were significantly more common among this age group compared with those aged ≥65 years included asthma (14.4% vs 9.4%), immunosuppressive condition (29.6% vs 20.2%), and obesity (35.7% vs 20.7%). Patients aged <65 years who died were significantly more likely to have a complicated hospital course compared with those aged ≥65 years, including the need for ICU admission (80.9% vs 45.8%), mechanical ventilation (68.3% vs 26.9%), extracorporeal membrane oxygenation (ECMO; 5.8% vs 0.2%), and a longer ICU length of stay (5 days vs 4 days; Table 2).

When comparing characteristics and clinical course of patients who died during their influenza-associated hospitalization vs within 30 days post discharge, a significantly greater proportion of those who died in the hospital were aged <50 years (10.9% vs 3.8%, respectively), non-White (31.7% vs 25.5%), unvaccinated (34.7% vs 29.1%), and had no previously diagnosed underlying medical conditions (6.8% vs 2.6%). Patients who died during their influenza-associated hospitalization were also significantly more likely to test positive for influenza A(H1N1)pdm09 (35.6% vs 22.2%), be diagnosed with pneumonia (49.7% vs 34.2%), be admitted to the ICU (74.5% vs 32.1%), be mechanically ventilated (57.4% vs 14.3%), receive ECMO (2.8% vs 0.2%), and be hospitalized for less than 1 day (3.0% vs 0.2%; Table 2).

Among the 3215 patients who died after hospital discharge, the median number of days from hospital discharge to death was 9 (IQR, 3–19; Figure 3).

Figure 3.

Days from discharge to death among patients who died within 30 days after hospital discharge, Influenza Hospitalization Surveillance Network, 2010–2019 (n = 3215).

Location of Patient Residence at Hospital Admission, Discharge, and Death

Among patients who died and had information on place of residence prior to admission (n = 5634), 61.8% lived in a private residence and 34.6% resided in a nursing home or long-term care facility (Supplementary Figure 1). Among patients who died post discharge, 36.9% were discharged to a nursing home or long-term care facility, 36.7% to hospice, and 23.3% to a private residence. Subsequently, 16.0% died in hospice care, 24.6% in a nursing home or long-term care facility, and 17.5% in a private residence, and 26.7% were subsequently readmitted and died in a hospital (Supplementary Figure 1).

Causes of Death

Among all patients who died, 36.8% had an influenza code, 51.1% had a P&I code, and 83.6% had an R&C code listed on their death certificate (Table 3). Patients who died during their influenza-associated hospitalization more often had an influenza (53.8% vs 18.5%), P&I (69.1% vs 31.8%), or R&C (90.6% vs 76.2%) code compared with those who died post-hospital discharge. Among all deaths, the most common other CODs included cardiovascular disease (36.8%) and sepsis (16.8%; Table 3). Compared with patients who died during their influenza-associated hospitalization, a higher proportion of those who died post-hospital discharge had chronic conditions listed as a COD, including cardiovascular disease, chronic lung disease, immunosuppressive conditions, chronic metabolic disease, and neurologic disease. The greater the length of time between hospitalization and death, the less likely a patient was to have influenza listed as a COD on their death certificate (53.8% of those who died during hospitalization, 26.9% of those who died within 0–9 days post discharge, and 9.7% of those who died within 10– to 30 days post discharge).

Table 3.

Causes of Death Listed on the Death Certificate Among Patients With an Influenza-Associated Hospitalization Who Died During Hospitalization Vs Within 30 Days Post Discharge: FluSurv-NET 2010–2019

Cause of Deatha	All Deaths (%)	In-Hospital Deaths (%)	Deaths Within 30 d Post Discharge
			All Post-Discharge Deaths (%)	Deaths by Days Post Discharge (%)		Deaths by Death Location (%)
				0–9 d	10–30 d	Hospital	Hospice	Other Facilityb	Private Residence
Respiratory and Circulatory	83.6	90.6	76.2	78.4	73.8	84.0	71.4	76.4	70.5
Pneumonia and Influenza	51.1	69.1	31.8	39.6	23.5	42.0	32.7	29.9	22.9
Influenza	36.8	53.8	18.5	26.9	9.7	23.2	20.4	17.6	14.6
Cardiovascular disease	36.8	32.4	41.5	38.2	44.8	41.3	35.3	46.6	41.1
Sepsis	16.8	24.0	9.2	9.5	8.8	17.7	9.2	5.3	3.2
Immunocompromised condition	11.1	8.3	13.8	13.0	14.8	11.7	18.6	9.1	22.3
Chronic lung disease	12.4	10.0	14.9	14.3	15.6	14.1	14.5	14.3	17.1
Kidney disease	7.8	8.0	7.5	7.5	7.5	7.5	8.4	7.6	7.0
Chronic metabolic disease	4.5	3.7	5.3	4.8	5.9	5.1	4.1	5.2	6.6
Liver disease	1.2	1.2	1.2	1.1	1.3	0.9	2.0	0.6	2.0
Neurological disease	4.4	1.4	7.5	6.8	8.3	2.4	6.5	12.9	8.6
Blood disorders	0.9	0.9	1.0	0.9	1.1	1.4	1.0	0.9	0.5
Otherc	5.0	3.1	7.1	6.8	7.5	4.8	8.4	8.2	6.3

Abbreviations: FluSurv-NET, Influenza Hospitalization Surveillance Network.

