Early Oseltamivir After Hospital Admission Is Associated With Shortened Hospitalization: A 5-Year Analysis of Oseltamivir Timing and Clinical Outcomes

Abstract

Background

Neuraminidase inhibitors (NAIs) are the only effective therapy for influenza, but few studies have assessed the impact of early NAI therapy on clinical outcomes or the patient-level factors that determine early NAI delivery in hospitalized patients.

Methods

We conducted a retrospective cohort study of all adults hospitalized in a metropolitan tertiary care hospital with confirmed influenza from April 2009 to March 2014. We performed logistic regression to determine patient-level factors that were associated with early NAI therapy. We analyzed the association of early NAI therapy with hospital lengths of stay (LOS) and in-hospital mortality rates using linear and logistic regression, respectively.

Results

In total, 699 patients were admitted with influenza during the 5 influenza seasons. Of those, 582 (83.4%) received NAI therapy; however, only 26.0% received the first dose within 6 hours of hospitalization (early NAI). Patients with diabetes mellitus or pregnancy were more likely to receive early NAI (P = .01, vs. P < .001 in those without these conditions), as were those reporting fever or myalgias at presentation (P = .002, vs. P = .005 without). Immunosuppressed patients were less likely to receive early NAI (P = .04). Early NAI was associated with shorter hospital LOS (P < .001). No patients died in the early NAI group, compared to 18 deaths in the 399 patients receiving NAI after 6 hours (4.5%) and 4 deaths in the 116 patients not receiving NAI (3.4%).

Conclusions

Over multiple influenza seasons, early NAI therapy was associated with shorter LOS in patients admitted with influenza. This suggests that efforts should focus on facilitating earlier therapy in patients with suspected influenza.

influenza, oseltamivir, neuraminidase inhibitors, hospitalization

(See the Editorial Commentary by de Jong on pages 59–60.)

Influenza viruses are segmented RNA viruses that are responsible for annual epidemics of respiratory illness, resulting in approximately 23000 excess deaths per year in the United States alone [1–3]. Periodic pandemics, such as the A/H1N1pdm09 pandemic, result in significant increases in hospitalization and mortality [4, 5]. Neuraminidase inhibitors (NAIs) are currently the only clinically-effective class of anti-influenza therapy [6, 7].

Meta analyses of all available randomized, controlled trials involving ambulatory patients with confirmed influenza has shown that oseltamivir shortens symptom longevity, decreases hospitalizations, and has few adverse side effects [8, 9]. Existing oseltamivir safety and efficacy data suggest that, among high-risk and hospitalized individuals, there are benefits to initiating antiviral therapy up to and potentially beyond 5 days after symptom onset, with the greatest benefits shown within 48 hours after symptom onset [10]. A comprehensive meta-analysis of hospitalized patients demonstrated that oseltamivir is associated with a mortality benefit in patients with A/H1N1pdm09 [11] and confirmed previous findings that the earlier initiation of therapy, typically defined as initiation within 48 hours of symptom onset, in hospitalized patients is associated with greater clinical benefits [12, 13]. However, less than half of A/H1N1pdm09-hospitalized patients were treated within 48 hours of symptom onset. In comparison, 73–79% of patients received antibacterial therapy, with 93–95% of antibacterial therapy given in the first 1–2 days of admission [4, 14].

Based on this growing, consistent body of evidence, the standard of care for all influenza patients is early initiation of NAI, optimally within 48 hours of symptom onset, with improved outcomes with empiric NAI therapy in hospitalized patients with respiratory symptoms [15]. Patients often present to the hospital over 48 hours after symptom onset. Approaches to minimize the time between symptom onset and therapy in hospitalized patients have not been well studied. In treating sepsis and pneumonia patients, similar links between the timely use of antibiotics and improved outcomes have led to a standard of giving antibiotics within 1 hour and 4–8 hours, respectively, after presenting for care [16, 17]. There is currently a paucity of data about the patient-level clinical factors that are associated with earlier NAI initiation and the impact on clinical outcomes of shorting delays in NAI initiation in hospitalized patients. Therefore, our objectives were to assess the patient-level factors associated with receipt of early NAI therapy among hospitalized patients and to assess the association of early NAI therapy with clinical outcomes over 5 years at a large, academic center.

