This study used the Clinical Practice Research Datalink (CPRD) Aurum database (February 2022 build), a curated set of patient-level electronic health records provided by 1,345 participating GP practices, which collectively deliver primary health care to one fifth (19.9%) of the UK population. CPRD Aurum provides pseudonymised, longitudinal SNOMEDCT-coded information on individuals’ demographic characteristics (e.g., sex, age), lifestyle behaviours (e.g., BMI, smoking status), symptoms and clinical diagnoses, medication and vaccination histories, as well as results of laboratory tests and details of referrals to secondary care32. The CPRD Aurum dataset has been shown to be representative of the UK population in terms of age, sex and ethnicity32. CPRD Aurum data are routinely linked to other health-related data sets, including the Index of Multiple Deprivation (IMD), a post-code level measure of social deprivation. Such linkages are currently limited to 80% of Aurum GP practices in England.

This study employed a retrospective cohort design. Our study population was drawn from a cohort of individuals registered at linkage-eligible CPRD Aurum GP practices who had been diagnosed with IPF on or after the start of our study period (1 January 2010). In order to be eligible for inclusion in our study, patients also had to have received their IPF diagnosis before the end of our study period (31 December 2019), be male or female, aged at least 40 years at the time of their IPF diagnosis, meet CPRD-defined data quality criteria and have at least one year of continuous registration with their current GP prior to their IPF diagnosis (baseline period). For each participant, the start of follow up was defined as the latest of the following dates: date of registration with their GP practice, start of the study period (i.e., 1/1/2010), date of their 40th birthday or date of IPF diagnosis. It concluded on the earliest of: last day of practice registration, the last day of practice data collection, date of death or the end of our study (31/12/2019).

Selection of risk factor variables was based on a combination of a literature review, a priori clinical knowledge and data availability. Included variables thus comprised demographic characteristics [sex, age at IPF diagnosis (40–59, 60–69, 70–79, 80+ years), socioeconomic status (IMD quintiles), region (nine), ethnicity (white, Asian, Black, mixed, other)], lifestyle characteristics [BMI (normal, underweight, overweight, obese), smoking status (current, ex, never)] and presence or otherwise of selected comorbid diseases. Data on patients’ demographic and lifestyle characteristics were obtained from primary care data; data on socioeconomic status was extracted from linked IMD records. BMI and smoking status were determined at the time of IPF diagnosis (using the closest recorded values to this date). Where GP-recorded values were missing, BMI values were calculated from height and weight data closest to the IPF diagnosis date. Comorbid chronic obstructive pulmonary disease (COPD), asthma (adult-onset), lung cancer, gastro-oesophageal reflux disease (GORD), hiatus hernia, dementia, depression/anxiety, diabetes and cardiovascular diseases (pulmonary arterial hypertension, ischaemic heart disease, heart failure) was determined on the basis of the presence at least one relevant clinical code in a patient’s record at any point prior to their IPF diagnosis.

Our primary outcomes were referral for PR and referral for PC. Patients were considered to have been referred if their primary care record included coded evidence of a referral after their IPF diagnosis and within their follow-up period. Any documented referrals before IPF diagnosis were discounted. The criteria for determining a referral were deliberately broad, accepting any code that signified that an individual had been considered and/or offered a referral to PR or palliative/end-of-life support services or hospice care. In the case of PR referrals, we made no distinction between being assessed/considered for PR, offered a referral and commenced/completed PR. Similarly, for PC, the selected codes encompassed referrals to generic (i.e., non-disease specific) end-of-life care support services and were not limited to those specifically aimed at people with IPF.

All code lists used in this study are available to download from: https://github.com/NHLI-Respiratory-Epi/IPF-non-pharmacological-treatment.

Baseline characteristics were reported for the cohort as a whole and for “referred” and “non-referred” patients (separately for PR and PC services), using frequencies for categorical variables and medians for continuous variables. In addition, we explored geographical and temporal patterns in patient referrals, again separately for PR and PC. Temporal trends were assessed by calculating the proportion of patients with a current IPF diagnosis who were referred in that year, for each year of the study period. Thus, for 2010 we divided the number of referrals made in 2010 by the number of people in our study cohort with an IPF diagnosis in 2010. Given ours was an incident cohort, the 2010 denominator was simply the number of newly-diagnosed patients in 2010; however, for 2011 onwards our denominator comprised the number of newly-diagnosed cases in that year plus any previous years, less those who died or were lost to follow up for some other reason in the preceding year. We estimated temporal trends in both PR and PC referrals stratified by sex. In the case of PR, we compared referral proportions in those with and without a diagnosis of COPD, and in the case of PC we estimated the referral proportion in the subset of people who also had comorbid COPD or lung cancer. We also estimated the proportion of people who according to their primary care records had commenced or completed PR post an IPF diagnosis.

We used Cox regression models to explore the association between patient characteristics and referral to PR and PC, estimating crude and adjusted hazard ratios (HR) for each risk factor listed above. Owing to a high level of missingness in BMI (~30%), our final adjusted model excluded BMI as a covariate. However, a comparison of models with and without BMI (complete-case analysis) revealed that inclusion of BMI had little impact on the HRs for the other covariates. In other sensitivity analyses, we added a cluster option to account for variations in referral decisions between different GP practices and additionally adjusted for year of diagnosis (given evidence that the likelihood of referral increased over the time of our study). We also conducted subgroup analyses whereby we estimated HRs for PR referral separately for people with and without COPD, and HRs for PC referral separately for people with and without lung cancer. We used Schoenfeld residuals to check that our final model did not violate the proportional hazards assumption. Finally, we constructed a multinomial logistic regression model to investigate whether the timing of a referral to PC services was influenced by the presence of a co-diagnosis of either lung cancer or COPD. This analysis was restricted to those who died during the study period, and we categorised referrals which occurred within a month of death as “late” and those which occurred more than 1 year prior to death as “early”.

All analyses were performed using Stata v17 (StataCorp LLC, TX, USA).

CPRD has NHS Health Research Authority (HRA) Research Ethics Committee (REC) approval to allow the collection and release of anonymised primary care data for observational research [NHS HRA REC reference number: 05/MRE04/87]. Each year CPRD obtains Section 251 regulatory support through the HRA Confidentiality Advisory Group (CAG), to enable patient identifiers, without accompanying clinical data, to flow from CPRD contributing GP practices in England to NHS Digital, for the purposes of data linkage [CAG reference number: 21/CAG/0008]. The protocol for this research was approved by CPRD’s Research Data Governance (RDG) Process (protocol number: 23_003015) and the approved protocol is available upon request. Linked pseudonymised data was provided for this study by CPRD. Data is linked by NHS Digital, the statutory trusted third party for linking data, using identifiable data held only by NHS Digital. Select general practices consent to this process at a practice level with individual patients having the right to opt-out.

This study is based in part on data from the Clinical Practice Research Datalink obtained under license from the UK Medicines and Healthcare products Regulatory Agency. The data is provided by patients and collected by the NHS as part of their care and support. The Office for National Statistics (ONS) was the provider of the ONS Data contained within the CPRD Data and maintains a Copyright © [2024]. Linked data were re-used with the permission of The Health & Social Care Information Centre, all rights reserved. The interpretation and conclusions contained in this study are those of the author/s alone.

Further information on research design is available in the Nature Research Reporting Summary linked to this article.