Article Text

Predictors of seasonal influenza vaccination among healthcare workers in hospitals: a descriptive meta-analysis
  1. Josien Riphagen-Dalhuisen1,2,
  2. Giedre Gefenaite1,2,
  3. Eelko Hak1,2
  1. 1Division of Pharmacoepidemiology and Pharmacoeconomics, Department of Pharmacy, University of Groningen, Groningen, The Netherlands
  2. 2Department of Epidemiology, University Medical Center Groningen, Groningen, The Netherlands
  1. Correspondence to Josien Riphagen-Dalhuisen, University of Groningen, Department of Pharmacy, Division of Pharmacoepidemiology and Pharmacoeconomics, PO Box XB45, Antonius Deusinglaan 1, 9713 AV Groningen, The Netherlands; j.riphagen{at}


Objective Vaccinating healthcare workers (HCWs) against influenza is one of the most important methods of decreasing influenza transmission among at-risk patients in healthcare facilities. However, despite recommendations, the rate of uptake of influenza vaccine among HCWs remains low. The objective of this meta-analysis was to determine the most important predictors of seasonal influenza vaccine acceptance among HCWs in hospitals.

Method A literature search of PubMed and Embase resulted in 4586 hits. Screening of the titles, abstracts and full text identified 13 studies eligible for inclusion in the meta-analysis. Based on the crude data, pooled risk ratios (Mantel-Haenszel risk ratios, mhRR) and their 95% CIs were calculated using Mantel-Haenszel analysis to estimate the associations of predictors with influenza vaccination status.

Results and conclusion Knowing that the vaccine is effective (mhRR 2.22; 95% CI 1.93 to 2.54), being willing to prevent influenza transmission (mhRR 2.31; 95% CI 1.97 to 2.70), believing that influenza is highly contagious (RR 2.25; 95% CI 1.66 to 3.05), believing that influenza prevention is important (mhRR 3.63; 95% CI 2.87 to 4.59) and having a family that is usually vaccinated (RR 2.32; 95% CI 1.64 to 3.28) were statistically significantly associated with a twofold higher vaccine uptake. We therefore recommend targeting these predictors when developing new influenza vaccination implementation strategies for hospital HCWs.

  • Influenza
  • healthcare workers
  • vaccination
  • epidemiology
  • health and safety
  • health promotion
  • immunisation
  • preventive medicine

Statistics from


Healthcare workers (HCWs) with influenza can transmit the virus to patients who are at increased risk of developing serious complications. Seasonal influenza vaccination reduces influenza-confirmed episodes among healthy adults by approximately 75% when matched with circulating strains1 and there is evidence that vaccinating HCWs against influenza reduces the number of respiratory tract infections among these workers.2 3 The main reason for vaccinating against influenza, however, is to prevent severe morbidity and mortality among patients, as shown by Hayward et al.4 In a recent Cochrane review,5 an overall reduction in all-cause mortality of 32% (95% CI 16% to 45%) was found in long-term care facilities in which some of the HCWs were vaccinated versus control homes. Although the included studies were heavily criticised by the authors, in most countries the evidence so far is still perceived as favouring vaccination. One of the studies from that review6 revealed that in the control homes 20% of a sample of 30 deaths were caused by influenza, while in the intervention homes none of the sampled deaths had evidence of influenza infection, giving a 100% reduction in deaths caused by influenza. In addition, Thomas et al estimated a 29% reduction (95% CI 10% to 45%) in influenza-like illness in intervention homes compared with control homes.5 It is well established that during influenza epidemics, the aetiological fraction of culture or PCR-confirmed influenza virus in elderly patients is high at between 55% and 67%.7 Recently, Van den Dool et al8 developed a mathematical model to predict the effects of increasing vaccine uptake among HCWs in hospitals. Assuming a 73% vaccine efficacy among HCWs, it was estimated that only seven of the workers needed to be vaccinated to prevent one influenza infection in a hospital patient. Another of their conclusions was that due to stochastic variations, more than 184 homes would be needed in each intervention arm to detect a statistically significant reduction in influenza episodes among patients between homes with zero and 50% vaccine uptake by HCWs. Therefore, a huge trial would be needed to confirm this assumption which in itself is less relevant from a clinical perspective. Meanwhile, current evidence supports the provision of large investments to improve vaccine uptake among HCWs, so waiting for more evidence is simply unethical.

