Discussion
Using primary care and secondary care-derived cohorts, we showed patient pathways for four chronic respiratory diseases in England. We demonstrated that there were commonalities, but also some notable disease-specific features.
Asthma (adults and children)
Patient pathways of adults and children diagnosed with asthma were broadly similar. Prior to diagnosis, healthcare utilisation was low (GP consultations, hospitalisations, A&E visits and OPs), especially in children. Only half of patients reported respiratory symptoms, and less than one-quarter of people with asthma had at least one NICE-prescribed diagnostic test. Despite low healthcare utilisation and diagnostic tests, relative to the other chronic respiratory diseases studied, time to diagnosis (from first presenting with symptoms) was short (between 6 and 8 months) and treatment initiation was rapid (whether in response to symptoms or in response to diagnosis). Little research has been done on asthma diagnostic pathways due to the vacillatory nature of the disease and symptom inconsistency,20 which can make diagnosis difficult. Current literature around asthma diagnosis currently suggests that both overdiagnosis (30%–40%) and underdiagnosis (19%–73%) of the condition is highly prevalent. However, we showed that treatment was instantaneous, and patients are most often treated before receiving a formal diagnosis.
Chronic obstructive pulmonary disease
Among people diagnosed with chronic obstructive pulmonary disease (COPD), only half reported respiratory symptoms, and the rate of GP consultations prior to diagnosis was low; but the majority underwent at least one NICE-prescribed diagnostic test in the lead-up to diagnosis. Despite the high proportion of diagnostic tests, time from first report of respiratory symptoms to diagnosis was high (between approximately 8 and 10 months), and 1 in 10 patients with COPD had a COPD hospitalisation prior to first primary care recorded diagnosis. Similar delays in the diagnosis of COPD have been reported in the literature; for example, in a small, cross-sectional, localised Chinese study, time to COPD diagnosis was reported to be 230 days (7.7 months),21 which aligns well with our findings. Our analysis suggests that diagnostic delays in COPD are dominated by the length of time taken to reach a diagnosis following the first diagnostic test rather than the time taken to refer patients for testing following first presentation of symptoms. Despite diagnostic delays, treatment initiation was rapid (less than 1 month), both in response to symptoms (prediagnosis treatment) or in response to diagnosis (postdiagnosis treatment).
Interstitial lung disease
The interstitial lung disease (ILD) patient pathway is characterised by low frequency of primary care encounters but high secondary care utilisation (1 in 10 was hospitalised prior to diagnosis and over 80% had numerous OPs) and delays in timelines. Furthermore, interpreting ILD findings presented some challenges. For example, an ILD diagnosis requires confirmation from a specialist respiratory clinician, and this information may be embedded within the ‘free-text’ section of the primary care record and/or within outpatient data, for which we do not have the granularity to know whether a patient does or does not have a specialist diagnosis. Although half of people diagnosed with ILD report symptoms to their GP prior to diagnosis, very few had a recorded CT scan (~0.2%) or evidence of a referral for a CT (~3.0%), as per NICE guidelines. However, it is unlikely that these ILD diagnostic test data reflect reality. First, 0.2%–3.0% may be large underestimate due to poor coding practices: CT scan referrals may be noted in the ‘free-text’ section, which we do not have access to. It is also likely that people with ILD underwent other diagnostic tests not analysed as part of the ILD pathway (such as chest X-rays or spirometry). The ILD patient pathway has delays at every interval: average time from first respiratory symptom report to diagnosis is approximately 10 months while time from symptoms to diagnostic tests is approximately 6 months. However, time from diagnostic tests to diagnosis (although in the minority of patients) is comparatively shorter (approximately 2 months), indicating that appropriate diagnostic tests do expedite a diagnosis. Furthermore, most people did not have evidence of supplemental oxygen or oral prednisolone treatments for ILD during follow-up. Among those who were treated, the time to treatment was up to 4 months. More recently, antifibrotic treatment has been incorporated into ILD treatment, however, as these medications are prescribed in secondary care, we were unable to include