Hospital Readmissions After Laboratory-Confirmed Influenza Hospitalization

Abstract

Background

Influenza infection causes substantial morbidity and mortality. However, little is known about hospital readmissions after an influenza hospitalization. The aim of our study was to characterize frequency of hospital readmissions among patients hospitalized with laboratory-confirmed influenza.

Methods

We conducted a retrospective study using Tennessee Emerging Infections Program Influenza Surveillance data from 2006 to 2016 and the concurrent Tennessee Hospital Discharge Data System. We analyzed demographic characteristics and outcomes to better understand frequency and factors associated with hospital readmissions.

Results

Of the 2897 patients with a laboratory-confirmed influenza hospitalization, 409 (14%) and 1364 (47%) had at least 1 hospital readmission within 30 days and 1 year of the influenza hospitalization, respectively. Multiple readmissions occurred in 739 patients (54%). The readmission group was older, female predominant, and had more comorbidities than patients not hospitalized. Pneumonia, acute chronic obstructive pulmonary disease/asthma exacerbation, septicemia, acute respiratory failure, and acute renal failure were the most common causes for readmission at 30 days. Underlying cardiovascular disease, lung disease, kidney disease, diabetes, immunosuppression, and liver disease were associated with increased risk of readmission during the subsequent year.

Conclusions

After an admission with laboratory-confirmed influenza, there is a high likelihood of readmission within 30 days and 1 year adding to the morbidity of influenza.

comorbidities, emerging infections network, hospital readmissions, influenza

Influenza is associated with an estimated 140 000 to 710 000 hospitalizations yearly [1]. This is a substantial burden to the healthcare system annually. In recent years, reducing readmission rates has been a goal of payers and hospitals. In addition, an increasing number of observational studies have suggested an association between respiratory infections and subsequent vascular events such as acute myocardial infarction and stroke [2–8]. However, few data exist about hospital readmissions after laboratory-confirmed influenza hospitalizations. The Centers for Medicare and Medicaid Services (CMS) publically reports 30-day mortality and readmissions for patients that have been admitted with pneumonia. Subsequently, this led to the establishment of the Hospital Readmissions Reduction Program in 2012, which requires CMS to reduce payments for excess readmissions [9]. We studied hospitalizations subsequent to laboratory-confirmed influenza hospitalizations to better determine their frequency and to characterize the patients requiring readmission.

METHODS

Study Design

This study was performed using data collected by the Tennessee (TN) Emerging Infections Program Influenza Surveillance Network (FluSurv-Net) and the TN Hospital Discharge Data System (HDDS). The FluSurv-Net prospectively identifies all laboratory-confirmed influenza hospitalizations among residents of Davidson County, TN (Nashville metropolitan area) and 7 surrounding counties occurring in a total of 18 hospitals. Data collected by trained surveillance personnel included general demographics, underlying medical conditions, influenza vaccine history, and influenza diagnostic testing. The TN HDDS contains claims data from all discharges that occur at any hospital in TN licensed by the TN Department of Health and includes number of hospitalizations and diagnosis codes from each hospitalization. All patients hospitalized during influenza seasons from 2006 through 2016 with a positive influenza laboratory test were identified through FluSurv-Net. Influenza seasons were defined as October 1 through April 30. The 2009 pandemic year was the exception, because influenza hospitalizations in TN began August 1, 2009, thus data collection for that year was adjusted. Laboratory confirmation of influenza infection was defined as a positive rapid antigen, respiratory viral culture, or nucleic acid detection.

Deterministic record linkage methods were used to link records in the FluSurv-Net database to data in the HDDS using various combinations of personal identifiers as linking keys (ie, first and last names, medical record number, date of birth, sex, race, ethnicity, and zip code). Linkages were attempted using 33 different combinations of personal identifiers. After each attempt, matches were manually reviewed to confirm that the FluSurv-Net record was matched to the correct individual in HDDS. After linkage, personal identifiers available in HDDS were used to identify subsequent hospitalizations. The HDDS was queried to identify all, all-cause, subsequent hospital admissions within 30 days and 1 year of the influenza hospitalization. The 2 Veterans Administration (VA) hospitals in the FluSurv-Net surveillance area do not contribute to HDDS; however, records were included in the dataset if the patients were transferred from a VA hospital. Patients who were transferred to another acute care facility or rehabilitation facility were analyzed based upon their last hospital discharge information. Participants were excluded from the analyses if their index hospitalization resulted in death. Observation stay admissions were included in the analyses and were identified through HDDS data.