Despite evidence in favour of vaccinating HCWs, the uptake rate of seasonal influenza vaccine among HCWs remains far below target. For example, the vaccination rates in the studies included in our meta-analysis ranged from as low as 2.1%9 to 62%.10 Many studies have examined possible predictors of influenza vaccination acceptance by HCWs, but due to small sample sizes, different study designs, settings, populations and assessed predictors, it is difficult to get an overall picture. Although current reviews provide some evidence regarding the most important predictors of acceptance of influenza vaccination in HCWs,11 12 systematic or pooled data analysis is missing. Additionally, no distinction has been made between evidence from intervention and evidence from non-intervention studies.

We therefore conducted a meta-analysis to assess the predictors of seasonal influenza vaccination in HCWs working in hospitals by pooling the crude data from non-intervention studies.


Search strategy

A literature search was performed using PubMed and Embase on 9 December 2009 using the following search strategy: (‘Vaccines’ (MeSH) OR ‘Vaccination’ (MeSH) OR ‘Immunization’ (MeSH) OR ‘vaccination’ (TIAB) OR ‘vaccin*’ (TIAB)) AND (‘influenza’ (TIAB) OR ‘influenza virus’ (TIAB) OR ‘influ*’ (TIAB) OR ‘flu’ (TIAB) OR ‘Influenza, Human’ (MeSH) OR ‘Influenza Vaccines’ (MeSH)) AND (‘HCWs’ (TIAB) OR ‘healthcare worker’ (TIAB) OR ‘hospital personnel’ (TIAB) OR ‘hospital staff’ (TIAB) OR ‘staff’ (TIAB) OR ‘personnel’ (TIAB) OR ‘Hospitals’ (MeSH) OR ‘Long-Term Care’ (MeSH) OR ‘Nursing Homes’ (MeSH) OR ‘Patient Care’ (MeSH) OR ‘Health Personnel’ (MeSH)) AND (English (LA) OR Dutch (LA)) NOT ‘child’ (MeSH Terms) NOT review(pt).

We used MeSH terms for the PubMed search and Emtree terms for the Embase search.

We limited our search to articles in the English or Dutch language. The studies were included in the meta-analysis if they were non-pandemic, non-intervention studies performed among HCWs working in hospitals, reported current influenza vaccination status and had available crude data on at least one predictor of interest for this study. Reference lists were scrutinised to identify other relevant studies, but none were found (figure 1).

Figure 1

Flow chart of literature search.

During the selection procedure, articles were also independently judged on their quality according to several quality criteria by two reviewers (JRD, GG), and discussed with EH in case of disagreement. The main quality criteria were response rate, sample size, number of included determinants, data collection methods, study design and vaccine uptake measures.

Predictors and outcome

We classified the predictors for influenza vaccination into several groups according to behavioural and implementation models (see table 1).

Table 1

Predictors of influenza vaccination

Current influenza vaccination status was chosen as the outcome.

Statistical analysis

We calculated the average response rates (%) of the included studies and their 95% CIs using SPSS V.16.0. After the raw data were pooled, the risk ratios (RR) and Mantel-Haenszel risk ratios (mhRR) and their 95% CIs were calculated using K. Rothman's Episheet.13 14 When information about a predictor was available from only one study, RR instead of mhRR were calculated.

Statistical heterogeneity between the studies was examined visually by comparing mhRR and 95% CIs in the forest plot across studies.


Description of the studies

Our search resulted in 4586 hits. Abstracts, titles and full texts were screened and duplicates excluded, resulting in 13 eligible studies.9 10 15–25

All 13 studies included in the meta-analysis were cross-sectional. The total sample included 84 880 HCWs and the average response rate was 56.9% (see table 2 for more details). Vaccination status was determined by questionnaire in 10 studies, and was known from clinical records in three studies.20 24 25 Twelve studies used a questionnaire during the same season that the vaccine was administered9 10 16–25 and one study used a questionnaire one season later.15 All studies included at least one demographic predictor (gender, age or occupation), eight studies included at least one behavioural predictor, and two studies included an organisational predictor. All studies were carried out in developed countries: six were performed in the USA,9 10 17 18 21 22 three in Europe,16 20 25 two in Canada19 23 and one in Australia,24 and one study was a multi-nationality survey.15 In most of the studies, influenza vaccines were provided for free by the hospitals on an annual, voluntary basis.