METHODS

Study Population and Data Collection

After institutional review board approval, we conducted a retrospective cohort study using the Northwestern Medicine Electronic Data Warehouse, an electronic healthcare record repository of all patients admitted to Northwestern Memorial and Prentice Women’s Hospitals. All patients ≥18 years old who were admitted from 1 April 2009 through 31 March 2014 with positive molecular tests for influenza A or B (xTAG Respiratory Virus Panel, Luminex Corp., Austin, TX, and Prodesse ProFlu+, Hologic, Marlborough, MA, respectively) performed at the Northwestern Memorial Hospital laboratory were included for analysis. Cases were limited to patients where influenza was diagnosed within 14 days of hospitalization. Demographics, admission vital signs and laboratory values, and antimicrobial therapy administration were extracted. For patients with incomplete data, charts were manually reviewed, resulting in no cases with missing data. The charts were manually reviewed for ventilatory support, intensive care unit admission, in-hospital mortality, comorbidities, symptoms and symptom onset, co-infections, and disposition.

Exposure, Outcome, and Covariate Definitions

We defubed 3 categories of NAI therapy timing a priori: documentation of NAI receipt <6 hours of hospital admission (early NAI group), NAI receipt from 6 to 24 hours after admission, and NAI delivery >24 hours after admission. In a second, separate analysis, 2 categories of NAI therapy were defined: empiric NAI, defined as documentation of drug delivery either before a positive result was documented or within 6 hours of hospital admission, and non-empiric NAI, defined as any administration of NAI not meeting the definition of empiric NAI. The timing of the first NAI was defined as the documentation of the first dose administration.

For the analysis of the association of patient-level factors with NAI timing, patient-level clinical factors were the exposures, including patient demographics and comorbidities, symptoms and symptom onset, and admission vital signs and white blood cell (WBC) counts. Early NAI therapy was the outcome of interest.

For the analysis of the association of early NAI therapy and empiric NAI with clinical outcomes, early NAI therapy was the exposure. The primary outcomes were hospital length of stay (LOS) and in-hospital mortality. LOS was defined as the difference in time (days) between discharge date and time and admission date and time.

Comorbidities were defined using the Charlson Comorbidity Index [18]. Additional comorbidities, including immunosuppression, pregnancy, and solid organ transplant, were also recorded. Obesity was defined as a body mass index (BMI) of ≥30 kg/m², and morbid obesity was defined as a BMI of ≥40 kg/m². Hypotension was defined as a systolic blood pressure of ≤90 mmHg and/or diastolic blood pressure of ≤60 mmHg. Hypoxemia was defined as oxyhemoglobin saturation of ≤90% on ambient air, or any supplemental oxygen requirement. Fever was defined as a core temperature of ≥100.4° Fahrenheit. Leukocytosis was defined as a WBC count of ≥12000/μL and leukopenia was defined as a WBC count of ≤4000/μL. Tachycardia was defined as a heart rate of ≥90 beats per minute and tachypnea as a respiratory rate of ≥20 breaths per minute.

Statistical Analysis

We performed standard descriptive statistics to summarize key characteristics. Differences in hospital LOS between NAI groups were evaluated using the Wilcoxon Rank Sum test.

Logistic regression was used to determine the independent associations of timing of NAI therapy with in-hospital mortality. Linear regression was used to assess the association of early NAI therapy with LOS. Because hospital LOS was right-skewed, it was log-transformed to satisfy the assumptions of linear regression, with a subsequent normal distribution. For all regression analyses, we first performed univariate tests of association with each outcome of interest. Gender and age were forced into each model. Additional variables were included in each multivariable model if they were associated with their respective outcomes with P ≤ .2, and they were retained in the final model if they remained significantly associated with the outcome at P < .05. Odds ratios (ORs) and relative risks (RRs) were estimated for logistic and linear regressions, respectively, and 95% confidence intervals (CIs) were estimated to evaluate the strengths of the associations.

