Comparing outcomes of ILD patients managed in specialised versus non-specialised centres

This study analysed claims data provided by the largest German statutory health insurance fund (Allgemeine Ortskrankenkasse—AOK), which insured around one third of the German resident population in 2020 (approximately 27 million people) [12]. Membership is open to everyone regardless of factors such as age, income, comorbidities or professional affiliation [13]. Insured persons have free access to a wide range of services with no (outpatient services) or only low co-payments (€10/day for inhospital care). According to national guidelines for secondary data analysis, ethical approval and consent to participate were  not required [14].

The initial data set included all insured adults with an ICD-10 diagnosis of ILD, including idiopathic interstitial pneumonia (IIP) [J84.1], other fibrosing ILD [J84.0, J84.8, J84.9, D48.1], sarcoidosis [D86.0–D86.9], drug-associated ILD [J70.2–J70.4], pneumoconiosis [J62.0–J62.8, J63.0–J63.8], radiation-associated pneumonitis [J70.1], eosinophilic pneumonia [J82], hypersensitivity pneumonitis [J67.9] and connective tissue-associated ILD (CTD) [J99.1], from 1 January 2013 to 31 December 2018. Information on survival was available until 31 December 2019.

Patient selection was performed according to established algorithms [6, 15, 16]. First, patients with an inpatient diagnosis of ILD or a confirmed outpatient diagnosis (outpatient ICD diagnoses in Germany have to be categorized as: ‘Z’ = condition after, ‘A’ = exclusion diagnosis, ‘V’ = suspected diagnosis, and ‘G’ = confirmed diagnosis) from a pulmonologist, an internal medicine specialist or a rheumatologist (the last for CTD only) were preselected. Only patients with at least one relevant diagnostic procedure (bronchoscopy, computerised tomography of the lungs, pulmonary function testing and assessment of antibodies) were considered further. Next, patients with implausible diagnostic patterns were excluded. Implausible diagnostic patterns referred to (1) patients receiving an exclusion diagnosis of ILD after a confirmed ILD diagnosis, (2) patients with radiation-associated pneumonitis, but without diagnosis of malignancy, (3) patients with CTD, but without diagnosis of autoimmune disease, and (4) patients simultaneously assigned to different ILDs. Additionally, patients not continuously enrolled with AOK and those with a baseline or outcome period of less than 1 year were discarded. As we aimed to compare ILD patients initially diagnosed and managed in either specialised ILD centres or non-specialised centres, patients were disregarded if they received their confirmed diagnosis only in an outpatient setting. Furthermore, patients dying during the initial hospital stay were excluded as their outcome period would start right after hospital discharge. Finally, we excluded patients with missing variables. Missing variables could only occur for patient characteristics such as area of residence or employment, but not for claims.

Specialised ILD centres were identified based on competency requirements of the European Reference Network on Rare Lung Diseases (ERN-Lung) valid until 2018 and including minimum patient numbers as well as procedures for expert centres [17]. More specifically, at least 400 patients treated per year were required, with at least 150 newly managed patients. In addition, conduct of at least 50 bronchoalveolar lavages (BAL), 30 endobronchial ultrasound-guided transbronchial needle aspirations (EBUS-TBNA) and 20 surgical lung biopsies or transbronchial cryobiopsies (SLB-TC) was necessary to be classified as a specialised ILD centre. The procedures were identified via the OPS codes (German version of the International Classification of Procedures in Medicine) presented in Additional file 1: Table S1. We adapted the numbers outlined above to the expected proportion of patients insured by the AOK using the market share of AOK in 2014 as reference. Therefore, 120 ILD patients (including 45 newly managed), 15 BAL, 9 EBUS-TBNA and 6 SLB-TC classified a hospital as a specialised ILD centre. In case a hospital did not fulfil the required criteria in a particular year but in both adjacent years, it was also considered a specialised ILD centre in the ‘gap year’. Hospitals not meeting these criteria were considered non-specialised centres. Following this definition, we included patients treated in specialized ILD-centres in the intervention group, while patients initially treated in a hospital not meeting these conditions belonged to the control group. Patients who were transferred directly from a non-specialised centre to a specialized ILD centre after initial diagnoses were also included in the intervention group, as this was considered an “uninterrupted” hospitalization.

Specialised ILD centre- and non-specialised centre-managed patients were compared regarding 2-year all-cause mortality, 1-year all-cause as well as respiratory-related hospitalisation and 1-year overall and respiratory-related health care costs. Outcomes were calculated on a date basis starting at the date of the initial diagnosis, which also represents the date of confirmed diagnosis.

