Determinants of influenza vaccination uptake among older adults in Catalonia using a longitudinal population study: the role of public health campaigns

Abstract

Introduction

This study examines the impact of influenza vaccination campaigns on the probability of immunisation among older adults in Catalonia, Spain.

Methods

A population-based cohort study was conducted using the Catalan administrative and health dataset. Longitudinal data on healthcare resource use for individuals born before 1965 in Catalonia were used to compute descriptive statistics and concentration measures. A Regression Discontinuity Design (RDD) was performed to calculate the jump in the probability of becoming vaccinated. The database covers administrative data from primary care, hospitalisations and emergency care in the national health system from January 2014 to October 2021.

Results

Significant differences were found across age groups, gender, drug copayment levels and nationality, and these differences were corroborated through concentration measures conditioned on health status. The RDD indicates a 4.78% increase in the probability of vaccination at the age at which vaccines were offered (60 years old), with the main differences observed among individuals from lower-income levels, specific health regions and nationalities. Age, a diagnosis of influenza in the previous vaccination campaign, and particular comorbidities were factors positively associated with a higher likelihood of vaccination.

Conclusions

Understanding the interplay of factors is crucial to addressing disparities and ensuring preventive measures reach vulnerable groups. Our findings have direct implications for influenza vaccination coverage among older adults, providing policymakers with valuable insights for enhancing outcomes.

WHAT IS ALREADY KNOWN ON THIS TOPIC

• Previous studies have identified demographic, socioeconomic, chronic conditions and cultural factors—such as age, income, gender and nationality—as key determinants of influenza vaccination among older adults. However, most evidence to date is descriptive, and there remains a lack of causal analysis using large-scale administrative data to assess the impacts of policies and health inequalities.

WHAT THIS STUDY ADDS

• This study uses longitudinal administrative data and a Regression Discontinuity Design to estimate the causal impact of age-based vaccine eligibility on uptake. It incorporates concentration and segregation indices to measure inequality across income groups, regions and nationalities.

HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY

• Findings underscore the role of socioeconomic and geographic disparities in vaccination coverage, supporting the need for equity-oriented policies and targeted outreach to improve preventive care delivery among underserved populations.

Introduction

Influenza remains a significant health challenge, particularly among vulnerable populations like older adults and those with chronic conditions.^{1 2} While most recover, influenza can lead to severe illness, hospitalisation and death, especially among those at higher risk.³ Vaccination is the most effective preventive measure for high-risk groups, reducing the severity of illness and the risk of fatality. However, due to the virus’s constant evolution, annual updates require public health coordination.^{4 5}

In Spain, healthcare policies are decentralised, and Catalonia takes a proactive approach to vaccination. Since 2008/2009, Catalonia has recommended universal vaccination for people 60 and older to reduce influenza-related morbidity and mortality. The timing of the influenza season varies by country and hemisphere. In Catalonia, as in much of the Northern Hemisphere, the annual vaccination campaign runs from late September to March. Since the 2008/2009 season, vaccination has been recommended for adults aged 60 and over (previously 65+). During the campaign, free vaccination is offered to those aged 60 or older and individuals at risk of influenza complications. Vaccination is also advised for high-risk groups and specific occupational categories, such as emergency service workers. Primary care centres lead the vaccination efforts, which vary by season but mainly use the adjuvanted trivalent-inactivated vaccine. Delivery methods depend on public engagement, outreach to care homes and day centres, proactive communication via text messages, and individuals proactively asking for new appointments or opportunistic vaccination during other healthcare visits. Despite these efforts, influenza vaccine uptake among older adults remains suboptimal, which aligns with Spanish and other international standards (75%), underscoring the need for a better understanding of the factors influencing vaccination behaviours.

Previous literature has indicated several factors associated with vaccination rates against influenza for older adults. Gender, age, household income, health status and nationality are key determinants of these vaccination rates.6,8 Most of these previous findings are linked to cultural and belief differences within the population. This study examines the social determinants of influenza vaccination among older adults in Catalonia. We contribute by providing findings using population administrative registers and evaluating concentration indices across characteristics, the slope of inequality and the change in vaccination probability due to national health system recommendations, disentangling these effects by individual characteristics. Through this analysis, we aim to inform the design of policies and interventions.