All analyses were performed using STATA version 14.1 (StataCorp LP, College Station, TX).

RESULTS

Influenza Incidence

In total, 699 unique cases of influenza were identified during the 5-year study period (seasonal incidence range 76–236; see Figure 1). H1N1 predominated in the 2009–10 (89.9%) and 2013–14 (75.9%) seasons, while H3N2 predominated in the 2012–13 (73.3%); mixed viruses were common in other seasons. Few cases had positive results within 6 hours of hospitalization (6.3%); most had positive results 6–24 hours (41.4%) or 24–48 hours (31.9%) after admission.

Figure 1.

Influenza subtype by influenza season.

Patient Characteristics on Admission

Over the 5-year period, the mean age of patients was 55.6 years (standard deviation 18.5 years; Table 1). The majority of patients had a medical comorbidity (75.7%), most commonly a pulmonary (26.9%) or cardiovascular (26.6%) comorbidity, and 38.6% of patients had multiple comorbidities (Supplementary Table 1). Over the 5-year period, 5.6% of patients were pregnant, with the highest admission rate during the 2009–10 influenza season (13.4%).

Table 1.

Demographics, Symptom Duration, and Clinical Findings of Presentation by Influenza Season and Subtype

	Study Year					Influenza Subtype
	2009–10	2010–11	2011–12	2012–13	2013–14	H1N1	H3N2	Untyepable A	B
Number of cases	127	127	76	236	133	340	233	24	106
Demographics
Age at admission	44.8 (±15.2)	54.8 (±19.7)	58.5 (±17.4)	61.2 (±18.6)	55.9 (±16.3)	51.4 (±16.8)	63.3 (±18.0)	37.8 (± 13.3)	56.8 (± 18.7)
Gender (male)	54 (42.5%)	59 (46.5%)	30 (39.5%)	119 (50.4%)	61 (45.9%)	158 (46.5%)	107 (46.5%)	9 (37.5%)	53 (50.5%)
BMI > 30 kg/m2	49 (38.6%)	48 (37.8%)	27 (35.5%)	68 (28.8%)	48 (36.1%)	124 (36.5%)	75 (32.6%)	9 (37.5%)	32 (30.5%)
BMI > 40 kg/m2	14 (11.0%)	15 (11.8%)	13 (17.1%)	13 (5.5%)	12 (9.0%)	33 (9.7%)	17 (7.4%)	3 (12.5%)	14 (13.3%)
Comorbidities
Cardiac	28 (22.0%)	31 (24.4%)	16 (21.1%)	76 (32.2%)	37 (27.8%)	87 (25.6%)	71 (30.9%)	2 (8.3%)	26 (24.8%)
Diabetes mellitus	19 (15.0%)	25 (19.7%)	21 (27.6%)	62 (26.3%)	31 (23.3%)	68 (20.0%)	61 (26.5%)	6 (25.0%)	22 (21.0%)
Immunosuppressed	21 (16.5%)	25 (19.7%)	18 (23.7%)	41 (17.4%)	29 (21.8%)	65 (19.1%)	39 (17.0%)	2 (8.3%)	27 (25.7%)
None	37 (29.1%)	30 (23.6%)	15 (19.7%)	62 (26.3%)	26 (19.5%)	85 (25.0%)	56 (24.3%)	6 (25.0%)	23 (21.9%)
Oncologic	8 (6.3%)	19 (15.0%)	20 (26.3%)	22 (9.3%)	22 (16.5%)	41 (12.1%)	26 (11.3%)	0 (0%)	24 (22.9%)
Organ transplant	7 (5.5%)	5 (3.9%)	5 (6.6%)	15 (6.4%)	9 (6.8%)	21 (6.2%)	13 (5.7%)	2 (8.3%)	5 (4.8%)
Pregnant	17 (13.4%)	5 (3.9%)	4 (5.3%)	11 (4.7%)	2 (2.3%)	23 (6.8%)	7 (3.0%)	7 (29.2%)	2 (1.9%)
Pulmonary	30 (23.6%)	45 (35.4%)	14 (18.4%)	59 (24.6%)	38 (28.6%)	95 (27.9%)	63 (27.4%)	6 (25.0%)	24 (22.9%)
