Prognostic factors for hospital mortality in patients with idiopathic pulmonary fibrosis

Background: The prognostic factors for hospital mortality in patients with idiopathic pulmonary fibrosis (IPF) have not been characterised well. We aimed to determine the impact of respiratory hospitalisation on patient outcomes and the prognostic predictors for hospital mortality. Methods: Data from patients with IPF was collected and analysed from two cohorts, TaiwanNational Health Insurance Research Database (NHIRD) between 2001 and 2011 and a national medical centre (NMC)between 2001and 2015. Results: 122 patients in NHIRD cohort were recruited. The median survival was lower in patients with respiratory hospitalisation (25/122, 29.8%) compared with non-respiratory hospitalisation(59/122, 48.4%) in NHIRD cohort (2.3 years vs nonrespiratory 9 years, p =0.007). In NMC cohort (N=47), patients with acute respiratory deterioration (ARD) hospitalisation (30/47,63.8%) compared with non-ARD hospitalisation (17/47, 36.2%) had poor survival (0.9 monthsvs 29.6 months, p <0.001). The differences in survival between two cohorts were associated with age, causes of hospitalisation, mechanical ventilation and comorbidities. The patients hospitalised for ARD had statistically higher GAP score, lower oxygenation, higher C-reactive protein levels, higher neutrophil counts and lower serum albumin level compared to those non-ARD hospitalisation. For both in-hospital and 1 year mortality, multivariate analysis adjusting for patient characteristics confirmed the independent predictive factors were steroid pulse therapy, acute exacerbation and mechanical ventilation. Conclusions: The respiratory hospitalisation negatively impacted patient outcomes, and the independent prognostic factors may be utilised to refine the

4 management in patients with IPF.

Background
Idiopathic pulmonary fibrosis (IPF) is a progressive and life-threatening interstitial lung disease with unknown etiology [1].The mortality burden attributable to IPF is higher than that of several cancer types [2,3].The median survival of patients with IPF is only 2-3 years [1,2].Acute respiratory deterioration (ARD) can occur at any time. Acute exacerbation of IPF (AE-IPF) is a serious condition of ARD and require hospitalisation which is associated with extremely poor outcome [4][5][6].However, the presently available literature cannot provide the sufficient markers to predict outcomes after hospitalisation in patients with IPF.
Respiratory hospitalisations have been recognised as an important endpoint in IPF outcomes [7][8][9][10].Although many individual clinical variables predict survival in IPF, the clinical characteristics of hospitalisation for ARD and the predictors of hospital mortality in patients with IPF have not been fully elucidated.The causes of respiratory hospitalisation include AE-IPF, infection and extra-parenchymal factors such as congestive heart failure, pleural effusion and pneumothorax. AE-IPF is the most common cause of death in IPF [11,12].The short-term mortality of AE-IPFis about 50% and typically exceeds 90% in patients admitted to an intensive care unit [13][14][15].Prognostic biomarker including serum lactate dehydrogenase [13]procalcitonin [16], circulating fibrocytes [17]havebeen proposed as the predictors of poor outcome in IPF.Whether these parameters can be applied topredict hospitalized patient outcomes in IPF remains unclear.
In the present study, we retrospectively analysed two IPF cohorts from Taiwan National Health Insurance Research Database (NHIRD) and a national medical centre 5 (NMC). This study aimed to determine the impact of respiratory hospitalisation on patient outcomes and to identify the independent prognostic factors for hospital mortality in patients with IPF.

Ethics statement
We consulted with the institutional review board of Taipei Veterans General Hospital (approval number: IRB2016-06-004CC) and waived informed consent for our retrospective observational cohort study.  (2). no NHI claim with diagnosis for any other interstitial lung diseases "see Supplementary Appendix S1, Additional File 1 "since the day of IPF diagnosis;(3). patient's age of ≥ 18 years old at the time of IPF diagnosis and (4). at 6 least one NHI procedure for surgical lung biopsy or computed tomography of the thorax "see Supplementary Appendix S1, Additional File 1 " before or on the day of IPF diagnosis. Hospitalisations of at least 1 day in duration subsequent to the IPF diagnosis were categorised as respiratory or non-respiratory in nature"see Supplementary Appendix S2, Additional File 1", "see Supplementary Figure  Part II: Electronic medical records from a national medical centre

Study sample
We retrospectively examined the electronic medical charts of all patients with IPF who were admitted to our institution for ICD-9-CM code 516.3 between January 2001 and January 2015. Each admission was reassessed by two independent pulmonologists for IPF diagnosis as defined by the 2011 American Thoracic Society/European Respiratory Society criteria [19].Of them, radiological confirmed to have usual interstitial pneumonia on high-resolution computed tomography (HRCT) by two independent radiologists "see Supplementary Figure S2, Additional File 2".

