Causes of acute respiratory hospitalizations predict survival in fibrosing interstitial lung diseases

Acute exacerbation of ILD (AE-ILD) is a common reason for hospitalization; it is also associated with significant mortality. Less is known about the prognostic significance of other events causing acute, non-elective hospitalizations in ILD patients. ILD patients hospitalized due to acute respiratory worsening were collected from medical records. Reasons for respiratory deterioration were classified into AE-ILDs and other causes. Clinical features and survival data of idiopathic pulmonary fibrosis (IPF) and other types of ILDs were evaluated and compared. In all, 237 patients (138 with IPF and 99 with other ILD) fulfilled the inclusion criteria. Of the non-IPF ILD types, the most prevalent subgroups were connective tissue disease-associated ILD (n = 33) and asbestosis (n = 22). The most common cause for hospitalization was AE-ILD explaining 41% of hospitalizations. Lower respiratory tract infection (22%), subacute progression of ILD (12%) and cardiovascular causes (7.2%) were other common reasons for hospital treatment. Patients with a lower respiratory tract infection had a more favorable prognosis compared with patients with AE-ILD. AE-ILDs were less fatal than cardiovascular or concurrent non-ILD-related causes for hospitalizations in non-IPF patients. High Gender-Age-Physiology (GAP) index was a marker for shortened survival and earlier AE-ILDs in all patients. IPF patients had a significantly shorter overall and post-hospitalization survival time compared with other ILDs. Most respiratory hospitalizations in ILD patients were related to causes other than AE-ILD, which highlights the importance of accurate differential diagnosis in order to target the appropriate treatment for each ILD patient.


Introduction
Idiopathic pulmonary fibrosis (IPF) and other progressive fibrosing interstitial lung diseases (ILD) are serious, often fatal conditions also causing a high burden for health care systems. In a Spanish study, the numbers of hospitalizations related to IPF almost doubled between 2004 and 2013 from 3.82/100 000 inhabitants to 6.98/100 000 inhabitants [1]. Based on the results of the epidemiological studies, a significant proportion of hospitalizations in ILD patients were related to ILD or other reasons causing respiratory symptoms [1][2][3][4], and the majority of ILD patients had experienced some kind of event requiring hospitalization during their disease history [4]. Most of the register-based studies have investigated solely patients with IPF, while those of other ILDs are sparse. Table 1 presents data from previously published epidemiological studies and other studies on hospitalizations of ILD patients [5][6][7][8][9][10][11].
Acute exacerbation (AE) is a severe complication of ILD, which seems to cause significant mortality in all types of ILDs [12,13]. The current diagnostic criteria for AE-IPF have been suggested to be applied for other ILDs as well [12,13]. The diagnosis of AE-ILD requires a detailed evaluation of clinical symptoms and high-resolution computed tomography (HRCT) images of chest [12]. One substantial problem in AE-ILD research is the fact that no specific diagnosis code exists for AE-ILD. As a consequence, the large epidemiological studies mentioned above do not provide detailed information on the number of AE-ILDs [1][2][3][4]. In other studies with access to the patient data, the proportion of AE-ILDs out of all the hospitalizations caused by an acute respiratory deterioration has ranged between 29-55% [5,6,10,11], when only two of these investigations included patients with non-IPF ILDs [5,11]. In the previous studies, which have not presented the exact data on the number of AE-ILDs, it was claimed that respiratory or cardiopulmonary hospitalizations increased the mortality of ILD patients more than hospitalization due to other illnesses [7,9].
The aim of this present study was to evaluate fibrosing ILD patients hospitalized non-electively due to an acute deterioration of respiratory symptoms in two Finnish hospitals, namely Oulu University Hospital (OUH) and Oulaskangas Hospital (OH). The causes for hospitalizations were evaluated and classified as either AE-ILDs or other reasons, the risk factors for mortality and AE-ILDs were compared between IPF and non-IPF patients.