Renal	13 (10.2%)	13 (9.4%)	16 (21.1%)	36 (15.3%)	15 (11.3%)	40 (11.8%)	36 (15.7%)	2 (8.3%)	15 (14.3%)
Rheumatologic	4 (3.1%)	1 (0.8%)	3 (3.9%)	5 (2.1%)	9 (6.8%)	12 (3.5%)	5 (2.2%)	1 (4.2%)	4 (3.8%)
Time from symptom onset to admission
<24 hours	15 (11.8%)	9 (7.1%)	9 (11.8%)	26 (11.0%)	19 (14.3%)	39 (11.5%)	21 (9.1%)	3 (12.5%)	14 (13.3%)
24–48 hours	25 (19.7%)	31 (24.4%)	15 (19.7%)	34 (14.4%)	18 (13.5%)	55 (16.2%)	40 (17.4%)	7 (29.2%)	20 (19.0%)
48–72 hours	19 (15.0%)	18 (14.2%)	11 (14.5%)	40 (16.9%)	21 (15.8%)	56 (16.5%)	38 (16.5%)	5 (20.8%)	11 (10.5%)
>72 hours	68 (53.5%)	69 (54.3%)	41 (53.9%)	136 (57.6%)	75 (56.4%)	190 (55.9%)	131 (57.0%)	9 (37.5%)	60 (57.1%)
Clinical findings on presentation
Fever	72 (56.7%)	60 (47.2%)	39 (51.3%)	112 (47.5%)	47 (35.3%)	155 (45.6%)	110 (47.8%)	14 (58.3%)	52 (49.5%)
Hypotension	26 (20.5%)	32 (25.2%)	20 (26.3%)	51 (21.6%)	20 (15.0%)	70 (20.6%)	48 (20.9%)	5 (20.8%)	25 (23.8%)
Hypoxemia	42 (33.1%)	46 (36.2%)	27 (35.5%)	94 (49.8%)	56 (42.1%)	128 (37.6%)	89 (38.7%)	7 (29.2%)	40 (38.1%)
Leukocytosis/leukopenia	29 (22.8%)	40 (31.5%)	30 (39.5%)	54 (22.9%)	41 (30.8%)	94 (27.6%)	60 (26.1%)	4 (16.7%)	36 (34.3%)
Tachycardia	97 (76.4%)	95 (74.8%)	52 (68.4%)	152 (64.4%)	93 (69.9%)	245 (72.1%)	153 (66.5%)	20 (83.3%)	71 (67.6%)
Tachypnea	76 (59.8%)	64 (50.4%)	41 (53.9%)	116 (49.2%)	69 (51.9%)	186 (54.7%)	112 (48.7%)	14 (58.3%)	56 (53.3%)

Age at admission is presented as mean age (±standard deviation). All other data are presented as number of subjects (%).

Abbreviation: BMI, body mass index.

Most patients (55.8%) presented to the hospital over 72 hours after symptom onset (median 3.6 days, interquartile range [IQR] 1.8–5.4 days). Only 11.0% of patients presented within 24 hours of symptom onset. Cough (79.3%) and fever (64.8%) were the most common presenting symptoms (Supplementary Table 2). On admission, 47.4% of patients had a fever, 37.8% of patients were hypoxemic, and 21.2% were hypotensive.

Antimicrobial Therapy

Most patients (83.4%) received NAI therapy, with seasonal rates ranging from 78.7% in 2010–11 to 86.8% in 2011–12 (Supplementary Table 3). The median time from admission to the first dose of NAI was 17.9 hours (IQR 4.0–33.9 hours), with only 26.0% of patients receiving the first dose within 6 hours of hospital admission and 32.6% of patients receiving the first dose after more than 24 hours of hospitalization. Empiric NAI was given to 52.2% of the patients who received NAI during hospitalization, with a median time to first dose of 4.32 hours (IQR -0.62–11.4 hours), compared to 33.5 hours (IQR 23.2–54.0 hours) in non-empiric NAI.