them. Although the percentage of people treated aligns well with previous observational research,22 the time frames for treatment are slightly shorter in our study (~4 months in our study, as opposed to ~9 months), possibly due to our using a 2-year follow-up prior to and after diagnosis, as opposed to a 5-year follow-up either side of a diagnosis. Indeed, ILD diagnostic delays of 7 months to over 2 years exist in smaller cohorts in the USA23 and Denmark.24 Additionally, a large USA-based study of over 44 000 idiopathic pulmonary fibrosis (IPF) patients showed diagnostic delays of 2.7 years between a final IPF diagnosis and a previous alternative respiratory diagnosis (including acute respiratory infections). The authors also expressed that 5 years of consecutive data were required to better understand diagnostic timelines.22 Therefore, it is possible that future studies should consider expanding the lead-up time for less commonly diagnosed respiratory diseases (such as ILD) when examining diagnosis and treatment timelines. One further recommendation for future ILD pathway studies would be to include antifibrotic therapy in treatment data, as has been done most recently in other pathways research.22
Bronchiectasis
The bronchiectasis patient pathway was characterised by missed opportunities to diagnose and treat. Over two-thirds of patients reported symptoms to their GP and over three-quarters were sent for at least one diagnostic test; moreover, healthcare utilisation (GP consultations and OPs) was high. Nevertheless, it took approximately 13 months from first symptom presentation to bronchiectasis diagnosis. Additionally, up to half of bronchiectasis patients were not prescribed NICE-recommended therapies by their GP during study follow-up; those who were treated faced delays in treatment initiation postdiagnosis of approximately 3 months. Diagnostic delays of over 2 years have been reported in previous research.25 Improvement of disease recognition as well as identification of at-risk groups for diagnostic delays (such as women25) could reduce the likelihood of missed opportunities to diagnose and treat bronchiectasis. While beyond the scope of this particular study, one element that we did not look into was bronchiectasis aetiology across different parts of the patient pathway, which we would recommend as an area of future research for bronchiectasis.
Temporal trends
We conducted our research over multiple time frames and found general stability over time, irrespective of disease, with a few exceptions. Although there were aspects of the pathway that did change over time, these were most often similar between diseases, suggesting that there may be general issues within the health system (at least regarding chronic respiratory disease). Furthermore, inclusion of a pandemic era patient pathway showed some changes in the 2020/2021 cohorts: although the number of new diagnoses increased over time in all diseases between 2008/2009 (likely due to a combination of population growth, an ageing population, better disease recognition and improved EHRs coding practices in primary care), the number of new diagnoses decreased again across all diseases in the very short time of 2018/2019–2020/2021, particularly for COPD. This sudden decrease is likely due to the COVID-19 pandemic and a lag in diagnoses yet to be made, particularly as our first two time frames as well as other work on COPD temporal prevalence indicate an increasing COPD prevalence.26 In addition, the proportion of people diagnosed with COPD who underwent diagnostic tests decreased—specifically, spirometry decreased done due being a high-risk procedure for transmission—while time from symptoms to diagnosis, as well as time from first diagnostic test to diagnosis, increased. The number of people diagnosed with asthma who underwent diagnostic tests also decreased to almost half in 2020/2021. Previous research has found that the pandemic saw fewer COPD exacerbation hospitalisations,27 primary care asthma exacerbations28 and bronchiectasis exacerbations.29 One proposed reason is that the pandemic lockdown resulted in COPD and asthma patients having fewer respiratory tract infections and consequential exacerbations, resulting in fewer people feeling ‘ill enough’ to seek care. However, it is equally possible that opportunities to diagnose may have been missed during the pandemic, possibly due patients not being able to see a GP or fears in the early weeks of the pandemic about whether COVID-19 was a respiratory disease. Future research should soon assess the number of new postpandemic asthma and COPD diagnoses; if this number was to increase quite dramatically, this would indicate a sizeable diagnostic lag.
