Thank you for inviting me to reply to the letter published in Pulmonology journal by H. Ito “Potential survival paradox in pneumonia”1 concerning the recent published article in Pulmonology “Pneumonia Mortality, comorbidities matter?”

Thanks for reading and commenting our article, “Pneumonia Mortality, comorbidities matter?”

Regarding the results described in your letter to the editor,one comment and some matters for concern raised about the results described in our paper.2 The comment refers to the possible high proportion of pneumococcal pneumonia identified in the etiology of pneumonia in Portugal, pointing to its possible relationship with vaccination policy. In fact, in Portugal the target populations for pneumococcal vaccination are people under five and older than 64, and patients with any kind of immunodeficiency. Interestingly, most of pneumococcal isolates we obtained in our study, were identified with young-adult patients with no comorbidities, which are not usually included in the target population for pneumococcal vaccination. This is probably one of the explanations for the fact that, that this etiology was not associated with a higher mortality risk.

The concerns are linked to the fact that in our study, smoking habits, obesity, COPD and diabetes are not associated with an increased risk of dying of pneumonia, whereas these comorbidities are usually associated with an increased risk of death from pneumonia. However, the studies3–5 used to justify the doubts raised in our research, have very different designs, populations, methodologies and analyses, so they can hardly serve for comparison.

In our study, all patients admitted to National Health Service (NHS) hospitals with pneumonia or sepsis with pneumonia, during 2015, were evaluated. The NHS hospitals are responsible for the overwhelming majority of hospitalizations in Portugal. The evaluation and classification of inpatient episodes is permanently carried out, prospectively, by medical doctors specialized in disease coding. This data was used in our study and, whenever necessary, electronic patient records of pneumonia episodes were evaluated. It is not possible, when the investigation is carried out on a retrospective basis, to eliminate all types of bias, due to information/recall or if a particular patient characteristic was included or not in the protocol. However, we believe this was controlled by the nature of our data and by the negligible number of missing values of the 36,366 patients we studied. The risk of misclassification, as we pointed out in the article, is impossible to eliminate completely and we assume that it existed. However, we think that these limitations do not explain the results. The prevalence of severe limiting comorbidities, such as the high number of bedridden patients, patients with stroke, sequelae, dementia, cachexia, cancer and patients living in Care Homes, accounting, in that year, for about 50% of the pneumonia mortality requiring hospital admission. The interpretation of “non concordant” results from different research designs, seeking the same goals, but using different data, is not unusual.6 One way of dealing with this, is using Bayesian approach for interpretation of to the data and results.7,8 This way of thinking, requiresnot only current data assessment, but also, use of “prior knowledge” of the context where the study was conducted, using the prevalence of the patients characteristics in the study population, the way of living and the access to medical care. The variable distribution and prevalence are determinant in the subsequent interpretation of the results and their external validity. The data we used included the great majority of the hospital admissions for pneumonia during 2015 in Portugal. The main goal of the study was to seek for explanations for the high mortality rate for pneumonia in our country,9,10 and in our view, comorbidities play an important role in this. Using the available data from lastyears,9,10 it is easy to conclude that chronic illnesses appear early, limiting quality of life. Our study allowed us to identify the coexistence, in a large number of patients, of multiple comorbidities (eg.: smoker + obese + diabetic + COPD + stroke sequelae + bedridden, …) highly limiting. The impact of any comorbidity, will always depend on their prevalence but also, if in each patient, it exist alone or in combination with others, leading to a different impact in the outcome we are studying.

The results we obtained must be interpreted according to the particularities of the studied population and the access to health care. The large number of elderly patients with concomitant multiple serious comorbidities and the universal access to NHS facilities, may justify the results we found. In fact in other study in our country, the authors found that hospital mortality was particularly related to the aging process and unfavorable socioeconomic conditions11.The risk model obtained will be tested prospectively in the near future, so we hope to be able to validate the results obtained for this population.