Methods

Data sources and linkage

An anonymised large dataset from the Agency for Health Quality and Assessment of Catalonia was used, considering longitudinal population data using the national health system. The dataset comprises data from multiple providers for the Catalan population born before 1 January 1965, allowing for a retrospective cohort study of 2 924 590 individuals. This observational retrospective dataset spans from January 2014 to October 2021, covering primary care, hospitalisations and emergency care, aggregated into seven seasonal periods (September–March). The data includes individual identifiers, visit dates, diagnoses, procedures and demographic information (age, gender, nationality, drug co-payment level and health region).

Individuals at risk based on their diagnoses, including hypertension, mental illness, neoplasms and diabetes, were identified. For consistency, diagnoses were converted to ICD-9 codes. The focus was on medically attended influenza, underestimating the total burden, as not all symptomatic cases are captured.

Reporting patient and public involvement in research

None are given, as the study corresponds to observational retrospective data containing the entire population that used public healthcare resources.

Vaccination dataset

Influenza, pneumonia and SARS-CoV-2 vaccinations, including injection dates, from all units of the Catalan Health Institute (283 primary care centres and eight hospitals) are included in our dataset. Vaccination campaigns in Catalonia typically run from late September to March, with universal vaccination recommended for individuals aged 60 and older since 2008/2009 (previously for those aged 65 and older). The Catalan programme offers free vaccination to eligible individuals aged 60 and above, as well as those at risk of complications.

Vaccinations are administered by primary care centres using trivalent inactivated vaccines. The process varies depending on citizens’ proactivity, outreach by nursing homes and daycare centres, and prevention activities (eg, text messages, letters, or in-clinic administration when individuals visit for other health issues).

Methods

Using a large dataset from Catalonia, vaccination rates across subgroups and the impact of health campaigns were evaluated. By leveraging techniques like Regression Discontinuity Design (RDD) and longitudinal analyses, our research provides evidence on how policy thresholds (eg, age 60) influence vaccination behaviour. The aim addressed the association between group membership (eg, income level, nationality, gender) and influenza vaccination. Inequality across these groups was measured using concentration indices to determine how their proportions in the vaccination condition deviate from the overall population distribution. Segregation occurs when a group is over-represented or under-represented in the vaccination group compared with the general population. Segregation measures using the Mutual Information Index (MI)⁹ were computed, which account for multilevel structures. Given that vaccination is a binary indicator, two-way tables and computed relative segregation indices were produced.

Inequality in vaccination was measured with concentration indices,¹⁰ which assess the covariance between vaccination status and socioeconomic status (proxied by income from copayment levels). The concentration index was modified for binary variables, following the approach that satisfies transfer, level independence, cardinal invariance and mirroring.¹¹ An alternative was also computed for sensitivity.¹² Copayment levels were used as a proxy for income, with higher copayment levels indicating lower income. Vaccination rates were compared across various factors, including nationality.

Then, the slope of inequality was computed, that is, the association between copayment levels and vaccination. A random-effects parametric survival model with an exponential distribution was estimated, using copayment levels as a proxy for income. Individuals who died during the period were excluded, and standard errors were clustered at the healthcare provider unit. The base category was the exempted group. For this purpose, our estimates were conditioned using further covariates: sex, nationality, smoking habits (recorded through medical visits with three categories: non-smoker, former smoker and current smoker), age in categories, the adjusted morbidity group, and the presence of specific comorbidities recorded through ICD codes (bronchitis, apnoea, respiratory infections, hypertension, diabetes, dyslipidaemia, Alzheimer’s, anxiety, depression, acute sinusitis and dementia). Decile fixed effects for income levels were also included, based on the postal code for each healthcare provider unit, as well as influenza season fixed effects.

A RDD, a quasi-experimental method widely used in health economics,¹³ was used to estimate the causal impact of health policy campaigns. RDD uses a forcing variable—age, in our case—to identify discontinuities in vaccination probability. The policy threshold in Catalonia for universal vaccination is set at age 60, which was used as the cut-off for analysis. This technique estimates the causal parameter τ, which captures the differential likelihood of vaccination between individuals above and below the age threshold.