Univariate and multivariable analyses of factors associated with the receipt of early NAI are outlined in Table 2. In a multivariable analysis, diabetes mellitus (OR 1.8, 95% CI 1.1–3.0; P = .01), pregnancy (OR 5.1, 95% CI 2.3–11.5; P < .001), and symptoms of fever (OR 1.9, 95% CI 1.2–3.1; P = .005) and myalgias (OR 2.0, 95% CI 1.3–3.2; P = .002) were associated with the receipt of early NAI. Tachypnea was associated with early NAI (OR 1.0, 95% CI 1.0–1.1; P = .047). Immunosuppression was negatively associated with early NAI (OR 0.6, 95% CI 0.3–1.0; P = .04), with only 16.5% of immunosuppressed patients receiving early NAI (see Supplementary Table 4). Surprisingly, presentation within 24 hours of symptom onset (P = .018) was associated with the receipt of non-empiric NAI and a delay in the initiation of NAI (Supplementary Table 5).

Table 2.

Univariate and Multivariable Analysis of Factors on Admission Associated With the Receipt of Early Neuraminidase Inhibitor Therapy

	Univariate Analysis		Multivariable Analysis
	OR (95% CI)	P Value	OR (95% CI)	P Value
Influenza Season (2009–10 reference)
2010–11	0.4 (0.2–0.8)	.01	0.6 (0.3–1.3)	.18
2011–12	1.4 (0.7–2.5)	.34	2.2 (1.0–4.8)	.05
2012–13	1.2 (0.8–2.0)	.38	1.4 (0.7–2.9)	.36
2013–14	0.8 (0.5–1.5)	.52	1.4 (0.7–2.7)	.32
Influenza subtype (H1N1 reference)
H3N2	1.6 (1.1–2.4)	.01	1.6 (0.9–3.0)	.10
Untypeable A	4.0 (1.7–9.4)	<.01	2.2 (0.8–5.6)	.11
B	1.0 (0.6–1.7)	.96	1.1 (0.6–2.1)	.78
Demographics
Gender (male)	1.4 (1.0–2.0)	.05	1.1 (0.8–1.7)	.58
BMI	1.0 (1.0–1.0)	.13	1.0 (1.0–1.0)	.54
Comorbidities
Diabetes mellitus	1.5 (1.0- 2.2)	.05	1.8 (1.1–3.0)	.01
Immunosuppressed	0.5 (0.3–0.8)	.006	0.6 (0.3–1.0)	.04
Pregnant	4.4 (2.2–8.7)	<.001	5.1 (2.3–11.5)	<.001
Symptoms
Arthralgias	1.9 (0.7–4.6)	.19	1.4 (0.5–4.1)	.52
Chills	1.4 (1.0–2.0)	.08	1.1 (0.7–1.7)	.67
Cough	1.5 (1.0–2.4)	.08	1.2 (0.7–2.0)	.49
Fever	2.1 (1.4–3.1)	<.001	1.9 (1.2–3.1)	.005
Myalgias	2.3 (1.6–3.4)	<.001	2.0 (1.3–3.2)	.002
Nasal congestion	1.4 (0.8–2.4)	.19	1.3 (0.7–2.3)	.36
Sore throat	1.4 (0.9–2.2)	.16	1.1 (0.6–1.8)	.85
Clinical findings on presentation
Febrile/hypothermic	1.6 (1.1–2.3)	<.01	1.4 (0.9–2.0)	.13
Heart rate	1.0 (1.0–1.0)	.03	1.0 (1.0–1.0)	.35
Leukocytosis/leukopenia	0.7 (0.5–1.0)	.07	0.9 (0.6–1.3)	.50
Respiratory rate	1.0 (1.00–1.1)	.04	1.0 (1.00–1.1)	.05
Systolic blood pressure	1.0 (1.0–1.0)	.09	1.0 (1.0–1.0)	.04

Abbreviations: BMI, body mass index; CI, confidence interval; OR, odds ratio.

