INTRODUCTION
The 23-valent pneumococcal polysaccharide vaccine (23vPPV) was licensed 25 years ago in the USA and is currently available in most developed countries. It is recommended in people aged ⩾65 years, and in children aged >2 years at high risk of invasive pneumococcal disease [1]. In 1999, pneumococcal vaccination programmes were introduced in several Spanish regions [Reference Salleras and Salleras2] following international recommendations [1]. During the last two decades several studies have proved the effectiveness of pneumococcal vaccination in the elderly [Reference Hedlund3–Reference Jackson, Neuzil, Plotkin, Orenstein and Offit6]. However, vaccination coverage of target groups is not optimal. Only a few Spanish studies have reported on vaccine uptake, finding rates of between 35% and 53·1% in different populations [Reference Batalla, Campins and Moraga7–Reference Vila-Córcoles9].
In Spain, influenza vaccination has been provided to people aged ⩾65 years and to established risk groups since 1992 [10]. Some studies have reported a steady increase in vaccination uptake during the last few years [Reference Batalla11–Reference Vazquez-Fernández del Pozo13], reaching 63·7% in 2003 [Reference Vazquez-Fernández del Pozo13]. Both influenza and pneumococcal vaccination are provided free of charge by the primary healthcare centres (PHCs) of the public Spanish Health Service.
Some studies have shown that differences in 23vPPV vaccination coverage are associated with sociodemographic characteristics and access to health services. The Centers for Disease Control and Prevention in the USA found lower 23vPPV coverage in African-Americans and Hispanics, while higher coverages were positively associated with educational levels and negatively with perceived health status [14]. Another study found higher coverage in people of middle or high socioeconomic status and those who had visited their physician during the last year [15]. Kamal et al. identified differences in 23vPPV coverage according to ethnicity, educational level, type of medical insurance and physician visits during the last year [Reference Kamal, Madhavan and Amonkar16]. A Dutch study of factors predictive of non-compliance in a programme introducing 23vPPV and influenza vaccination in people aged ⩾65 years found that factors related to the pattern of use of health services associated with vaccination coverage included private medical insurance and difficulties in visiting the physician, as well as age (<75 years) [Reference Opstelten17]. 23vPPV is normally offered together with the influenza vaccine to patients in whom both are indicated. Studies have also found differences in influenza vaccination coverages which are associated with socioeconomic factors and access to health services [Reference Kamal, Madhavan and Amonkar16, Reference de Andres18–Reference Jiménez-García20]. Analysing these factors in a Spanish population could provide useful information for the future planning of vaccination campaigns. Therefore, the objective of this study was to evaluate the association between specific socioeconomic factors and the use of health services and pneumococcal and influenza vaccination in hospitalized people aged ⩾65 years.
MATERIAL AND METHODS
This was a cross-sectional, descriptive study that formed part of a larger case-control study designed to evaluate 23vPPV effectiveness [Reference Domínguez21]. Study subjects were non-institutionalized patients aged ⩾65 years admitted to three acute hospitals in two Spanish autonomous communities (Catalonia and Galicia) between May 2005 and January 2007: The Hospital Clinic and Hospital de Bellvitge (Greater Barcelona, Catalonia) and Hospital Juan Canalejo, La Coruña (Galicia). All three are general hospitals serving essentially urban populations, with 861, 920 and 871 beds, respectively, and 45 144, 35 747 and 42 499 annual discharges, respectively. The methods of the case-control study have been previously reported [Reference Domínguez21]. Briefly, patients aged ⩾65 years hospitalized due to community pneumonia were included. Three control subjects matched by sex, age, date of hospitalization and underlying disease were selected for every case. Patients were located and asked to cooperate in the study on hospital wards [Reference Domínguez21]. The research protocol was approved by the Ethics and Scientific Committee of each hospital.
Independent variables
Age, sex and underlying disease were obtained from medical records. A short interview with the patient (or with the family when the patient's ability to respond was impaired) collected information on sociodemographic variables and the use of health services. The variables included were: last occupation, educational level (no formal education, complete or incomplete primary, secondary, university and other), type of household (‘Who do you live with?’ – alone, with partner only, with children, others), the number of physician visits during the last year and the length of time with the same general practitioner. Socioeconomic status was classified according to the last job as unqualified or manual workers (classes IVa, IVb, V) and others (classes I–III) [Reference Domingo-Salvany22].