Data collection
Demographic and pulmonary function test data were collected during IPF diagnosis.
The gender, age, physiology index for IPF (GAP index) was calculated for each patient [20]. The clinical characteristics that were collected within 24 h at the first hospitalisation were considered as physical and laboratory findings, such as the use of steroidsor long-term oxygen therapy prior to hospitalisation, HRCT scan and bronchoalveolar lavage. Hospitalisations of at least 24 h following IPF diagnosis were categorised as ARD and non-ARD ones for survival analysis. The primary outcome was the survival status after the first hospitalisation. The secondary outcome was to identify the predictor of in-hospital and within 1-year follow-up mortality. Survival status that was obtained from electronic medical records and telephone were reviewed by the investigator.

Definition
Hospitalisation forward was defined, as shown in "Supplementary Table S1, Additional File 3". ARD was further categorised as AE-IPF, parenchymal cause of ARD other than AE (e.g., pneumonia and subacute progression), or extraparenchymal cause of ARD (e.g., pneumothorax and pleural effusion) [21] "see Supplementary Figure S2, Additional File 2". AE-IPF was diagnosed according the revised diagnostic criteria [22]. Radiation pattern on HRCT was classified as peripheral, multifocal and diffuse types by one radiologist and one pulmonologist.

Statistical Analysis
On the basis of Taiwanese NHIRD, we counted the number of patients diagnosed with IPF and calculated the incidence rate as the number of newly diagnosed patients per 100,000 persons in the NHI system in each year from 2001 to 2011.Survival time was calculated from diagnosis to death or the end of 2011. We estimated the median time on the basis of a Kaplan-Meier survival estimator.
Survival time for patients with respiratory or non-respiratory hospitalisation was calculated with Kaplan-Meier survival curves and compared by the log-rank test.
For data from the NMC, Chi-squared test or Fisher's exact test was used for categorical variables, and t-test or Mann-Whitney U test was used for continuous variables. Median survival time was analysed from initial consultation until death or the end of 2015. Survival time for patients with ARD or non-ARD hospitalisation was analysed based on the Kaplan-Meier estimator and the log-rank test. A multivariate 8 Cox regression model performed by the method of enter technique was applied using the variables associated with in-hospital mortality and 1-year mortality as a univariate analysis. A two-sided p ≤ 0.05 was considered statistically significant. All data were analysed using IBM SPSS 19 statistical package.

IPF from NHIRD
A total of 348 patients with ICD-9 code 516.3 were evaluated between 2001 and 2011. Amongthese patients, 122 were confirmed to have IPF ( Figure S1).The mean age was 65.0±16.7 years old, and the male-to-female ratio was2.5:1.The median survival from diagnosis to death was 5.2 years. A total of 84 patients underwent hospitalisation after diagnosis and 29.8% (25/84) were admitted due to respiratory cause. The median survival was lower in patients with respiratory hospitalisation than those admitted due to non-respiratory hospitalization(n=59)(2.3years vs 9 years, p=0.007; Figure 1a).

Impact of ARD hospitalisation on survival of patients with IPF in cohort from a NMC
A total of 129 hospitalised patients with ICD-9 code 516.3 were evaluated between 2001 and 2015.Amongthese patients, 53 patients with definite IPF were enrolled, including ARD hospitalisation (n=30), extra-parenchymal causes (n=6) and non-ARD hospitalisation (n=17). The median survival from initial diagnosis to death was 27 months. The median survival was lower in patients with ARDhospitalisationthan those in non-ARD hospitalisation (0.9 months vs. 29.6 months, p<0.001; Fig. 1b).For ARD hospitalisation, 60 % of them died on their first hospitalisation compared with 9 0% of non-ARDhospitalized patients (p<0.001) ( Table 1).Particularly, AE-IPF accounts for 53.3 %(16/30) of ARD hospitalization ( Figure S2).9 of the 16 AE-IPF patients needed ICU care with mechanical ventilation. Furthermore, the hospitality mortality for patients in ICU with MV was 84.6% (11/13) with median survival time was 23 days.The median survival for AE-IPF was much lower than hospitalization for parenchymal cause(0.8 monthsvs12.0 months; p =0.006).Mortality at 1yr after hospitalisation for ARDwas significantly worse than in non-ARD (73.3 % vs 5.9%,p<0.001)( Table 1). 29 patients survived from ARD hospitalization were followed and median time to death after discharge was 5.7 months.

Comparison of characteristics of two cohorts
Patients in NMC cohort were statistically older at diagnosis(74. p<0.001,respectively).The incidence of respiratory hospitalisation,mechanical ventilation, and co-existing comorbidity in gastroesophageal reflux, diabetes mellitus, chronic kidney disease, pulmonary hypertension or emphysema were statistically higherfor patients in NMC compared to those in NHIRD (Table 3).