Antibacterial therapy was given to 80.1% of patients, with 66.0% of patients receiving the first dose within 6 hours of hospitalization.

Clinical Outcomes

The median LOS over the 5 influenza seasons was 3.6 days (IQR 2.2–6.6 days; Supplementary Table 6). The median LOS was shorter for those receiving early NAI (2.8 days, IQR 1.8–4.1 days) than those receiving NAI between 6 and 24 hours (3.9 days, IQR 2.1–6.6 days) or for those who received NAI after 24 hours (5.6 days, IQR 3.6–11.0 days; P = .0002 and P < .00001, respectively). In a multivariable analysis, cardiac comorbidity (RR 1.4, 95% CI 1.2–1.7; P < .001), oncologic comorbidity (RR 1.6, 95% CI 1.3–2.0; P < .001), and hypoxemia on admission (RR 1.3, 95% CI 1.2–1.5; P < .001) were associated with longer LOS. Early NAI was associated with shorter LOS than NAI from 6–24 hours (RR 1.4, 95% CI 1.2–1.7; P < .001) or over 24 hours (RR 2.3, 95% CI 1.9–2.7; P < .001; Table 3). Empiric NAI was associated shorter median LOS (3.2 days, IQR 2.0–5.4), as compared to non-empiric NAI (5.0 days, IQR 3.0–9.2; P < .001), in univariate and multivariable analyses (P < .001; Supplementary Table 7).

Table 3.

Univariate and Multivariable Analysis of Factors Associated With Length of Stay

	Univariate Analysis		Multivariable Analysis
	RR (95% CI)	P Value	RR (95% CI)	P Value
Influenza subtype (H1N1 reference)
H3N2	0.9 (0.9–1.1)	.57	0.98 (0.84–1.14)	.82
Untypeable A	0.7 (0.5–1.1)	.13	1.15 (0.81–1.63)	.44
B	1.0 (0.8–1.2)	.79	0.9 (0.7–1.1)	.19
Demographics
Age at admission	1.0 (1.0–1.0)	.03	1.0 (1.0–1.0)	.86
Gender (male)	0.9 (0.8–1.0)	.07	0.9 (0.8–1.1)	.34
Comorbidities
Cardiac	1.4 (1.2–1.7)	<.001	1.4 (1.2–1.7)	<.001
Immunosuppressed	1.2 (1.0–1.4)	.10	0.9 (0.8–1.1)	.52
None	0.7 (0.2–0.9)	<.001	0.9 (0.8–1.1)	.26
Oncologic	1.7 (1.4–2.1)	<.001	1.6 (1.3–2.0)	<.001
Pregnant	0.7 (0.5–0.9)	.20	1.1 (0.9–1.5)	.68
Clinical findings on presentation
Febrile/hypothermic	0.8 (0.7–1.0)	.03	0.9 (0.8–1.1)	.40
Leukocytosis/leukopenia	1.2 (1.0–1.4)	.09	1.0 (0.9–1.2)	.92
Heart rate	1.0 (1.0–1.0)	.05	1.0 (1.0–1.0)	.93
Systolic blood pressure	1.0 (1.0–1.0)	.10	1.0 (1.0–1.0)	.44
Hypoxemia	1.4 (1.2–1.6)	<.001	1.3 (1.2–1.5)	<.001
Time to NAI (<6 hours reference)
6–24 hours	1.5 (1.2–1.8)	<.001	1.4 (1.2–1.7)	<.001
>24 hours	2.3 (2.0–2.7)	<.001	2.3 (1.9–2.7)	<.001

Abbreviations: BMI, body mass index; CI, confidence interval; NAI, neuraminidase inhibitor therapy; RR, relative risk.

