Time lapses between distribution of influenza vaccines to health authorities and their administration by General Practitioners (GPs) to older adults: a retrospective study over five influenza seasons in Italy

ABSTRACT

Background

Delays in influenza vaccine delivery and administration can hinder vaccine coverage and protection. This study examines the differentials in distributing and administering adjuvanted trivalent (aTIV) and quadrivalent influenza vaccines (aQIV) to older adults in Italy’s primary care setting and its potential impact on hospitalization risk over 5 epidemic seasons.

Methods

Using a primary care database, individuals aged ≥ 65 years were selected. The proportion of vaccine distribution to regional authorities and subsequent administration by GPs was estimated using census data. Using quantile (median) regression, we examined the relationship between velocities of vaccine distribution and administration (doses/week) and the incidence of hospitalizations.

Results

Over the 5 influenza seasons, the velocity of distribution and administration of aTIV/aQIV ranged 341–833 and 152–270 median doses/week; no trend was yielded for the difference between these velocities (p = 0.189) or vaccine coverage (p = 0.142). An association was observed for each differential dose/week between distributed and administered vaccines and all-cause hospitalizations with a 10% increase in 2017–2018, 54% in 2018–2019, and 12% in 2020–2021 season.

Conclusions

These findings highlight the importance of minimizing the time lapse between vaccine distribution and administration to mitigate the impact of influenza and address factors that contribute to vaccination barriers.

KEYWORDS:

• Influenza vaccine
• older adults
• elderly
• influenza hospitalizations
• velocity
• distribution
• administration
• time lapse

1. Introduction

Influenza causes one of the greatest vaccine-preventable disease burdens in Europe, accounting for an estimated annual average of 4–50 million symptomatic cases, approximately 15,000–70,000 deaths and 150,000 influenza-related hospitalizations [1–3]. However, these complications are not equally distributed across populations, with some vulnerable subgroups, such as older adults, young children, pregnant women, and persons with chronic diseases or immunosuppressive conditions, experiencing an increased burden compared to the general population [4]. Elderly individuals have an increased susceptibility to infectious diseases, as a consequence of the high prevalence of comorbidities and immunosenescence [5,6]. In Italy, excess mortality rates are more than 6 times higher among the elderly than in the general population, with influenza- and pneumonia-related mortality counting among the top 10 causes of deaths nationally [7,8]. Vaccination in older adults is therefore strongly recommended in European Economic Area (EEA) countries. In Italy, among older adults, a 1% increase in influenza vaccine coverage has been estimated to reduce the incidence of influenza from 2 to 4%. A statistically significant association was found between vaccine coverage declines observed between in the trienna 2005–2008 and 2014–2017 and an increase in influenza-like illnesses from 2.7% to 4.2% between the same 2 periods [9].

The window of opportunity for influenza vaccination is relatively brief, typically starting in September/October and ending in January or beyond in the Northern Hemisphere [8,10]. In Italy, annual influenza vaccination is recommended from the 1^st of October continuing throughout the whole season. Vaccine coverage refers to the population proportion that has received the recommended doses of a vaccine. The Italian Ministry of Health has established a minimal effective target of 75% for influenza vaccination coverage for older adults and at-risk individuals of all ages, however, during the last 23 seasons (from 1999–2000 to 2021–2022), this objective has never been achieved [8].

The time lapse between vaccine delivery and administration can have an impact on vaccine coverage [4]. In general, the shorter the time lapse between vaccine delivery and administration, the higher the vaccine coverage is likely to be, because delays in vaccine delivery or administration can create barriers to vaccine access and may cause people to miss their scheduled vaccine appointments [11–13]. Additionally, delays in vaccine delivery or administration can increase the risk of vaccine wastage or expiration, which can reduce overall vaccine coverage [4]. In the 2000–2001 season in the United States, a 6- to 8-week delay of influenza vaccine delivery led to a sharp drop of 16% in vaccine coverage compared to the previous year. An analogous delay in the 2014–2015 season, along with antigenic drift, resulted in a shortage during the epidemic peak and the greatest numbers of hospitalizations and deaths over the previous 5 seasons [14]. Hence, both the possibility of an intraseasonal waning in vaccine effectiveness and the risk of a decrease in influenza vaccine uptake due to late vaccine distribution must be taken into account to determine the best period of distribution and administration of the influenza vaccine.

