Background
Vaccination is the most effective medical intervention ever introduced [
1,
2]. A recent study from the US Centers for Disease Control and Prevention reported that vaccination would have prevented 322 million illnesses and 732,000 premature deaths in children born during the period 1994–2013 [
1,
3]. The smallpox vaccination has been responsible for the total eradication of the disease since its implementation in the 20th century [
2]. Similarly, vaccines prevent millions of cases of infectious diseases worldwide, allowing for an increase in life expectancy from 59 to 70 years from 1970 through 2010 [
2]. Nevertheless, vaccination coverage is lower than desired. From 2000 to 2010, the percentage of French respondents who declared they were unfavourable towards vaccination increased from 9 to 38 %, possibly due to the 2009 influenza A (H1N1) pandemic [
4].
Vaccination experts acknowledge that public confidence in vaccination is waning [
5,
6]. This “erosion in trust” is illustrated by the decline of vaccination coverage and is observed in most developed countries. Other evidence of the vaccine confidence gap is the increase in controversies surrounding various vaccines: the measles, mumps and rubella (MMR) vaccine in the United Kingdom, the hepatitis B vaccine in France and the H1N1 vaccine in many countries in 2009–2010 [
5].
The relevance of the vaccination is also questioned by healthcare professionals. A study reported that French primary care physicians (PCPs) and paediatricians applied loosely the 2010 recommended vaccination schedule [
7]. The tendency of clinicians reconsidering vaccination is not limited to France. One study showed that only 42 % of Italian paediatricians knew about the recommended vaccination schedule, and that only 10 % of Italian clinicians had positive opinions regarding the usefulness of recommended vaccinations [
8]. In Canada, 37 % of healthcare professionals considered that there were too many vaccines and 36 % of them considered that a good lifestyle could eliminate the need for vaccination [
9].
PCPs play a key role in the vaccination system in France. The relationship between the patient and the PCP and more generally healthcare professionals is the cornerstone of maintaining public confidence in vaccination, including addressing parents’ concerns on vaccines [
9‐
11]. Several studies have shown PCPs’ influences in the recommendations on vaccination to their patients [
12‐
14].
It is therefore crucial to understand the factors that influence PCPs’ commitment to vaccination. To address this need, the French Society of General Medicine (‘Société Française de Médecine Générale’ -SFMG-) developed a tool, the Determinants of Intentions to Vaccinate (DIVA
©) questionnaire, following a rigorous qualitative stepwise methodology [
15,
16]. The aims of the present study were to define the scoring rules and to assess the measurement properties of the DIVA tool.
Materials and methods
The DIVA questionnaire
The hypothetical conceptual framework of the DIVA questionnaire included 56 items, grouped into six thematic domains and one commitment domain: ‘Disease characteristics and expected benefits’ (9 items), ‘Properties of the vaccine’ (10 items), ‘Information about the vaccination’ (8 items), ‘Practical and organisational aspects’ (6 items), ‘Adaptation to the patient’s profile’ (11 items), ‘PCP’s individual experience’ (5 items) and ‘PCP’s commitment to the vaccination approach’ (7 items).
Study setting and design
An observational, cross-sectional and multi-centre study involving French PCPs was conducted. All PCPs from the mailing list of the SFMG were invited by email to participate in the study. PCPs who agreed to enter the study were asked first to complete the socio-demographic form on a website. PCPs were then randomly assigned to one of the six vaccine-prevented diseases (VPDs) (measles, pertussis, pneumococcus infection, seasonal influenza, human papillomavirus -HPV- infection and tetanus) and were asked to complete the DIVA questionnaire. As tetanus is a compulsory vaccine in France and was very well accepted by PCPs during the qualitative phase, the tetanus disease group was considered as a control group in the study and thus included by design the smallest number of PCPs (the allocation ratio was one PCP for tetanus for three PCPs for each of the five other VPDs). There was no eligibility criterion to enter the study.
A modified Delphi panel, composed of the members of the DIVA scientific committee who are PCPs and specialists in vaccination, was carried out using the RAND/UCLA method in the early stage of the study [
17]. It was conducted to appraise the position of the panellists on the six thematic domains of the DIVA questionnaire according to the six VPDs. The Delphi Panel was carried out into 3 phases. In the first phase, panellists were asked to complete a questionnaire with six propositions corresponding to the six thematic domains of the DIVA questionnaire for each VPD. Each proposition was presented in a numerical scale graduated from 1 to 9, with 1 indicating that the thematic domain was not a determinant of decision-making in vaccinating against the concerned VPD and 9 indicating that the thematic domain was a determinant of decision-making in vaccinating against the concerned VPD. In the second phase, panellists discussed online their responses. In the third phase, panellists were asked to complete the questionnaire again. At the end, each proposition was ranked according to the agreement between the panellists and the strength of the agreement.