There were 22 patients (3.1%) who died during hospitalization. Yearly in-hospital mortality rates ranged from 1.3% in 2012–13 to 7.1% in 2009–10. The majority (68.2%) of the deaths occurred in patients infected with H1N1 (Supplementary Table 3). Of the 22 deaths, 4 occurred in the 116 patients (3.4%) who did not receive NAI therapy during hospitalization. No deaths occurred in the patients who received NAI within 6 hours of admission, while 3 deaths (1.7%) occurred in patients who received NAI from 6–24 hours and 15 deaths occurred in the 228 (6.6%) patients receiving NAI over 24 hours after admission. In a multivariable analysis, admission during either the 2012–13 (OR 0.05, 95% CI 0.01–0.4; P = .003) or 2013–14 (OR 0.1, 95% CI 0.02–0.8; P = .03) influenza season was associated with a decreased risk of an in-hospital death, as compared to admission in the 2009–10 influenza season (Table 4). An increased BMI (OR 0.8, 95% CI 0.7–1.0; P = .008) was associated with a lower risk of in-hospital mortality. NAI given <6 hours from admission was omitted from the multivariable model, due to predicting survival perfectly. Older age (OR 1.1, 95% CI 1.0–1.1; P = .005) was associated with an increased risk of in-hospital mortality, and NAI given >24 hours from admission was omitted from the model due to collinearity. Empiric NAI therapy was associated with a lower risk of in-hospital mortality in a multivariable analysis (P = .026; 3 [1.0%] deaths with empiric therapy vs. 16 [5.7%] with non-empirc therapy; see Supplementary Table 8).

Table 4.

Univariate and Multivariable Analysis of Factors Associated With Mortality

	Univariate Analysis		Multivariable Analysis
	OR (95% CI)	P Value	OR (95% CI)	P Value
Influenza season (2009–10 reference)
2010–11	0.5 (0.1–1.8)	.27	0.2 (0.4–1.3)	.10
2011–12	0.7 (0.2–2.8)	.63	0.3 (0.6–1.9)	.21
2012–13	0.2 (0.3–0.8)	.02	0.1 (0.0–0.4)	.003
2013–14	0.4 (0.1–1.6)	.20	0.1 (0.0–0.8)	.03
Demographics
Age at admission	1.0 (1.0–1.0)	.12	1.1 (1.0–1.1)	.005
Gender (male)	0.4 (0.2–1.1)	.09	0.7 (0.2–2.1)	.49
BMI	0.9 (0.8–1.0)	.003	0.8 (0.7–1.0)	.008
Comorbidities
Immunosuppressed	3.4 (1.3–8.7)	.01	2.0 (0.6–7.1)	.27
Oncologic	4.0 (1.5–10.9)	.005	2.0 (0.5–8.3)	.35
Pulmonary	0.3 (0.1–1.4)	.13	0.6 (0.103.4)	.55
Renal	2.5 (0.9–7.3)	.09	2.4 (0.6–9.1)	.21
Symptoms
Time from symptom onset	0.9 (0.8–1.0)	.02	0.9 (0.8–1.1)	.18
Clinical findings on presentation
Febrile/hypothermic	2.1 (0.8–5.8)	.13	2.0 (0.6–6.4)	.23
Leukocytosis/leukopenia	2.6 (1.0–6.7)	.05	3.2 (1.0–10.5)	.06
Time to NAIa
6–24 hours	0.3 (0.1–0.9)	.03	0.2 (0.6–1.0)	.05

Abbreviations: BMI, body mass index; CI, confidence interval; OR, odds ratio; NAI, neuraminidase inhibitor therapy.

^aTimes <6 hours were omitted from the model due to predicting survival perfectly. Times >24 hours were omitted from the model due to collinearity.