Although there is limited qualitative information about how General Practitioners (GPs) judge vaccine delivery delays and how it affects their practice [15,16], there have not been any studies that use actual data to measure the extent of this public health concern. We therefore assessed the time it takes for vaccines distribution to regional health authorities (i.e. usually occurring between September-October) and subsequent administration by GPs (i.e. typically starting in October, albeit there is no contraindication in administering influenza even as the epidemic curve rises). Additionally, we investigated how the differential between the time of distribution and time of administration might increase the risk of hospitalizations due to all causes or respiratory issues. This analysis spanned across 5 consecutive epidemic seasons.

2. Methods

This study was designed to examine the relationship between the rates of vaccine distribution and vaccine administration in doses/week to persons aged 65–95 years and to analyze the association between the differential speed of vaccine distribution and administration and rates of all-cause and respiratory-related hospitalizations in this population over 5 seasons from 2017 to 2022. The study protocol was approved by the Scientific Committee of the Italian College of General Practitioners and Primary Care. This study followed the principles of the Declaration of Helsinki and was compliant with the European Network of Centres for Pharmacoepidemiology and Pharmacovigilance (ENCePP) Guide on Methodological Standards in Pharmacoepidemiology.

2.1. Data sources

We used the overall numbers of doses of adjuvanted trivalent (aTIV [Fluad®]) or quadrivalent influenza vaccines (aQIV [Fluad® Tetra]) as provided by the marketing authorization holder (MAH; i.e. Seqirus) to regional health authorities between 2017 and 2021. These data were provided on a regional and daily basis. Numbers of influenza vaccines administered to older adults aged ≥65 years by GPs were obtained from the Italian College of GPs health search database (HSD) over the same 5 influenza seasons. For both data sources, data were from 20 Italian regions of residence.

The HSD is a longitudinal observational database established in 1998 that contains electronic healthcare records (EHRs) of almost 1.2 million subjects under the care of approximately 1000–1200 GPs distributed throughout Italy. The present study included computer-based patient records collected by a selected group ranging 1064–1129 GPs, over the 5 influenza seasons, who met standard quality criteria regarding the levels of data entry (i.e. levels of coding, prevalence of selected diseases, rates of mortality, and years of recording). These GPs were selected on a geographical basis to include patients that would be representative of the whole Italian population. All diagnoses were coded according to the International Classification of Diseases, 9^th revision, clinical modification (ICD-9-CM). To complement the coded diagnoses, GPs have the ability to add a free text. Information on drug prescriptions includes the name of the prescribed drug (i.e. active substance and/or brand name), the corresponding anatomical therapeutic chemical (ATC) code along with the related defined daily dose (DDD), the date of prescription, and the number of supply days. Vaccinations are registered in a dedicated section. HSD has been extensively used for retrospective and longitudinal research, including effectiveness investigations on influenza vaccines [17–19].

2.2. Study population

From the HSD, we identified all individuals aged 65–95 years and the regions where their GPs operated over the period between the 2017–2018 and 2021–2022 influenza seasons. A subject could be included in one or more seasons. Those aged ≥96 years were not considered, given the high rate of hospitalizations and/or institutionalization of these subjects, which suggests a reduced completeness of their EHRs in these settings. We therefore estimated the proportion of aTIV and aQIV (as reported in MAH data sources) doses distributed to regional Local Health Authorities (LHAs) and expected to be delivered to GPs (who may also have obtained doses from authorized pharmacies), weighting the population size of older adults served by each physician for the ratio between the actual number of delivered doses of aTIV/aQIV and the total number of older Italian residents (source: https://demo.istat.it/). Such an approach was adopted given that we were able to collect data on vaccine aggregated distributions referring to resident population. On the other hand, HSD is formed by a sample (representative) of Italian GPs. These data were compared with the aTIV/aQIV doses being actually administered (as per EHRs) to older adults in the care of GPs belonging to the HSD network.

