Healthcare utilisation preceding the diagnosis of interstitial lung disease

Authors: Thomas Prior, Claus Høstrup Vestergaard, Sissel Kronborg-White, Elisabeth Bendstrup, Anders Prior

Abstract

Background

Diagnosis of interstitial lung disease (ILD) is frequently delayed, as symptoms are misattributed to other lung or cardiac diseases. Identifying prediagnostic patterns in healthcare utilisation may highlight areas of intervention leading to earlier diagnosis and treatment, potentially improving prognosis. We aimed to investigate healthcare utilisation in primary and secondary care in ILD patients compared to matched references over 10 years preceding the ILD diagnosis, using nationwide Danish registers and adjusting for comorbidities.

Methods

We conducted a nationwide matched cohort study using data from national registers. The study population was patients with a first-time diagnosis of ILD between 1 January 2011 and 31 December 2019. Cases were matched with 10 references. Outcomes included number of hospital contacts, chest radiographs or computed tomography, and general practitioner consultations, spirometries, and C-reactive protein measurements. Incidence rate ratios (IRRs) were calculated using multivariate negative binomial regression models, adjusted for socioeconomic factors and comorbidities.

Results

We identified 13 161 cases with ILD, matched with 131 620 references. Mean±sd age at diagnosis was 68.4±13.9 years, with a slight male predominance. Most adjusted IRRs were elevated for up to 10 years preceding diagnosis, steadily increasing and sharply peaking 1 year prior to diagnosis. Similar patterns were seen when stratifying analyses for comorbidities. The adjusted IRRs were generally higher for patients without comorbidities.

Conclusions

This nationwide study shows an increase in healthcare utilisation in both primary and secondary care in ILD patients up to 10 years before diagnosis, indicating a diagnostic delay longer than previously reported.

Introduction

Interstitial lung diseases (ILDs) encompass a spectrum of inflammatory to fibrotic pathological processes in the lungs. The fibrotic ILDs such as idiopathic pulmonary fibrosis (IPF) and fibrotic hypersensitivity pneumonitis (HP) especially have a poor prognosis [1, 2]. Antifibrotic treatments slow down disease progression [3–5], but initiation of treatment is hampered by a diagnostic delay from initial symptoms to final diagnosis. In IPF, a median delay of up to 2.2 years has been reported, but up to 25% of patients had >5 years delay [6–9]. A long diagnostic delay was shown to negatively impact the prognosis of patients with IPF [10].

The diagnostic delay was primarily due to delays among patients, primary care physicians and community hospitals [6]. Patients with fibrotic ILD often suffer from many comorbidities [11–13], and symptoms of ILD including dyspnoea, cough and fatigue may be attributed to more common diseases such as respiratory infections, COPD, asthma and cardiac disease. This may explain some of the delay in the diagnosis of ILD due to overshadowing of symptoms and result in numerous healthcare contacts and examinations in the diagnostic period.

Many studies have focused on healthcare utilisation in patients with ILD after the diagnosis, but only a few studies have focused on the period preceding the diagnosis. Two studies have shown an increase in healthcare utilisation in the year preceding the diagnosis, and two other studies have shown an increase in dyspnoea and cough consultations in primary care during 2–5 years before the diagnosis of IPF [9, 14–17]. However, these studies have not investigated the use of spirometry or C-reactive protein (CRP) as an indicator of general diagnostic activity, and radiological examinations and comorbidities have only been studied sparsely in the pre-diagnostic period. Thus, a comprehensive assessment of healthcare utilisation in primary care and hospitals during the years preceding the diagnosis of ILD is relevant. Description of specific patterns of healthcare utilisation in the pre-diagnostic period can help identify areas of intervention to achieve an earlier diagnosis of ILD, ensure timely treatment and thus potentially improve the prognosis of patients with ILD.

We hypothesised that healthcare utilisation was higher in patients with ILD than in matched references for several years preceding the diagnosis. The aim of this study was to investigate healthcare utilisation in primary and secondary care in patients with ILD compared to matched references during the 10 years prior to the ILD diagnosis using nationwide Danish registers.

