Microbial aetiology, outcomes, and costs of hospitalisation for community-acquired pneumonia; an observational analysis

Patients with CAP above 18 years of age admitted to the St. Antonius Hospital in Nieuwegein or the Gelderse Vallei Hospital in Ede (both teaching hospitals in the Netherlands), between October 2004 and August 2006 (n = 201), and between November 2007 and September 2010 (n = 304), who participated in two consecutive clinical studies were enrolled. The first study was a prospective cohort study on clinical characteristics and polymorphisms in innate immunity genes in patients with CAP; the second study was a placebo controlled double blind randomized clinical trial evaluating dexamethasone as adjunctive therapy (NCT00471640). In both studies, the same clinical inclusion criteria were used and patient characteristics of the patients studied resembled data from another large CAP cohort (over 20,000 patients admitted to hospital for pneumonia) from the same time period [15]. CAP was defined as a new pulmonary infiltrate on chest radiograph, in combination with at least two of the following criteria: cough, sputum production, temperature above 38°C or below 35°C, auscultatory findings consistent with pneumonia, C-reactive protein concentration of more than 15 mg/L, and white blood cell count of above 10 × 10⁹ cells/L or below 4 × 10⁹ cells/L, or >10% of rods in leukocyte differentiation. Patients who were immunocompromised, had been directly admitted to the intensive care unit (ICU), or who had received immunosuppressive therapy (including the use of >20 mg prednisone equivalent per day for >3 days) were excluded. More detailed inclusion and exclusion criteria are described elsewhere [10, 16]. Comorbidities were recorded of each patient and pneumonia severity index (PSI) score was calculated on admission. The present study has been approved by the Medical Ethical Committees of the St. Antonius Hospital (Nieuwegein) and the Gelderse Vallei Hospital (Ede), both in The Netherlands.

At least two sets of separate blood and sputum samples of each patient were Gram stained and cultured. Streptococcus (S.) pneumoniae cultured from either sputum or blood was serotyped by the Quellung reaction. Moreover, sputum samples were analysed with TaqMan real-time polymerase chain reactions (PCRs) in order to detect DNA of Mycoplasma (M.) pneumoniae, Legionella (L.) pneumophila, Coxiella (C.) burnetii, and Chlamydophila species. Antigen testing of S. pneumoniae and L. pneumophila was performed in urine samples. Furthermore, pharyngeal swabs were taken for viral culture and viral PCR. Finally, patients were analysed for a serotype specific rise in S. pneumoniae antibodies when two blood samples (one drawn at admission and one after discharge) were available. Antibodies against pneumococcal polysaccharides were measured on a Luminex platform (Luminex Corporation, Austin, TX), using a quantitative multiplex immunoassay: the xMAP pneumococcal immunity panel. More detailed information can be found elsewhere [17].

If both a bacterium and virus were detected in a patient, the bacterial species was classified as the causative pathogen. If two different bacterial species were identified, the pathogen known to most likely cause CAP was considered causative. For the purpose of this study, aetiological agents were classified into ten groups: the first seven groups consist of the most frequently identified bacteria (S. pneumoniae, C. burnetii, Haemophilus (H.) influenzae, L. pneumophila, Chlamydophila species, M. pneumoniae, and Staphylococcus aureus), group eight contains remaining bacteria (‘Other pathogen’), group nine comprises viruses (‘Viral pathogen’), and the last group consists of CAPs with unidentified aetiology (‘No pathogen identified’).

ICU admission during hospitalisation, length of stay, in-hospital mortality, 30-day and one-year mortality were documented for each patient.

Hospital administrative databases were extracted for all resource utilisation on a patient level. Resource items were grouped in seven categories: general ward nursing, nursing on ICU, clinical chemistry laboratory tests, microbiology exams, radiology exams, medication drugs, and other. Except for nursing, only resources plausibly related to pneumonia treatment were selected. For example, medication drug use only included antibiotics, analgesics, bronchodilators, sedatives, blood products and antithrombotic drugs. The category “other” comprised physical therapy sessions, electro and echocardiograms, bronchoscopy and laryngoscopy and invasive empyema diagnostic and treatment procedures. For duration of nursing, the unit of measurement was number of days.

