European Respiratory Society


The objective of our study was to evaluate the effectiveness of the 23-valent pneumococcal polysaccharide vaccine (PPV) in preventing hospital admission for community-acquired pneumonia (CAP) in people ≥65 yrs of age.

We conducted a matched case–control study in patients with CAP admitted to five Spanish hospitals. Cases were persons aged ≥65 yrs admitted to hospital through the emergency department, who presented a clinical and radiological pattern compatible with pneumonia, assessed using established criteria. We matched each case with three control subjects by sex, age (±5 yrs), date of hospitalisation (±30 days) and underlying disease. The study period was May 1, 2005 to January 31, 2007. The PPV immunisation status of cases and controls was investigated. Adjusted ORs for vaccination were calculated using logistic regression analysis.

A total of 489 cases and 1,467 controls were included in the final analysis. The overall adjusted vaccination effectiveness for all patients was 23.6% (95% CI 0.9–41.0). The adjusted vaccination effectiveness for immunosuppressed patients was 21.0% (95% CI -18.7–47.5).

Our results suggest that the PPV may potentially reduce hospitalisations for pneumonia in the elderly and supports vaccination programmes in this age group.

Community-acquired pneumonia (CAP) is an important cause of morbidity and mortality in elderly people and those of any age with underlying diseases 1, 2. In Spain, the overall incidence in adults varies between two and 10 cases per 1,000 persons per yr in all ages and between 14 and 35 per 1,000 persons per yr in persons aged >70 yrs 3, 4. In a Spanish study, the incidence increased dramatically by age in elderly people (9.9/1,000 in people aged 65–74 yrs versus 29.4 in people aged ≥85 years) 4. Hospitalisations due to CAP increase with age and may reach 61% for all ages, and 67% in people aged >65 yrs 5, 6. Case-fatality rates may reach 17% in patients aged >75 yrs 5, with higher rates in those with underlying disease 1, 3, 5. A substantial proportion of CAP cases requiring hospitalisation are caused by Streptococcus pneumoniae: 30–50% according to most reports 1, 711. Bacteraemic pneumococcal pneumonia, the most severe disease form, accounts for only 10–20% of adult cases of CAP caused by S. pneumoniae, with non-bacteraemic pneumococcal pneumonia being much more frequent 1.

The 23-valent pneumococcal polysaccharide vaccine (PPV) has been available in the USA for 25 yrs and is currently licensed in most developed countries. Vaccination is usually recommended for people aged ≥65 yrs and for high-risk persons aged >2 yrs 1, 12, 13. There is a general consensus that observational studies have shown vaccination to be effective in preventing invasive pneumococcal disease 1416. However, vaccination rates are not high in most countries, partly due to doubts about the vaccine's efficacy and vaccination effectiveness in preventing non-bacteraemic pneumococcal pneumonia 1417.

Laboratory methods for diagnosing non-bacteraemic pneumococcal pneumonia have a low sensitivity and specificity, and are difficult to carry out in clinical practice. Therefore, all-cause pneumonia has been proposed as a more appropriate outcome measure for evaluating vaccination effectiveness (VE) 1, 15. If a substantial proportion of hospital admissions for CAP are of pneumococcal origin and vaccination is effective against non-bacteraemic and bacteraemic disease, this should be reflected in a decline in admissions for all-cause pneumonia.

In 1999, pneumococcal vaccination programmes for the elderly and high-risk individuals were introduced in several Spanish regions 12 according to international recommendations 13. Vaccination coverage in some regions reached 35% in 2001, and reached >40% in later years 18. This coverage and the large number of hospitalisations for CAP in Spain facilitated the objective of this study: to evaluate the effectiveness of PPV in preventing hospital admission for CAP in people aged >65 yrs by assessing whether the proportion of vaccinated subjects was lower in hospitalised patients with pneumonia than in those without pneumonia.


Study design

We conducted a matched case–control study in patients with CAP admitted to five hospitals in three Spanish regions. The study period was May 1, 2005 to January 31, 2007.

Case selection

We defined a case as a person aged ≥65 yrs admitted to hospital through the emergency department, who presented with an infiltrate on chest radiograph compatible with pneumonia and one or more of the following symptoms or signs of acute lower respiratory tract infection: cough, pleuritic chest pain, fever >38°C, hypothermia <35°C or dyspnoea within the past 24 h 1, 5, 10. Exclusion criteria were: institutionalised patients, patients with nosocomial pneumonia (onset ≥2 days after hospital admission), patients whose initial diagnosis of pneumonia was not confirmed during the hospital stay, and cases of CAP in whom the pneumococcal and influenza vaccination status could not be determined.

Selection of controls

We selected three hospital controls for each case: two medical patients and one surgical patient. Controls aged ≥65 yrs admitted through the emergency department with a diagnosis other than pneumonia were selected from the admission lists of each participating hospital. On selection, the vaccination status of controls was not known and, if the status could not be determined later, they were excluded.