The following primary diagnoses constituted respiratory-related hospitalisations [18]: respiratory infection [A481, B250, J09–J22, J40], pneumothorax [J93], pulmonary embolism [I26], pulmonary hypertension and right heart disease [I50, I270, I272, I278, I279], respiratory insufficiency [J96], other chronic and acute lung diseases [J40–J47] and the ICD-10 codes for the ILD of the respective patients as presented above.

1-year health care costs following the confirmed diagnosis were calculated based on outpatient physician care, inpatient care and pharmaceuticals. Outpatient physician costs refer to all costs in the outpatient sector provided by physicians including all specialties. Inpatient costs comprise all services provided by hospitals, while pharmaceutical costs refer to all medications retrieved by patients from a pharmacy. Respiratory-related outpatient and inpatient costs referred to the respiratory-related ICD codes outlined above. Respiratory-related drug costs referred to filed prescriptions of antifibrotic drugs (pirfenidone, nintedanib), steroids (glucocorticoids, corticosteroids), immunosuppressants, acetylcysteine, sildenafil and drugs for pulmonary arterial hypertension identified by the Anatomical Therapeutic Chemical (ATC) codes presented in Additional file 1: Table S2. As outpatient physician costs were available per quarter of the year, we assigned costs incurred in the quarter of the initial hospital discharge proportionally to days before and after discharge. Within the last observation quarter, costs were proportionally allocated to days under observation. In contrast, inpatient costs were available on a daily base. If a hospital stay exceeded the 1-year observation period, costs were distributed proportionally to in-hospital days within the observation period. Costs for pharmaceuticals were calculated for the day the prescription was filled.

In addition to the main outcomes, differences in pharmaceutical treatment patterns were investigated. Therefore, the onset of ILD-relevant drugs in the 3 months after initial hospital discharge was explored. Here, we considered antifibrotic drugs, immunosuppressants, steroids, anti-clotting drugs, anti-acid drugs, anti-depressants, anti-diabetic drugs, drugs for obstructive airway disease, drugs for pulmonary arterial hypertension, drugs for heart insufficiency/cardiac arrhythmia and drugs for cardiovascular disease (Additional file 1: Table S2).

We used stabilised inverse probability of treatment weighting (IPTW) based on covariate balancing propensity scores (CBPS) [19, 20] to adjust for baseline variables. Based on clinical expertise and pre-existing literature, the following covariates were considered: age in years, sex (male/female), residential area (major city, urban, rural, remote rural) [21], the 2015 version of the German Index of Multiple Deprivation (GIMD 2015) as a proxy for individual socioeconomic background [22, 23], information on nursing home residency (yes/no), care level reflecting need for nursing care, the patient’s employment status (yes/no) and comorbidity burden. Comorbidities were captured based on the Elixhauser Index [24]. To reflect ILD-specific comorbidity more comprehensively, pulmonary hypertension and lung cancer were separated from the corresponding Elixhauser categories [6, 25] and gastro-oesophageal reflux disease, obstructive sleep apnoea syndrome, ischaemic heart disease and thromboembolism were also considered [6]. The Elixhauser score was calculated by summing up all Elixhauser categories without the ILD-specific comorbidities, which were included separately in dummy-coded format. Dummy-coded Elixhauser categories were additionally included in the propensity score model, if they had a prevalence of at least 5% in the treatment group.

Furthermore, we adjusted for medical treatment with ILD-relevant and comorbidity-related pharmaceuticals (Additional file 1: Table S2) in the 3 months prior to the initial hospitalisation. With respect to the utilisation of health care services, the respective time frame was accounted for by including information on all-cause and respiratory-related hospitalisation as well as contacts with all outpatient physicians and a dummy-coded variable for treatment by a pulmonologist. Moreover, we incorporated outpatient, inpatient and pharmaceutical costs in the year prior to the initial hospital admission. Finally, the year of the confirmed diagnosis was accounted for.

To achieve a further reduction in group differences, restricted cubic splines with five knots for the continuous variables were applied [26]. Balance of covariates between the groups was examined via standardised mean differences (SMD) and the Kolgomorov–Smirnov test [19], with SMDs less than 0.1 indicating a good balance [27].

We depicted IPTW-weighted Kaplan–Meier plots for survival and hospitalisations. Cox Proportional Hazard models with sandwich estimators were used to analyse the related group differences as hazard ratios (HR) with 95% confidence intervals (CI). Depending on the share of zero costs, we performed IPTW-weighted one-part and two-part Generalized Linear Gamma models to account for the skewedness of cost data [28]. Euro amounts were calculated based on recycled predictions [29] with 95% CIs obtained from 1000 bootstrap replications.

