In this retrospective population-based study, we found that the annual incidence rate of radiologically adjudicated CAP requiring hospitalization was ranging from 259 to 337 cases per 100,000 py, depending on the use of age-standardization and the inclusion of ICD codes in non-primary positions. Hospital utilization for CAP was 19 times higher in the age group ≥80 years compared to 18–64 years (IRR 18.6) and the case-fatality risk was 5.5%.
Our crude incidence rate of 259 cases per 100,000 py is in line with the prospective studies conducted by Jain et al., and Marston et al. with 248 and 267 per 100.000 py [7, 21]. Other studies indicate higher figures with Ramirez et al. and Hyams et al. demonstrating almost identical incidence numbers of 649 resp 645 per 100,000 py [22, 23]. The comparison of studies is complicated by the lack of a reference standard for diagnosis, different study designs and settings, with variations in local routine, diagnostic accuracy, and access to hospital care. In addition, population demographics are important and may complicate comparisons across settings. Our catchment area has a population younger than national average as indicated by the increase in IR from 259 to 294 upon age-standardization, emphasizing the need to provide standardized estimates. Including pneumonia diagnoses in non-primary positions resulted in an increase in crude IR from 259 to 293 cases per 100,000 py and in ASR from 294 to 337. On the other hand, only 18% of the reviewed cases with pneumonia diagnoses in non-primary positions fulfilled radiology criteria, suggesting that non-primary codes should not be included per se but that reviews of diagnostic accuracy are necessary in this group.
In line with previous studies, we demonstrated that older persons had a significantly increased hospital utilization rate, especially after 80 years of age [7, 19, 22, 24]. When combining age and sex there was no difference in ages 18–79 but males aged ≥80 years had a more than 50% higher IR compared to women of the same age. This finding, that has been previously described, may have implications on focused preventive strategies, as this age group is expected to increase substantially [25]. The majority of included patients also had comorbidities associated with risk of CAP. These comorbidities varied with age; among the population aged 18 to 64 years, immunocompromising conditions and organ transplantation were dominating, emphasizing the need for targeted prevention in this group. In the group aged over 80 years, cardiovascular conditions, heart failure and COPD were the most common.
The overall in-hospital case-fatality risk was 5.5%, which is higher than the 3.7% in the prospective ECAPS study [12]. Other prospective studies have reported in hospital mortality ranging between 2.0%-6.5% and in retrospective studies between 7.9% -17.2% [7, 18, 22, 26]. The differences in mortality rate could be explained by the fact that the frailest or most severely diseased patients may not be enrolled in prospective studies due to exclusion criteria.
We included codes representing COPD with acute exacerbation (ICD-10 codes J44.0 or J44.1) in our case-finding. Among these, 18% had CAP according to our definition. Nevertheless, this was a large diagnostic group and among the patients with radiology-verified CAP, one third of CAP was misclassified as COPD exacerbation in our study. Previous studies of ICD-10 discharge diagnosis codes usually do not include COPD codes as an indicator of CAP, which could lead to an underestimation of CAP incidence. [7, 11, 18, 19, 22]. However, the low specificity suggests that these codes should not be included without review of radiology.
In our study 45% of patients diagnosed with pneumonia or COPD exacerbation fulfilled the CAP criteria after review, indicating that these ICD codes alone may overestimate CAP incidence. Compared to previous studies of diagnostic accuracy, this proportion was lower, but our case definition also included J44 as a non-pneumonia ICD-10 code [27]. Among the most common reasons for not meeting CAP criteria was lack or infiltrates on radiology. Adding radiology as a diagnostic CAP criterion could underestimate CAP incidence since the sensitivity of chest X-ray is only 44 to 77% and is often negative in the first couple of days in older patients with pneumonia [28]. Moving from the traditional chest X-ray to CT could increase sensitivity; a study by Claessen et al. found that a CT scan improved diagnostic accuracy in suspected CAP in 59% of cases [29]. In addition, the lack of reference standard enables the diagnosis of pneumonia without an evident finding on radiology. Furthermore, as mentioned above, the addition of COPD codes, with only 18% representing CAP cases after review, reduced specificity.
The strengths of this study are that it was population-based, and that all episodes with an ICD diagnosis of pneumonia were confirmed by a thorough review of radiology. The retrospective design made it possible to include a full population sample, including frail patients, or those unable to give consent, a patient group that are often excluded from prospective studies. In addition, the review of radiology justified the inclusion of COPD diagnoses, which enabled the identification of a subset of CAP patients often missed in previous studies. Limitations also exist. The retrospective study design is reliant on the treating physicians´ assessments, and documentation, with hospital discharge diagnoses being dependent on local routines. As stated above, patients treated for pneumonia without radiologic assessment or findings were not included as CAP cases in our analysis. Sweden has one of the lowest numbers of hospital beds per capita (2.1 per 1,000 inhabitants) as compared to other countries in Europe limiting the external validity [30]. This may entail a higher hospital admission threshold, with more CAP patients being treated in ambulatory care and consequently a lower hospitalization rate of CAP, compared to settings with more hospital beds per capita.
Implications of findings
Our results highlight the lack of an established reference standard to define CAP, complicating the comparison of CAP epidemiology across settings. The use of age-standardization and diagnoses in non-primary positions as well as the selection of ICD diagnoses representing pneumonia have all been inconsistent in previous studies. This also makes the evaluation of the effect of additional interventions challenging. From a public health perspective, our results indicate a substantial burden of CAP. This burden is likely to increase due to the expected demographic change, with an increasing proportion of the population being aged over 80 years. We found an especially high hospital utilization with CAP in this group, highlighting the need for targeted preventive measures. After this study was performed, national vaccine strategies for risk groups and/or the elderly, targeting respiratory pathogens, including S. pneumoniae, COVID-19, and Respiratory Syncytial Virus have been updated or added. Vaccine coverage in the prospective ECAPS study were only 15 and 49% with pneumococcal or influenza vaccines, respectively. Therefore, we believe that further improvement of the current implementation should be emphasized. Data obtained in the present study could potentially also be used to define specific target groups to add preventive measures. However, cost-benefit needs to be evaluated to clarify if a more targeted vaccination program should be launched for these groups.