Demographic and other variables

For each case and control we obtained information on age, sex, dates of hospitalisation and discharge (alive or dead), smoking, risk-consumption of alcohol and the presence or absence of underlying diseases or conditions. We stratified each case according to the level of risk and the degree of immunosuppression associated with the underlying disease. Stratum I (high risk) included all patients with conditions associated with immunocompromisation: solid organ or haematologic neoplasia with activity in the past year, solid organ or bone marrow transplant, radiotherapy within the past 3 months, immunosuppressive therapy or treatment with corticosteroids ≥20 mg daily in the past month, asplenia, autoimmune disease, chronic renal failure requiring haemodialysis, active nephrotic syndrome and AIDS. We also included those with neurological disease impeding daily activities. Stratum II (moderate risk) included immunocompetent patients with one or more high-risk medical condition: heart failure grade 3 or 4, chronic obstructive pulmonary disease (COPD), diabetes mellitus, chronic renal failure not requiring haemodialysis, and chronic liver disease. Stratum III included patients not included in strata I or II.

To guarantee the true value of the overall effectiveness of vaccination in preventing all-cause pneumonia hospitalisation, the numbers of cases in the three strata were selected to reflect the real proportions of the corresponding strata in hospitalisations for all-cause pneumonia in Catalonia, Spain (J. Carratalà, personal communication). When the number of subjects required for each stratum was reached, recruitment for this stratum was stopped.

Matching cases and controls

We matched each case with three control subjects by sex, age (±5 yrs), date of hospitalisation (±30 days) and underlying disease. If the case had more than one high-risk medical condition and was immunosuppressed (stratum I), control subjects were matched using the immunosuppressive disease of greatest duration (if recorded) or another immunosuppressive condition suffered by the case, if disease duration was not available. If controls with the same underlying disease were not found, we sought controls with diseases from the same stratum.

If the case had more than one high-risk medical condition but was not immunosuppressed (stratum II), controls were matched using the disease of greatest duration (if recorded) or by some other condition of risk of the case if information on duration was not available.

If the case had no high-risk condition (stratum III), we selected controls with no such conditions.

If no adequate controls were found, the intervals for age and the date of hospitalisation of the case were extended.

Information collection

Patient information was obtained through two sources: 1) review of written hospital medical records (underlying diseases, alcohol consumption, history of pneumonia and vaccination status) and 2) interview of the patient or close relatives (spouse or offspring) for visits to the doctor in the past year, alcohol consumption and vaccination status using a questionnaire completed by qualified staff. Vaccination status was also obtained from the vaccination card and healthcare centre vaccination registers.

Ascertainment of pneumococcal and influenza vaccination status

We sought information on the vaccination status in all health centres that each patient had visited during the 5 yrs before hospitalisation. The vaccination status was ascertained by staff blinded to whether the patient was a case or control. As vaccination status may be recorded in different documents, we searched all relevant sources and considered PPV as administered only when confirmed by the patient's hospital record, adult vaccination card or primary healthcare vaccination record. Patients were considered vaccinated when the vaccine had been given ≥15 days before the onset of pneumonia, for cases or ≥15 days before the date of hospitalisation for controls. The same criteria were used to determine prior influenza vaccination (IV) status.

Sample size

We calculated the minimum required sample size according to standard criteria 19. We assumed a prevalence of vaccination in the control group of 0.35 20 and VE against all-cause pneumonia of 35%. With an α error of 0.05 (two-tailed), a β error of 0.20 and three controls per case, we calculated that 269 cases and 807 controls would be needed. Because vaccination coverage was estimated to be lower in some of the participating regions, we increased the number of cases to 405 and controls to 1,215.

Statistical analysis

We analysed the differences observed between cases and controls according to the study variables using paired tests. The McNemar Chi-squared test or binomial distribution test, when appropriate, were used for categorical variables and the paired t-test for continuous variables. We assumed a two-tailed distribution for all p-values and considered p<0.05 to be statistically significant.

We used conditional logistic regression (CLR) to account for the effects of confounding variables. The variables introduced in the CLR analysis were influenza vaccine status, variables potentially related to the vaccination response and those with a p-value <0.1 in the univariate analysis. In the final analysis, variables with a significance of p<0.05 were included in the model. We calculated adjusted ORs for immunosuppressed (strata I) and immunocompetent patients (stratum II and III) separately and for all three strata combined.

VE was estimated using the following:Embedded ImageThe study was approved by the ethics committee of each participating hospital.


Recruitment of cases and controls

Of the 598 cases recruited, 35 (5.9%) were excluded because their vaccination status (PPV or IV) could not be determined. We recruited 1,605 controls, of which the PPV or IV status could not be determined in 38 (2.4%).