Regional and socioeconomic status trends
Previous research in England has found differences in relative risk of death from chronic lung diseases in England.9 We, however, did not find stark discrepancies in patient pathways by region, within any disease or across time. One minor exception to our findings was a slight fluctuation in secondary care. Regional disparities may only appear as disease progresses in the lead-up to death, rather than in the lead-up to first treatment; or regional differences could be found in secondary care only and not primary care. Regarding socioeconomic status, our findings were mostly stable, except for a small gradient in percentage change in the proportion of people who visited secondary care: hospitalisation and A&E visits increased slightly as deprivation increased. Welsh Secure Anonymised Information Linkage (SAIL) data, which investigated healthcare utilisation and treatment by socioeconomic status, modelled (age-adjusted and sex-adjusted) socioeconomic status and emergency visits and suggested an inverse relationship between socioeconomic status and emergency asthma care.8 Although we did not model our data, our small gradient in percentage change supports these models.
We found a very large range in times between events in the patient pathway. For example, although (on average across the three periods), COPD diagnosis time was 8.6 months, the IQR was between 1.8 and 17.2 months. This trend is echoed across time and diseases, and between most event time frames. Although stratified by time, socioeconomic status and geographical region, the large range in event time frames remained. There may, therefore, be other factors driving differences in the times between key points in the patient pathways, leading to diagnostic delays, outside the scope of this study. It may even be appropriate for there to be a large range in the data: an ILD patient pathway for a multimorbid older patient with concurrent COPD may look different than the asthma pathway of an otherwise healthy 10-year-old child. However, there may be other aspects of the pathway that we have not detected, particularly as respiratory symptoms are often non-specific and could be attributable to pathway ‘noise’ from investigating other diseases (such as breathlessness with heart failure). Additionally, previous observational research demonstrated differences in hospitalisation by age and sex.3 One suggestion for future research, therefore, would be to explore events and time frames in patient pathways by age, sex and ethnicity.
Limitations
Regarding healthcare utilisation, we did not have data for 2020/2021 for OPs or A&E visits. Knowing what occurred within secondary care during COVID-19 is crucial for understanding the potential lag-related burden that could occur in the health system in England. A second limitation was the data granularity for A&E and outpatient data: A&E data are only as granular as ‘respiratory related’, and the outpatient data could be any OP for any condition. We suspect that there may, therefore, be noise from comorbidity pathways among outpatient visit results. Our results also relied on GP coded events (symptoms, diagnostic tests or diagnoses), and it is possible that, during the lead-up to a diagnosis, GPs may be writing notes in the ‘free-text’ section of consultation records, which we do not have access to, and we may, therefore, have some missing information along the pathway. Furthermore, CPRD data are not originally recorded for research purposes, however, we used respiratory disease defining methodology that has been validated by previous studies.30 Additionally, one limitation to treatment findings is that a patient may have become ill with an acute respiratory infection and was prescribed an inhaler or oral corticosteroid, which would have appeared as a treatment record in our analysis. Finally, there is the limitation of misdiagnosis. We used the first diagnosis of a specific chronic respiratory disease in a patient’s medical record as our date of diagnosis. Although calculated percentages of misdiagnosis to interpret results contextually (online supplemental material), it is possible that patients do go on to be diagnosed with another disease. One avenue for future research is to follow patients up to assess how the pathway changes as different disease possibilities are recorded, whether respiratory or otherwise. Except for acute respiratory infections in primary care for patients with COPD, we did not investigate other acute respiratory infections in primary or secondary care, in the lead-up to diagnosis. Other research demonstrated acute respiratory diseases may be prevalent in primary and secondary care in the lead-up to an IPF diagnosis22; it may be useful to contextualise acute respiratory diseases trends in the lead-up to chronic disease diagnosis in future research. Despite limitations, our research aimed to take a broad view of chronic respiratory disease pathways to lay groundwork for repeatable methods, as well as for more detailed byproduct research to be conducted on individual sections of the pathways.