2.3. Outcome definition

During the course of the 5 influenza seasons, we computed the velocity of vaccine doses distributed to regional LHAs and administration by GPs (or nurses belonging to the same clinics) on weekly basis. The term velocity describes the identification of doses of influenza vaccines delivered in a specific period of time (i.e. weekly). Specifically, the term distribution velocity refers to the doses of ordered influenza vaccine vials delivered from the MAH to the regional LHAs on a weekly basis, whereas the term administration velocity is used to define the number of influenza vaccines administered by GPs (or nurses belonging to the same clinics) each week. The mathematical difference between the distribution and administration velocities is used to indicate the lapses between distributed and administered doses on a weekly basis, thus providing a unique dynamic measure of the vaccine doses actually administered instead of remaining stocked by the regional LHAs and/or GPs. We investigated the potential relationship between these 2 velocities over the 5 seasons and determined the difference between the 2 velocities for each epidemic season and analyzed the correlation between this difference and the rates of all-cause/respiratory-related hospitalization being registered during the epidemic seasons. We identified each hospital admission recorded in the database during the follow-up period. We also used free text to identify hospitalizations by searching for terms such as ‘recover*,’ ‘admiss*,’ and ‘hospit*’ within 3 months before and/or after the exit date (i.e. death, end of the specific season, end of data availability). All records were reviewed and validated by an expert clinician to ensure the classification accuracy of the events, using the same approach as in prior studies [17,19]. To maintain biological plausibility regarding the impact of the influenza vaccine on the risk of hospitalization, events occurring within the first 15 days of aTIV/aQIV administration were excluded from consideration [20].

2.4. Data analysis

For each epidemic season, descriptive statistics were calculated. We plotted the trends of the cumulative vaccine doses expected to be received and those actually administered (according to EHRs) by GPs to immunize older adults over the 5 influenza seasons. To be consistent with the time frame of influenza season, weeks were the units of observation [4]. To calculate vaccine coverage, we divided the number of older adults who received the influenza vaccine by the total number of persons aged 65–95 years under the care of GPs. The skewness of distributed and administered vaccine doses over seasonal weeks was proven through the Belanger and D’Agostino test [21]. We therefore tested the potential presence of trend for both distribution and administration velocity among the 5 seasons using quantile (median) regression. With the same approach, we investigated the possible relationship between the difference in velocities of vaccine distribution and vaccine administration (doses/week) and the occurrence of all-cause or respiratory-related hospitalizations. Operationally, the beta coefficients, with related 95% confidence intervals (CI), were calculated to quantify the proportional increase (or decrease) in the outcome for an increase of 1 dose/week of influenza vaccine (or differentials between doses distributed and administered; in this case, an increase indicates a slower allocation and administration of vaccines). Every regression analysis was clustered by region of residence after identifying the presence of intraclass correlation (p < 0.001). The estimates, which were obtained for every epidemic season, were pooled through meta-analysis. Heterogeneity was evaluated and tested using I-square and Q test.

To test the robustness of the results, we conducted two sensitivity analyses. First, our estimate of the number of doses expected to be delivered to GPs was derived by combining distribution data (sourced from the MAH) and regional census information. However, certain GPs who were affiliated with specific regional LHAs but who could not be identified in the HSD may have received additional aTIV/aQIV doses. Moreover, prior to the COVID-19 pandemic, some older adults may have independently purchased doses in pharmacies. As a result, the number of administered doses exceeded the expected quantities that were delivered to, or collected by, GPs during the initial 3 seasons. Second, we tested the effect of vaccine under-registration (i.e. false-negative vaccinations) on the results [17,22], which can lead to bias because GPs are required to register vaccination twice: in a public regional registry and in their own EHRs. The additional GP workload might reduce the completeness of data collection, leading to underestimation of the ratio of doses/week. The primary analysis was therefore recalculated by limiting GPs to those reporting a vaccine coverage of ≥ 55%, which is consistent with the lowest coverage reported in official reports of Italian public health authorities for 1 of the included seasons [8]. Utilizing a subset of GPs with more precise vaccine recording methods enabled the first sensitivity analysis to be revised as well, given that the number of older adults purchasing the influenza vaccine themselves should be minimized.

3. Results

Table 1 displays the characteristics of older adults forming the study population over the 5 epidemic seasons along with the actual number of residents in Italian regions where the overall number of aTIV/aQIV doses were delivered by MAH. The proportions of older adults registered in HSD were consistent with those calculated for the resident populations (ranging 26–27% for general population and 27–28% for HSD population). This information enabled us to quantify the anticipated number of vaccine doses for administration to older adults: 78455, 73226 (2017–2018), 78973 (2018–2019), 137,787 (2019–2020), and 120,820 (2020–2021) doses. More women than men received vaccines.

