In the past years, many studies have been conducted aiming at finding new strategies to lower incidence and improve clinical outcomes of CAP. To determine cost-effectiveness of these strategies, knowledge about causing microorganisms, clinical outcomes, and related costs is needed. To our knowledge, this is the first study that studies the potential associations between costs of hospitalisation for CAP and its microbial aetiology. The main finding in the present study is that costs related to hospitalisation for CAP show great variation between patients, and CAP caused by S. pneumoniae and Staphylococcus aureus is associated with significantly higher costs, mainly due to longer duration of hospital stay.
In this study, S. pneumoniae was confirmed as the most prevalent causative pathogen in CAP (24.6%). Compared to other aetiological groups, median LOS (8.5 days), rate of ICU admission (8%), and one-year mortality (9.7%) were relatively higher for pneumonia caused by S. pneumoniae, despite the relative younger age of patients of this aetiological group (60.4 ± 19.0 years versus 64.4 ± 17.6 years, p:0.033). These findings are in accordance with other CAP studies that also reported higher disease severity and increased need for ICU admission in S. pneumoniae pneumonia [20, 21]. In agreement with these findings, we showed S. pneumoniae to be an independent cost-driving factor (on average plus 18% per hospitalisation).
Interestingly, Staphylococcus aureus could also be identified as an independent cost driving factor. CAPs caused by this pathogen were associated with a longer LOS and a higher mortality rate as well. This unfavourable outcome might be explained by the difficulty of treating Staphylococcus aureus pulmonary and systemic infections. Recently, Restrepo et al. have reported that late ICU admission versus early ICU admission is more prevalent in cases of CAP caused by Staphylococcus aureus, which aligns with the higher mortality rate observed in our study .
In our study, median total costs of hospitalisation were almost €4,000 per patient. These expenditures are higher compared to similar studies performed in Germany and Spain (median costs of €1,362 , €1,683  and €1,553 , respectively), but lower than reported in a study from the United Kingdom (£1,700-5,100, depending on length of stay ) and a European study (US$6,530 in a secondary-level hospital in the Netherlands and US$8,444 in a teaching hospital) . The most likely explanation for these discrepancies in hospital costs are expected to be differences in registration, and individual resource item prices. Furthermore, diagnostic and treatment standards might differ between countries, leading to other price calculations. The recent study of Ostermann et al., however, showed no large differences in mean total duration of hospital stay for CAP between several EU countries (range 9.6-15.0 days) . Unfortunately, most published studies do not indicate prices of individual resource items, which makes detailed comparisons between studies very difficult. Besides this, none of the available studies in literature included aetiological groups in their analyses, further limiting the possibility of a relative comparison with our study findings at this moment.
A further relevant finding in our study was that 57% of the total costs of hospitalisation is due to general ward nursing. This finding is in accordance with other costs studies [27, 28]. The latter is also reflected by C. burnetii, causing a relatively milder course of the disease and a significant shorter duration of hospital stay, being identified as an independent cost limiting factor in the multivariable model.
In the present study, costs of medication represented a very small part of the total costs of hospitalisation (on average 3.2%). This means that policies aiming at an early intravenous to oral switch of antimicrobial treatment will not result in substantial cost-savings by reducing drug-expenses; costs might be reduced if the switch resulted in earlier hospital discharge. Medication costs for pneumonia caused by Legionella pneumophila appeared significantly higher compared to other aetiological groups. This is most likely caused by a higher ICU admission rate for these pneumonias and linked to the use of specific drugs such as fresh frozen plasma and sedatives.
This study has several strengths. First, we were able to identify the causative pathogen in a large number of patients enabling comparisons between aetiological groups. Second, we analysed resource utilization on an individual patient level. Third, data of two hospitals were studied (showing no differences) adding to the external validity of the findings. Besides this, the characteristics of the patients studied resemble data from another large nationwide CAP cohort from the Netherlands further adding to the generalisability of the findings .
There are also limitations that need to be addressed. First, due to missing data in some resources categories, not all 505 patients could be included in the overall cost analyses. This was due to being unable to retrieve some resource use from the years 2004 until 2006. We consider, however, that this has no impact on the validity of the findings because the more recent years are fully included , making the total costs of hospitalisation representative for the present standard of care for CAP. A further reassuring factor is that the comparison of patient characteristics and clinical outcomes of the 361 patients included in the analyses with the 144 patients not included, showed no large differences (see Additional file 1: Table S2). However, the lower number of patients available for analysis resulted in some aetiological subgroups becoming rather small.
Another limitation is that patients directly admitted to the ICU were absent in the study cohorts used. In the most recent cohort, 25 of the 817 eligible patients (3%) were not included due to direct ICU admission. This phenomenon could have lead to an underestimation of the absolute costs of hospitalisation for CAP. However, given this low percentage, we expect this effect to be rather small. Furthermore, it is very unlikely to have biased the relative costs per pathogen.
Finally, we cannot rule out that the costs related to microbiology exams are overestimated (9% share of total costs of hospitalisation). We studied patients who had participated in clinical studies in which a large panel of microbiological tests had been performed to maximize pathogen identification. However, presuming this resulted in a 50% increase in microbiology costs, decreasing these costs by 50% influences the total costs by less than 5%. In the present study, 58.2% of the causative pathogens could be identified, which is relatively high as compared to other studies .