Total costs per patient were calculated by summing the number of resources multiplied by the costs per item. Costs per resource item were based on the National Diagnosis Treatment Combination rates valued in 2011 or 2012 [18], except for nursing costs during hospital stay and costs of medication drugs. Nursing costs for general ward and ICU stay were based on mean costs per hospital unit prices belonging to diagnostic treatment combination code 401 (‘pneumonia’) for the year 2011. Costs of drugs were based on the lowest medication drug price according to the College for Health Insurance website valued in 2012 [19]; if not available on this website, the hospital’s purchase price was recorded.

Overall, descriptives were stated as number (%), mean (standard deviation (SD)) or median (interquartile range (IQR)), and compared using independent samples T-test, Chi-square test, or Mann–Whitney U test, where appropriate. Kruskall Wallis test was used to assess overall differences in length of stay and costs between aetiologic groups.

In 294/505 (58.2%) patients, a causative pathogen was identified. Table 2 lists the microbiological test results of most frequently identified pathogens. Overall, S. pneumoniae was most prevalent (124/505, 24.6%). In 51 of these 124 patients, S. pneumoniae serotyping could be performed. Type 1 was the most common serotype. A complete overview of the pneumococcal serotypes is given in Additional file 1: Table S1. In 43/505 patients a mixed infection was found. No penicillin resistant S. pneumoniae, methycillin-resistant Staphylococcus aureus or multi-resistant gram-negative pathogens were identified.

Clinical outcomes categorized by aetiology group are listed in Table 3. Overall, LOS differed significantly between the major aetiological groups (p < 0.001). CAP caused by either M. pneumoniae or C. burnetii was associated with a significantly shorter LOS compared to the other aetiologic groups (p:0.007 and p < 0.001, respectively), while S. pneumoniae CAPs resulted in a significantly longer duration of hospital stay (p:0.03).

For 361/505 (71.5%) of the patients complete resource utilization data were available for analysis. The clinical characteristics of the 144 patients who could not be included, as compared to the included patients can be found in Additional file 1: Table S2.

Table 4 lists the top 10 most frequent and the top 10 most expensive resource items. In the Additional file 1, the top 5 most frequent used items for each individual category can be found in Table S3.Figure 1 shows the total distribution of hospital costs per patient. Total median hospital costs per patient were €3,899 (IQR 2,911-5,684) with minimum costs of €901 and maximum costs of €112,634. Figure 2 shows the share per category in the total costs: general ward nursing represented the largest share (56.5%), followed by nursing on ICU (16.4%) and diagnostic microbiology exams (9.4%).

Overall, total hospital costs differed between the 10 aetiological groups (p:0.002); costs for hospitalisation of CAP caused by C. burnetii were significantly lower (p < 0.001), while hospitalisation of patients with S. pneumoniae as causative agent represents significantly higher costs (p:0.03) compared to other aetiologies. For M. pneumoniae and Staphylococcus aureus a trend towards respectively lower and higher costs (p: 0.10 and p:0.08, respectively) was observed.

Figure 3 shows median costs of each aetiologic group subdivided into the seven resource categories. Raw numbers of this figure can be found in Additional file 1: Table S4. Overall, costs for general ward nursing, microbiology exams, clinical chemistry laboratory tests, medication drugs, and radiologic exams all differed between the aetiological groups. On an individual pathogen level, CAP caused by C. burnetii was lower in costs for nursing, clinical chemistry tests, and radiological examinations compared to most of the other aetiological groups. Costs of medication were specifically high in patients with L. pneumophila pneumonia. CAP caused by Staphylococcus aureus was higher in costs for nursing and CAP caused by S. pneumoniae was more expensive in radiological examinations. Additional file 1: Table S5A to S5G shows details of this aetiological subgroup analysis.

As S. pneumoniae is the most frequent identified pathogen in CAP, costs of serotypes were explored grouped per pneumococcal vaccine available in the European Union (results presented in Additional file 1: Table S6). Total costs of hospitalisation were not higher for patients with CAP caused by the serotypes present in the different vaccines compared to patients infected by pneumococcal serotypes not included in these vaccines.

To identify cost driving factors, a multivariable linear regression model was constructed. Table 5 lists the variables included in the final model together with their corresponding regression coefficients. Staphylococcus aureus, high PSI score (classes IV and V), and Streptococcus pneumoniae were all independent cost driving factors, increasing total costs of hospitalisation by 98%, 43%, and 18% respectively. Coxiella burnetii decreased total costs of hospitalisation by 35%.