Background
Community-acquired pneumonia (CAP) causes a heavy burden of illness with high morbidity, mortality, and health-related costs [
1‐
5]. In European countries with reliable coding systems, respiratory diseases are responsible for 15% of in-hospital deaths, with pneumonia being the second most important cause, and for 7% of hospital admissions, with pneumonia being the leading cause (2%). A recent review across Europe reported a still high incidence of CAP, of 68–7000 per 100,000 populations [
4].
Among hospitalized CAP, the proportion of those requiring intensive care unit (ICU) admission ranges from 5 to 40% [
6]; it was 22.7% in a recent US study [
7]. Some scoring systems have been specifically set up to assist clinicians to identify patients who will require ICU admission [
8,
9]. However, these scores are focused to identify patients with short-term mortality. The factors associated with an increased risk of long-term mortality, and the magnitude of the associated increase are poorly known.
In addition, the burden of in-hospital management of CAP, due to pneumococci (P-CAP) and other etiologies, as well as management of associated comorbidities, generates high costs. Although the short-term mortality and in-hospital costs of P-CAP in France have already been estimated [
10], little is known about the associated 1-year survival, burden of hospital care, and costs. However, the evaluation of the overall burden associated with CAP, and more specifically P-CAP, is instrumental to guide public health strategies for CAP prevention, such as targeting pneumococcal or influenza vaccination to high-risk groups. This is of notable importance in France as a French study conducted in 2007–2012 showed only 2% of the patients admitted in the ICU for severe P-CAP were vaccinated against pneumococcal infections [
11].
The objectives of this study were to ascertain the overall burden associated with ICU-admitted P-CAP in France in terms of in-hospital mortality and the associated direct costs, during both the initial hospital stay and within the year following the P-CAP onset.
Discussion
This study describes a comprehensive survey of all patients with pneumococcal pneumonia hospitalized in the ICU in France for 1 year and the associated burden in terms of in-hospital mortality and direct costs. Although many studies on patients with pneumococcal pneumonia are focused on the risk factors for ICU admission and the 28-day mortality rates, this study broadened the scope to the risk factors for increased mortality and costs during both the initial and all the subsequent hospital stays within the year following the ICU admission. We showed that the early mortality was correlated with the initial illness severity at ICU admission, age, and two comorbidities, malignant diseases and chronic liver diseases, whereas the initial illness severity was no longer a risk factor for 1-year mortality. The in-hospital costs were decreased in patients with early mortality, i.e., the most severely ill at admission. Unlike mortality, the comorbidities that impacted in-hospital costs were chronic cardiac and respiratory diseases.
The 28-day mortality rate of our study is consistent with that reported by others. In a multicenter French study on 614 ICU-admitted patients with severe P-CAP, the mortality rate was 18.9% [
11]. Age, gender, and organ failures at ICU admission were more strongly associated with hospital mortality than comorbidities as expressed through the Charlson index [
11]. This population was slightly less severely ill, as their median SAPS II at ICU admission was at 43 [IQR 32–57], whereas our patients had a median SAPS II at 48, and more than 60% were mechanically ventilated and required vasopressor agents. The mortality rate of 29% observed in another French study on invasive pneumococcal infections, mainly P-CAP, comprised 48% severely ill patients, of which 31% were admitted in ICUs [
12]. Another French study on ICU patients with invasive pneumococcal infection reported a similar 28-day mortality rate of 19.8% [
13]. Diabetes mellitus was the only comorbidity independently associated with increased mortality (odds ratio 1.91, 95% CI [1.23–3.03],
p = 0.006), in that study that contrasts to our study where diabetes had no prognostic value in multivariate analysis. The frequent complication of diabetes with chronic cardiovascular diseases may explain why diabetes itself is not eventually associated with prognosis in our multivariate model.
After the initial ICU stay, the 1-year in-hospital mortality was 12% and increased with age and comorbidities including malignant and liver diseases. Our observed mortality is similar to the overall 1-year post-ICU mortality reported by a Dutch study [
14] but lower than that of a US study after ICU stay for sepsis [
15] and a Finnish study in ICU patients with severe CAP [
16]. The prognostic value for the long-term mortality of any type of preexisting comorbidities was also reported by others [
17‐
22]. More specifically, our study showed that the 1-year mortality is correlated with some comorbidities rather than with the illness severity at ICU admission, as already highlighted by others [
23‐
26]. A French study on the 1-year survival of ICU survivors reported an OR at 1.65 [95% CI 1.13–2.42] for patients with active malignant disease [
24]. Another study reported a hazard ratio of 1.98 [95% CI 1.17–3.37] for the long-term mortality in hospitalized patients with CAP with cancer [
25]. Similarly, the independent impact of chronic liver diseases on the mortality of ICU patients was already described [
23,
26]. However, the impact of these severe complications on the 1-year mortality might be more related to their own severity than to the CAP episode that triggered the initial hospital admission. Of note, cardiac events, frequently reported as jeopardizing the medium- and long-term outcomes of hospitalized CAP patients [
17,
21,
27], were not a major cause of readmissions.
