Between-practice variation in chronic obstructive pulmonary disease diagnosis guideline compliance: an observational study

Summary

Overall, approximately two in three patients received pre-diagnosis spirometry. The proportion improved in cohort 2 on cohort 1 but fell back for the early COVID-19-era cohort. In contrast, chest X-ray, FBC, and BMI all improved after cohort 1 and were maintained for the COVID-19-era cohort. Almost all patients received one of these recommended tests and measurements, but only a small minority received all four. We found much non-random variation in spirometry use between practices.

Strengths and limitations

CPRD is much used in research. Its main strengths include relatively easy access, large sample size, broad representativeness of the UK population, and the almost universal coverage by primary care in the UK. Its longevity means that many groups have experience with it and have conducted validation studies on the accuracy of disease and date coding. The QOF incentivises the recording of chronic diseases, including COPD, GP actions that are important for diagnosis, and follow-up such as spirometry and blood tests, and key risk factors such as BMI and smoking.

Our COVID-19-era cohort was fairly small. As with the other two cohorts, it was defined according to the date of diagnosis (first recording) of COPD. This means that many patients would have had their tests before the pandemic onset, so it does not fully reflect the situation during the pandemic. Owing to subsequent COVID-19 disruptions of the health service, there may have been more misdiagnosis for this cohort than for the earlier two.

While CPRD codes for COPD may be mostly accurate, misdiagnosis for the condition is a well-known problem. A 2018 review suggested that approximately 70% of COPD worldwide may be underdiagnosed, whereas between 30% and 60% of patients with a previous physician diagnosis of COPD may not actually have the disease and have therefore been overdiagnosed.⁹ Requesting the NICE-recommended tests is one key element to diagnosis. However, we did not investigate the GPs’ ability to correctly interpret those tests. Previously, our group found that the GPs correctly spotted obstruction in only three in four cases, less often if the patient had asthma.¹⁰

Comparison with existing literature

The most direct comparison with our study is the national audit, although it relates to practices in Wales rather than in England and did not assess between-practice variation, as in our study. The most recent report⁶ is based on patients coded with the disease between April 2020 and July 2021, which subsumes the timeframe for our cohort 3 and extends it further into the COVID-19 era. Its key findings for COPD relevant here are that just 1.9% had post-bronchodilator spirometry in the previous 2 years, and, overall, no improvement since the previous audit. We were unable to assess specifically post-bronchodilator spirometry but also found low levels of this test overall. Spirometry use in other countries has also been found to be low. A US study of 101 patients from two community clinics found a low rate (21%) of lung function testing.¹¹ In a German primary care network, record-keeping was not standardised and spirometry in the year before diagnosis was just 29%.¹² In an Italian study of 300 patients sent for spirometry, of whom only 75 had COPD, the authors made a link between the low use of spirometry to help diagnose asthma and COPD and the low proportion of doctor-diagnosed COPD with their concordant spirometry findings.¹³

QOF has facilitated studies on the association between the quality and organisation of primary care and outcomes for long-term conditions. For COPD, these studies cover aspects of quality of care other than the four NICE-recommended tests that we investigated. A London study using data from 2006–2010 used these QOF indicators for COPD: proportion of patients who had received the influenza vaccination in the preceding 12 months; confirmation of diagnosis using spirometry; recording of FEV1 in the previous 15 months; recording of good inhaler technique; and patients who had had a review by a health professional in the preceding 15 months. Emergency hospital admission rates were found to be related to COPD prevalence and small-area deprivation but not to GP QOF performance, GP supply, or nurse supply.¹⁴ Another QOF study looked at the influence of primary care nursing levels on QOF performance in chronic disease management, including that for COPD. This made use not only of the usual clinical indicators but also those describing organisational performance. The many organisational variables were reduced by factor analysis. Although the relation with staffing levels was statistically significant, the strongest predictors of quality of care were found to be the organisational factors of clinical recording, education and training, and use of patient experience surveys.¹⁵ An interaction term suggested that the adverse effects on quality of the lowest levels of nurse staffing were mitigated by good organisational quality.

The between-practice variation in NICE guideline compliance for COPD is in line with our previous CPRD analysis for patients with HF in which we focused on following the steps of serum natriuretic peptide test and/or echocardiogram, and referral to a specialist; a timeframe of 6 months since presentation with a HF symptom was used. In that study, 13% of practices were identified as 3 SD funnel plot outliers, with an ICC of 8.6% (4.9% following adjustment for patient factors).¹⁶

A Danish study on 2008 data for ‘patients over 18 years receiving first-time prescriptions for medication targeting obstructive lung disease’ found that practices with partnerships were more likely than single-handed, and older GPs less likely, to order spirometry; practices with training status were more likely to order spirometry.¹⁷ Unfortunately, this level of detail on practices is not available within CPRD.

Implications for research and practice

A 2010 King’s Fund review of the evidence of the quality of diagnosis and referral in UK primary care noted the challenges facing GPs such as vague symptoms, overlap in symptoms for serious common conditions, and the modest predictive value of diagnostic tests in primary care. These are complicated by limited consultation times, differences in GP knowledge, skills, and attitudes, access to diagnostics, and the time taken to receive test results.¹⁸ To mitigate these challenges, the 2019 NHS Long Term Plan set out plans for more joined-up and coordinated care, including the creation of primary care networks that would work together to deliver services such as spirometry. Earlier diagnosis would lead to early interventions for smoking cessation, better management, and improved health outcomes.

Access to spirometry is clearly improved when the kit is present in the practice, but ownership of a spirometer was already 60%–80% in the UK by 2005.^19,20 The use of spirometry in primary care stopped during the early COVID-19 waves as spirometry-associated cough can generate aerosol droplets, but it has resumed with published guidance on restarting it with risk mitigation.²¹

The shift towards remote consultations in primary care, which has persisted post-pandemic, may have implications for the investigation and diagnosis of a condition where an important element of assessment is physical examination, although improved access through online consultations may result in earlier investigation. Unfortunately, whether the consultations are remote or face to face is currently poorly coded and not reliably identifiable in routinely collected primary care datasets such as CPRD. In general, it is important that coding is consistent and accurate, as high-quality patient records are the foundation of good clinical care delivery and not just useful for research.

In conclusion, the use of NICE-recommended diagnosis tests, particularly spirometry and chest X-rays, warrants further improvement. The COVID-19 pandemic hit many diagnostic and preventive services hard, and it will be important to monitor these effects; our analysis can serve as baseline data for this monitoring.