Of the 563 cases in whom vaccination status was determined, three appropriate controls were not found for 58. Of the remaining 505 complete sets, 16 were excluded because one or more control subject exceeded the age interval by more than 8 yrs. Therefore, 489 complete sets were included in the final analysis: 200 (41%) in stratum I, 190 (39%) in stratum II and 99 (20%) in stratum III.

Characteristics of study subjects

The characteristics of cases and controls are shown in table 1. The distribution of study variables was similar in the two groups, although more cases than controls had experienced a previous episode of pneumonia. The only significant differences in the distribution of underlying diseases between cases and the three controls were in the proportions with solid organ and haematologic neoplasia and COPD; diabetes mellitus and corticosteroid therapy showed significant differences between cases and two of the controls.

View this table:
Table 1– Characteristics of cases and controls for all patients#

Of 489 sets, 200 were immunosuppressed and 289 were immunocompetent. The characteristics of cases and controls according to immune status are shown in table 2.

View this table:
Table 2– Characteristics of cases and controls according to immune status

Vaccination effectiveness

The history of pneumococcal vaccination in cases and controls, the unadjusted and adjusted ORs and the unadjusted and adjusted VE according to immune status are shown in table 3. The overall adjusted VE for all three strata combined was 23.6% (95% CI 0.9–41.0). For overall effectiveness the significant variables included finally in the model were history of pneumonia, solid organ neoplasia, haematologic neoplasia, chronic obstructive pulmonary disease and diabetes mellitus.

View this table:
Table 3– Effectiveness of 23-valent pneumococcal polysaccharide vaccination in preventing hospitalisation for pneumonia

The adjusted VE for immunosuppressed cases was 21.0% (95% CI -18.7–47.5). For immunosuppressed patients, the significant variables included in the model were history of pneumonia, solid organ neoplasia, haematologic neoplasia, and COPD.

When strata II and III were combined into one group of immunocompetent patients, the adjusted VE was 23.6% (95% CI -7.2–45.6). For immunocompetent patients, the significant variables included in the model were a history of pneumonia, diabetes mellitus and smoking.


We have found that the PPV has an effectiveness of 23.6% (CI 0.9–41.0) in preventing hospitalisations due to pneumonia. Although evidence is limited, some observational studies have shown a protective effect of PPV against hospitalisation for CAP. Nichol and co-workers 21, 22 and Wagner et al. 23 found that vaccination reduced hospital admissions for all-cause pneumonia. Protection was observed both against cases of disease and against deaths from all-cause pneumonia 22, 23. Protection against pneumonia was also confirmed by a prospective cohort study by Vila-Córcoles et al. 24. However, Jackson et al. 25 found no reduction in hospitalisation for pneumonia, despite noting significant reductions in immunocompetent patients in the occurrence of both pneumococcal bacteraemia (54%) and all-cause mortality (12%) 26. A historical cohort study by Ansaldi et al. 27 and a case–cohort study by Skull et al. 28 also failed to show that vaccination reduced hospital admission for CAP.

The effectiveness in preventing hospitalisations due to pneumonia (23.6%) in our study was close to that found by Nichol 22 (27%) and Vila-Córcoles et al. 24 (26%) but lower than that found by Wagner et al. 23 in a study carried out in a long-stay geriatric hospital (72.1%).

A recently published meta-analysis of randomised clinical trials carried out in an elderly population failed to show protection of the 23-valent pneumococcal vaccine against all-cause pneumonia 29. The Cochrane Collaboration has recently published a systematic review of English-language studies evaluating the efficacy and effectiveness of the 23-valent pneumococcal vaccine 30. The review evaluated the effectiveness of the vaccine in reducing all-cause mortality but not the prevention of hospitalisations due to pneumonia. Moberlley et al. 30 also reviewed the results of clinical assays designed to evaluate the efficacy of the vaccine in preventing all-cause pneumonia, and found a global result of 29% (95% CI 3–48), similar to the results of our study (23.6% 95% CI:0.9–4).

Only 30–50% of cases of CAP are thought to be due to S. pneumoniae 1, and thus the effectiveness of PPV against all cases of pneumococcal pneumonia (non-bacteraemic and bacteraemic) would be expected to be much higher. In the study by Austrian et al. 31 of a 13-valent pneumococcal polysaccharide vaccine in South African gold miners, vaccine efficacy was 82% against bacteraemic pneumococcal pneumonia and 78.5% against putative (bacteraemic and sputum culture-positive) pneumococcal pneumonia caused by vaccine serotypes. Observational studies have shown that PPV prevents ∼50–70% of hospitalisations for invasive pneumococcal disease (all serotypes) 1, 15. If 30–50% of all cases of CAP in our population were caused by vaccine-type S. pneumoniae, our findings suggest that if the level of vaccination-induced protection against all CAP cases was 23.6% (table 2), the level of protection against vaccine-serotype pneumococcal pneumonia was close to the level of protection (50–70%) found in observational studies of invasive disease alone 16.

