Repeated influenza vaccination and hospitalization outcomes among older patients with cardiovascular or respiratory diseases

ABSTRACT

Influenza vaccination in a single season protects against hospitalization outcomes among older adults hospitalized for cardiovascular or respiratory diseases, but the effectiveness of repeated influenza vaccination is less clear. Four hospitalization outcomes (in-hospital death, re-admission, length of stay, and direct medical costs) were extracted from the Beijing Urban Employee Basic Medical Insurance database in 2015–2016 for adults aged ≥60 years hospitalized for cardiovascular or respiratory diseases. Vaccination status during three influenza seasons (2013/2014-2015/2016) was ascertained through linkages to the Beijing Elderly Influenza Vaccination database. The summer months (June–August) were used as a reference period to control unmeasured confounders during the influenza season. There were 99,135 periods of observation in the analysis, with 8.3% participants receiving influenza vaccination in all three seasons. After adjusting for confounders, influenza vaccination in all three seasons was associated with a lower risk of re-admission among patients with cardiovascular diseases (odds ratio 0.71 [95% CI 0.53–0.96]) and a lower risk of death among patients with respiratory diseases (0.68 [0.46–0.98]) compared with those unvaccinated in any season. Among patients with cardiovascular diseases, influenza vaccination in all three seasons was also associated with a non-significant lower risk of death (0.66 [0.44–1.03]) in addition to shorter hospital stays and lower direct medical costs. When stratified by history of vaccination, the effectiveness of current season vaccination was similar among patients with cardiovascular or respiratory diseases (p-value for heterogeneity all >0.05). Repeated influenza vaccination protected against hospitalization outcomes among older adults with cardiovascular or respiratory diseases.

Plain language summary

• We found that influenza vaccination in three seasons (2013–2016) was associated with lower risk of in-hospital death and re-admission among older adults hospitalized for cardiovascular disease and with lower risk of in-hospital death among those hospitalized for respiratory disease. In addition to the effectiveness of influenza vaccination on preventing influenza itself, this study suggested that there were protective effects of repeated influenza vaccination on cardiovascular or respiratory diseases.

KEYWORDS:

• Repeated influenza vaccination
• older adults
• cardiovascular disease
• respiratory disease
• hospitalization

Background

The World Health Organization (WHO) recommends annual influenza vaccination for everyone six months and older.¹ Adults ≥60 years and those with chronic conditions (e.g. cardiovascular disease [CVD], diabetes) are at higher risk of death when contracting influenza, and therefore these two groups are the targets of influenza vaccination.² In 2007, Beijing became the first municipal government in China to promote a “free influenza vaccine” policy.³ Our previous study has reported vaccine coverage rates between 17.9% and 22.4% during 2013–2015 influenza seasons among adults ≥60 years in Beijing.⁴ A survey in Beijing showed that the predominant barrier to influenza vaccination was the perception of not expecting to contract influenza.⁵ Our previous study has shown that influenza vaccination was associated with lower risks of in-hospital death and re-admission among older adults hospitalized for CVD and a lower risk of in-hospital death among older adults hospitalized for respiratory diseases.⁶

Despite the concern about the negative impact of serial influenza vaccination on vaccine effectiveness, several previous studies have suggested protective effects of repeated influenza vaccination.^7–10 A meta-analysis involving 20 studies has shown that the vaccination effectiveness for vaccination in the current season only was higher compared with no vaccination in either the current or previous season for all subtypes of influenza.⁷ A case–control study in Spain including 2554 older adults has reported that repeated vaccination was associated with lower risks of influenza and in-hospital deaths.⁸ However, there is limited evidence on the effectiveness of repeated influenza vaccination among older adults with major chronic diseases.

Therefore, we examined the effect of repeated vaccination for influenza on hospitalization outcomes among older adults admitted for cardiovascular and respiratory diseases, including in-hospital death, readmission, length of stay, and direct medical costs. We also assessed the effect of current season vaccination by the history of vaccination on hospitalization outcomes among older adults admitted for cardiovascular and respiratory diseases.

