Discussion
Our study showed that, in the 15 years under analysis, despite a trend of decrease in CAP hospitalizations and associated mortality after 2012, the number of admissions and in-hospital mortality increased steadily over time. In-hospital CAP mortality was mainly associated with the very old individuals (> 75 years), males, and patients’ parish unemployment rate.
This transversal large-scale study, over a period of 15 years, based on a retrospective data analysis of hospital admission episodes and associated mortality is one of the largest studies of this nature carried out in Portugal, covering all ages, all public hospitals, and the entire mainland population, allowing robust results and comparisons.
Reports on CAP incidence focus mainly on hospitalizations and show wide variability across countries, ranging from 1.48 per 1000 population in England, 2.75 in Germany to 6.27 in Spain [
6,
22‐
24]. Despite our hospitalization rates (2001/2011) are higher than other European countries [
25], our data are in line with a previous study in our country [
13] and close to reports from Spain [
23,
24]. Furthermore, our age group-specific incidence rates increased with advancing age, with incidence rates in older adults very similar to those reported in Spain [
24]. This cross-country variability may be related not only to climate differences and vaccine coverage for influenza and pneumococcal vaccines, but also to differences in health system structure and other factors such as heterogeneity in admission criteria [
26].
The steady increase of CAP hospitalizations and associated mortality is particularly related to the very old people (Figs.
2 and
3) and reflects most probably the demographic changes that occurred in the Portuguese population during the large study period, with a more aged population according to 2001 and 2011 census data. However, if we look at the numbers associated with the growth of the elderly population and the number of CAP hospitalizations, the increments are disproportionate, taking that in the correspondent period, the growth of population over 75 years was 37.5%, contrasting with an increment of 115% of the hospitalization episodes. This suggests that there are other responsible factors regardless of the change in the population structure. Actually, this disproportionate increase may reflect not only, the weight of ageing, but also the overall impact of the associated comorbidities in the elderly, as documented in several studies [
25‐
28]. Furthermore, apart from the known health consequences of the ageing process, it has been observed that the older Portuguese adults, report a low household income and are a vulnerable group in terms of unhealthy lifestyle behaviors and poor socioeconomic conditions [
29]. In particular, older people over 75 years of age who live alone represent a group that is subject to great economic and social vulnerability [
30].
During the analyzed period, there was a decrease in hospital admissions and corresponding mortality in the group of children under 4 years, which can be explained by the introduction in Portugal of 7- and 13-valent pneumococcal conjugate vaccine in 2001 and 2009, respectively. Although not included in the National Immunization Program until 2015, their use was widely recommended by pediatricians for all healthy children, allowing increasing vaccination coverages, 33–72% between 2001 and 2013 [
31], eventually justifying the reduction of hospital admissions and their mortality taking into account that
Streptococcus pneumoniae is the main etiological agent of CAP [
4].
Regarding the in-hospital mortality associated with CAP, our exploratory model of the factors associated with the probability of death revealed that the increase in age was the factor that contributed most to the rise in death probability, mainly in the very old (> 85 years). In fact, 91.7% of all deaths occurred in the age group ≥65 years old.
The overall mortality rate was 18.5% over the 15 years analyzed. This value is higher than that found in several studies ranging from 4.8 to 14.4% [
3,
25,
27,
28], but is similar to other series of various studies conducted in Europe [
4]. A possible explanation the observed higher in-hospital mortality rate of CAP may be related to the progressive ageing of Portugal elderly population where very older people (aged 85 and over) are observed to be growing at a faster rate than any other age group. Indeed, Portugal stands out from the rest of the EU countries, as it is one of the fastest ageing countries, considering that by 2050 those aged 55 and over should represent almost half (47.1%) of the total population [
32]. In addition, Teixeira-Lopes hypothesized that the fear of dying at home, of patients and family members, could explain the inclusion of many patients with end-of-life pneumonia, especially in older age groups, and we believe this may have impact on observed hospital mortality [
33].