^aCauses are not mutually exclusive and could be listed anywhere on the death certificate. Denominator (n = 6574) excludes 113 cases because they have no cause of death information.

^bNursing home or long-term care facility.

^cOther causes of death include all other diseases not listed as well as nondisease causes of death (eg, accidents, intentional self-harm, assault).

DISCUSSION

Among more than 100 000 patients hospitalized with laboratory-confirmed influenza over 9 influenza seasons in the United States, 5.5% died either during hospitalization or within 30 days after discharge, ranging from <1% among children aged <18 years to 12% among adults aged ≥85 years. Among patients hospitalized with influenza who died after discharge, almost half of all deaths occurred within 9 days after hospital discharge. The proportion of patients who died after hospital discharge increased with age; among adults aged ≥85 years, 60% of deaths occurred after discharge. A little more than one-third of patients had influenza listed as a COD on death certificates, and this decreased to <20% for patients who died after hospital discharge. These results highlight a large burden of all-cause mortality among patients with an influenza-associated hospitalization, though not all deaths in this study can be attributed to influenza. Deaths after hospitalization disproportionality affected older persons and those with underlying medical conditions. Incorporation of post-discharge outcome monitoring into influenza hospitalization surveillance can improve national estimates and help to better quantify immediate and short-term (within 30 days of hospitalization) morbidity and mortality associated with influenza.

Our findings of the large underestimated burden of all-cause mortality among patients hospitalized with influenza that occurred within 30 days after discharge are similar to estimates that have been previously reported among patients hospitalized with community-acquired pneumonia or from severe acute respiratory infection surveillance [10, 16–18]. However, by focusing on laboratory-confirmed influenza-associated hospitalizations, our study provides pathogen-specific data to refine our understanding of mortality associated with severe influenza illness. Of note, we found that three-quarters of deaths that occurred within 30 days after discharge occurred within the first 19 days, and we also identified season-to-season variations in the proportion of deaths after discharge. Generally, seasons predominated by influenza A(H3N2) viruses [19] yielded higher proportions of patients who died after discharge, likely due in part to the fact that these seasons disproportionately affected older adults who were also more likely to die post discharge. Further investigation into factors that contribute to seasonal variations in in-hospital vs post-discharge deaths is warranted, including exploration of factors such as circulating influenza virus type/subtype and vaccine effectiveness.

We observed notable differences in the characteristics of patients who died during their influenza-associated hospitalization compared with those who died after discharge. Those who died in the hospital were younger, more likely to be non-White, had no documented underlying medical conditions, were unvaccinated, and were more likely to receive ICU care or mechanical ventilation and less likely to receive antiviral treatment compared with those who died post discharge. These results likely reflect the differences in the age and severity of clinical illness of these groups, though differences could also be due to varying admission criteria. Older adults may also be less likely to receive ICU care or mechanical ventilation due to advanced care directives, though these data were not available in FluSurv-NET for the included seasons. Those who died post discharge often died in a nursing home or long-term care facility, which could be a potential source of underestimated mortality burden in older adults. A notable percentage also died in hospice care upon discharge.

We noted several interesting findings when we examined COD data from death certificates. Consistent with findings from other studies, we found that only 37% of deaths had influenza listed as a COD in any position on the death certificate. Since all patients in the study population had recent laboratory-confirmed influenza, this finding underscores the limitations of relying on death certificate codes alone for estimating mortality that might be attributable to influenza. We also found differences in the types of CODs recorded for in-hospital deaths vs deaths that occurred after discharge. In particular, the in-hospital deaths had higher proportions of acute codes including influenza, P&I, R&C, and sepsis CODs compared with post-discharge deaths, which had higher proportions of chronic underlying condition codes including cardiovascular disease, chronic lung disease, and neurological disease. These findings are likely related to the fact that the further out from an acute illness a patient is, the greater role chronic conditions may play in contributing to death. These findings may also reflect the natural evolution of the clinical course of influenza virus infection and its influence on the exacerbation of preexisting chronic conditions. Our finding that a high proportion of post-discharge deaths had cardiovascular disease listed as a COD is similar to findings reported in other studies describing an increased risk of cardiovascular disease outcomes within 7 days after influenza virus infection [20, 21].