Materials and methods

Study population

We extracted hospitalization outcomes among older adults admitted for cardiovascular and respiratory diseases from the Beijing Urban Employee Basic Medical Insurance (UEBMI). The UEBMI provides comprehensive healthcare for employees and retirees from enterprises and public institutions and includes hospital records of 14 million people in Beijing (more than 60% of the total population).¹¹ Individuals were included in the study if they were 60 years or older and were hospitalized with a primary discharge diagnosis of cardiovascular (ICD-10 code, I20-I25 and I60-I69) or respiratory diseases (J00-J99). These diseases were selected because previous studies have shown that influenza vaccination has protective effects on these diseases.^12,13 Outcomes were included if they occurred during the influenza season or within two weeks after its end. In the current study, 97% participants in the UEBMI database had complete data for analysis.

Influenza vaccination

We ascertained annual status of influenza vaccination for hospitalized older adults by linkages to the Beijing Elderly Influenza Vaccination (EIV) database. The Beijing EIV database was launched by the Center for Disease Control and Prevention (CDC) in Beijing in 2013 and records adults aged 60 years and older who have received influenza vaccines in Beijing since 2013.⁴ In the current study, 98% participants in the EIV database had complete data for analysis. The linkage between UEBMI and EIV was done via the unique national identifier (complete for 99% participants) and the data was delivered to researchers at Peking University without individual identifiers. This study was exempted from the Institutional Review Board (IRB) approved by the Ethics Committee of Beijing Center for Disease Control and Prevention. The need for informed consent was also waived by the IRB.

According to influenza surveillance data from Beijing CDC, we defined influenza season as October 1 to March 31. In order to examine the effect of repeated vaccination on hospitalization outcomes in the current season, we used the EIV database for three continuous influenza seasons from October 1, 2013 to September 30, 2016. We used the UEBMI database from October 1, 2015 to September 30, 2016 to match the influenza season. Within the current influenza season (October 1, 2015 to September 30, 2016), participants were considered immunized if they received the vaccine at least 14 days before hospitalization, while those who were unvaccinated or received the vaccine <14 days before hospitalization were considered as unvaccinated. To assess vaccination history, vaccination status in the previous two influenza seasons (October 1, 2013 to September 30, 2015) was ascertained regardless of the timing of vaccination.

From October 1, 2015 to September 30, 2016, trivalent influenza split vaccine was administered to 99.998% of older adults, with only 9 individuals taking trivalent influenza subunit vaccine.

Patients were classified into five categories of repeated vaccination according to influenza vaccination status in the current and two previous influenza seasons: (1) unvaccinated in the current and previous two seasons (none, reference); (2) vaccinated in the current season only (current only); (3) vaccinated in the current and previous season and unvaccinated in the penultimate season (current and previous); (4) vaccinated in the previous and penultimate seasons and unvaccinated in the current season (penultimate and previous); and (5) vaccinated in the current and previous two seasons (all three).

In total, 99,135 periods of observation were included in the current study, with 8.3% of participants receiving influenza vaccination in all three seasons. There were 62,715 periods of observation for CVD patients and 36,420 periods of observation for patients with respiratory diseases.

Hospitalization outcomes

We extracted four primary outcomes from the UEBMI database: (1) in-hospital death: defined as death from any cause; (2) hospital re-admission: defined as re-admission for the same disease diagnosis within 14 days of discharge, which is used as an indicator of quality of care; (3) the length of hospitalization; and (4) direct medical costs: defined as single-time medical costs including all the expenditures on medications, medical consumables, physical and biochemical tests, non-medication treatments, surgeries, bed fees, and other services.

Statistical analysis

We compared the baseline characteristics among participants by categories of repeated vaccination using chi-square tests for categorical variables and Student’s t-tests for continuous variables. The vaccination rate for each category of repeated vaccination was calculated among patients hospitalized for CVD and respiratory diseases separately as the number of vaccinated patients divided by the total number of patients.

We examined the effect of repeated vaccination on hospitalization outcomes using multivariate logistic regression analysis for in-hospital death and re-admission and using multivariate linear regression for length of hospital stay and direct medical costs. We log-transformed direct medical costs because of the skewed distribution. We exponentiated the regression coefficients for log-transformed direct medical costs to obtain adjusted ratio of geometric means (geometric means ratio, GMR).