Table 1. Characteristics of the overall older adult population and the amount of influenza vaccine doses delivered and administered over 5 influenza seasons in Italy.

	Influenza season
N	2017–2018	2018–2019	2019–2020	2020–2021	2021–2022
Total resident population (aged 15–95 years)^b	51,862,355	51,833,341	51,794,400	51,479,434	51,540,338
Resident older adults (aged 65–95 years)^a	13,487,719	13,581,770	13,739,556	13,821,297	13,927,447
Weeks	29	31	31	32	17
GPs in HSD	1129	1111	1096	1085	1064
Total population under care of GPs in HSD	1,346,465	1,287,476	1,223,344	1,174,423	1,058,289
Older adults under care of GPs in HSD	362,023	350,440	337,929	322,907	288,401
Male	160,320	155,634	151,480	145,928	133,103
Female	201703	194806	186449	176979	155298
Influenza vaccines
Delivered doses	3,004,521	2,847,156	3,235,771	6,035,439	5,947,126
Expected doses^b	78,455	73,226	78,974	137,787	120,821
Administered doses	87,316	89,955	99,384	114,253	85,815

Abbreviations: GP, general practitioner; HSD, health search database.

^aAs per https://demo.istat.it/.

^bThe calculation involves, for each season, multiplying the number of older adults under the care of GPs by the ratio of doses delivered to the total number of older adults as indicated in official demographic reports on regional bases.

Figure 1 shows the patterns of distribution and administration of influenza vaccines for older adults in the 5 seasons under study. The pattern of distribution to the regional LHAs started from the first available dates related to the deliveries. The trend of the distribution and administration pattern represents the velocity (i.e. doses/week) of these actions in the weeks just preceding and overlapping the epidemic phase. In all influenza seasons, vaccine administration (presumably the first dose being injected) started 2–4 weeks after the first delivery. Over the five subsequent seasons, the difference between the maximum (peak) of vaccine distribution and vaccine administration was equal to 1, 1, 2, 4, and 3 weeks. With the exception of the 2020–2021 season, in which there was an extended delay between distribution and administration patterns, the other 4 seasons were consistent in terms of differentials between distribution and administration patterns.

Figure 1. Trends in influenza vaccine distribution to regional authorities and vaccine administration by general practitioners by influenza season.

Across the 5 seasons, the median distribution velocity ranged from 341 to 833 doses/week, the median administration velocity from 152 to 270 doses/week, the median differences between distribution and administration velocities from 289 to 622 doses/week, and vaccine coverage from 23.65% to 36.53% (Table 2). The distribution velocity exhibited a rising trend over the 5 seasons, except for the 2018–2019 season, which decreased relative to its predecessor. A similar pattern was observed for the administration velocity across the 5 seasons, but no variance was noted during the initial 3 seasons. The fourth (2020–2021) season was marked by a more substantial increase in the differential between distribution and administration velocity. There was no trend for both distribution (quantile regression, p = 0.421) and administration velocity (quantile regression, p = 0.07) over the 5 seasons. In addition, no apparent trends were observed in the difference between distribution and administration velocities (quantile regression, p = 0.189) or in vaccine coverage over the 5 seasons (quantile regression, p = 0.142).

Table 2. Vaccine distribution and vaccine administration velocities and vaccine coverage over 5 influenza seasons in Italy.

Influenza season	Velocity, median doses/week (IQR)			Vaccine coverage (%)
	Distribution	Administration	Difference^c
2017–2018	518 (235–925)	176 (106–408)	395 (108–777)	23.65
2018–2019	341 (102–675)	209 (88–348)	289 (110–551)	24.51
2019–2020	466 (159–952)	152 (96–299)	304 (200–669)	28.49
2020–2021	740 (261–1360)	203 (102–351)	622 (216–971)	36.53
2021–2022	833 (242–1514)	270 (112–499)	431 (167–1068)	30.75
p value over 5 seasons^c	0.421	0.065	0.189	0.142

Abbreviations: IQR, interquartile range (25th-75th percentile); MAH, marketing authorization holder; SD, standard deviation; ‘week’ is the observation time unit for each influenza seasons.

^aFrom MAH data weighted by official demographics.

^bDifference between distribution and administration rates (analyses are clustered by region of residence and then inferred to the Italian cohort).