In one study where three world regions were analyzed, CAP mortality was 13.3% in Latin America, 9.1% in Europe and 7.3% in US/Canada [
34]. After adjustment for confounding variables, estimated differences in mortality between the three regions were significantly reduced. The factors that contributed to the differences found were the incidence of H1N1 infection, comorbidities (cerebrovascular disease), abnormal laboratory values (elevated blood urea nitrogen), antimicrobial therapy (macrolide use, fluoroquinolone use) and vaccinations (influenza, pneumococcal). Some of these factors, may explain the numbers observed in our country, however, given the limitations related to the use of an administrative database we were unable to include accurate data concerning the above-mentioned factors.
After 2012, the overall trend of growth in CAP hospitalizations and associated mortality changed. This change, in particular the decrease in the number of deaths in the last 2 years of the study and a reduction in the risk of death between 2010 and 2014, may be related to the creation and implementation of a national program for chronic respiratory diseases in 2012 [
35,
36]. This program under the guidance of the Directorate-General of Health was responsible and encouraged the creation and implementation of respiratory guidelines and health policies in our country. We highlight the free introduction of the influenza vaccine in the population over the age of 65 and in other at-risk groups since the 2012/2013 winter season. In fact, with this public health policy influenza vaccination coverage rate in the population aged 65 and older increased from 43 to 50%, between 2012 and 2014 [
37,
38]. Previous studies have documented an association between influenza vaccination and a less severe clinical course of CAP and an improved survival in hospitalized patients with this disease, during influenza season [
39,
40]. Moreover, this reduction in CAP hospitalizations associated with the observed risk of death reduction in the five-year period 2011–2014 compared to the previous five-year periods, may also be related to the introduction in Portugal of the 13-valent pneumococcal conjugate vaccine, as referred in the study of Kislaya, where a reduction of pneumococcal pneumonia admissions in Portuguese elderly was attributed to the hypothetical indirect effect (Herd effect) of pneumococcal child vaccination [
31].
Our results showed a sex gap, with men presenting higher numbers of CAP hospitalizations, mortality and a higher risk of death during hospitalization. These results are in accordance with other studies that show a sex inequality in health indicators associated with hospitalizations [
3,
4,
12,
29], with men presenting consistently higher hospitalization rates for CAP, which may be related to a great number of risk factors, such as smoking, alcoholism, and exposure to toxic occupational exposure [
3,
29]. The greater ratio of male to female mortality may also reflect higher rates of smoking in males pre-disposing to respiratory disease [
3,
16].
Our study revealed a mean LOS of 10.6 days and a median of 8 days which was in line with other reports [
3,
29,
34]. Eventual differences between studies and countries are related to the availability of continuing care outside the hospital environment [
41]. In our model for each day of hospitalization, the risk of death decreased 2.3%, which we interpret as a reflex of the interruption of hospitalization by death, given that more than one-third of the deaths (35.9%) occurred within the first 4 days of admission.
There are few reports on the association between the unemployment rate and health status [
42]. In our study, patients’ parish unemployment rate was positively associated with CAP mortality. This is the first study that links CAP mortality with unemployment rate as a proxy for the regional economic situation. Given that a higher unemployment rate may generally be related to a more adverse economic condition and this has been associated with increased bacterial pneumonia hospitalization rates [
28], our hospital mortality rates may have been inflated by the impact of the economic crisis. This relationship points to the need for specific socioeconomic, regional and population policies in the most vulnerable groups, such as the elderly in social deprivation. It also hints that common spatial processes (environmental-level factors) may be involved in pneumonia outcome, determining patterns and geographic variations and justifying future investigations.
There are limitations to this study. First, it was conducted using retrospective administrative data, conditioning the lack of information about comorbidities, microbiologic etiology, the process of care and influenza/pneumococcal vaccination status. Also, it was not possible to exclude patients who lived in nursing homes or in long-term care institutions in which it was possible, that pneumonia cases were healthcare related. Second, the ICD-9 codes did not allow for the exclusion of nosocomial infections as no code specific to this diagnostic category exists, with the consequent overestimation of CAP hospitalization rates. Third, changes, across the time period in analysis, concerning admission guidelines, diagnosis and coding practices might be probable sources of bias.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.