Patient and data collection
The study material consists of ILD patients who were non-electively hospitalized due to acute respiratory symptoms in OUH or OH between 1/1/2008 and 31/12/2017. A total of 128 of the 237 patients included were also included in our previous study [14]. The flowchart of the study and criteria used to define AE-ILD are presented in Fig 1. For definition of AE-ILD, we applied the current criteria by Collard et al., thus including both triggered and idiopathic AE-ILDs [12]. We performed a search with International Classification of Diseases version 10 (ICD-10) codes J84.1, J84.8 and J84.9 to identify the patients and their treatment periods from medical records as described previously [14]. Another search was performed with ICD-10 codes J61, J99, J99.0 � and J99 � M05.1 to find the hospitalizations related to asbestosis and rheumatoid arthritis-associated ILD (RA-ILD). However, there were no matches for the search with J99 codes, because J84 codes had been used for all RA-ILD patients in our cohort. Most ILD patients had undergone a multidisciplinary diagnosis (MDD) evaluation by pulmonologists, thoracic radiologists and a lung pathologist. For this particular study, the type of ILD was reevaluated according to the current international criteria assessing all clinical data from patient records and death certificates, pathologists' reports and the radiologists' reports [15,16]. The re-evaluation was performed by a specialist physician in pulmonary medicine (JS). Ambiguous cases were re-assessed by another specialist physician in pulmonary medicine (RK) participating in this study. All patients included (n = 237) had HRCT images of chest available for evaluation of ILD type. The clinical information was collected systematically from electronic patient records dating back about 20 years in OUH and OH in a form designed for the present study. We collected the date of birth, age at diagnosis and at hospitalization, gender, smoking status, date of ILD diagnosis, pulmonary function test (PFT), HRCT reports and images, use of longterm oxygen treatment, pharmacological therapy prior to hospital admission and the possible trigger for AE-ILD. Patients with less than a 5 years' smoking history were considered nonsmokers. Recorded PFT values were measured maximally 6 months prior to or after the diagnosis or hospitalization date. PFT was performed according to the ATS/ERS guidelines [17,18] and Finnish reference values were used for interpretation of the results [19]. We categorized the reasons for hospitalizations to AE-ILDs and other reasons having caused respiratory symptoms, as presented in Fig 1. The detailed data on hospitalizations was collected related to the first non-elective hospitalization after or at the time of ILD diagnosis, if the diagnosis of ILD was made during a non-elective hospitalization. Out of the 237 patients included in the study, there were 96 patients whose first respiratory hospitalization was caused by an AE-ILD (Table 2). We also evaluated the subsequent respiratory hospitalizations of those patients who had their first respiratory hospitalization due to a cause other than an AE-ILD to identify the patients who experienced an episode of AE-ILD later. Almost half, i.e. 109 patients out of 237 did not experience an AE-ILD at all during the follow-up time. Thirtytwo patients out of 237 who were first hospitalized due to some reason other than AE-ILD, were discharged and experienced a hospital treatment caused by an AE-ILD later during the follow-up time. The date of the first AE-ILD of each patient having experienced this episode was recorded and utilized for the evaluation of the risk for AE-ILD. Retrospective data on IPF patients from one center between 1997 and 2012.
The subacute progression of ILD was defined as increased fibrotic changes either in HRCT or chest x-ray compared with previous images. A subclass "respiratory symptoms without explanatory findings" was used for patients hospitalized non-electively due to acute respiratory symptoms, but whose investigations did not reveal any acute findings in clinical status, HRCT, chest x-ray or laboratory tests. A lower respiratory tract infection, e.g. pneumonia or acute   bronchitis, was recorded as the reason for hospitalization if the patient had the typical symptoms of a lower respiratory tract infection (increased cough and mucus production above the normal without other reasons), with or without an elevated CRP level or a new infiltrate in chest x-ray or HRCT. The treatment periods caused by heart failure, coronary atherosclerosis, myocardial infarction or atrial fibrillation were considered to be cardiovascular (CV). Heart failure was documented as the reason for hospitalization if at least two of the following criteria were met: congestion in HRCT or chest x-ray, elevated B-type natriuretic peptide (BNP), swollen limbs or a finding compatible with heart failure in the echocardiogram. High Gender-Age-Physiology (GAP) index and GAP stage were defined as in the previous study [20].  linear trend for factor levels was used, when the survival differences of patients were evaluated according to the cause of the hospitalization. Cox regression model was utilized for the assessment of risk factors for death and AE-ILD.

Ethics
As this was a retrospective study and the majority of study subjects were deceased, no patient consent forms were gathered in accordance with Finnish legislation. The study protocol was approved by the Ethical Committee of the Northern Ostrobothnia Hospital District (statement 2/2015). Permission to use death certificates was given by Statistics Finland (Dnro: TK-53-515-15). The study was conducted in compliance with the Declaration of Helsinki.

