Background
Community-acquired pneumonia (CAP) is one of the most common infectious diseases in the world [
1‐
4]. In Europe, the incidence of all-cause CAP (hospitalized and outpatient) in the elderly population (≥65 years) is estimated at 14 per 1000 person years [
5,
6]. Approximately 56 % (equal to €5.7 billion) of CAPs economic burden in Europe is ascribed to hospitalization expenses [
7]. The other costs can be attributed to outpatient care, loss of workdays and drug expenses. Increasing CAP incidence due to growth of the elderly population in the upcoming years [
8‐
11], more severe episodes due to increasing comorbidity within the aging population [
12], and limited economic resources demand for new preventive measures [
13]. In Western countries like the Netherlands, new healthcare interventions will be evaluated for their costs and effects and detailed cost estimates are indispensable in this process. However, most larger studies describing CAP costs were conducted retrospectively and based on insurance claims with limited clinical information [
4,
14‐
19]. The few prospective studies on CAP hospitalization costs covered relative small cohorts only [
20‐
22].
The Community Acquired Pneumonia Immunization Trial in Adults trial (CAPiTA), a nation-wide study assessing the efficacy of the 13-valent pneumococcal vaccine (PCV13) in the elderly, created a unique opportunity for quantifying hospitalization costs of CAP [
23,
24]. A parallel study was designed to estimate costs for hospitalized CAP patients, including patients with and without invasive pneumococcal disease (IPD), stratified by age- and risk-groups for developing pneumococcal infections and by outcome (i.e. mortality). Secondary objectives were to analyze disease severity differences between PCV13 and placebo cohorts and to identify cost drivers in hospitalized CAP cases.
Methods
Study subjects
The etiology and prognosis of community-acquired pneumonia (Etio-CAP) study is a prospective cohort study conducted in parallel to CAPiTA [
23,
24]. In this double blind randomized clinical vaccination trial in the Netherlands, the efficacy of PCV13 in adults was evaluated by vaccinating 84,496 subjects of ≥65 years of age with either placebo or PCV13 with an average follow-up period of 3.97 years [
24]. Detailed information, in- and exclusion criteria and outcome measures have been previously described [
23,
24]. Follow-up was organized in 59 sentinel centers (58 hospitals and 1 outpatient clinic) between September 2008 and August 2013, where the 58 hospitals participated in the Etio-CAP study. When a CAPiTA-participant was registered to one of these hospitals with a suspicion of pneumonia, regardless of the causing pathogen, clinical data were collected according to the CAPiTA protocol. When admitted, additional data, like use of resources and complications, were collected as part of the Etio-CAP-study.
Definitions
CAP was defined, regardless of the causing pathogen, as the presence of two or more clinical signs of pneumonia together with a chest x-ray consistent with pneumonia [
23]. IPD was defined as
Streptococcus pneumoniae cultured from a sterile site (e.g. blood). The analysis was conducted for three groups: ‘total CAP’ (i.e. all confirmed CAP cases), non-IPD-CAP (i.e. confirmed CAP cases, no IPD) and ‘IPD-CAP’ (i.e. cases with confirmed CAP and IPD). Subjects with IPD without CAP were not included in the analyses.
Admissions within 30 days post-discharge of a suspected CAP hospitalization episode were labeled as a readmission and treated as a continuation of the previous admission with regard to a) diagnosis (CAP or IPD) and b) costs (i.e. resources used in first and following admissions, and associated costs were added up).
The study was approved by the Central Committee on Research Involving Human Subjects (CCMO11.0810/TV/NL23014) as a sub-study of the CAPiTA-trial. Therefore no separate informed consent was required for the Etio-CAP study.
Data collection
Data collected included length of stay (LOS) and intensive care unit (ICU) admission, comorbidities at moment of admission (see Additional file
1: Table S1), date of symptom onset and date of symptom relieve. Based on the presence of comorbidities, subjects were stratified into high (i.e. immunocompromised patients), medium (i.e. presence of other chronic conditions) and low risk for pneumococcal infections (Additional file
1: Table S1).
Data included for the Etio-CAP-study concerned: mortality (with minimum follow-up of 6 months after hospital admission, i.e. confirmation of vital status by the GP of the participant), laboratory results, number of hospital admissions 1 year prior to index CAP hospitalization episode, diagnostic and therapeutic interventions for both pneumonia and its complications, additional “waiting days” in the hospital (i.e. before home care could be organized), and location of discharge (own home or nursing home). Pneumonia severity was assessed using the Pneumonia severity index (PSI), as described by Fine et al. [
25], and CURB-65 [
26]. PSI score was categorized into low (score 1 and 2), medium (score 3 and 4) and high (score 5). Based on postal codes, information on socioeconomic status (continues variable from low to high, scores ranging from -5.5 to 3, with lower scores reflecting lower socio-economic status) and urbanity of neighborhood (from high to low; score 1, 2, 3, 4 and 5) was obtained from Statistics Netherlands and referred to the situation in 2010 [
27].