^cTest on median clustered by regions of residence.

In Table 3 are reported the results on testing the variation between distribution and administration velocity and number of all-cause/respiratory-cause-related hospitalizations among older adults being vaccinated with aTIV/aQIV. We captured statistically significant association with a greater effect size for three of the examined seasons, nominally 10% (2017–2018), 54% (2018–2019), and 12% (2020–2021) increase versus the median number of all-cause hospitalizations for each differential dose/week between distributed and administered vaccines. Although with a reduced effect size, we found consistent results when the analysis was restricted to hospitalizations likely due to respiratory causes. Indeed, we obtained statistically significant association for three of the examined seasons, nominally 1% increase versus the median number of respiratory-related hospitalizations for each differential dose/week between distributed and administered vaccines.

Table 3. Association between difference of distribution and administration velocities (doses/week of influenza vaccines) and hospital admissions over 5 influenza seasons according to quantile (median) regression.

	All-cause hospitalizations					Respiratory-related hospitalizations
Influenza season	Hospital admissions (n)	Vaccinees (n)	Cumulative incidence (%)	β^a (95% CI)	p value	Hospital admissions (n)	Vaccinees (n)	Cumulative incidence (%)	β^a (95% CI)	p value
2017–2018	3303	87,316	3.78	0.10 (0.01–0.20)	0.034	179	87,316	0.21	0.01 (0.1–0.2)	0.014
2018–2019	3899	89,955	4.33	0.54 (0.10–1.20)	0.030	211	89,955	0.22	0.03 (0.01–0.06)	0.034
2019–2020	4373	99,384	4.40	0.45 (−0.12 to 1.01)	0.185	212	99,384	0.21	0.01 (−0.02 to 0.03)	0.524
2020–2021	4283	114,253	3.75	0.12 (0.01–0.22)	0.030	263	114,253	0.23	0.01 (0.001–0.013)	0.016
2021–2022	899	85,815	1.05	0.01 (−0.1 to 0.05)	0.854	21	85,815	0.02	—^b	—
Pooled estimate^c				0.10 (0.01–0.20)

Abbreviations: CI, confidence interval.

^aProportional (%) increase or decrease versus the median value of the outcome for each individual vaccine dosage/day.

^bReduced analysis power for low number of hospital admissions (n = 21).

^cMeta-analysis using random effect model.

When the results of the individual season were meta-analyzed, a random effect model (I² = 71%; Q test, p = 0.008), the pooled estimate was equal to a 10% increase (beta coefficient = 0.10 [95% CI, 0.01–0.20]) in all-cause hospitalization.

For the sensitivity analyses, when we adopted the subset of GPs reporting a vaccine practice coverage of ≥ 55% (No. of GPs ranging 89–317), we found significant associations for both distribution (quantile regression, p < 0.001) and administration velocity (quantile regression, p = 0.009) over the 5 seasons (Table S1). In addition, the difference between the distribution and administration velocities (quantile regression, p = 0.001) was significantly different over the 5 influenza seasons (Table S2). The association between the median number of all-cause hospitalizations and each differential dose/week between distributed and administered vaccines was statistically significant, with 9% (2019–2020) and 7% (2020–2021) increases, in two seasons (Table S3). The pooled estimate stemming from the meta-analysis showed a 6% significant increase (beta coefficient = 0.06 [95% CI, 0.02–0.09]; I² = 68%; Q test, p = 0.014) in all-cause hospitalization for each differential dose/week of influenza vaccine between distributed and administered doses.

4. Discussion

To our knowledge, this is the first study quantifying the time lapses between distribution of influenza vaccine to regional health authorities and vaccine administration to older adults in a primary care setting. Over 5 epidemic seasons, we found a difference between velocities of vaccine deliverables to regional authorities and injections by GPs, although there was no significant trend. Instead, we found statistically significant increases with velocity variations between vaccine distribution and vaccine administration and all-cause/respiratory-related hospitalization rates in 3 out of 5 influenza seasons. Interestingly, when the analyses were restricted to GPs with greater accuracy in registering vaccine administration, there was a significant difference over the 5 seasons for both distribution and administration velocities. Overall, the findings from the sensitivity analyses were consistent with the results from the primary analyses. It may be expected that data from GPs with higher influenza vaccine coverage among their patients would corroborate the seasonal variations in the effective execution of the vaccination campaign.