Methods

Study population and setting

We conducted a nationwide matched cohort study based on linked data from national registers [18, 19]. The source population included all adults in Denmark (aged ≥18 years) listed with a general practitioner (GP) (99% of the population) and residing in Denmark for the past 10 years. Denmark has a universal healthcare system with free access to primary and secondary care. The general practitioner acts as a gatekeeper to secondary care. Private practice specialists in pulmonary medicine exist, but ILD is diagnosed and managed in public hospitals. The study population was patients with a first-time diagnosis of ILD between 1 January 2011 and 31 December 2019. A wash-out period from 1 January 1988 to 1 January 2011 was applied to exclude prevalent cases from the analysis. Cases were matched on their first diagnosis date (index date) with 10 references without an ILD diagnosis on age, sex and listing with the same GP. If matching could not be achieved on GP listing, the required references were assigned randomly from the source populations. A reference could be sampled only once per case but could act as reference for several cases and was allowed as a case later in the study period.

Data sources

Individual level data were retrieved from registers on demographic and socioeconomic factors; age, sex, cohabitation status, educational attainment and labour market attachment. This was linked with data on GP contacts and services, and outpatient clinic visits, radiological procedures, hospital admissions, and ICD-10 diagnoses from public and private hospitals [18, 20–22]. The Danish Multimorbidity Index algorithm provided information on 39 long-term physical and mental conditions from hospital diagnoses and redeemed prescription [23]. The Patient List Database created the link between patients and their GP listing.

Exposure

Cases were defined with ILD based on an ICD-10 diagnosis (ICD-10:DJ84X or DJ67X). For specific subgroup analyses, IPF was defined as either a diagnosis (ICD-10:DJ84.1A) or first hospital treatment with nintedanib (procedure BWHA449, nintedanib for non-IPF fibrotic ILD was first approved in February 2022) or pirfenidone (procedure BWHB85) and an age of ≥50 years. HP was defined by diagnosis (ICD-10:DJ67X). Exclusion from the subgroup analysis occurred if a diagnosis incompatible with IPF or HP occurred after the index date (supplementary Table E1).

Similar approaches have been used in previous studies based on Danish registers, as patients with ILD are only diagnosed and treated in selected, highly specialised centres in Denmark [24, 25].

Outcomes

The outcomes of interest included the number of hospital contacts (any contact, hospital contact due to respiratory symptoms (planned and unplanned), or unplanned hospital contacts (all discharge diagnoses)), hospital radiological procedures (chest radiographs (CXR) or chest computed tomography (CT)), standard GP consultations, out-of-hours GP consultations and GP diagnostic procedures (spirometry or point-of-care CRP measurement) (coding definitions are given in supplementary Table E2).

Statistical analysis

We estimated the mean number of outcomes per year for up to 10 years prior to the index date with 95% confidence intervals (CI). Incidence rate ratios (IRRs) with robust 95% CIs, comparing cases with references for each outcome, were estimated using multivariate negative binomial regression models to account for overdispersion, adjusted for age group, sex, cohabitation status, educational attainment, labour market attachment and 39 comorbidities as indicator variables (levels in table 1). Chronic pulmonary disease from the multimorbidity index was modified and included hospital-based diagnoses of COPD and asthma. Stratification was made on the number of concurrent comorbidities. A pre-planned subgroup analysis assessed patients with IPF and HP separately. Sensitivity analyses included adjustment for pulmonary nodule control or lung cancer schemes to account for higher radiological activity. Analyses were performed in Stata version 18.

Results

We identified 13 161 incident cases with a diagnosis of ILD. Among these, 1419 (11%) had a diagnosis of IPF, and 735 (6%) had a diagnosis of HP. Cases were matched with 131 620 references. The mean±sd age at ILD diagnosis was 68.4±13.9 years with a small predominance of males; cases had more comorbidities than references, but similar levels of education and labor market attachment (table 1 and supplementary Table E3). In the IPF cohort, there was a majority of males and cases had a mean±sd age of 73.0±8.0 years, whereas the HP cohort had an equal sex distribution and a mean±sd age of 60.0±15.5 years (supplementary Table E4).