Patient demographics
The most common ILD type was IPF, accounting for 58% of the 237 patients. The median follow-up time of the patients was 4.7 years. The hospital mortality was 11% with no statistically significant difference between IPF and non-IPF cases. The majority of the patients (87%) had died at the end of the follow-up time, with more survivors among the non-IPF patients than in IPF. Median survival time from diagnosis in IPF patients was significantly shorter, e.g. 3.5 years, compared with non-IPF patients who had 7.8 years' median survival (p<0.001, log rank-test). The post-hospitalization survival was also shorter in IPF patients, e.g. 1.1 years, compared with non-IPF patients with a survival time of 1.9 years (p = 0.001, log rank-test). Vital capacity (VC), but not forced vital capacity (FVC), and diffusion capacity for carbon monoxide (DLCO) in the non-IPF patients at diagnosis were higher than those of IPF patients; however, these differences could no longer be observed in the PTF results examined near the hospitalization date. Slightly more than every fourth patient (26%) had ILD diagnosed during the non-elective hospital treatment period. The median time from the ILD diagnosis date to the hospitalization date in those with a previous ILD diagnosis was more than double in non-IPF patients as compared to those with IPF, namely 5.9 years compared with 2.4 years (p<0.001). Characteristics of the study subjects are shown in Table 2 and ILD types and the number of patients having experienced an AE-ILD are presented in Table 3. Medical therapy for ILD prior to hospitalization is shown in S1 Table. Most hospitalizations were due to reasons other than AE-ILD AE-ILD accounted for 96 (41%) of all hospitalizations with no statistically significant difference between IPF and non-IPF cases ( Table 4). The AE-ILD was fatal in 20 patients (IPF n = 15, other ILDs n = 5). A lower respiratory tract infection (22%) and CV causes (7.2%) were the most common single causes of non-ILD-related hospitalizations. A subacute ILD progression was a more common reason for acute hospitalization among IPF patients than in their non-IPF counterparts. Non-ILD-related hospitalizations for various reasons (subcategory "other" in Table 4) were more common in non-IPF patients than in IPF. The cause for hospitalization was associated with post-hospitalization survival (Fig 2). Patients with a lower respiratory tract infection had a more favorable prognosis compared with patients with an AE-ILD (Table 5). In the subgroup of non-IPF patients with a CV or a multifactorial cause for hospitalization, the post-hospitalization survival was shorter than in patients experiencing an AE-ILD (Table 5).

High GAP index was a risk factor for mortality and earlier AE-ILD
A high GAP index was associated with a shorter overall survival and an earlier occurrence of AE-ILD in all patients (Fig 3, Table 6). The occurrence of deaths and AE-ILDs at different time points from diagnosis according to the GAP stages is presented in Figs 4 and 5 and in S2

Discussion
We have investigated data in 237 ILD patients that were hospitalized non-electively due to acute worsening of respiratory symptoms. All of the data of each patient was re-assessed in a very detailed manner. Acute exacerbations of ILD explained about 40% of the hospitalizations whereas the majority of the hospitalizations were explained by other causes. The cause for hospitalization was associated with survival time in all patients, patients with lower respiratory infection having the most favorable prognosis. A high GAP index was associated with both a shortened survival and an earlier occurrence of AE-ILDs and was a more specific indicator for earlier AE-ILDs than for deaths.   Non-IPF patients were found to have a lower GAP stage and higher VC and DLCO levels at the time of diagnosis as compared with IPF patients, indicative of a less progressed disease of non-IPF patients at baseline. However, at the time of hospitalization, the differences in VC and DLCO between IPF and non-IPF patients had disappeared. As can be seen in S4 Table, in comparison with most other studies, here the median time from diagnosis to hospitalization seemed to be longer [5,6,9] and patients were older [5][6][7]9], even though the baseline pulmonary function test (PFT) values in our study and those of others were approximately at the same level. A Japanese study has reported higher PFT results in their investigated subjects near the first hospitalization date in comparison with our material, even though the patients in that study and our own were approximately the same age [11]. These differences may be attributable to the different genetic and ethnic backgrounds of the study subjects and also to the heterogeneity of the non-IPF patients in the above publications [5,7,11]. There are also some previous studies suggesting that ethnic background might influence the clinical features and disease course of the ILD patients [21,22].