Our analysis is subject to several limitations. First, FluSurv-NET surveillance is based on clinician-directed testing and may underestimate the true burden of influenza-associated hospitalizations and mortality since not everyone hospitalized with influenza was tested for influenza. Second, FluSurv-NET covers approximately 9% of the US population, and findings may not be generalizable to the overall US population. Third, matching approaches used to link FluSurv-NET cases to death certificates may have underascertained some post-discharge deaths; furthermore, patients who died more than 30 days post discharge were not included in the analysis. Fourth, certain CODs may have been misclassified while recoding data entered into free text fields [22]. Fifth, as the objective of this study was to capture all-cause mortality, we were not able to conclude that all deaths in this study were directly attributed to influenza. Finally, while it is helpful to monitor post-discharge deaths in order to characterize morbidity and mortality associated with influenza, death certificate data from vital statistics can be delayed up to 18 months or longer, and collection of these data requires time and resources.

National mortality surveillance systems that rely solely on death certificates could miss up to two-thirds of patients in whom influenza may have contributed to death. Furthermore, surveillance for mortality among influenza-associated hospitalizations may miss up to 50% of deaths if restricted to in-hospital mortality alone. Supplementing national mortality surveillance with data from laboratory-confirmed influenza-associated hospitalization surveillance, including capture of post-discharge outcomes, can improve annual estimates of influenza-associated disease burden as well as the economic burden of influenza, which may help target prevention strategies and risk communication. Medical providers should be aware of the residual risk for mortality soon after discharge from an influenza-associated hospitalization, particularly among older adults. Influenza surveillance systems should consider capture of post-discharge outcomes to better characterize immediate and short-term impacts of influenza on morbidity and mortality.

Supplementary Data

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

Notes

Acknowledgments. The authors thank the following individuals and organizations for their assistance collecting FluSurv-NET data: Sherry Quach, Gretchen Rothrock, Jeremy Roland, Joelle Nadle, Ashley Coates, Monica Napoles (California Emerging Infections Program); Sharon Emmerling, Breanna Kawasaki, Madelyn Lensing, Sarah McLafferty, Jordan Surgnier, Millen Tsegaye (Colorado Department of Public Health and Environment); Maria Correa, Amber Maslar, Tamara Rissman, (Connecticut Emerging Infections Program, Yale School of Public Health); Katelyn Ward, Andrew Martin, Jeremiah Williams (Foundation for Atlanta Veterans Education and Research, Georgia Emerging Infections Program, Georgia Department of Public Health, Atlanta Veterans Affairs Medical Center); Maya Monroe, Alicia Brooks (Maryland Department of Health); Jim Collins, Shannon Johnson, Sue Kim, Libby Reeg, Val Tellez Nunez, Elizabeth McCormick, Sarah Rojewski, Genevieve Palazzolo, Amber Brewer (Michigan Department of Health and Human Services); Craig Morin, Cynthia Kenyon, Team Flu (Minnesota Department of Health); Chad Smelser, Daniel M. Sosin, Salina Torres, Emily Hancock, Chelsea L. Esquibel, Sarah Shrum Davis, Marisa Bargsten, Carla Young, Jeremy Espinoza (New Mexico Department of Health); Sarah Lathrop, Kathy M. Angeles, Sarah A. Khanlian, Lisa Butler, Robert Mansmann, Marjorie McConnell, Katherine Gienger, Lori Morrow, Joan Baumbach (New Mexico Emerging Infections Program); Kerianne Engesser, Adam Rowe (New York State Department of Health); Sophrena Bushey, Christina Felsen, Kevin Popham, Katherine St. George, Erin Licherdell, Christine Long (University of Rochester School of Medicine and Dentistry); Laurie Billing, Nicholas Fisher, Julie Freshwater, Denise Ingabire-Smith, Ann Salvator, Maya Scullin, Rebekah Sutter (Ohio Department of Health); Sam Hawkins (Public Health Division, Oregon Health Authority); Tiffanie Markus, Katie Dyer, Karen Leib, Terri McMinn, Danielle Ndi, Gail Hughett, Bentley Akoko, Kathy Billings, Anise Swain, Manideepthi Pemmaraju (Vanderbilt University Medical Center); Andrea George, Andrea Price, Andrew Haraghey, Ashley Swain, Diana Thurston, Melanie Crossland, Ryan Chatelain (Salt Lake County Health Department); Sandra Chaves, Carmen Sofia Arriola, Craig McGowan, Danielle Iuliano, Melissa Rolfes (Influenza Division, Centers for Disease Control and Prevention [CDC]).

Financial support. The Influenza Hospitalization Surveillance Network is a collaboration of state health departments, academic institutions, and local partners and is funded by the CDC. This work was supported by the CDC through an Emerging Infections Program cooperative agreement (grant CK17-1701) and through a Council of State and Territorial Epidemiologists cooperative agreement (grant NU38OT000297-02-00). For 1 site (Tennessee), this project was supported by Clinical and Translational Science Award (CTSA) UL1 TR002243 from the National Center for Advancing Translational Sciences.

Disclaimer. The findings and conclusions presented here are those of the author(s) and do not necessarily represent the official position of the CDC, National Center for Advancing Translational Sciences, or the National Institutes of Health.