Statistical analyses
Item selection and definition of the scoring rule were based on the quality of completion of the DIVA questionnaire and on the distribution of item responses.
The DIVA questionnaire included both psychometric and composite domains. A psychometric domain is composed of items correlated with each other and measuring a single concept. In the DIVA questionnaire, the psychometric domain was the ‘PCP’s commitment to the vaccination approach’ domain. A composite domain is composed of items that are a combination of indicators of a common concept but not necessarily correlated with each other. In the DIVA questionnaire, composite domains were the six thematic domains. The statistical methods that are described below were adapted to the different natures of the concepts measured.
Principal component analysis (PCA) was performed to challenge the hypothetical conceptual framework of the DIVA questionnaire. This analysis aims at defining the grouping of items into domains by examining the self-organization of items [
18]. Interpretable components with an eigenvalue greater than one represented the empirical domains.
An extension of the Rasch model, the Rating Scale Model (RSM) used for polytomic response scale [
19], was performed on all items of the ‘PCP’s commitment to the vaccination approach’ domain to validate the existence of a unique underlying construct (called the “latent trait”) measuring the PCP’s commitment to vaccination [
20]. All items were scaled with four categories (“Totally disagree”, “Somewhat disagree”, “Somewhat agree” and “Totally agree”). The Rasch model utilizes person and item parameters to determine the probability of an item score [
21]. Raw scores for both persons and items are transformed into measures (known as “locations”) using a logistic mathematical function [
21]. Model fit statistics were computed to indicate whether all items met the unidimensional measurement criterion using the residuals (range of acceptability from −2.5 to 2.5 [
21]) and the item location to look at the deviation of the observed data from the model expectation [
22]. The person-item threshold map, which is a graphical representation of person and item locations (measured in logits) plotted on the same latent trait along the x axis was also used.
The measurement properties of the DIVA questionnaire were assessed, including reliability and validity.
Reliability is the degree to which the instrument is free from measurement error. Internal consistency reliability (the extent to which items within a domain are consistent with each other and measure a single underlying concept) was assessed for the ‘PCP’s commitment to the vaccination approach’ domain, using Cronbach’s alpha [
23].
Validity is the degree to which the instrument measures what it is supposed to measure. Construct validity was conducted to assess the degree to which DIVA thematic domains can distinguish the determinants of intentions to vaccinate among the six VPDs groups. The hypotheses for this analysis were formulated by the Delphi panel. The ranks of DIVA thematic domains for each VPD calculated from the Delphi panel were compared to those obtained from descriptive analyses.
Ethical consideration
The present study was conducted in compliance with national ethical principles and the list of the members of the SFMG was submitted to the French National Commission for Data Processing and Privacy (‘Commission Nationale de l'Informatique et des Libertés’ -CNIL-).
Discussion
In this study, the DIVA questionnaire confirmed its reliability and validity in a large sample of PCPs and across a variety of VPDs. The quality of completion of the DIVA questionnaire was very good. A few missing data were observed for items. The analyses on the hypothetical conceptual framework of the DIVA questionnaire led to some minor adjustments: one item removed and a few changes in three item labels. The Rasch model validated the commitment domain and its six items as a measure of the level of commitment of PCPs in the vaccination against a specific disease [
20].
The final 55-item DIVA questionnaire includes the six original thematic domains, covering the concepts measuring a large set of factors that influence PCPs’ attitudes and beliefs toward vaccination including: disease characteristics, vaccine properties, scientific and non-scientific communication, practical and organizational aspects, the patient’s profile and the PCP’s individual experience. The DIVA questionnaire also includes a specific domain assessing the commitment of the PCP to vaccination against the specific disease of interest.
To our knowledge, our study was the first to present simultaneously all the six thematic domains plus a commitment domain in a measurement approach applicable to a large variety of VPDs.
