Although there is a high variability in published data on annual incidence of CAP, the incidence in the present series (3.0/1,000 adults) is within the range described by others [2, 6, 13, 19, 38] but two-times higher than the annual incidence in a previous study in a similar area . Although hospitalization rates depend on the structure of the primary and secondary healthcare system of the studied area, the percentage of patients admitted in hospitals found in our study (41.5%) is in accordance with published rates [7, 13, 14, 36, 39] but lower than the percentage described in other studies [37, 40] where probably underestimation of ambulatory cases could have occurred. The multivariate analysis performed to investigate variables associated with hospitalization identified liver disease, stroke, dementia, COPD, diabetes mellitus and age as significant variables. Some of them had been previously described as being significantly different between inpatients and outpatients . The length of hospitalization showed linear positive correlation with PSI score, associated in turn with age and Charlson comorbidity index. Of interest is the low percentage of patients requiring ICU admission in our series (0.8%), markedly lower than in other studies [1, 37].
No two studies of the etiology of CAP are the same. Differences in frequency of pathogens may be due to healthcare delivery (primary vs. secondary care), hospital and ICU admission practices, population factors (comorbidities, alcoholism…) and study factors . In our series microbiological studies were performed in 62% patients, with great differences in relation to site of care (100% inpatients vs. 35% outpatients), probably because for patients managed in the community microbiological investigations are not recommended routinely . In nearly half of cases with microbiological tests, the etiological agent could not be identified in accordance with results of previous studies [9–15], showing again the need for improving microbiological diagnostic tools for CAP. As expected, S. pneumoniae was the most frequent etiological agent among patients with identified pathogen, accounting for approximately one-third of outpatients and two-third of inpatients. Vaccination with the 23-valent pneumococcal vaccine has been reported as cost-effective in individuals aged ≥45 years in our area , however, up to 23.1% of patients with Pn-CAP in our series had been previously vaccinated. This finding suggests the need for improving pneumococcal vaccination strategies, an important point since in Spain nonsusceptibility rates to β-lactams and macrolides in S. pneumoniae are among the highest in the world [15, 42]. However, regardless antimicrobial susceptibility, the link between outcome and serotypes has been described in a published meta-analysis . The fact that levofloxacin was the compound most frequently used as treatment in our series (with macrolides or β-lactam plus macrolides combinations used in <5% cases) contrasts with data from a previous study in our area carried out in 1993–95 where figures were completely different with 65.5% use of macrolides , and could be associated with the high non-susceptibility rates to β-lactams and macrolides in S. pneumoniae in our country. In the present study, Legionella pneumophila was only identified in 7% patients with etiological filiation, without differences in relation to the site of care. This suggests that when establishing empirical antimicrobial therapy in our region, coverage of Legionella should be considered even in outpatients.
Pneumonia is the fifth to ninth leading cause of death in developed countries [39, 44, 45]. Despite mortality of CAP varies depending on the series and site of care, mortality of CAP managed in the community is <1% and from 4% to 10% for hospitalized CAP . In our series, mortality (2.5% for inpatients and 0% for outpatients) was low and length of stay short, probably related to PSI distribution of patients at admission. However, other indices of patient evolution as readmission (19.1%) or time to recovery (29.9 days) were similar or slightly higher than those published [7, 37, 46, 47]. As in a previous study , readmissions were associated with comorbidities.
It has been reported that costs of CAP requiring hospital admission are eight-times higher than those managed in the community . This ratio was lower (3.3-times) in our study where the adjusted mean total cost was €698.6 per-outpatient and €2,332.4 per-inpatient. However, costs in the present study were higher than those in previous studies focused on direct hospital costs, both in our country (€1,210 , and €1,847 ), Germany (€1,201)  or Italy (€1,587) . Higher costs in our country are probably related with the more recent study period, the higher rate of readmissions, and mainly with the higher number of analyzed variables, including lost of working days. However, it should be considered that indirect costs in this study could be underestimated since they were calculated based on the minimum interprofessional salary in Spain instead of mean salary amount. In addition, 39.4% patients were ≥65 years, the majority probably retired and thus, without lost of working days. In this sense, the fact that indirect costs were higher in outpatients than inpatients may be related with the significantly higher percentage of patients ≥65 years among inpatients vs. outpatients (59.0% vs. 25.6%). Interestingly, hospitalized patients with Pn-CAP showed significantly higher overall mean costs and direct costs, due to longer time to recovery in relation to significantly higher percentage of comorbidities (COPD, diabetes mellitus and asthma) and PSI score.
Several limitations can be identified in our study making difficult extrapolation of results. It was limited to a specific geographical area, with a specific healthcare system, and costs calculated with local data. In addition, limitations derived from the retrospective nature of the observational study design are also applicable.