The primary analysis was conducted in influenza seasons. Participants could be hospitalized more than once during a single influenza season, and thus, each participant could contribute more than one period of observation to the study. We used the generalized estimating equations (GEE) model to estimate the influence of within-person dependency with the independent correlation structure. To address confounding, we constructed a propensity score from several demographic and clinical factors that may be associated with a decision to vaccinate, including age, age squared, sex, prior healthcare use (i.e. number of inpatient visits in the past 12 months), Charlson Comorbidity Index (CCI), surgical operation status (defined as whether the participant received operation during the hospitalization), hospital level (tertiary or secondary), area of residence, alcohol- and obesity-related diagnoses. The propensity score can be used to reduce or eliminate selection bias in observational studies by balancing covariates (i.e. the characteristics of participants) between treated and control groups. When the covariates are balanced, it become much easier to match participants with multiple characteristics. The CCI included 17 different groups of diseases (Supplementary Table 1). This CCI predicts the ten-year survival, therefore informing the clinicians about how the comorbidities the patient is suffering from impact their life expectancy, thus helping them decide upon the short- and long-term benefits of the treatment they are about to recommend.¹⁴ Alcohol- and obesity-related diagnoses were used as proxy measures of healthy living because lifestyle factors were unavailable in medical records. Alcohol-related diagnoses were mental and behavioral disorders due to use of alcohol (ICD10: F10), alcoholic liver disease (K70), alcoholic polyneuropathy (G62.1), alcoholic cardiomyopathy (I42.6), alcohol-induced acute pancreatitis (K85.2), alcohol-induced chronic pancreatitis (K86.0). Obesity-related diagnoses were type 2 diabetes (E11), nonalcoholic fatty liver disease (K76.0), and dyslipidemia (E78.5). Propensity score was constructed separately for CVDs and respiratory diseases. The effect estimates of repeated vaccination was adjusted for age, age squared, sex, prior healthcare use, CCI, surgical operation status, hospital level, area of residence, alcohol- and obesity-related diagnoses and stratified by quintile of propensity score.

We calculated effect estimates in influenza seasons and in summer months (June-August), separately. Summer months were used to adjust the effect estimates in influenza seasons for unmeasured confounding. As the influenza vaccine was not expected to be effective during summer months (expected OR 1.0),^15,16 deviations of the associations in summer months from one were used to quantify bias arising from unmeasured confounding. We adjusted effect estimates in influenza seasons for unmeasured confounding in summer months with the formulas below:

1. for in-hospital death and re-admission: OR_adjusted = OR_{influenza season}/OR_summer = exp(β_{influenza season –} β_summer),

2. for length of stay and log-transformed direct medical costs: β_adjusted = β_{influenza season –} β_summer.

where β denotes the regression coefficient for influenza vaccination. To estimate a 95% CI of the estimate (ratio of ORs or difference in βs), we sampled 1000 times from the distributions of effect estimates for influenza seasons and summer months separately using bootstrap techniques. From the 1000 pairs of effect estimates (1000 for influenza seasons and 1000 for summer months), we calculated the ratio (for in-hospital death and re-admission) or difference (for length of stay and log-transformed direct medical costs) of the two sampled numbers each time. The 2.5% and 97.5% percentiles of this distribution indicated the lower and upper bound of the 95% CI of the estimate, respectively.

Results

Baseline characteristics by status of repeated vaccination

For CVD, there were 41,363 and 21,352 periods of observation (i.e. records of hospitalization) in the influenza seasons and summer months, respectively. During the influenza seasons, 1188 patients with CVD died, and 3892 were re-admitted. During the summer months, 513 patients with CVD died, and 2051 were re-admitted. For respiratory diseases, there were 24,639 and 11,781 periods of observation (i.e. records of hospitalization) in the influenza seasons and summer months, respectively. During the influenza seasons, 2399 patients with respiratory diseases died, and 3111 were re-admitted. During the summer months, 1049 patients with respiratory diseases died, and 1564 were re-admitted.

The vaccination rates among patients with CVD were 1.8% for the current season only, 1.7% for the current and previous seasons, 3.9% for the penultimate and previous seasons, and 8.6% for all three seasons. The corresponding vaccination rates among patients with respiratory diseases were 1.6%, 1.4%, 4.6%, and 7.9%. For CVD patients, participants vaccinated in all three seasons tended to be older and male, had fewer inpatient visits during the past 12 months and more surgical operation, and were more likely to have obesity-related diagnosis (Table 1). For respiratory diseases patients, participants vaccinated in all three seasons tended to be male, had fewer inpatient visits during the past 12 months and fewer surgical operation, and were more likely to have an obesity-related diagnosis (Table 1).