In our study, the mean cost of the initial ICU stay was slightly above €19,000. In a previous study conducted in France in 2011–2014 which evaluated the direct costs associated with P-CAP, the cost for the subset of the ICU-admitted patients was €14,385, that is, lower than observed in our study [
10]. With regard to age, the costs in our study ranged approximately €18–20,000 for the initial stay and were slightly lower for patients older than 75 years, which may be explained by the higher early mortality in this age-group. Interestingly, the costs of the subsequent stays for the younger adults (18–54 years) were twice as high as those for the oldest adults (> 75 years). Importantly, on its own, the high age was not an independent risk factor for increased costs in our study. Because of the differences in health systems, costs differ from one country to another, and direct comparisons must be made with caution.
The 1-year hospital costs were closely linked to rehospitalizations and were experienced by one patient out of three in our study. This rate is lower than the 40% readmission rate within 3 months observed in the USA for sepsis [
16], and the overall 72% readmission rate observed for hospitalized CAP patients [
28]. It may reflect the French policy for ICU admission of patients that selects patients who will benefit the most from their ICU stay with a better intermediate-term prognosis. Importantly, chronic cardiac and respiratory diseases were associated with a substantial increase in 1-year costs, and readmissions were by far mostly due to these two comorbidities. Of note, the proportion of readmission due to chronic respiratory diseases was higher than the proportion of this comorbidity in our population, suggesting that the ICU admission for P-CAP may be associated with a worsening of the underlying respiratory diseases. The respective impact of P-CAP and chronic illness by itself on the readmissions was not addressed by our study.
In our study, specific comorbidities but not age were independent prognosis factors for the pneumococcal disease burden, that supports French recommendations, which are solely comorbidities driven for adult patients [
29]. This is in contrast to other guidelines for which age directs vaccination recommendations [
30,
31]. The four comorbidities that we identified as drivers for an increased CAP burden are included in most immunization recommendations. However, malignant diseases, which are all included in the recent French recommendations [
32], are often limited elsewhere to some subpopulations, such as metastatic stage, or some specific hematological malignancies. Patients with cancer, although at higher risk of increased 1-year in-hospital mortality, as observed in our study and others [
33], exhibit a low rate of pneumococcal vaccination [
34,
35]. A recent French study showed a 10% rate among patients with comorbidities at risk [
12]. A single-institution study focusing on 99 patients with gastrointestinal cancer yielded a pneumococcal vaccination rate of 10.1% (95% CI [4.1–16]) [
36]. As shown in several studies, childhood vaccination programs contribute to herd immunity, thus partly protecting non-vaccinated adults [
37,
38]. However, this protection of adults has been deemed insufficient in France, despite a high coverage of vaccination among children [
39]. Vaccination among adults including those older than 65 years has been shown to be efficient [
40]. Healthcare professionals, cancer societies, and other societies advocate for increasing this vaccination rate, which some vaccination programs increased successfully [
36]. Cardiovascular diseases are the leading cause of mortality of diabetic patients [
41], which may explain the lack of a signal due to diabetes mellitus in our model. Beyond specific comorbidities, the immunization coverage reported for elderly residents from French long-term nursing facilities was low at 27% (95% CI [21–34]) among those with targeted comorbidities and 17% (95% CI [14–20]) overall [
42].
Despite its strengths, this study has some limitations. It relies on an administrative database that mainly has budgetary purposes. Therefore, the coding of some diseases such as the P-CAP diagnosis might be suboptimal. The reliability is not that of a clinical database monitored against source data. However, we used ICD-10 of the coding system, which showed more reliability for the pneumonia diagnosis, and we combined codes for an improved accuracy [
43,
44]. In addition, some types of data were not available, such as the intensity of signs and symptoms at admission; the use of some resources, such as the dialysis procedures; or the immunization status of the patients. However, as discussed earlier, the rate of vaccinated people is low in France, even among at-risk patients. It is another limitation that we did not have the tests used and their results and microbiological data related to the susceptibility profile of the serotype of the pneumococcal strains, as well as the adequacy of antimicrobial therapy.
Another limitation is that this database collects solely in-hospital data. Therefore, we do not have any insight regarding what happened outside of the hospital, such as the out-of-hospital mortality rate. However, a French study showed that the rate of in-hospital deaths in 2008 ranged approximately 60% of deaths for the patients aged between 40 and 89 years and that these rates were stable over 15 years [
45]. Therefore, we can evaluate our rate of missing death-related data at approximately 40% of our known decedents. In addition, solely data on acute hospital stays were collected, and we did not collect whether the patient was discharged home or to a rehabilitation facility. Another limitation is the lack of indirect cost data, e.g., those related to ambulatory consultations, drugs, and medical procedures. Finally, while potentially several control groups would have been of interest, such as patients admitted in the ICU with CAP not due to
S. pneumoniae, or pneumococcal invasive infections other than CAP, or other reasons besides CAP patients for being transferred into the ICU.
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.