DISCUSSION

In this large, retrospective study of hospitalized patients, the administration of NAI either within 6 hours of hospital admission or given empirically was independently associated with a decreased LOS. Empiric NAI therapy was also associated with a lower risk of in-hospital mortality. Previous papers have defined early therapy as the initiation of NAI within 2 days of symptom onset; however, only 30–40% of patients during the A/H1N1pdm09 season received NAI during this window [11, 12]. We found that only 28.9% of patients over the 5-year study period were admitted within 2 days of symptom onset, and that 16.0% of patients met the conventional definition of receipt of early therapy. A recently-published work showed that empiric therapy delivered prior to an influenza diagnosis was associated with improved outcomes [15]. We are the first to analyze the timing of therapy as defined by the time from hospital admission to the initiation of antimicrobial therapy, a definition based on the management of the hospitalized patient and not the time to the hospital presentation.

Previous papers have shown that 50–70% of patients received NAI therapy within 24 hours of hospitalization during the 2009–10 A/H1N1pdm09 season, but they did not assess the admission factors associated with NAI timing [12, 19]. We found that patients reporting the classic influenza symptoms of fever and myalgias were more likely to receive early therapy. While pregnant patients—a population at risk for severe influenza, with significant public health awareness regarding influenza [20]—were more likely to receive early therapy, we were surprised to find that immunosuppressed patients were less likely to receive early therapy. Whereas the vast majority of patients received antibacterial therapy within 6 hours of hospitalization, less than a third received NAI in this time frame. The early initiation of antibacterial therapy in sepsis has become widely accepted as the standard of care [16]. Our findings show a delay in initiating early empiric therapy with NAIs, which has not garnered the same policy attention as early antibacterials for sepsis. This may also reflect the lack of clear guidance advocating for preemptive anti-influenza antivirals for patients with community-acquired pneumonia, despite increasing evidence suggesting that viruses, including influenza, are a major cause of pneumonia [21].

Despite the Centers for Disease Control’s recommendation that all patients with proven influenza receive NAI unless a contraindication exists, we found that 16% of patients did not ever receive NAI therapy. While we did not assess how many of these patients had contraindications to receiving NAIs, these patients presented, on average, more than 5 days after symptom onset, so it is possible that clinicians did not feel that therapy was indicated or beneficial. We found a consistent rate of NAI prescribing over the 5-year period, ranging from 78.9–87.4%, which differed from other reports, which found a decrease among hospitalized patients in the years following 2009–10, down to 65.7% [22].

A strength of this study is the inclusion of the 2009–10 A/H1N1pdm09 and 4 subsequent influenza seasons, thus adding to a literature largely based on the 2009–10 season and A/H1N1pdm09. Further, we were able to collect comprehensive, patient-level, clinical data without any missing data points, allowing us to analyze clinical decision-making factors. Limitations include the single health system experience and the retrospective analysis of NAI timing on clinical outcomes, with the potential for residual confounding.

In conclusion, we have identified poor adherence to recommendations for early and empiric initiation of NAI therapy for hospitalized patients with influenza, and found that over multiple influenza seasons, early NAI therapy was associated with decreased hospital LOS and mortality. Our findings suggest that efforts should be focused on facilitating earlier initiation of therapy for the management of severe influenza.

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

Author contributions. The study concept and design were developed by J. L. H., M. G. I., and J. K. All authors participated in the acquisition, analysis, or interpretation of data. The manuscript was drafted by M. G. I., J. K., and R. K. All authors participated in the critical revision of the manuscript for important intellectual content. Statistical analyses were performed by J. K. and R. K. Funding was secured by M. G. I. Administrative, technical, or material support was provided by M. G. I. and J. L. H. Supervision was provided by M. G. I.

Disclaimer. M. G. I. and J. K. had full access to all of the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.

Potential conflicts of interest. M. G. I.’s institution received research support from Astellas, Beckman Coulter, Chimerix, Emergent BioSolutions, Gilead, Jansen/Johnson, and Johnson and Shire. M. G. I. is a paid consultant for Celltrion, Genentech/Roche, Toyama/MediVector, Seqirus, Shionogi, and VirBio; an unpaid consultant for GlaxoSmithKlein, Romark, and Vertex; and a paid member of data and safety monitor boards related to research activity conducted by GlaxoSmithKlein and Shionogi. 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.