Table 1. Patients’ characteristics by history of repeated vaccination

	Vaccination history
	None		Current only		Current & previous		Penultimate & previous		All three
Cardiovascular diseases
Periods of observation	77,741		1768		1707		3911		8667
Age (year)	72.8 (8.8)		71.4 (7.9)		73.5 (7.8)		75.6 (6.8)		75.1 (6.5)*
Female, %	43.3		38.3		42.7		42.1		40.4*
Tertiary hospital, %	64.9		64.3		64.9		66.9		65.2
Inpatient visit	3.4 (4.4)		2.5 (2.6)		2.4 (2.5)		3.3 (3.9)		2.4 (2.5) *
CCI	1.6 (1.0)		1.5 (1.0)		1.5 (1.0)		1.6 (1.0)		1.5 (1.0)
Surgical operation, %	30.7		36.1		32.9		30.2		33.7*
Alcohol-related diagnoses, %	0.09		0		0.10		0.10		0.10
Obesity-related diagnoses, %	63.7		68.6		67.6		62.4		65.2*
Respiratory diseases
Periods of observation	44,767		931		823		2635		4579
Age (year)	76.9 (9.1)		74.4 (9.0)		76.1 (8.5)		78.2 (7.1)		77.3 (6.9)
Female, %	45.2		36.7		43.1		36.6		39.1*
Tertiary hospital, %	64.0		69.0		65.4		64.1		64.5
Inpatient visit	4.2 (5.8)		2.7 (2.5)		2.3 (2.2)		4.8 (6.1)		2.6 (2.6) *
CCI	1.6 (1.0)		1.4 (1.0)		1.5 (0.9)		1.5 (0.9)		1.5 (0.9)
Surgical operation, %	15.9		15.4		14.0		15.5		13.9*
Alcohol-related diagnoses, %	0.2		0		0.1		0.1		0.07*
Obesity-related diagnoses, %	30.7		34.3		35.8		31.3		36.2*

Values are mean (SD) for continuous variables or proportions for categorical variables.

Abbreviations: CCI, Charlson Comorbidity Index.

*For all variables, p-values are <0.05 for differences across 5 repeated vaccination categories. An asterisk denotes that p-value is <0.05 for the difference between patients vaccinated in all three seasons and in none of the three seasons.

Effect of repeated vaccination on in-hospital death and re-admission

For CVD, vaccination in all three seasons was associated with lower risks of in-hospital death and re-admission during the influenza season, compared with unvaccinated in any season (Table 2). There were no associations of other vaccination categories with risks of in-hospital death and re-admission during the influenza seasons. After adjustment for unmeasured confounders with the summer months as the reference, vaccination in all three seasons was associated with a non-significantly lower risk of in-hospital death and a significantly lower risk of re-admission (ratio of ORs 0.66 [0.44, 1.03] and 0.71 [0.53, 0.96]).

Table 2. The effectiveness of repeated vaccination among patients hospitalized for cardiovascular diseases

Outcome, metric	Current only	Current & previous	Penultimate & previous	All three
	Effect size (95% CI)	Effect size (95% CI)	Effect size (95% CI)	Effect size (95% CI)
In-hospital death, OR
Influenza seasons	0.76 (0.43, 1.34)	0.62 (0.36, 1.07)	0.79 (0.59, 1.06)	0.48 (0.36, 0.64)
Summer months	0.47 (0.19, 1.15)	0.56 (0.25, 1.27)	0.44 (0.22, 0.91)	0.73 (0.52, 1.01)
Ratio of ORs	1.71 (0.60, 8.32)	1.17 (0.42, 5.64)	1.77 (0.92, 4.96)	0.66 (0.44, 1.03)
Re-admission, OR
Influenza seasons	0.94 (0.62, 1.42)	0.77 (0.44, 1.35)	1.06 (0.85, 1.32)	0.78 (0.61, 1.01)
Summer months	0.92 (0.58, 1.47)	1.17 (0.70, 1.95)	0.87 (0.57, 1.32)	1.09 (0.86, 1.38)
Ratio of ORs	0.99 (0.55, 1.86)	0.65 (0.36, 1.20)	1.23 (0.81, 1.83)	0.71 (0.53, 0.96)
Length of hospital stay, β
Influenza seasons	−1.14 (−1.68, −0.61)	−1.41 (−1.97, −0.85)	−0.43 (−0.95, 0.08)	−1.33 (−1.61, −1.04)
Summer months	−0.66 (−1.26, −0.06)	−0.94 (−1.62, −0.25)	−0.58 (−1.38, 0.23)	−0.80 (−1.16, −0.45)
Difference in βs	−0.47 (−1.23, 0.26)	−0.47 (−1.27, 0.33)	0.13 (−0.61, 0.97)	−0.51 (−0.91, −0.11)
Direct medical costs, GMR
Influenza seasons	0.74 (0.57, 0.96)	0.63 (0.45, 0.88)	1.11 (0.97, 1.27)	0.76 (0.67, 0.88)
Summer months	0.86 (0.60, 1.25)	0.64 (0.43, 0.97)	0.78 (0.56, 1.09)	1.00 (0.82, 1.21)
Ratio of GMRs	0.84 (0.53, 1.32)	0.98 (0.61, 1.59)	1.42 (1.04, 1.94)	0.77 (0.60, 0.95)