The International Classification of Diseases, Clinical Modification, Ninth and Tenth Revisions, were used to define primary and secondary diagnoses associated with each hospitalization. The primary diagnosis was defined as the first diagnosis associated with that admission. Hospital discharge codes were classified as cardiovascular, pulmonary, cerebrovascular, geriatric, renal/genital-urinary, infectious diseases, and psychiatric conditions (see Supplementary Material).

In addition to influenza hospitalizations, HDDS was used to generate all hospitalization admissions from 2006 through 2016 with associated all-cause readmissions within 30 days and 1 year. These data were used to generate overall general readmission rates within 30 days and 1 year.

Statistics

Categorical variables were summarized as frequencies and percentages and compared using χ ² tests. Distributions of continuous variables were summarized as medians and interquartile ranges (IQRs). Statistical significance was defined as P < .05. Multivariate logistic regression model was used to determine the association between readmission and risk factors at index admission. Risk factors modeled included age, sex, race, influenza vaccination status, cardiovascular disease, atrial fibrillation, congestive heart failure, chronic kidney disease, lung disease, stroke, diabetes, immunosuppression, hemoglobinopathies, dementia, liver disease, and smoking status. The restricted cubic spline with 4 knots was applied on age and interaction terms between age and vaccination status. A mixed-effects survival model was used to evaluate time from index admission to readmission and included risk factors similar to those included in the multiple logistic regression model. The survival time was measured as the time from the index admission to the first readmission or the time from the previous readmission to the next readmission for all subsequent readmissions. A restricted cubic spline with 3 knots was applied to the age covariate. All analysis was done using R version 3.4.1.

RESULTS

A total of 3166 patients with hospitalizations for laboratory-confirmed influenza from 2006 through 2016 were identified. Of those, 172 (5%) patients were excluded because they were unable to be matched with the HDDS database. A total of 97 (3%) patients died during their index influenza hospitalization and were excluded from analyses (Figure 1). Of the remaining 2897 patients, 1533 (53%) were not readmitted within 1 year; they had a median age of 56 years (IQR, 26–73 years) and 55% were female.

Figure 1.

Flow chart of patient inclusion.

There were 1364 patients (47%) who had subsequent readmissions within 1 year after their index hospitalization; they had a median age of 62 (IQR, 42–78 years) and 61% were female. Of the patients who were readmitted, 88 (7%) died during one of their readmissions; 32 (36% of the deaths) died during their first readmission. Patients 65 years of age and older accounted for 56 deaths (64%), but mortality occurred in all age groups, with the exception of those younger than 6 years of age.

The readmission group was older, predominantly female, and had more comorbidities in comparison to the non-readmitted group (Table 1). Of the comorbidities investigated, a history of cardiovascular disease, congestive heart failure, chronic kidney disease, lung disease, diabetes, stroke, immunocompromised state, hemoglobinopathy, dementia, and liver disease occurred more often in those who were readmitted. Smoking status and race were not statistically significant between the 2 groups. Patients who were readmitted were more likely to have been vaccinated than those who were not (48% vs 37%, P < .01).

Table 1.

Demographics of Patients Hospitalized With Laboratory-Confirmed Influenza

Demographic	No Readmissions (%) N = 1533	Readmissions (%) N = 1364	P Value
Age (mean ± standard deviation)	49 ± 29	58 ± 27	< .01
Age group (years) <2	171 (11)	42 (3)
2–5	62 (4)	17 (1)
6–18	91 (6)	48 (4)
19–49	330 (22)	319 (23)
50–64	317 (21)	319 (23)
65–74	196 (13)	215 (16)
75+	366 (24)	404 (30)
Sex			< .01
Male	695 (45)	535 (39)
Female	838 (55)	829 (61)
Race			< .01
White	1115 (73)	1008 (74)
Black	327 (21)	312 (23)
Hispanic	29 (2)	13 (1)
Other	62 (4)	31 (2)
Cardiovascular disease	498 (32)	668 (40)	< .01
Congestive heart failure	160 (10)	250 (18)	< .01
Atrial fibrillation	18 (1)	11 (1)	.32
Chronic kidney disease	145 (9)	259 (19)	< .01
Lung disease	526 (34)	648 (48)	< .01
Diabetes	257 (17)	343 (25)	< .01
Stroke	117 (8)	171 (13)	< .01
Immunocompromised	128 (8)	206 (15)	< .01
Hemoglobinopathy	25 (2)	54 (4)	< .01
Dementia	86 (6)	110 (8)	< .01
Liver disease	25 (2)	51 (4)	< .01
Smoker	247 (16)	237 (17)	.36
Vaccinated	565 (37)	650 (48)	< .01