A fixed-effects logit model was also used to examine vaccination-associated factors over time. The dependent variable was vaccination status, which was time-varying (eg, smoking behaviour and comorbidities registered across medical visits), and fixed characteristics (age, gender, income and nationality) were included. Campaign-fixed effects were incorporated to account for variations in vaccination outreach each season. Age, gender, co-payment level and comorbidities (eg, hypertension, diabetes) were included as explanatory factors. Furthermore, previous influenza experiences recorded through ICD codes were treated as explanatory variables, following the concept of ‘learning by suffering’,¹⁴ where individuals who had previously experienced influenza infection may be more likely to seek vaccination in subsequent campaigns.

Results

Descriptive information for the population cohort

Approximately 1 200 000 vaccines are administered annually, with a spike to 1.45 million in 2020/2021 due to the SARS-CoV-2 pandemic. The dataset corresponds to 2 924 590 individuals. Influenza vaccination rates remained relatively stable across campaigns, with women and men reporting rates of approximately 57.0% and 58.5%, respectively (SD: 3.6 and 3.2). Table 1 depicts population characteristics for our dataset.

Table 1. Descriptive statistics on sociodemographic characteristics for the population (administrative registers for Catalonia, campaigns 2014–2021).

Female	0.541 (0.50)
Passed away during the period	0.172 (0.38)
Age at 31 December 2020
Age (55–60)	0.183 (0.39)
Age (60–65)	0.163 (0.37)
Age (65–70)	0.140 (0.35)
Age (70–75)	0.130 (0.34)
Age (75–80)	0.111 (0.31)
Age (80–85)	0.083 (0.28)
Age (85–90)	0.083 (0.28)
Age (90–95)	0.059 (0.24)
Age (95–100)	0.047 (0.21)
Drugs co-payment level
Exempted	0.185 (0.39)
10% co-payment	0.523 (0.50)
40% co-payment	0.150 (0.36)
50% co-payment	0.100 (0.30)
60% co-payment	0.013 (0.11)
Exempted from co-payment	0.029 (0.17)
Nationality
Spanish	0.945 (0.23)
Maghreb	0.009 (0.10)
South America	0.011 (0.11)
East Europe	0.008 (0.09)
Europe	0.026 (0.16)
Health region
Lleida	0.049 (0.22)
Camp de Tarragona	0.077 (0.27)
Terres de l’Ebre	0.027 (0.16)
Girona	0.112 (0.32)
Catalunya Central	0.072 (0.26)
Alt Pirineu i Aran	0.010 (0.10)
Barcelona	0.653 (0.48)

Note: average values or percentages expressed as decimals and SD in parentheses are displayed.

Figure 1 presents descriptive statistics for key sociodemographic characteristics, illustrating overall vaccination rates by campaign across age groups, gender, drug copayment levels, nationality and health area. There were significant gender differences in vaccination rates, with men generally exhibiting higher rates. Statistically substantial variations in vaccination rates were observed based on nationality, subregion of residence and copayment levels. Copayment levels, a reliable proxy for income, indicated a positive correlation with vaccination rates, except for those exempt from copayments. Spanish individuals consistently demonstrated the highest vaccination rates regardless of the campaign. In contrast, those born in Eastern Europe, the Middle East and Eastern Asia exhibited the lowest rates. The SARS-CoV-2 pandemic had a notable impact on increasing vaccination rates. Additionally, figure 1 highlights variations across health subregions, with Girona and Camp de Tarragona initially reporting the lowest rates.

Figure 1. Vaccination rates across age, sex and seasonal influenza campaigns in Catalonia (2014–2021): heatplots.

Measures of concentration

Negative values for the concentration index indicate that vaccination rates are concentrated among less wealthy individuals (see figure 2). A more negative value signifies a higher concentration among lower-income households. Conversely, a concentration index closer to zero suggests a lower concentration among less affluent individuals. The baseline concentration index was −0.215, slightly decreasing across campaigns (from −0.225 in 2014/15 to −0.192 in 2020/21). The alternative Wagstaff index was −0.257, somewhat lower than the Erreygers index. Two complementary indices were calculated with different weights (beta=1.5 and beta=5). A higher beta value places more weight on the extremes of the income distribution (−0.231), while a lower beta value emphasises the middle (−0.155).