The reference group was those not vaccinated in any of the three seasons. The model adjusted for age, age square, sex, number of inpatient visits in the past 12 months, Charlson Comorbidity Index, operation status, hospital size, alcohol-related diagnosis, and obesity-related diagnosis and was stratified by quintiles of propensity score. Effect estimates in influenza seasons were adjusted for unmeasured confounding in summer months. Abbreviations: OR, odds ratio; β, regression coefficient; GMR, geometric means ratio.

For respiratory diseases, vaccination in the current season only, in the current and previous seasons, in the penultimate and previous seasons, and in all three seasons were each associated with a lower risk of in-hospital death during the influenza season (Table 3). After adjustment for unmeasured confounders with the summer months as the reference, these vaccination groups were still associated with a lower risk of in-hospital death, but the magnitude of associations attenuated toward the null. The ratios of ORs were mostly significant and were 0.44 (0.20, 0.95), 0.63 (0.24, 2.02), 0.64 (0.44, 0.95), and 0.68 (0.46, 0.98) for the current season only, the current and previous seasons, the penultimate and previous seasons, and all three seasons, respectively. In contrast to the ORs for in-hospital death during influenza seasons, the OR for re-admission was only protective for the current and previous seasons (0.35 [0.16, 0.76]).

Table 3. The effectiveness of repeated vaccination among patients hospitalized for respiratory diseases

Outcome, metric	Current only	Current & previous	Penultimate & previous	All three
	Effect size (95% CI)	Effect size (95% CI)	Effect size (95% CI)	Effect size (95% CI)
In-hospital death, OR
Influenza seasons	0.39 (0.23, 0.66)	0.34 (0.20, 0.60)	0.72 (0.58, 0.89)	0.32 (0.25, 0.41)
Summer months	0.87 (0.51, 1.48)	0.55 (0.27, 1.11)	1.13 (0.81, 1.58)	0.47 (0.34, 0.65)
Ratio of ORs	0.44 (0.20, 0.95)	0.63 (0.24, 2.02)	0.64 (0.44, 0.95)	0.68 (0.46, 0.98)
Re-admission, OR
Influenza seasons	0.68 (0.42, 1.10)	0.35 (0.16, 0.76)	0.99 (0.79, 1.24)	0.95 (0.76, 1.18)
Summer months	0.57 (0.31, 1.05)	0.69 (0.34, 1.42)	0.56 (0.37, 0.85)	0.53 (0.38, 0.74)
Ratio of ORs	1.17 (0.57, 3.73)	0.49 (0.18, 1.43)	1.69 (1.15, 2.59)	1.76 (1.24, 2.51)
Length of hospital stay, β
Influenza seasons	−1.04 (−1.82, −0.26)	−1.44 (−2.19, −0.69)	−1.56 (−2.14, −0.98)	−1.71 (−2.13, −1.28)
Summer months	−0.55 (−1.54, 0.44)	−0.60 (−1.58, 0.38)	−0.36 (−1.33, 0.62)	−1.21 (−1.64, −0.78)
Difference in βs	−0.52 (−1.70, 0.62)	−0.84 (−2.05, 0.39)	−1.23 (−2.15, −0.35)	−0.52 (−1.07, 0.07)
Direct medical costs, GMR
Influenza seasons	0.89 (0.62, 1.28)	0.59 (0.36, 0.98)	0.95 (0.77, 1.16)	0.66 (0.54, 0.82)
Summer months	0.86 (0.52, 1.43)	0.92 (0.49, 1.72)	0.94 (0.62, 1.43)	0.81 (0.61, 1.07)
Ratio of GMRs	1.02 (0.57, 1.84)	0.66 (0.31, 1.48)	1.00 (0.69, 1.50)	0.84 (0.61, 1.13)

The reference group was those not vaccinated in any of the three seasons. The model adjusted for age, age square, sex, number of inpatient visits in the past 12 months, Charlson Comorbidity Index, operation status, hospital size, alcohol-related diagnosis, and obesity-related diagnosis and was stratified by quintiles of propensity score. Effect estimates in influenza seasons were adjusted for unmeasured confounding in summer months. Abbreviations: OR, odds ratio; β, regression coefficient; GMR, geometric means ratio.