Overall, IRRs adjusted for socioeconomic status and comorbidities (aIRRs) were elevated for all outcomes from 10 years preceding the ILD diagnosis, steadily increasing and sharply peaking in the year prior to diagnosis. This held true for primary and secondary care outcomes (figures 1 and 2). The only notable exceptions to this pattern were CRP measurements and spirometries, where aIRRs mainly were higher from 8 years before diagnosis, as well as unplanned hospital contact and out-of-hours GP contacts with higher aIRRs observed 4–6 years before diagnosis (supplementary Figure E1).

The same pattern was seen when stratifying analysis for comorbidities (figures 3 and 4). The aIRRs were generally higher for patients with no comorbidities, except for hospital contacts due to respiratory symptoms.

The aIRRs for chest CT scans remained almost unchanged after adjustment for pulmonary nodule control or lung cancer schemes (supplementary Figure E2).

In subgroup analyses of patients with IPF and HP, a similar pattern was observed for GP consultations, spirometries, CRP measurements, CXR and chest CT scans up to 9 years before diagnosis (supplementary Figures E3–E6).

Discussion

Summary of main findings

Our study shows that patients with ILD have more contacts with both the primary and secondary healthcare systems compared to references as early as 10 years before diagnosis. This was evident despite matching with references on age, sex and general practice listing as well as adjustment for comorbidities and socioeconomic factors, all of which could confound the association between diagnosis and healthcare utilisation.

Comparison with existing literature

To the best of our knowledge, this is the first study to assess healthcare utilisation 10 years preceding an ILD diagnosis using nationwide registers, references matched on age, sex and primary care listing, as well as adjustment for comorbidities and socioeconomic factors. No previous study has examined the rate of CRP measurements and spirometry in primary care in the years preceding the ILD diagnosis. Also, the rate of CXR has not been compared to matched references in patients with ILD prior to diagnosis.

Only a few studies have investigated healthcare utilisation prior to the ILD diagnosis. Previous studies using primary care databases to identify patients with pulmonary fibrosis or IPF reported increasing rates of primary care consultations due to respiratory symptoms 2–5 years before the diagnosis [9, 15–17]. However, their study period was limited to 5–10 years before diagnosis and only one study used matched references. In contrast, in our study, patients with ILD had higher relative rates of primary care contacts for 7–10 years before diagnosis compared to matched references and after adjustment for confounders.

Even though our data does not contain reasons for primary care contacts, the increased rates of spirometries, CRP measurements, CXR and chest CT scans may indicate increased respiratory symptoms. Hoyer et al. [6] reported misdiagnosis of respiratory symptoms in 41% of patients before the diagnosis of IPF and an increased diagnostic delay in patients treated with inhalation therapy irrespective of an obstructive lung disease. Jones et al. [9] found an increase in CXR as well as antibiotic and acute steroid prescriptions 2–3 years prior to diagnosis. Higher use of antibiotics and corticosteroids before an IPF diagnosis was also found by Davidsen et al. [24]. Herberts et al. [26] also reported respiratory infections as common diagnoses preceding an IPF diagnosis. Higher healthcare utilisation prior to an IPF diagnosis was found by Farrand et al. [17] in the year preceding the diagnosis but also higher rates of chest CT and spirometries for up to 5 years before the diagnosis. Likewise, we found higher relative rates of spirometries, CRP measurements, and CXR up to 10 years before diagnosis with a peak of all outcomes in the last year before diagnosis. Spirometries and CXR indicate higher rates of respiratory symptoms being investigated many years before the ILD diagnosis, whereas CRP measurements indicate a higher general diagnostic activity, including suspicion of respiratory infections causing the symptoms. The peak of all visits and examinations during the last year before the diagnosis is probably a result of increasing symptom severity causing an accelerated diagnostic work up leading to the final ILD diagnosis. This increase in healthcare utilisation during the last year before diagnosis has been reported in previous studies [7, 9].