Figure 2. Concentration and normalised mutual information indices for vaccination across sociodemographic characteristics and Catalan influenza campaigns (2014–2021).

When comparing nationalities, Spanish individuals exhibited a concentration index close to the baseline (−0.211). The remaining nationalities demonstrated lower concentrations among less wealthy individuals. Except for Europeans (−0.159), all other nationalities had concentration values much smaller than the overall average, with the lowest value for Central/Caribbean America (−0.037).

Next, segregation within health provider units and a decomposable segregation index were calculated.⁹ The MI enables the decomposition of desired properties compared with alternative entropy or inequality indices, such as the Theil or Atkinson indices. The normalised index represents the MI value relative to its maximum, expressed as a percentage.

Each campaign’s overall MI and individual values were initially computed (figure 2). The overall value was 3.65%, with a decreasing trend across vaccination influenza seasons, except for the most recent campaign, which coincided with the SARS-CoV-2 outbreak. The entropy value (Theil/MI) was 0.04.

Then, the possibility of certain nationalities exhibiting significantly different behaviour within health provider units compared with their average vaccination rates was explored. Specifically, the deviation of the vaccination rate within a health provider unit from the population-wide vaccination rate was investigated. Heterogeneous behaviour was observed across different nationalities. A higher MI value indicated greater segregation in vaccination rates for the specific nationality. The overall MI value was similar for Spanish and European individuals but notably higher for African, North American and East Asian/Oceanian nationals. In contrast, individuals from Eastern Europe and South America displayed the lowest normalised MI values.

Slope of inequality

Our HR results showed that, compared with exempted individuals (those with a lower income level), all categories were less likely to be vaccinated. The good news was that the lower copayment level (10%) indicated a slight difference in vaccination rates (15.96%) compared with those exempted from paying the drug copayment. On the contrary, 40% and 50% copayment levels showed 40.54% and 45.37% lower probabilities of being vaccinated through the universal public health system, respectively. The group with a 60% copayment level showed a 32.38% lower likelihood of being vaccinated, which was larger than that of those with similar copayment levels (40% and 50%). Figure 3 presents these estimates, along with their corresponding confidence intervals. Results hardly changed, except for the 50% copayment level during the last season, which coincided with the SARS-CoV-2 outbreak. This group also showed a higher vaccination rate than the 40% copayment group.

Figure 3. RDD estimates for the jump in probability around the recommended age for influenza vaccination and heterogeneous results for Catalonia (campaigns 2014–2021). RDD, Regression Discontinuity Design.

Probability of vaccination because of campaigns: RDD results

The change in vaccination probability using the RDD technique was estimated, leveraging the discontinuity at age 60 to determine the causal impact. Given that vaccination likelihood declines from a certain age and excessive controls were available, only individuals under 62 and campaigns with participants under 60 were evaluated. As shown in figure 3, there is a jump in vaccination likelihood around age 60, likely due to targeted vaccination campaigns. Additionally, an RDD analysis was conducted for each vaccination campaign, focusing on individuals aged 58–62 to create more comparable groups. The first campaign (2014/2015) was considered a control. Using a first-order polynomial, our estimates indicate that in the 2015/2016 campaign, the vaccination probability was higher for the age group above 60 by 0.005, representing a 4.77% increase for the control group (58–60 years). It was also computed for each campaign. The difference was 0.005 in 2016/2017, 0.002 in 2017/2018, 0.006 in 2018/2019, and 0.001 and 0.002 (not statistically significant) in 2019/2020 and 2020/2021.

Next, heterogeneous RDD results were investigated, following up¹⁵ to identify the factors driving the jump in vaccination probability. The impact was estimated for the entire period, disaggregated by copayment level, health region and gender. Figure 3 summarises these results. Most copayment levels, except 60%, showed a significant increase in vaccination rates, particularly among lower-income individuals. All health sub-regions, except Girona and Alt Pirineu, showed increases, although some were not statistically significant (such as Camp de Tarragona or Terres de l’Ebre). The impact on females was higher, although not statistically significant, than on males. Finally, the results were mainly driven by Spaniards, as there was no statistically significant impact on influenza vaccination for other nationalities.