Effect of repeated vaccination on length of stay and direct medical costs

For CVD, vaccination in all three seasons was associated with shorter length of hospital stays and lower direct medical costs compared with unvaccinated in any season, both during the influenza season and when controlling for unmeasured confounders (Table 2). For respiratory diseases, although vaccination in all three seasons was associated with shorter length of hospital stays and lower direct medical costs compared with unvaccinated in any seasons during the influenza season, there was no association when controlling for unmeasured confounders (Table 3).

Effect of current season vaccination on hospitalized outcomes stratified by vaccination history

For both CVD and respiratory diseases, the effectiveness of current season vaccination did not differ by history of vaccination (Figure 1, p-values for heterogeneity all >0.05). For CVD, current season vaccination was associated a non-significant lower risk of in-hospital death regardless of vaccination history. Similarly, current season vaccination was non-significantly associated with a lower risk of re-admission, but was significantly associated with shorter length of hospital stays and lower direct medical costs in each stratum of vaccination history. For respiratory diseases, current season vaccination was associated with a significantly lower risk of in-hospital death and shorter length of hospital stays, and was associated with non-significantly lower direct medical costs regardless of vaccination history. In contrast, current season vaccination was associated a non-significant higher risk of re-admission regardless of vaccination history.

Figure 1. The effectiveness of current season vaccination among patients hospitalized for CVDs or respiratory diseases stratified by vaccination history.

The model adjusted for age, age square, sex, number of inpatient visits in the past 12 months, Charlson Comorbidity Index, operation status, hospital size, alcohol-related diagnosis, and obesity-related diagnosis and was stratified by quintiles of propensity score. Effect estimates in influenza seasons were adjusted for unmeasured confounding in summer months. Abbreviations: OR, odds ratio; β, regression coefficient; GMR, geometric means ratio.

Discussion

This is one of the studies in China that examine the effectiveness of repeated vaccination on hospitalized outcomes in older adults with major chronic diseases. We reported low rates of influenza vaccination among older adults hospitalized for major chronic diseases, with a particularly low rate of repeated vaccination. In older adults hospitalized for CVD, vaccination in all three seasons was associated with a non-significantly lower risk of in-hospital death and a significantly lower risk of re-admission as well as shorter length of hospital stays and lower direct medical costs, compared with unvaccinated in any season. In older adults hospitalized for respiratory diseases, vaccination in all three seasons was associated with a significantly lower risk of in-hospital death, but not with other hospitalized outcomes, compared with unvaccinated in any season. In addition, the effectiveness of current season vaccination on hospitalized outcomes was not altered by history of vaccination. Our study adds to the evidence base by showing non-significant protective effects of repeated vaccination on hospitalization outcomes among older adults hospitalized for CVDs and respiratory diseases. These findings provide preliminary evidence to support the promoting of repeated vaccination among older adults. Despite the “free influenza vaccine” policy, more efforts are needed to promote influenza vaccination among older adults.

Previous studies conducted mainly in Western countries have shown that repeated influenza vaccination is associated with lower risks of influenza and in-hospital death.^7–10 A meta-analysis published in 2019 including 20 studies has supported current season vaccination regardless of prior season vaccination.⁷ Current season vaccination was associated with lower risk of H1N1, H3N2, and influenza B, compared with no vaccination in either the current and previous season. For H1N1, no differences were observed between vaccination in both seasons and the current season only. A case-control study in Spain included 728 influenza cases and 1826 controls (aged ≥65 years) and assessed influenza vaccination over three influenza seasons between 2013 and 2015.⁸ This Spanish study showed lower risks of influenza and 30-day mortality in older adults vaccinated in the current (2015) and any previous season (2013 or 2014) compared with unvaccinated in any of the three seasons (2013–2015). In contrast, there were no apparent effects among those vaccinated in either the current season or any previous season. Likewise, a historical cohort study in Taiwan involving approximately 200,000 adults aged ≥65 years assessed the impact of repeated vaccination on influenza-associated hospitalization over two influenza seasons between 2007 and 2009.⁹ This study reported a lower risk of influenza-associated hospitalization among older adults vaccinated at that season (2009) and sequential vaccination compared with unvaccinated in any season. There were no effects among other vaccination groups, including current season vaccination and no vaccination as well as no current season and sequential vaccination.