Costs
Hospitalization costs were calculated by multiplying recorded units of healthcare resources used with corresponding unit prices (see Additional file 1: Table S2) [
28]. Unit costs were either taken from the Dutch manual for healthcare costing research [
28,
29] or were calculated from a unit price medium of different hospital price lists available. All costs were expressed for the year 2012, and if necessary, updated using Dutch consumer price indices [
30].
Statistical Analysis
Data were analyzed using IBM SPSS Statistics for Windows, Version 20.0. The one-way ANOVA test, Mann-Whitney U-test and the Kruskal-Wallis test were used to compare quantitative variables and the Chi-square test was used for categorical variables. Intubation days at the ICU were recorded in a dichotome (yes or no) and a quantitative way (e.g. number of days of intubation). There were no missing data with the exception of number of intubated ICU days (0.6 %) and number of previous admissions (0.2 %). Missing data were imputated using multiple imputation. A stepwise multiple regression model was used to identify cost drivers with log-transformed hospitalization costs as a dependent variable. Explanatory variables were gender, age, previous admission (categorized as: none, one, two, three or more admissions), smoking status (non-smoker, previous smoker, current smoker), risk-level, socioeconomic status and urbanity, and CURB-65. The latter was used instead of PSI scoring, because of overlapping definitions between PSI and risk levels as both are estimated using comorbidity information. The 95 % confidence interval (CI) was determined using bootstrapping with 1000 iterations. A p-value of <0.05 was considered significant.
Discussion
Among those who participated in the CAPiTA-study, the mean costs of hospitalization for CAP were €8301. Costs for a hospitalized CAP episode with IPD were €13,611, or 168 % of the mean costs for a non-IPD-CAP episode (i.e. €8081). A high clinical severity of disease at the time of hospitalization and having medium risk profile for pneumococcal disease were identified as independent drivers of high costs, whereas male gender, increasing age, previous admission(s) in the past year, low risk profile for pneumococcal CAP, lower urbanity and higher socio-economic status were associated with lower overall costs. Vaccination status was not associated with clinical severity of CAP or costs of hospitalization.
Quantification of costs associated with CAP are usually based on retrospective studies using national insurance claims databases. In a previous retrospective study in Netherlands by Spoorenberg et al. [
20], median costs based on ICD recordings were estimated at €3899. In the current study median costs were €4809, and the difference could be explained by the younger cohort in the Spoorenberg et al. study (mean age 63.4 versus 77.4) potentially leading to a shorter LOS (8.5 versus 12 days) and a lower case-fatality rate (30-day mortality: 5.1 versus 13.1 %). Furthermore, in the current study, non-survivor costs were €13,526 and appeared to be 177 % of the hospitalization costs of survivors (i.e. €7638). Also, Spoorenberg et al. did not consider the additional costs of readmissions and included patients of only two general hospitals in the Netherlands, as compared to 58 hospitals, of which seven are tertiary care centers, were included in the current study. Indeed, the proportion of patients needing ICU admission was 21.2 % in tertiary care (24/113) and 12.1 % in general hospitals (202/1672).
Another nationwide ICD-based cost analysis by Rozenbaum et al. [
19] reported medium costs of CAP of €6060 and €5937 in those 65–74 and 75–84 years of age, respectively, which is also lower than the medium costs for these age cohorts in the current study (€8674 and €8770, respectively). The study by Rozenbaum et al. was based on reimbursement of hospital expenses, as regulated by Diagnosis Related Groups (in Dutch: Diagnose BehandelCombinatie or DBC). By selecting solely DBCs related to pneumonia, other costs (e.g. treatment of complications) were not considered, possibly explaining the lower costs. Also readmissions were not taken into account, resulting in shorter length of stay in the 65–74 and 75–84 age cohorts (7.7 versus 12.4 and 8.1 versus 12.3, respectively). Furthermore, the use of ICD-coding in both Dutch studies [
19,
20] may have caused important misclassification. In another Dutch study across seven Dutch hospitals, the sensitivity of ICD-coding was estimated at 79.5 % [
31]. The current prospective study with 1933 patients with confirmed CAP based on standardized diagnostic criteria may, therefore, be more precise and generalizable than previous studies [
16,
19‐
22].
Different settings and healthcare systems makes a comparison between countries nearly impossible at the level of cost estimates [
32]. Length of Stay (LOS) has been identified as a major contributor to healthcare resource utilization. We found a mean of 12 days which is within the range of other European studies, [
21,
22,
33‐
40], ranging from 8 up to 13.2 days.
There are also limitations of this study that need to be addressed. An exclusion criterion for the CAPiTA-trial was the presence of immunodeficiency (i.e. individuals at high risk for developing pneumococcal infections), except for patients with asplenia. However, there were 330 (18.5 %) patients with CAP that had developed immune deficiencies after enrollment in the CAPITA-trial. Hospital costs might be slightly underestimated as we did not have exact information on the percentage of patients that received in-hospital rehabilitation services, or because some diagnostic or therapeutic procedures, such as incidental pleural punctures, may not have been reported. Additionally, in the current cohort 10.1 % of the survivors were discharged to a nursing home or rehabilitation center. As information on the total length of stay in nursing home or rehabilitation center was lacking, we could not add these costs to our cost estimates.