Factors associated with vaccination: longitudinal logit results

A fixed-effects logit analysis was conducted to identify factors related to vaccination. The results are shown in figure 4. Variables such as age and having been diagnosed with influenza/pneumonia in the previous campaign were statistically significant and positively related.

Figure 4. Logit longitudinal fixed effects: ORs. Catalan vaccination campaigns (2014–2021).

Discussion

This study provides important insights into the factors influencing influenza vaccination uptake among older adults in Catalonia, a population at heightened risk for severe influenza outcomes. A large, population-based dataset has shown that demographic characteristics—such as age, gender, socioeconomic status and health history—play a crucial role in determining vaccination decisions. Our findings suggest that demographic factors (such as gender and socioeconomic status) strongly correlate with influenza vaccination decisions and are aligned with previous research,16,21 corroborating that age, health status and physician quality are key to influenza vaccination decisions. We found further specific vaccination rates across age, gender and nationality compared with earlier findings for the Spanish population.²² Additionally, prior influenza outbreaks have been confirmed to influence individuals’ vaccination decisions, a phenomenon known as ‘learning by suffering’.¹¹

Our analysis also highlights the effectiveness of public health campaigns. Specifically, a 4.77% increase in vaccination probability was observed among individuals aged 60 and older, indicating the positive impact of targeted vaccination campaigns. This finding is consistent with previous studies demonstrating the effectiveness of well-designed public health campaigns in improving vaccine uptake among high-risk groups.²³

However, the study reveals areas for improvement. Despite overall increases in vaccination rates, specific subgroups—particularly those from non-Spanish backgrounds and those with lower socioeconomic status—remain underserved. This highlights the need for more tailored interventions considering cultural, socioeconomic and geographical factors. The lower vaccination rates among individuals from Eastern Europe, the Middle East and Eastern Asia suggest a need for targeted messaging and outreach strategies to address cultural and language barriers to vaccination. The variation in vaccination uptake across different health subregions in Catalonia further supports this argument. Although some areas showed significant improvements, others (such as Girona and Camp de Tarragona) exhibited persistently low vaccination rates despite similar policy efforts. This indicates a need for a more granular, region-specific approach to intervention, where healthcare providers collaborate with local community organisations to identify and overcome barriers specific to each area.

Our findings suggest that future vaccination campaigns should adopt a more nuanced approach. Policymakers must continue broad public health campaigns while also focusing on addressing vaccination inequities. Given the strong correlation between socioeconomic status, nationality and previous health experiences with vaccination uptake, interventions should focus on reaching groups with historically low vaccination coverage. For example, communities with high proportions of non-Spanish nationals may benefit from targeted communication strategies that consider cultural attitudes toward vaccination and address language barriers.

The study also found that copayment levels—used as a proxy for income—had a significant impact on vaccination rates. Individuals with lower copayment levels were more likely to be vaccinated, reinforcing that economic barriers are a key obstacle to vaccine uptake. Although this study primarily focuses on older people, the results have broader implications for universal vaccination programmes. Reducing financial barriers, even marginally, could enhance vaccination coverage.

This study has some limitations. First, the used dataset only includes vaccinations administered through public healthcare centres in Catalonia, which likely leads to an underestimation of overall vaccination coverage, particularly among those who seek care from private providers. Additionally, our analysis was restricted to individuals born before 1965, excluding younger at-risk groups. The longitudinal dataset only captures vaccination decisions within the study period, which limits our ability to assess long-term trends or the cumulative effects of repeated vaccination campaigns. Moreover, since the data is based on a single autonomous community in Spain, the findings may not be generalisable to other regions with different healthcare systems or demographic characteristics. Finally, although RDD was used to estimate causal impacts, unmeasured confounders may still affect the results. Future research should incorporate more granular data on individual behaviour and healthcare utilisation and consider the role of physician-patient communication and healthcare professionals’ influence on vaccination decisions.

In conclusion, this study provides valuable insights into the factors influencing influenza vaccination uptake among older adults in Catalonia. It emphasises the importance of targeted public health campaigns to overcome social and economic barriers to vaccination. To improve vaccination coverage and reduce health disparities, policymakers should focus on more tailored strategies that reach underserved groups, particularly those related to the aforementioned cultural differences that influence vaccination uptake.

Data availability statement

No data are available.