AE-ILD has been reported to be the most common individual cause for a respiratory deterioration in different types of ILDs accounting for 29-55% [5,6,10,11] of respiratory hospitalizations, values consistent with our own findings. Lower respiratory tract infections were observed in 20-33% of respiratory hospitalizations in previous studies, which is in line with our result of 22% [5,6,10,11]. Furthermore, a subacute progression was the reason for hospitalization in 12% of patients in our study, again in accordance with earlier investigations reporting proportions of 15% and 17% [5,10]. Given the challenges concerning the diagnostic practices of ILD patients with acute respiratory worsening, which make retrospective studies extremely demanding to perform, the results of the previous and the present study are amazingly consistent. To be able to identify the reasons for hospitalizations retrospectively, the clinical evaluation of the patients during the hospital treatment had to be thorough e.g. involving HRCT imaging in most cases. The consistent findings of the studies might also be related to the fact that researchers in European, American and Asian hospitals have utilized the current criteria for AE-ILD, allowing them to differentiate between AE-ILDs and other causes of hospitalizations based on the same principles. So far, our study has been the only one conducted in Europe, since other studies have investigated patients originating from either the USA [5,7,9] or Asia [6,11]. Thus, it does seem that the ethnic background did not exert a major impact on the reasons for respiratory worsening of ILD patients.
AE-ILD has been reported to be associated with a poor outcome in various types of ILDs [5,6,10,[23][24][25][26]. The treatment of AE-ILDs usually includes high doses of corticosteroids and/ or other immunosuppressive therapies, even though the evidence of their efficacy is limited [12,13]. Our results together with earlier investigations suggest that it is crucial to be meticulous in the differential diagnostics in ILD patients with acute respiratory symptoms. Based on the present results and those of other investigators, most hospitalizations were due to some reasons other than AE-ILD. Patients deserve to be treated according to the actual cause, thus avoiding the potentially harmful effects of the treatments targeted at AE-ILD.
In our study population, those patients hospitalized due to AE-ILD had a shorter survival compared with patients with a lower respiratory tract infection, the difference was more distinct in IPF patients in comparison to those with other ILDs. A similar result has also been presented by Teramachi et al. in a recent study of IPF patients, in which a respiratory tract infection was associated with a reduced 90-day post-hospitalization mortality as compared with other causes for hospitalization [10]. It can be speculated that the occurrence of a respiratory tract infection, if not fulfilling the criteria for a triggered AE-ILD, may be a marker of an ILD phenotype associated with a slow disease progression.
Interestingly, in the subgroup of non-IPF patients with a CV cause for hospitalization or a lower respiratory tract infection concurrent with some other cause, the post-hospitalization survival time was shorter than in non-IPF patients with AE-ILD. Moua et al. have observed that in non-IPF patients, an AE-ILD did not cause increased mortality in comparison with other types of hospitalizations in a multivariate model [5]. These findings may reflect the fact that AE-ILDs are less fatal for non-IPF patients than for those patients with IPF, which was also supported by our previous study, where IPF patients had a significantly shorter survival after an AE in comparison with non-IPF patients [14].
In our study, the GAP index was a risk factor for both death and AE-ILD. At least one of these findings has been described in earlier investigations [18,[27][28][29][30][31][32], even though this has not been verified in all reports [9,33]. We also demonstrated that AE-ILDs can occur at any phase of the disease in a cohort involving both IPF and non-IPF patients, a finding which is a known feature of IPF, making the disease course difficult to predict [12,34].
The occurrence of AE-ILD was relatively high, e.g. 54%, in our study population. In the previous studies presenting data on GAP and its effect on mortality, the prevalence of AE-ILD has varied between 20-40% [18,22,23], these discrepancies may be caused by different types of study protocols, e.g. we included only non-electively hospitalized patients. Our results indicate that the GAP index is a functional prognostic marker even in a cohort enriched with AE-ILDs. Nonetheless, almost half of the patients survived less than one year from diagnosis had GAP stage I or II at baseline, suggesting that GAP alone did not provide an exact prognosis for these patients, a result consistent with earlier studies presenting data on the accuracy of GAP in the prognostic evaluation of IPF patients [29,31]. However, GAP stage III seemed to be quite a specific predictor of early AE-ILDs, since all the patients with GAP stage III disease at diagnosis experienced an AE-ILD within 1 year after the diagnosis, whereas those GAP stage III patients who survived the first year without an AE-ILD did not experience the episode later. To our knowledge, there is no earlier published data demonstrating similar findings.