To contrast pharmaceutical treatment patterns in the 3 months after discharge exploratively, we applied weighted logistic regression models.

Furthermore, we performed three subgroup analyses addressing patients with IIP, sarcoidosis and all other ILD. A new propensity score model was calculated for each subgroup, although we used the same statistical methods as described above.

An additional sensitivity analysis considered that patients initially managed in non-specialised centres might have received treatment in specialised ILD centres in the year after the confirmed diagnosis. Patients who were also managed in specialised ILD centres were censored for the time-to-event analyses at the day they were admitted to the specialised ILD centre and excluded for the cost analyses.

Results were considered significant if the 95% CIs of the differences did not contain ‘1’ in the Cox models and ‘0’ in the cost analyses. All analyses were conducted in R Software version 4.0.3 and SAS version 9.4.

Of the 30,793 patients who met the inclusion criteria, 2022 (6.6%) were initially managed in one of the seven specialised ILD centres classified according to the ERN-Lung requirements. The remainder were managed in 1156 non-specialised centres (Fig. 1). Specialised ILD centres treated on average 154 ILD patients per year, whereas non-specialised centres treated about 10 ILD patients. Additional file 2: Table S3 provides the numbers of procedures per clinic type. Patients managed in specialised ILD centres were younger, less often diagnosed with IIPs and more often employed (Table 1). They lived in less deprived areas and had fewer comorbid conditions. After IPTW weighting, the groups were very well balanced with all SMDs at 0.001 or lower and no difference in the distribution of the continuous variables (Additional file 2: Figure S1). The subgroups are characterised in Additional file 2: Tables S4–S6.

In all, 29% of patients managed in non-specialised centres, but only 21% of patients managed in specialised ILD centres died during the 2-year follow-up. The mean follow-up time was 594 days for patients managed in non-specialised centres and 635 days for patients managed in specialised ILD centres. Hence, the 2-year survival probability for patients initially managed in specialised ILD centres was improved (Fig. 2). Accordingly, there was a significant survival benefit (Table 2). The positive association observed between specialised ILD centre management and mortality was most pronounced for IIPs (Table 2). However, the effect sizes obtained for patients with sarcoidosis and other ILDs were similar to the main analysis. The unweighted estimates are presented in Additional file 3: Table S7.

During the 1-year follow-up, 64% of patients managed in non-specialised centres and 58% of patients managed in specialised ILD centres were hospitalised. In both groups, 29% and 27%, respectively, were hospitalised for respiratory-related reasons. Hence, there were fewer all-cause hospitalisations in specialised ILD centre-managed patients (Fig. 3a), although the pattern for respiratory-related hospitalisations was rather similar (Fig. 3b). The hazard for all-cause hospitalisation was significantly reduced in specialised ILD centre-managed patients, whereas there was no difference regarding respiratory-related hospitalisations (Table 2). In the subgroup analyses, similar estimates below the significance threshold were observed for all-cause hospitalisations. In sarcoidosis patients, a lower, non-significant effect was demonstrated regarding respiratory-related hospitalisation. The unweighted estimates are presented in Additional file 3: Table S7.

Overall costs were similar between both groups but, when focusing on the distinct cost categories, we observed lower inpatient and higher pharmaceutical costs for patients managed in specialised ILD centres (Table 3). Respiratory-related costs were higher in specialised ILD centre-managed patients, as higher respiratory-related pharmaceutical costs overcompensated for lower respiratory-related inpatient costs. In the subgroup analyses, costs for patients with IIP were similar to those in the main analysis (Additional file 3: Table S8). For patients with sarcoidosis and other ILD, overall costs for specialised ILD centre-managed patients were significantly lower on account of reduced inpatient costs (Additional file 3: Tables S9, S10).

Patients managed in specialised ILD centres more often received prescriptions of antifibrotics, immunosuppressants and drugs for pulmonary hypertension in the 3 months after initial discharge (Table 4). Follow-up treatment with anti-clotting drugs, medication against obstructive airway disease and medication for heart insufficiency/cardiac arrhythmia was less frequent for patients managed in specialised ILD centres. The crude percentages for the periods before and after the initial diagnosis are presented in Additional file 4: Table S11.

Of the 28,771 patients initially managed in non-specialised centres, 450 (1.6%) received follow-up treatment in a specialised ILD centre within 12 months. Censoring/exclusion of these patients hardly affected the results obtained in the main analysis (Additional file 5: Table S12).