All DIVA thematic domains were identified in the literature, and several studies have explored specific determinants in specific diseases [
24]. The most frequently studied domain was related to the ‘Adaptation to the patient’s profile’ domain, which showed the lowest mean score across our study sample. A survey was conducted among pediatricians and family medicine physicians in 2010, before the quadrivalent HPV vaccine (HPV4) was recommended for 11- to 12-year-old boys [
25]. The most commonly reported barrier to HPV4 was financing (out-of-pocket cost and lack of adequate reimbursement for vaccine). The second barrier was related to parents’ attitudes, not thinking that HPV4 was necessary for their sons [
25]. In 2008, a French survey with 1,000 family physicians showed that physicians who did not think that HPV vaccine had a negative effect on the image of sexuality were more in favour of vaccination [
26].
Previous studies on PCPs’ attitudes and practices in vaccination showed that French PCPs’ concerns were focused on the safety of the vaccine [
10,
12,
14,
27]. This is partially consistent with our finding, as the ‘Disease characteristics and expected benefits’ domain showed high scores across all studied diseases. However, our study showed that ‘Properties of the vaccine’ domain was not the major concern. Rather, we found that the ‘Information about vaccination’ and the ‘Practical and organizational aspects’ domains constituted the main barriers to vaccination for PCPs. Other studies confirmed that the lack of information on vaccination and the recommendations for vaccination might impact vaccine uptake rates in the general public [
10,
12,
14,
27]. The level of acceptance of vaccination by patients was shown to be also an important factor that encouraged PCPs to vaccinate [
28‐
31]. To summarize, it seems that PCPs are fully aware of the advantages of vaccination, generally confident in the vaccines properties, but face difficulties when communicating to patients and addressing practical organization of vaccination, which negatively impact their commitment to vaccination.
In this study, six VPDs were chosen according to their level of acceptance: the vaccines against measles, pertussis, pneumococcus infection and tetanus (compulsory in France) diseases have not been subject to much debate; the vaccines against the seasonal influenza and HPV infection diseases have created controversies [
4,
32]. The expected barriers toward vaccination for each of the diseases were assessed with a Delphi Panel approach, and were generally confirmed by our findings in the surveyed population of PCPs. However, a couple of inconsistencies appeared between the rank of DIVA scores given by descriptive analyses and those given by Delphi panel in the pertussis, seasonal influenza and tetanus diseases. Indeed, PCPs seemed less convinced than panellists that the disease characteristics of the seasonal influenza and the expected benefits of vaccination were sufficient to engage them in vaccination. This could have been influenced by the debate in the A/H1N1 outbreak in 2009 [
27]. The observed discrepancies between panellists and surveyed PCPs about pertussis and tetanus vaccination are more compelling and need further exploration.
In addition to the assessment of the determinants of PCPs’ attitudes toward vaccination, the DIVA questionnaire specifically measures their commitment toward vaccination against each of the six selected VPDs. The ‘PCP’s commitment to the vaccination approach’ domain had very good internal consistency reliability overall and in the six VPDs, showing its accuracy in measuring the commitment of a PCP to vaccination. In addition, it was validated by a Rasch model, encouraging future researchers to use the ‘PCP’s commitment to the vaccination approach’ as a stand-alone tool either as an outcome measure of intervention programmes encouraging vaccination, or as an explanatory factor in public health research.
We can assume that our survey has not reached PCPs who would have strong beliefs against vaccination. However, the issue is not to convince a minority of opponents about the relevance of vaccination for public health. Rather, the variations observed across VPDs clearly show that the commitment to vaccine is explained by well-identified factors. The DIVA questionnaire is the first standardized tool examining the barriers to vaccination that specific diseases may face and informing where efforts should be concentrated in order to increase adoption of positive attitudes toward vaccination by PCPs.
Conclusions
The DIVA questionnaire was found to be reliable and valid, allowing PCPs’ attitudes and beliefs toward vaccination to be comprehensively assessed.
In a long-term perspective, the DIVA questionnaire could help develop fine-tuned, efficient interventions to overcome the barriers that prevent PCPs from engaging in vaccination. Ultimately, these interventions will improve vaccine coverage rates in the general population in France.
Furthermore, the findings on the ‘PCP’s commitment to the vaccination approach’ domain of the questionnaire suggest the use of this domain as a stand-alone questionnaire.
Acknowledgments
The study received a grant from Pfizer. We thank all the PCPs who agreed to participate in the study. We also thank Isabelle Guillemin for providing support in the manuscript writing and Hayley de Freitas, who is a native English speaker, for reviewing the manuscript for the English quality.