Some studies suggest that the PPV reduces rates of intensive care unit (ICU) admission and in-hospital CAP mortality 3234. Moreover, even if the proportion of non-bacteraemic pneumococcal pneumonia admissions prevented by vaccination were much lower than suggested by our results, preventing these additional hospital admissions and reducing ICU admissions and in-hospital CAP mortality would still dramatically increase the cost-effectiveness of a vaccine that is already very cost-effective in preventing invasive disease alone 35.

Our study, like other observational epidemiological studies, has strengths and limitations. One strength was the large sample size (489 cases and 1,467 controls), which allowed statistically significant results to be obtained for the whole population studied. The overall adjusted VE (all cases and controls) was 23.6% (95% CI 0.9–41.0). The lack of significance in immunocompetent subjects may be due to the small sample size of this group.

In case–control studies of vaccination there is always the possibility that bias can distort the results and decrease the validity of the findings 36. One source of bias is incomplete or inaccurate ascertainment of the vaccination status. This did not occur in our study because information on vaccination status was obtained retrospectively by blinded investigators using common records for both cases and controls, and these records were completed before the study period began.

To control for confounding variables, controls were matched with cases for sociodemographic and medical variables (risk factors) that could have influenced disease incidence. Even so, statistically significant differences between cases and controls were observed for six medical variables (history of pneumonia, solid organ neoplasia, haematologic neoplasia, corticosteroid therapy, diabetes mellitus and COPD). We adjusted for the possible confounding effect of these variables using conditional logistic regression. Influenza vaccination could have been a possible confounding factor, although we believe it had no effect because the variable was introduced into the conditional logistic regression analysis and because the proportion of vaccinated cases and controls was similar.

Introduction of the seven-valent conjugated vaccine in children aged <2 yrs in the first decade of this century does not seem to have caused any bias. In the USA, the incidence of invasive pneumococcal disease in the elderly has fallen since conjugate vaccination programmes were introduced. This is largely because decreased rates of nasopharyngeal colonisation by vaccine serotypes in children have reduced rates of transmission to older individuals 3739. In Spain, the seven-valent conjugated pneumococcal vaccine has not been included in the official routine vaccination schedules of the Ministry of Health or those of the three regions participating in this study. Nonetheless, it is estimated that vaccination coverage with the conjugated vaccine in Spain during the period of this study was between 30–40% 40, 41. In the USA, rates of invasive disease in adults began to fall soon after conjugate vaccination of children was introduced, although vaccination rates were low and within the range of those reported in Spain 37, 38. It is conceivable that conjugate vaccination of children in Spain had already reduced absolute rates of invasive pneumococcal disease in older adults. Nevertheless, the relative reduction in rates of CAP observed in our study can be considered to have occurred independently of the effects of conjugate vaccination of children. The only effect conjugate vaccination of children might have had on our estimate of effectiveness (i.e. relative risk reduction) of PPV in older adults would have been to reduce overall rates of CAP, thus leading to a requirement for larger sample sizes to detect an effect of the polysaccharide vaccine.

Current recommendations for PPV vaccination are based on studies of vaccination effectiveness against invasive pneumococcal disease. Our results reinforce these recommendations and suggest that the cost-effectiveness of PPV is greater than reported, since all economic studies of PPV carried out until the present have only considered its protective value against invasive pneumococcal disease.


The other members of the working group for the Study of Prevention of CAP in the Elderly are: A. Manzur (Hospital Universitari de Bellvitge, L'Hospitalet de Llobregat, Barcelona, Spain); S. Sugrañes (Hospital Clínic, Barcelona, Spain); A. Terren (Hospital Royo Villanova, Zaragoza, Spain); S. Rivera, I. Justo and A. Arevalo (Hospital Juan Canalejo, Coruña, Spain); C. García and E. Clemente (Hospital Ernest Lluch, Calatayud, Spain); X. Sintes (Public Health Agency of Barcelona, Spain); J. Batalla (Department of Health, Generalitat of Catalonia and CIBER Epidemiología y Salud Pública (CIBERESP), Barcelona, Spain).

The authors thank D.S. Fedson (Sergy Haut, France) for his encouragement in undertaking this study and his comments on the manuscript.

The authors also thank all health centres that provided vaccination information on their patients. None received any economic compensation.


  • Support Statement

    This study was supported by the Instituto de Salud Carlos III (project numbers: 04/1835; 04/0151; 04/2516; 04/1573; 04/2303; 04/2351), CIBER Epidemiologı´a y Salud Pu´ blica (CIBERESP).

  • Statement of Interest

    None declared.

  • Received October 29, 2009.
  • Accepted December 29, 2009.


View Abstract