Table 2

Characteristics of included studies

Predictors of influenza vaccination in HCWs

The results from the pooled analyses of the associations between influenza vaccination status and different predictors are shown in table 3. The forest plot in figure 2 presents the results graphically.

Table 3


Figure 2

Forest plot (number of studies with this specific predictor, total number of persons per predictor). CDC, Centers for Disease Control and Prevention.

Male gender, being aged 40 years and older and being a physician were the demographic predictors that were positively associated with being vaccinated, while being a nurse was negatively associated with vaccine uptake. Statistically significant factors resulting in a twofold higher vaccine uptake were knowing that the vaccine is effective, being willing to prevent influenza transmission, believing that influenza is highly contagious, believing that influenza prevention is important and having a family that is usually vaccinated.


With this meta-analysis we aimed to determine the most important predictors of acceptance of seasonal influenza vaccination by HCWs in hospitals. Our results showed that the strongest predictors were knowing that the vaccine is effective, being willing to prevent influenza transmission, believing that influenza is highly contagious, believing that it is important to prevent influenza and having a family that is usually vaccinated. Other predictors had a weaker association but might be helpful in developing future influenza vaccination campaigns. Our findings are consistent with the results of previous reviews which also suggested that raising awareness about vaccine effectiveness and the risks of influenza makes vaccination more likely to be accepted.11 12

Although it seemed that there were many more predictors of influenza vaccination acceptance in HCWs, the risk ratios of many predictors were close to 1. This suggests that interventions focusing on these predictors most likely would not achieve a significant increase in vaccination uptake rates in HCWs as these predictors have little impact on vaccine acceptance.

We also looked at demographic characteristics to determine if they predicted vaccination status. We found that being male, being older than 40 years and being a physician increased the chances of being vaccinated, while being a nurse was associated with less vaccination. However, the risk ratios only differed very slightly from 1 for these predictors. Nevertheless, these population characteristics could be used in implementation programs to define specific target groups according to age, gender and occupation.

In addition, in our study we assessed the predictors of influenza vaccination in non-intervention studies performed in similar settings, namely (teaching) hospitals. We deliberately excluded intervention studies because merging the results from intervention and non-intervention studies might have introduced bias: knowledge derived from an educational campaign used in an intervention study could have influenced a particular predictor, which would not have occurred in a non-intervention study.

Regarding heterogeneity, most predictors point in the same direction with overlapping risk ratios. Therefore, these predictors are fairly homogeneous.

The fact that our study was unsuitable for multivariate analysis should be taken into account when interpreting the results. It might be that some predictors contain other information. For example, male gender was found to be a predictor of influenza vaccine acceptance. However, it is likely that there are more male than female physicians, and therefore the relative risks might be overestimated. We were not able to address the issue of how the predictors interacted with one another.

Finally, this meta-analysis was performed with the underlying purpose of providing evidence for use in future influenza vaccination campaigns among HCWs in hospitals during common epidemics, so we described the factors that were associated with seasonal influenza vaccine uptake. Pandemic influenza is different from seasonal influenza in many ways, but some of the described factors might also be important in a pandemic. Some reports on pandemic influenza vaccination have shown predictors similar to those described here, although other predictors also might play a role in vaccine uptake.26 27 Influenza vaccination will only be successful in HCWs if they are properly educated and if the vaccine is easily accessible. Therefore, we recommend targeting these predictors when developing new influenza vaccination implementation strategies for hospital HCWs.


Our main aim was to assess the predictors for seasonal influenza vaccination in HCWs in hospitals. Our meta-analysis provided information on the strength of the predictors of current influenza vaccination status. Future studies could use this information for their interventions and target the predictors that seem to have the most influence on vaccine uptake, and also focus on educating HCWs in order to prevent misinformation.


View Abstract


  • Funding The Netherlands Organization for Health Care (ZONMW) sponsored this study (grant number 125030012).

  • Competing interests None.

  • Provenance and peer review Not commissioned; externally peer reviewed.

Request permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.