A total of 3545 readmissions were observed among the 1364 readmitted patients. The median number of readmissions within 1 year was 2 (range, 1 to 34 readmissions); 740 of the readmitted patients (54%) had more than 1 readmission. Patients age 18 years and older accounted for 92% of the readmissions within 1 year; those aged 65 years and older accounted for 46% (Figure 2A).

Figure 2.

Patients hospitalized with laboratory-confirmed influenza. A, Proportion of readmissions within a year by age. The total number of readmissions in each age category are depicted by the grey bars and the vertical axis on the left. The solid black line and the vertical axis on the right represents the proportion of patients readmitted compared to total number of patients within each individual age group. B, Readmissions from time from index hospitalization. The total number of readmissions from the time of the index hospitalization are depicted by grey bars and the primary vertical axis on the left. The solid black line represents the proportion of patients readmitted within the time frame designated compared to the total number of people readmitted.

There were 469 (13%) readmissions that occurred less than 30 days from index hospitalization in 409 patients (14%) (Figure 2B). Of the patients readmitted within 30 days, 28% had an infectious disease diagnosis, 14% had a pulmonary disease diagnosis, and 11% had a cardiovascular diagnosis as their principal diagnosis. Pneumonia, acute chronic obstructive pulmonary disease (COPD)/asthma exacerbation, septicemia, acute respiratory failure, and acute renal failure were the most frequent principal diagnosis within 30 days. Other diagnoses such as sickle cell crisis, disorders of urethra/urinary tract, acute myocardial infarction, atrial fibrillation, and cellulitis were also seen within 30 days. Likewise, acute COPD/asthma exacerbation, septicemia, and pneumonia were the most frequent within 1 year. However, congestive heart failure exacerbation, respiratory failure, diabetic ketoacidosis, and acute renal failure also were associated with readmissions within 1 year in all age groups. Patients aged 65 years and older accounted for 1380 total readmissions, 215 of which occurred within 30 days among 188 patients (6%). Of patients aged 65 years and older, 27% of readmissions were due to a primary infectious diseases diagnosis, 17% with cardiovascular disease diagnosis, and 14% with pulmonary disease diagnosis within 1 year.

Pre-existing cardiovascular disease (odds ratio [OR], 1.6; 95% confidence interval [CI], 1.30–1.98), lung disease (OR, 1.6; 95% CI, 1.34–1.86), kidney disease (OR, 1.7; 95% CI, 1.34–2.15), diabetes (OR, 1.3; 95% CI, 1.05–1.56), immunosuppression (OR, 1.6; 95% CI, 1.27–2.10), hemoglobinopathy (OR, 2.4; 95% CI, 1.44–4.02), and liver disease (OR, 2.1; 95% CI, 1.23–3.41) were associated with increased risk of readmission (Table 2).

Table 2.

Multivariable Analysis of Hospital Readmission Within 1 Year

Variable	OR	95% CI	P Value
Age	0.7	.50–.96	.04
Male	0.8	.70–.96	.01
Race			.22
Black	1.1	.88–1.30
Hispanic	1.0	.50–2.20
Other	0.6	.40–1.01
CVD	1.6	1.30–1.98	.01
CHF	1.1	.84–1.41	.51
Atrial fibrillation	0.4	.19–.90	.03
Stroke	1.1	.85–1.51	.39
Diabetes	1.3	1.05–1.56	.02
Lung disease	1.6	1.34–1.86	< .01
Kidney disease	1.7	1.33–2.15	< .01
Immunosuppression	1.6	1.27–2.10	< .01
Hemoglobinopathy	2.4	1.44–4.02	< .01
Dementia	1.3	.93–1.81	.12
Liver disease	2.1	1.23–3.41	< .01
Smoker	1.0	.79–1.24	.95

Abbreviations: CI, confidence interval; CVD, cardiovascular disease; CHF, congestive heart failure; OR, odds ratio.