In Finland pirfenidone received a recommendation for reimbursement in 2013 and nintedanib in 2015 for the treatment of the patients with IPF by the Social Insurance Institution of Finland (Kela). According to the open database of Kela, about 90 IPF patients had received a reimbursement for antifibrotic drugs in OUH and OH by the end of 2017 when only 13 patients in our study had been taking pirfenidone or nintedanib prior to hospitalization (S1 Table) [35]. Since only the patients with acute hospitalizations, but not all, were included in our study, the results of Kela database are more trustworthy about the use of antifibrotic drugs in our hospital area than the numbers of the present study. Thus, one can speculate that perhaps the patients who were not being treated with antifibrotic therapy were more often hospitalized than those receiving this form of drug treatment.
Our study has some limitations that are common in retrospective studies, namely some missing patient data and the possible incorrect use of ICD-10 diagnosis codes by clinicians, which might have led to the exclusion of a few potential cases. Furthermore, a radiologist did not re-evaluate the HRCT scans of the patients especially for this research, which can also be regarded as a limitation of this study. Nevertheless, the re-evaluation of the HRCTs was performed for the majority of the patients in MDD by thoracic radiologists as a part of the routine clinical practice. The small numbers of patients with a certain non-IPF ILD subtype can also be considered a limitation of our study. However, combining different progressive ILD subtypes together to a group called "progressive fibrosing interstitial lung diseases" (PF-ILD) has recently become more common [36][37][38][39][40] and we believe that our study may provide relevant novel important information on the disease course of PF-ILDs.
The careful and systematic evaluation of the data as well as the re-classification of ILD types according to the current international guidelines are the strengths of our study. It should be noted that differential diagnostics in hospitalized patients with progressed ILD would be challenging even in a prospective study protocol, because AE-ILD, respiratory infections and heart failure often share similar and even concurrent clinical and radiological features.

Conclusions
To conclude, although AE-ILD was the most common single cause for hospitalization, a relatively large proportion of hospitalizations, e.g. 59%, was attributable to causes other than AE-ILD. In view of the potentially harmful effects of the therapies targeted at managing an AE-ILD, clinicians should be aware that there are multiple possible etiologies in addition to an AE-ILD when treating ILD patients with acute respiratory worsening; some of them may be potentially even more serious for patients than an AE.
Supporting information S1 Table. Pharmacological therapy for idiopathic pulmonary fibrosis (IPF) and other interstitial lung disease (ILD) prior to the hospital treatment. Data are presented as numbers of patients (%) or median (interquartile range). a The total number of patients treated with corticosteroid monotherapy or combination therapy either because of ILD or other disease. Cs, corticosteroid; ILD, interstitial lung disease; IPF, idiopathic pulmonary fibrosis; NAC, N-acetylcysteine. (PDF) S2 Table. Deaths and patients with an acute exacerbation of ILD in different Gender-Age-Physiology (GAP) stages. Data are presented as the number of patients (% of all patients within the GAP stage / % of cases (either deaths or AE-ILDs) within the follow-up time category). Only the patients for whom GAP data was available were included in the analysis; GAP data of 24 patients was missing. There were 16 patients in GAP I, 13 patients in GAP II and 0 patients in GAP III alive at the end of the follow-up time. Of these still living patients, 8 had a follow-up time less than 10 years. a Of the patients having experienced multiple AE-ILDs, only the first AE-ILD is recorded. Abbreviations: AE-ILD, acute exacerbation of interstitial lung disease. AE-ILD, acute exacerbation of interstitial lung disease; GAP, gender-age-physiology index, ILD, interstitial lung disease. (PDF) S3 Table. Underlying and immediate causes for death in idiopathic pulmonary fibrosis (IPF) and other interstitial lung disease (ILD) patients. Seven patients with lung transplantation were excluded from the analysis. a 2 patients had Morbus Alzheimer, 1 diabetes mellitus, 1 Morbus Parkinson, 1 chronic obstructive lung disease and 1 patient was drowned. b 1 patient had staphylococcus aureus septicemia and 1 patient had septicemia caused by a gram-negative rod bacterium. c 1 patient had Morbus Alzheimer, 1 an epileptic seizure, 1 an acute kidney injury. (PDF) S4 Table. Clinical features of ILD patients in the previous and the present studies at baseline and during hospitalization. Abbreviations: AE, acute exacerbation of ILD; c-IIP, chronic idiopathic interstitial pneumonia; CTD-ILD, connective tissue disease-associated interstitial lung disease; DLCO, diffusion capacity for carbon monoxide; FVC, forced vital capacity; ILD, interstitial lung disease; IQR, interquartile range; IPF, idiopathic pulmonary fibrosis; NA, not applicable; SD, standard deviation. (PDF) S1 Appendix. Anonymized patient data. (XLSX)