Independent factors associated with in-hospital and 1 year mortality for ARD hospitalisation
The patients hospitalised for ARD had statistically higher GAP score (Table 2), lower oxygenation (P/F ratio; partial pressure of oxygen/fraction of inspired oxygen), higher C-reactive protein(CRP) levels, higher neutrophil counts and lower serum albumin levels compared to those non-ARD hospitalisation ( Table 1). The predictors of in-hospital mortality (Table 4) and 1 year mortality (Table5) for patients with IPF are presented. For both in-hospital and 1 year mortality, multivariate analysis adjusting for patient characteristics confirmed the independent predictive factors were steroid pulse therapy, AE-IPF and ICU admission with MV.

Discussion
The current study confirmed the impact of respiratory hospitalisation on patient outcomes and identify the prognostic factors for mortality in patients with IPF. In the NHIRD cohort,the median survival was lower in patients with respiratory hospitalisationcompared with non-respiratory hospitalisation. The finding was also observed inthe NMC cohortfor patients with ARD hospitalisation. The median survivalfrom IPF diagnosis to death was 5.2 and 2.3 years for NHIRD and NMC cohorts respectively. The differences in survival betweenthese two cohorts were associated with age of diagnosis, causes of hospitalisation, mechanical ventilation and comorbidities. The patients hospitalised for ARD had statistically higher GAP score, lower oxygenation, higher C-reactive protein levels, higher neutrophil counts and lower serum albumin levels compared to those non-ARD hospitalisation. For both in-hospital and 1 year mortality, multivariate analysis adjusting for patient characteristics revealed that steroid pulse therapy, acute exacerbation and mechanical ventilation are the independent predictive factors. The respiratory hospitalisation negatively impacted patient outcomes and the independent prognostic factors may be utilised to refine the management in patients with IPF. ARD hospitalisationis commonly due to AE-IPF which hasextremely poor outcome andhigh mortality rate reported in some studies [4][5][6].Teramachi et al. [21]showed and acute disease process [22].The major histological findings in AE-IPF were a diffusing alveolar damage(DAD) pattern superimposed on ausual interstitial pneumonia pattern [23,24]. DAD is similar to the pathological finding of acute respiratory distress syndrome (ARDS) that can occur in patients without chronic lung disease following a large number of triggers [25].The key driver of the pathogenesis was dysregulated inflammation both in the epithelial and endothelial sites, recruiting activated macrophages and neutrophils which release a variety of inflammatory mediators [26].Acute worsening of IPF may result from various viral infections in an immune-mediated manner [27,28] Corticosteroids may have some beneficial effects on AE-IPF and high-dose corticosteroids are commonly applied for the treatment of AE-IPF [24,33,34].However, there were no controlled trials on which to judge efficacy. In our observation, steroids pulse therapy for AE-IPF was associated with the increased hospital mortality which may suggest progressive alveolar damage superimposed on interstitial fibrosis and the necessity of oxygen supplement and even MV support.
This could explain why steroid pulse therapy and ICU admission with MV predict poor outcome.
Our study has several limitations. Firstly, the NHIRD database lacks detailed clinical information. Therefore, we cannot investigate the impact of other factors on patient survival, given that we use NMC clinical record to identify the clinical characteristics in patients with IPF. Secondly, NMC cohort was a retrospective study based on clinical records, and there were some items with missing values included in the statistical analysis. Third, this study included the period when antifibrotics were not approved in Taiwan. It is possible that differences in treatment plan influenced the incidence of AE-IPF. Fourth, we were limited in our ability to collect complete information on hospitalizations that may have occurred outside of NMC.

Conclusion
In summary, hospitalisation for ARD is an ominous predictor of subsequent mortality in patients with IPF. Clinical independent predictors for mortality would be useful in counselling patients, making treatment decisions and considering lung transplantation promptly. Declarations SN.Y.,HK. K and DW. P. were responsible for the study design and were involved in data analysis. SN.Y., PC. L. and CC. H.were responsible for data collection and interpretation. SN.Y and HY. H. performed the statistical analyses. SN.Y.,HK. K and YL. C wrote the first draft of the manuscript, which was revised by all authors. MI. C. and DW. P.are the guarantors of the paper, and they take responsibility for the integrity of the work as a whole, from inception to the published article.

Funding Sources
This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors.

Availability of data and materials
The datasets used and analysed during the current study are available from the corresponding author on reasonable request.

Ethics approval and consent to participate
This study was performed in accordance with the Declaration of Helsinki. This human study was approved by IRB of Taipei Veterans General Hospital, Taipei, Taiwan -approval: 2016-06-004CC. Adult participant consent was not required for our observational cohort study.

Competing interests
The authors declare no conflict of interest concerning this article