Readmission for all hospitalizations regardless of reason was calculated to compare to the readmission rate associated with an influenza hospitalization. From 2006 to 2016, the proportion of patients readmitted for any cause was 8% at 30 days and 29% at 1 year. In contrast, the proportion of patients diagnosed with influenza who were readmitted at 30 days was 14% and 47% at 1 year (Figure 3A and B).

Figure 3.

Hospital readmissions 2006–2016. A, The total number of readmissions within the TN EIN network for the years 2006–2016 are depicted by the grey bars and the primary vertical axis on the left. The solid black line represents the proportion of patients readmitted within 30 days of their index hospitalization. The dotted black line represents the proportion of patients readmitted within one year of their index hospitalization. B, The total number of influenza-associated readmissions within the TN EIN network for the years 2006–2016 are depicted by the grey bars and the primary vertical axis on the left. The solid black line represents the proportion of patients readmitted within 30 days of their index hospitalization. The dotted black line represents the proportion of patients readmitted within one year of their index hospitalization.

Discussion

Our study found that readmissions within 30 days of an influenza-associated hospitalization were high and were more frequent at both 30 days and 1 year in comparison to overall readmission rates. Our 30-day readmission rate was 14% for all patients, and 47% of patients were readmitted within 1 year of influenza diagnosis. During the same period, among all patients admitted to the same hospitals, the all-cause readmission rate was 8% within 30 days and 29% within the subsequent year. Of the patients readmitted after their hospitalization for influenza, 54% had more than 1 readmission within the subsequent year (range, 1–34 readmissions) and 7% died during one of the readmissions. Recent pediatric studies have found that 30-day readmission rates after an influenza-associated hospitalization range from 4.7% to 10.2%, similar to our 14% [10, 11].

Patients with underlying chronic conditions were more likely to be readmitted than patients without those conditions. The principal diagnoses during the readmission hospitalizations reflected these underlying conditions (eg, cardiovascular disease, chronic renal or lung disease, diabetes, liver disease, stroke, and dementia, among others). Future studies will be needed to determine whether these readmissions were a direct consequence of the previous influenza illness admission or due to underlying health conditions.

Of those readmitted within 30 days, 11% had a cardiovascular diagnosis. An association between influenza and cardiovascular complications has been speculated since the early 1900s [12]. However, only recently has cardiovascular disease been more closely associated with influenza and its immediate aftermath. Kwong et al [13] found a 6 times increased risk of acute myocardial infarction after laboratory-confirmed influenza in the first 3 days after influenza than among controls.

Approximately half (46%) of readmissions occurred among patients aged 65 years and older. As the population continues to age, this will contribute an increasing burden on the healthcare system. Our results suggest that patient comorbidities are likely factors predisposing to readmission and raise the question of whether an influenza infection serious enough to require hospitalization can precipitate an exacerbation of underlying illnesses and a functional decline. An alternative possibility is that influenza is known to cause more serious disease in those who already are frail and that the subsequent readmissions are indicative of that predisposing frailty.

Influenza vaccination was not associated with a lower risk of readmission. It may be that patients with underlying conditions have higher healthcare exposure increasing the likelihood of influenza vaccination. Hence, the underlying conditions may be the reason for readmission. Further studies into influenza vaccination and hospital readmission are needed.

There are limitations to our study. We may have underestimated the number of influenza-related hospitalizations because laboratory testing was based on the provider’s discretion. If readmissions occurred in hospitals located outside the study surveillance counties, they would not have been detected. Finally, influenza vaccination status was obtained through review of hospital medical records, and they may have been incomplete.

Conclusions

Influenza infection results in a large number of hospitalizations annually. When admitted with laboratory-confirmed influenza, there is a high likelihood of readmission within 30 days and 1 year. Patient comorbidities could be an important link to understanding these readmissions. Further studies are needed to elucidate conditions that could be targeted to decrease readmissions and to determine the role of vaccination in prevention of readmission.

Supplementary Data

Supplementary materials are available at The Journal of 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

Financial support. This work was funded by the Centers for Disease Control and Prevention Emerging Infections Program Cooperative Agreement with Tennessee (Grant/Award Number U50CK000198).

Potential conflicts of interest. H. K. T. has received research funding from Sanofi Pasteur and has served on a safety board for Seqirus. W. S. is a member of a Data Safety Monitoring Boards for Merck and Pfizer and has served as a consultant for Roche Diagnostics. 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.

Presented in part: IDWeek 2018, October 3–7, 2018, San Francisco, CA.