Predictors of accuracy of diagnosis of chronic obstructive pulmonary disease in general practice

Chronic obstructive pulmonary disease (COPD) is a major cause of mortality, morbidity and health service use, including hospitalisation. It is the seventh leading cause of burden of disease in the Australian population and the fourth and sixth leading causes of death for men and women, respectively.1 The prevalence of COPD in the population attending general practitioners in Australia is estimated to be 2.6%.2 In patients aged 65 years or older, COPD is managed at a rate of 21 per 100 encounters with patients in this age group,3 making it a common problem in older people.

Both international4 and Australian5 clinical practice guidelines for COPD state that its diagnosis rests on the demonstration of airflow obstruction that is not fully reversible. Spirometry is therefore essential for accurate diagnosis of COPD. Although there is evidence that office spirometry improves early detection of COPD in general practice, and despite the majority of general practices in Australia owning a spirometer,6 it is infrequently used in primary care in Australia and internationally, and the diagnosis is usually made on clinical grounds.7-9 Several studies in primary care settings have demonstrated inaccuracy of COPD clinical diagnosis compared with spirometric diagnosis.10-13

We know little about the practitioner, practice and patient factors that influence the accuracy of the diagnosis of COPD in general practice in Australia or comparable countries. Understanding these factors could be of use in developing and targeting interventions aimed at improving COPD diagnosis. As part of the baseline assessment for an intervention study involving patients considered by their GPs to have COPD, we examined the accuracy of the GPs’ diagnoses in relation to the gold standard of spirometric diagnosis. We also examined practitioner, practice and patient factors that predicted agreement between the diagnostic label and the finding of post-bronchodilator airflow obstruction.

This study was done in the context of a randomised controlled trial of a nurse-based intervention to improve COPD management. The data are from the cross-sectional baseline assessment conducted for this trial between November 2006 and April 2008. Ethics approval was granted by the University of New South Wales Human Research Ethics Committee.

Fifty-six GPs from 44 general practices in Sydney participated in the study. The mean age of GPs was 52.3 years, and 47 were male. Of the 56 GPs, 51 were vocationally registered (recognised as having specialty qualification in general practice), 30 were graduates of an Australian or New Zealand university, and they had a mean of 15.6 years’ (range, 1–35) experience in practice. Forty-seven of the GPs worked in accredited practices, 40 had a spirometer in the practice, 25 had a practice nurse and 45 had a practice manager.

The record search and review of the list generated by the GPs identified 1144 patients who were eligible and invited to participate. Of these, 445 patients (38.9%) were recruited and provided baseline data. The mean age of participants was 65.0 years, and 217 (48.8%) were male. The mean number of comorbidities was 3.9 (range, 0–14). More than half (55.1%; 244/443) were born in Australia, and 30.5% (134/440) were current smokers.

Of the 445 patients, 257 (57.8%) had post-bronchodilator spirometry showing COPD with or without asthma, 16 (3.6%) had asthma only, 82 (18.4%) had normal spirometry, and 90 (20.2%) had other spirometric diagnoses such as restriction. Of the patients with COPD, 33 (12.8%) had GOLD Stage 1, 118 (45.9%) had Stage 2, 79 (30.7%) had Stage 3 and 27 (10.5%) had Stage 4, by post-bronchodilator spirometry. The characteristics of all participating patients are shown in Box 1 by spirometric diagnosis. Patients who had COPD on spirometry tended to be older, male and have fewer comorbidities. There were significant differences between those with and those without COPD in terms of treatment received, disease-related quality of life and overall quality of life.

The associations between practitioner, practice and patient factors and having airflow obstruction on spirometry are shown in Box 2. There were three statistically significant associations. Having a practice manager in the general practice was associated with a higher chance of misdiagnosis. Older patients were more likely to be correctly diagnosed, and those with higher numbers of comorbidities were less likely to be correctly diagnosed.

We found that there are substantial rates of misdiagnosis of COPD in primary care, with less than 60% of patients having the clinical diagnosis confirmed on spirometry. This has important implications for management, including using medicines to treat COPD in patients who do not have the condition. In this study, 57 of the 188 patients who did not have COPD on spirometry were being treated with tiotropium. This exposes patients to unnecessary risk of adverse effects of medications and has cost implications for the patients and the health system. It also means that the true cause of a patient’s respiratory symptoms may not be correctly identified and treated.