Patients were classified into three strata according to the risk of pneumonia due to the underlying disease using medical records: high risk (solid organ neoplasia, haematological neoplasia, solid organ transplant, bone marrow transplant, radiotherapy, immunosuppressive therapy, corticosteroid therapy, splenectomy, autoimmune disease, chronic renal failure requiring haemodialysis, nephrotic syndrome, disabling neurological disease or AIDS); moderate risk (asymptomatic HIV infection, diabetes mellitus, heart failure, chronic obstructive pulmonary disease, chronic liver disease, renal failure not requiring haemodialysis) and low risk (none of the conditions mentioned).
Vaccination status
Patients were considered to have received 23vPPV when the information was recorded in the hospital record or adult vaccination card. Vaccination status was verified by the PHC. Patients were considered as vaccinated only if the vaccine had been given ⩾15 days before the onset of pneumonia for cases or ⩾15 days before the date of hospitalization for controls. The same criteria were used to determine influenza vaccination status. In most cases, information was obtained within 15 days after first contact. The deadline for receiving information was 4 weeks.
Statistical analysis
Associations between study variables and 23vPPV and influenza vaccination were evaluated using bivariate logistic regression analysis. Logistic regression analysis was performed to account for the effects of confounding variables, introducing variables with a value of P<0·25 in the bivariate analysis and those potentially related to the response. The analysis was performed with SPSS software version 12.0 (SPSS Inc., USA).
RESULTS
A total of 2130 patients were initially screened in the emergency room, of whom 73 were excluded because vaccination status could not be ascertained, leaving 2057 patients, of whom 355 women were excluded due to lack of information on social class. Therefore, 1702 patients were finally included, in whom 23vPPV uptake was 59·9% and 65·6% had received the influenza vaccine during the previous season (Table 1). Patients' characteristics are shown in Table 1.
Patients' characteristics and influenza and pneumococcal vaccination coverage
Bivariate analysis
An association was found between 23vPPV coverage and a moderate risk of pneumonia (OR 1·41, 95% CI 1·02–1·80), age (OR 1·03, 95% CI 1·02–1·05), lower social class (OR 1·40, 95% CI 1·07–1·84), and living with a partner (OR 1·96, 95% CI 1·44–2·67) or with children (OR 1·54, 95% CI 1·12–2·11) (Table 2). In contrast, a negative association was found between 23vPPV coverage and high-medium educational levels compared to those with low levels (OR 0·70, 95% CI 0·55–0·89). 23vPPV was associated with visiting the physician during the last year, the numer of physician visits and the last seasonal influenza vaccination (OR 6·95, 95% CI 5·57–8·69).
Association between pneumococcal vaccination in people aged ⩾65 years and study variables in bivariate and multivariate analyses
Influenza vaccination coverage was associated with a moderate risk of pneumonia (OR 1·87, 95% CI 1·45–2·43), male sex (OR 1·66, 95% CI 1·31–2·10), living with a partner (OR 1·70, 95% CI 1·24–2·31) and the number of physician visits during the last year (Table 3).
Association between influenza vaccination in people aged ⩾65 years and study variables in bivariate and multivariate analyses
Multivariate analysis
In the adjusted multivariate analysis, factors independently associated with 23vPPV were age (OR 1·04, 95% CI 1·02–1·06), living with a partner (OR 1·72, 95% CI 1·20–2·46) and influenza vaccination during the previous influenza season (OR 6·64, 95% CI 5·25–8·41) (Table 2).
Factors associated with influenza vaccination (Table 3) were male sex (OR 1·44, 95% CI 1·11–1·87), age (OR 1·02, 95% CI 1·01–1·04), a moderate risk of pneumonia compared to a low risk (OR 1·58, 95% CI 1·20–2·07), living with a partner compared to living alone (OR 1·52, 95% CI 1·09–2·13), and frequent physician visits during the last year [1–6 annual visits (OR 2·65, 95% CI 1·81–3·88) or >6 visits (OR 3·83, 95% CI 2·58–5·69)]. In constrast, a high risk of pneumonia was negatively associated with influenza vaccination compared to a low risk (OR 0·77, 95% CI 0·59–0·99).
DISCUSSION
The 23vPPV and influenza vaccine coverages obtained in this study (59·9% and 65·6%, respectively) are similar to the mean coverage found in previous studies [Reference Kissling23–Reference Christenson and Lundbergh25]. Older age was associated with higher vaccination coverage of both vaccines, as in previous studies [14, Reference de Andres18]. Higher 23vPPV and influenza vaccination coverages were found in patients living with partners compared with those living alone. However, the factor most closely associated with 23vPPV was influenza vaccination in the vaccination season before hospital admission. This might be explained by PHC vaccination strategies, which use seasonal influenza vaccination as an opportunity to promote and administer 23vPPV in susceptible patients [Reference Opstelten17, Reference Kyaw26]. Women had lower influenza vaccination coverage, as previously reported [Reference Sarria and Timoner27].