Interestingly, Jones et al. [9] found no increase in respiratory hospital contacts until 2 years prior to diagnosis. This differs from our results showing an almost three times higher rate of respiratory hospital contacts 10 years before diagnosis. Likewise, Herberts et al. [26] found that many patients had multiple respiratory hospitalisations in the years preceding their IPF diagnosis. Combined with more spirometries, CXR and chest CT scans, there is a potential diagnostic delay in both primary and secondary care. As ILDs are rare and share many risk factors with other more common respiratory diseases such as COPD and other causes of respiratory symptoms such as cardiac diseases, these conditions are often examined or even treated first, apparently both in primary and secondary care [8]. The high rates of CXR and especially chest CT scans for 10 years prior to diagnosis could indicate a lack of knowledge or attention to early radiological findings in ILD among radiologists.

Strengths and limitations

A major strength of our study was the prospectively collected, high-validity, national data over a long period of time on both diagnoses and healthcare utilisation. This ensured strong confounder control and virtually no loss to follow-up. The fact that we were able to match references on GP listing minimised the possible confounding due to GP treatment preferences and referral patterns, geographical location, and the healthcare resources available locally in both practices and local hospitals. In total, this increases the generalisability and robustness of the results. Residual confounding may be present, and observational studies should always be interpreted with caution.

Our study did not have access to smoking habits. This could potentially confound our results, as many patients with ILD are active or former smokers. On the other hand, increased healthcare utilisation due to smoking would probably be caused by more comorbidities, which we have adjusted for in our analyses, thus reducing the influence of smoking on the results. Hospital-based diagnoses were used to identify patients with asthma and COPD. However, most of these patients are diagnosed and followed in primary care, and primary care diagnosis codes are not available in the registers. Inhalation therapy medications could be used as a proxy for these diseases, but many patients with ILD are often misdiagnosed with obstructive lung disease. Our approach may overestimate IRRs a little, but inclusion of inhalation therapy would probably underestimate the IRRs to a higher degree due to a high occurrence of misdiagnosis of obstructive lung disease among patients with ILD [8, 27]. Also, CRP measurements are not specific for ILD and should as such be interpreted with caution as a marker of ILD diagnostics but are still relevant as an indicator of general diagnostic activity.

Implications

The increased healthcare utilisation shown in our study could be an indicator of patients with ILD presenting with significant respiratory symptoms many years before their final diagnosis and even longer than the previously reported delays [6–9]. Thus, there is a potential for earlier identification of these patients. Early diagnosis is key to early treatment and specialised care. Our study indicates that increased attention to ILD in both primary and secondary care is needed. In early phases of ILD, spirometry and CXR may be normal, and primary care physicians should refer patients with unexplained respiratory symptoms for further diagnostic work up at hospitals. Likewise, increased attention to and knowledge on ILD symptoms in admitted patients and ILD patterns on CXR and especially chest CT scans would decrease the diagnostic delay indicated in our study. Depending on the type of ILD, early treatment can result in stabilisation or slowing down disease progression of fibrotic ILD and might also lead to reversal of symptoms and remission of inflammatory ILDs.

Attention to ILD should be raised in patients with recurrent healthcare contacts for respiratory symptoms, both with and without comorbidities. Our results showed the highest IRRs of contacts and examinations in the subgroup of patients with no comorbidities. Still, even though the relative difference is large, these results should be interpreted based on the low absolute number of contacts and examinations in both cases and references. The lower IRRs in patients with comorbidities could be caused by symptoms of ILD being mistaken for or overshadowed by other conditions also requiring attention, healthcare contacts and examinations. Still, the IRRs in patients with comorbidities were also higher for many years before diagnosis in these patients, and ILD could potentially be diagnosed earlier.

Conclusion

In conclusion, this nationwide study shows an increase in healthcare utilisation in both primary and secondary care as well as respiratory-related examinations in patients with ILD up to 10 years prior to diagnosis with a peak in the last year before diagnosis. This could indicate a diagnostic delay in both primary and secondary care even longer than previously reported.