Our findings confirm those from previous studies in primary care. In a study in Greece of 319 patients aged over 40 years diagnosed as having COPD, 50% had a post-bronchodilator FEV1/FVC ratio < 0.7.10 A similar rate was found in a study in primary care practices in Scotland and the United States,11 where 48% of patients with a prior diagnosis of chronic bronchitis or emphysema were confirmed as having COPD on post-bronchodilator spirometry. In a study of 580 patients on a general practice-based COPD register in England,12 79% had obstruction on pre-bronchodilator spirometry. Reversibility testing was not done for all patients in that study, so this may be an overestimate of the prevalence of COPD. In a Canadian study, where spirometry was more often used in clinical practice, the rate of misclassification was lower (12% of patients aged over 40 years).13

Our study adds to previous research in that we looked for characteristics of practitioners, their practices and patients that are associated with diagnostic accuracy. The striking finding is that presence of a spirometer in the practice was not associated with diagnostic accuracy. It may be that spirometers are not being used because of time limitations or lack of adequate financial reward, or that there are problems with the quality of spirometry performed in general practice. Qualitative research in Tasmania found that GPs preferred to diagnose COPD on clinical grounds and there were both organisational and technical barriers to greater use of spirometry.7 A study in the United Kingdom found that, despite incentives to perform spirometry, lack of adequate training in its use and interpretation meant that diagnosis was more likely to be made on clinical grounds.17

The only practice factor significantly (and negatively) associated with diagnostic accuracy was the presence of a practice manager. We are unable to explain this, as practice managers have no role in the diagnostic process, and believe it must be a chance finding. Although practice managers were more likely to be working in larger practices, which are in turn likely to have more doctors in training, this is unlikely to explain the association as there was no difference found in diagnostic accuracy between solo and group practices.

The other important finding was that presence of comorbidities was significantly associated with diagnostic inaccuracy. As accurate spirometry should be possible despite comorbidities, this has implications for practice as it suggests that patients with comorbidities should be prioritised for thorough investigation. Older age was associated with higher diagnostic accuracy, presumably due to increased prevalence of COPD with age.

A recent study found that the likelihood of misclassification increased with overweight or obesity and self-reported allergic rhinitis or hayfever,18 but did not report a relationship with the number of comorbidities. Consistent with our study, it found decreased likelihood of misclassification with increasing age.18 The investigators did not examine for practice and practitioner factors associated with misclassification.

A limitation of our study is that we did not have information on how the diagnosis of COPD was made in the patients. It could be that patients had been diagnosed on the basis of clinical features and had not had spirometry, particularly as there are barriers to the use of spirometry in primary care. Another possibility is that the spirometry done by the treating clinicians or by the study team was incorrect or involved poor patient technique. We are confident in the quality of the spirometry done for the study as the project officers had substantial training and expert oversight. The ATS criteria that require three acceptable expirations were achieved for the patients in our study, so poor technique was not a factor.

While it was possible to withhold treatment with bronchodilators before spirometry, it was not possible to withdraw inhaled corticosteroids, and some patients treated with these medicines may have met the criteria for COPD if they had not been using this treatment. Similarly, tiotropium was not withheld for 24 hours or more in every patient, so it is possible that a small number of patients with mild obstruction when not being treated may have had normal spirometry.

Until an alternative to spirometry is found, there is a need to develop and evaluate strategies to encourage its use and improve its quality in the diagnostic process of patients seen in general practice. A policy issue is the low level of remuneration received for performing spirometry (the current Medicare rebate is $16.85 for pre- and post-bronchodilator spirometry). Another issue is the availability of appropriately trained workforce to perform spirometry in primary care. Up-skilling of practice nurses is an option, as there is evidence that nurses can take on this role,19-21 but training and quality assurance are needed, along with adequate funding.9

An approach arising from this study could be to prioritise patients with comorbidities for thorough diagnostic assessment, including spirometry, as this group may be more likely to be incorrectly diagnosed. As demonstrated in research in Tasmania,22 patients are presenting to their GPs with respiratory symptoms, but a substantial number do not have either COPD or asthma. Thus, there is also a need to better understand the nature and causes of respiratory symptoms in patients in general practice.