Influenza vaccination may be protective of respiratory diseases because of the protective effects on influenza like illness, influenza-related respiratory infections, and exacerbations of chronic obstructive pulmonary disease (COPD).^13,17 Consistent with previous studies, the current study showed that repeated influenza vaccination was associated with lower risk of in-hospital death among older adults hospitalized for respiratory diseases.^8–10 There was an apparent effect among those vaccinated in all three seasons, but no effects among other vaccination categories. The lack of effects among older adults with non-sequential vaccination was consistent with findings for influenza. A possible explanation is that there were a limited number of participants in non-sequential vaccination categories, and therefore there is little power to detect protective effects, if there are any. For CVD, influenza vaccination may be protective of hospitalized outcomes because of the protective effects on infections subsequent to influenza, endothelial dysfunction, atherosclerosis, and inflammation.^18,19 Indeed, we observed lower risks of in-hospital death and re-admission among older adults vaccinated in all three seasons compared with those unvaccinated in any season. Similar to the findings for respiratory diseases, no apparent protective effects were observed among non-sequential vaccination groups, also possibly because of the limited sample size.

When stratifying by vaccination history, our study showed protective effects of current season vaccination on hospitalization outcomes among older adults hospitalized for CVDs or respiratory diseases. Taken together with our findings on repeated vaccination, our study suggests that previous vaccination did not interfere on current season vaccination. This is consistent with several previous studies on the effectiveness of influenza vaccination.^7–9 From a public health perspective, the overall evidence that repeated influenza vaccination is associated with a lower risk of influenza as well as a better profile of hospitalization outcomes among older adults hospitalized for CVD or respiratory diseases strengthens the recommendation of annual influenza vaccination among older adults.

The strengths of the current study include the large sample size, information on repeated vaccination, and careful adjustment for both measured and unmeasured factors associated with vaccination and hospitalization outcomes. Specifically, we used multivariable regression and propensity score to adjust for observed confounders and used reference periods (i.e. summer months) to adjust for unmeasured confounders. However, there are several limitations. First, the number of participants was small in non-sequential categories of influenza vaccination, which greatly reduced the power of analysis. This is possibly because of the vaccination habits among older adults. As suggested by previous studies, adults who had been vaccinated in the past seasons were more likely to get vaccinated in the current season.^20–22 Second, our sample size was still limited to assess the protective effect of influenza vaccination on disease subtypes, including influenza like illness and hemorrhagic stroke. However, our previous study showed that the effectiveness of influenza vaccination on hospitalization outcomes was similar for ischemic CVD and hemorrhagic stroke as well as for influenza/COPD and respiratory diseases.⁵ Future studies pooling data from more influenza seasons are warranted to replicate our findings.

In conclusion, repeated influenza vaccination was associated significantly with lower risk of in-hospital death among older adults hospitalized for respiratory diseases and lower risk of re-admission among those hospitalized for CVD. The effectiveness of current season vaccination on hospitalization outcomes was not modified by vaccination history. Our findings support the recommendation of annual influenza vaccination among older adults.

Author contributions

YP, M Lv, M Lu, and YH had full access to the data. YP, M Lv, and YH conducted data analysis and are responsible for accuracy of the results and the decision to submit for publication. All authors were involved in study design, conduct, long-term follow-up, review and coding of disease events, interpretation of the results, or writing the report.

Acknowledgments

We thank 444 vaccination clinics for their contributions to the acquisition of the immunization data.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Supplementary material

Supplemental data for this article can be accessed on the publisher’s website at https://doi.org/10.1080/21645515.2021.2007012.

Additional information

Funding

This work was supported by grants from Beijing Natural Science Foundation (19L2055), Beijing Municipal Health Commission, Pilot Projects of Public Welfare, Development and Reform for Medical Research Institutes in Beijing [grant number 2018-1], and the Capital Health Research and Development of Special.