The present study is the first quantitative evidence supporting qualitative observations reported in prior work. For example, a survey conducted among 448 primary care physicians in the US showed that the majority of physicians experienced delays in vaccine delivery during the 2009 H1N1 pandemic. Delays were attributed to insufficient vaccine supply, communication problems with vaccine distributors, and logistical issues related to vaccine distribution. The study also revealed that physicians who reported earlier delivery of vaccines had higher vaccination rates among their patients. The authors concluded that improving the vaccine delivery system is crucial to achieve higher vaccination coverage rates and reduce the impact of influenza on public health [23]. Other authors claimed that multifaceted interventions that address both system-level and patient-level barriers are likely to be the most effective in improving influenza vaccination coverage in primary care practices [16]. Our findings did not show a direct association between velocity of distribution and velocity of administration and vaccine coverage. We did not formally test such a relationship given the ecological nature of our study. Indeed, we did not have enough adjustment factors (e.g. those measuring vaccine hesitancy and/or other related socioeconomic features) to test a hypothesis including ‘vaccine coverage’ as response variable. However, the time to reach older subjects remains a relevant component to ensure vaccine protection at patient and population levels. This dimension is clearly related to proper and on-time provision of vaccine supplies [24].

Our findings may provide valuable information on the efficiency of the administration and distribution of the substance or medication, allowing for further analysis and potential improvements in the process. In particular, we observed heterogeneity over the 5 influenza seasons concerning the time of allocation of doses to vaccinators, which was clearly identified among GPs with better recordings of influenza vaccines. In this respect, variation in how vaccine deliveries are organized by different regional and local health authorities may benefit from further study. After proving the presence of intra-class (region) correlation, we clustered by region for each regression analysis. Notably, as shown in Figure 1, starting in the 2019–2020 season, there appears to be a growing trend toward greater delays between the maximal distribution and maximal administration of vaccines. The difference between the distribution and administration peak passed from only 1 week in the 2017–2018 and 2018–2019 influenza seasons to 2 (2019–2020), 4 (2020–2021) and 3 weeks (2021–2022) for the other three seasons. The results of the present study provide evidence regarding the positive impact of minimizing the time period between vaccine distribution and vaccine administration. Therefore, the most favorable strategy is to order vaccines from MAH closer to the time regional LHAs deliver vaccines to GPs. Efficient vaccine distribution systems play a vital role in achieving high vaccination coverage rates and reducing the spread of vaccine-preventable diseases. In this respect, Manzoli and coworkers reported a significant association between influenza-like illness increase and decline in vaccine coverage. They also demonstrated that each 1% rise in vaccine coverage could prevent roughly 2,690 influenza-like illness cases among older adults in Italy [9]. Along this line, study by Stockwell et al. found that the implementation of an EHR-based vaccine reminder system in New York City increased influenza vaccination coverage rates among adolescents from 5% to 34% [25]. Similarly, a study by Fiks et al. indicated that the use of an EHR-based reminder system for pediatric influenza vaccination resulted in a 41.2% increase in vaccination coverage rates [26].