The association between a moderate risk of pneumonia due to the underlying disease and higher influenza vaccination coverage may be explained by the fact that these patients also have more indications for influenza vaccination [Reference Bayas and Campins28]. As expected, a positive relationship between influenza vaccination and the number of physician visits during the year prior to hospital admission was also found. Other studies have also linked a higher probability of influenza vaccination with greater use of health services [Reference Kamal, Madhavan and Amonkar16, 19, Reference Jiménez-García20].
Although some studies have found an association between vaccination coverage and higher socioeconomic and educational levels [14, Reference Kamal, Madhavan and Amonkar16, 19, Reference Schoefer29], our results do not support this association. We found higher 23vPPV coverage in patients with both low socioeconomic and educational levels, although the association was not retained in the ajdusted models.
De Andrés et al. [Reference de Andres18] and Sarria & Timoner [Reference Sarria and Timoner27] found higher influenza vaccination coverages in municipal areas with <10 000 inhabitants. Our results do not support this association, since coverage in patients from rural (<10 000 inhabitants) and urban (⩾10 000 inhabitants) areas were very similar both for 23vPPV (59·2% and 60·1%, respectively) and influenza vaccination (63·9% and 65·8%, respectively). However, this variable was not included in the study design and sampling procedures.
Strengths and limitations
A potential limitation of our study is the exclusion of women who had not worked or who identified themselves as housewives, as there was not sufficient information to categorize the social class. 23vPPV and influenza vaccination coverages were analysed in these women before exclusion (23vPPV 60·3%, influenza vaccine 60·0%) and were compared to the remaining female patients (23vPPV 55·9%, influenza vaccine 56·4%), with no significant differences being found (23vPPV, P=0·225; influenza vaccine, P=0·322).
Another possible limitation is the potential information bias associated with self-report or proxy reporting (interview with patient or relatives on admission), especially with respect to the validity of sociodemographic factors and the use of health services. To minimize this bias we obtained clinical data and other demographic data from medical records. In addition, the vaccination status was checked using vaccination cards or PHC records.
This study was based on hospitalized patients. Therefore, the cases are not representative of the general population, although they are representative of hospitalized cases in the areas studied, because they were recruited from the reference hospitals for each area. In addition, from the public health perspective, one of the most important objectives of vaccination is to avoid hospitalization due to influenza and pneumococcal disease in this particular age group. Our results are consistent with other studies carried out in samples from the general population [14, Reference Kamal, Madhavan and Amonkar16, Reference de Andres18, 19, Reference Christenson and Lundbergh25, Reference Sarria and Timoner27], suggesting that selection bias did not compromise the study results. Knowledge of the factors associated with vaccination in hospitalized patients is crucial in order to plan vaccine strategies for this age group.
The main strengths of the study are its size and the reliability of data collection, as previously reported [Reference Domínguez21]. In summary, we found that influenza and pneumococcal vaccination coverages were influenced by sociodemographic factors and the use of health services. Efforts should be made to increase vaccination in population groups not sufficiently reached by current vaccination schedules. This will require better coordination between public health services and PHC staff. The seasonal influenza vaccination campaign could provide a good opportunity to increase both influenza and pneumococcal vaccination coverages.
APPENDIX
Members of the working group are: Joan Batalla [Department of Health, Generalitat of Catalonia and CIBER Epidemiología y Salud Publica (CIBERESP), Spain], Lluís Salleras [Department of Public Health, University of Barcelona, and CIBER Epidemiología y Salud Publica (CIBERESP), Spain], Adriana Manzur (Bellvitge University Hospital–IDIBELL, L'Hospitalet de Llobregat, Spain), Silvia Sugrañes (Hospital Clínic–IDIBAPS, Barcelona, Spain), Wenceslao Varona, Ángel Ramón Terren (Hospital Royo Villanova, Zaragoza, Spain), José Miguel Celorrio, Carmen García, Esperanza Clemente (Hospital Ernest Lluch, Calatayud, Spain), Isabel Justo, Ana Arévalo, Susana Rivera (Hospital Juan Canalejo, A Coruña, Spain).
ACKNOWLEDGEMENTS
This study was financed by the Fondo de Investigaciones Sanitarias (Projects 04/1835, 04/0151, 04/1573, 04/2516 and 04/2351), CIBER Epidemiología y Salud Pública: Project (CIBERESP) [PD07-006], and the Red de Centros de Epidemiología y Salud Pública of the Carlos III Health Institute (exp C03/09), Madrid, Spain. The authors thank all primary healthcare centres who freely provided vaccination information.
DECLARATION OF INTEREST
None.