Our study has several limitations. In the first 3 seasons, the number of administered doses was higher than the number estimated by distribution data. Given that we estimated the number of doses expected to be allocated to GPs using regional census data, some GPs operating under certain LHAs, not coded in HSD, could have received additional doses of aTIV/aQIV. In addition, in the period preceding the COVID-19 pandemic, some older adults might have autonomously purchased the vaccine from pharmacies, which were then injected by GPs. Nevertheless, this bias should have artificially reduced the difference between distribution and administration velocity, thereby further emphasizing the importance of logistic management in vaccine allocation to the final vaccinators. Furthermore, when we conducted the meta-analysis, the pooled estimate of beta coefficients was consistent with those obtained for seasons in which a significant association was captured. Based on this finding, we recalculated the analyses by limiting the GPs to those reporting a vaccine practice coverage of ≥ 55% – a group that administered an average of 79.8% of doses over the 5 seasons – the results of the quantile regression still captured an association for those seasons in which the differential velocity was particularly relevant. It appears that for this subset of GPs, vaccination is a priority in terms of both administration Elk Grove Village (IL)and record-keeping. The effect exerted by differential velocity in the 2021–2022 season is difficult to assess, given the poor circulation of influenza virus. Irrespective of the fact that vaccine effectiveness is not an issue in case of absence of viral circulation, the two related curves of distributed and administered doses/week were those showing the minimum overlap, so indicating the need to reduce this gap. A second limitation is that the administration velocity might be underestimated, given the availability of other vaccines in older individuals (excluding the high-dose vaccination which was not yet available in the included seasons). Nevertheless, the fact that our analysis of GPs who vaccinated ≥ 55% of their patients still captured a statistically significant association was reassuring. Furthermore, the specific use of aTIV/aQIV in the older adult population has become predominant over recent influenza seasons, and the assessment of distribution velocity should not be biased by the residual presence of other vaccines. Third, we used an ecological design that may have led to ecological fallacy. Confirmative studies using individual patient records are therefore needed. Nevertheless, we compared the subgroups of older adults being vaccinated for influenza, and they were similar in terms of comorbidities, as demonstrated in our previous studies using HSD [18,27]. As such, the effect of confounding on our results should be modest. Fourth, some hospitalizations might have not been necessarily occurred during the epidemic period. Nevertheless, a reduced number (0.6% (n = 101/16757); to be conservative we counted them in September and October of the related seasons) of the hospitalizations occurred in the period usually preceding or overalpping (not necessarily excluding) the epidemic phase. In any case, the results should be biased toward the null while we yielded statistically significant associations. Fifth, for the last 3 seasons, the effect on the results of COVID-19 and its related vaccination cannot be excluded. Nevertheless, given that we examined cohorts of older adults being vaccinated for influenza, there is no reason to think about a differential effect exerted by concurrent vaccinations. Along this line, the absence of some associations in the 2021–2022 season is reassuring given the low circulation of influenza virus. Fifth, our definition of events was based on EHRs and may not have specifically captured influenza-related hospital admissions. However, the operational definition of this outcome has been largely adopted in prior investigations with consistent results [17–19]. Finally, the influenza vaccine uptake is also influenced by a variety of other factors, such as vaccine hesitancy, vaccine availability, vaccine accessibility, and public health policies related to vaccine distribution and vaccine administration [28]. Therefore, reducing the time lapse between vaccine delivery to regional LHAs and actual allocation to GPs may not necessarily guarantee the greatest effectiveness of a vaccine campaign. A comprehensive approach that addresses all these factors is necessary to achieve high vaccine coverage rates.

In conclusion, we reported a difference between velocities of vaccine deliverables to regional authorities and vaccine administrations by GPs, although no significant trend was obtained over the examined seasons. On the other hand, we found a statistically significant increase with variations of velocity between vaccine distribution and vaccine administration and all-cause/respiratory-related hospitalization rates in 3 out of 5 seasons. These findings emphasize the significance of a reduced time period between vaccine distribution and vaccine administration for reducing the incidence of all-cause and respiratory-related hospitalizations, in addition to other factors that influence vaccination barriers and hesitancy.

Declaration of interest

F Lapi and E Marconi provided consultancies in protocol preparation for epidemiological studies and data analyses for Seqirus, GSK, Sanofi, Pfizer and MSD. A Rossi and C Cricelli provided clinical consultancies for Seqirus, GSK, Sanofi, Pfizer, and MSD. E Fallani, M Salvatore, and M Cambiaggi are full-time employees of Seqirus S.r.l. The present study was, however, conceived and carried out during the PhD program at University of Siena by E Fallani and M Salvatore, and outside working hours at Seqirus S.r.l. The authors have no other relevant affiliations or financial involvement with any organization or entity with a financial interest in or financial conflict with the subject matter or material discussed in the manuscript apart from those disclosed.

Reviewer disclosures

A reviewer on this manuscript has disclosed that they receive funding from an investigator initiated grant from Sanofi. Additionally, a reviewer on this manuscript has received honoraria for their review work. Peer reviewers on this manuscript have no other relevant financial or other relationships to disclose.

Acknowledgments

Medical consultant C. Gordon Beck and Amanda M. Justice provided editorial support in the preparation of this article, which was funded by CSL Seqirus.

Supplementary material

Supplemental data for this article can be accessed online at https://doi.org/10.1080/14760584.2023.2291184.

Additional information

Funding

This study was funded by Seqirus S.r.l., a pharmaceutical company that manufactures and commercializes influenza vaccines.