The primary finding of our study was that in patients with a doctors’ diagnosis of COPD, spirometry performed within six months of diagnosis increased from 59% in the first PRAXIS cohort of 2000–2003 to 88% in the second PRAXIS cohort of 2004–2010. Factors associated with having a COPD diagnosis without a diagnostic spirometry were current smoking, low educational level and being managed in primary care. Secondary findings were that the FEV1/FVC or FEV1/VC ratios were incorrectly interpreted regarding COPD diagnosis in 18% of the cases. This finding was more likely in females, patients with concomitant hypertension or diabetes and those who were managed in primary care. In 94% of the patients with a correct COPD diagnosis, and where a follow-up spirometry was available in the records, airway obstruction was persistent over time.

Previously reported results from the first PRAXIS cohort14 are consistent with other studies reporting that diagnostic spirometry was performed in about half to two-thirds of patients with a COPD diagnosis4,5. However, the present study from the second PRAXIS cohort shows that the frequency of performed diagnostic spirometries has clearly increased over time. There was also an increase in the proportion of correctly interpreted spirometries. We find it encouraging that the management of COPD in Sweden has improved and now complies with international and national diagnostic guidelines to a higher degree. The most important independent factor associated with not having performed a diagnostic spirometry was current smoking. We speculate that the clinical diagnosis of COPD in current smokers with respiratory symptoms may seem more obvious to physicians and thus explain the lower degree of confirmation with spirometry. Our finding is in line with a qualitative study by Joo et al. that presented a similar explanation of physicians´ motives towards diagnosis of COPD17. Feng et al. conclude that people with multiple unhealthy lifestyles, including smoking, are less prone to consult primary health care18. Consequently, they would therefore not be referred to secondary care. This could explain the difference in primary and secondary care concerning the proportion of patients that did not undergo a diagnostic spirometry.

Smoking, poverty and low education are important factors associated with a higher burden of disease19. This is in line with our result of low education being associated with not having performed a spirometry. Low socioeconomic status can be a reason why patients refrain from seeking care20. In Sweden, however, healthcare is financed with taxes and is equally available for everyone, thus the financial cost of healthcare cannot explain the lower frequency of spirometry testing in patients with low education. We speculate that our results could in part be due to a different care-seeking behaviour. For instance, smoking is found to be associated with reduced likelihood of care-seeking21. This highlights the importance of using a holistic approach and being aware of health inequalities when managing potential COPD patients, in order for individualised care and effective smoking cessation to be delivered. We believe that this result follows a pattern evident in other studies that have demonstrated associations between higher education and a higher degree of adherence to smoking cessation interventions and greater interest in learning self-management skills22,23.

The proportion of spirometries not consistent with a correct diagnosis of COPD was 18%. Similar misdiagnosis of COPD has been described previously, yet to a larger extent than in our study population7,24. In this particular group, we found a significant change of diagnosis from COPD to asthma over the study period (OR 9.90, 95% CI 3.09–31.78), indicating that these patients may have had asthma and not COPD from the beginning. Distinguishing asthma from COPD may be difficult, as untreated asthma may also have a temporary or persistent airway obstruction. The complexity of this differentiation was recently shown in a large global study25. We thus believe that some of the patients with a COPD diagnosis in our study may have had a suboptimally treated asthma where the diagnosis was changed from COPD to asthma after treatment and follow-up. Patients could also have had both asthma and COPD. A post hoc analysis showed 8.8% of patients with FEV1/FVC ≥ 0.70 had both a diagnosis of asthma and COPD recorded. In the group with FEV1/FVC < 0.70 the corresponding number was 9.6%, a non-significant difference. Another potential explanation of a COPD diagnosis in spite of a normal ratio is presence of “preserved ratio impaired spirometry” (PRISm), which can increase the risk of COPD, as well as nonpulmonary conditions in the future26. In a post hoc analysis, some 60% (n = 61) of the patients with an incorrect COPD diagnosis actually had PRISm. This may have contributed to a clinical diagnosis of COPD. Of these, 43 patients had overweight or obesity where the restrictive impairment of high BMI could have masked obstruction26. Furthermore, 6% of the patients did not have a persistent airflow obstruction when the first spirometries were compared to later ones. This may indicate an initial misdiagnosis, as these patients could have had asthma instead. This finding is consistent with a large UK study in which patients with an established COPD diagnosis did not have persistent airflow obstruction in 11.5% of cases27. Aaron et al. conclude that a single spirometric assessment may not be reliable for diagnosing patients with COPD, especially in patients with spirometry results close to the FEV1/FVC threshold28. An important clinical implication of our and others´ findings is that spirometry should be repeated after treatment has been initiated in cases of newly diagnosed COPD.

A COPD diagnosis despite an FEV1/FVC or FEV1/VC ≥ 0.70 was more common in patients who had concomitant hypertension and diabetes. We speculate that this may indicate that the focus of the consultation was on these conditions rather than COPD. On the other hand, a post hoc analysis adding a merged index of all comorbid conditions listed in Table 1 to the multivariable model did not change the results significantly (data not shown).

We find it reasonable that the number of performed spirometries and the proportion of correct interpretations is higher in specialised pulmonology care than in primary care, and that the identification of airway obstruction is easier when COPD is more severe.

We believe the increase in the proportions of performed and correctly interpreted spirometries over time is a result of an extensive national educational effort to update and implement recommendations on the assessment of COPD. Furthermore, access to spirometers in primary care in Sweden is high. At the sites of our PRAXIS II cohort, 98% of the participating healthcare units reported that they had access to spirometers. Although the proportion of performed spirometries has increased, there is still room for improvement. In Denmark, general practitioners who participated in an educational programme showed substantial improvement in the assessment of patients with COPD29. This is in conformity with Sandelowsky et al. who concluded that educational interventions enhanced knowledge of COPD management in primary care in Sweden30. In Finland, a national programme for COPD prevention and treatment had significant positive consequences, including an increase in the use of spirometry31,32. Since most patients with COPD are diagnosed and managed in primary care, we believe continuous education addressed to primary care physicians is of great importance.

A finding related to the interpretation of spirometries was that female sex was significantly associated with a COPD diagnosis despite a ratio ≥0.70. The reason for this is unclear. A potential explanation could have been that FEV1/FVC or FEV1/VC were closer to 0.70 for women. However, a post hoc analysis showed that ratios did not differ significantly between sexes (data not shown). Our result is in contradiction to a Spanish study where women were found not to have a COPD diagnosis despite fulfilled spirometric COPD criteria33. We speculate that, as symptoms of other diseases can have different clinical manifestations in women than in men34,35, the higher degree of spirometries with a ratio ≥ 0.70 in women with a diagnosis of COPD may mirror a more difficult differential diagnostic situation.

A major strength of our study is that it is a real-world study with a large sample size. This contributes to a high external validity and generalisability. Another strength is that, to the best of our knowledge, there are few studies with a follow-up of diagnostic assessment in two different cohorts from the same geographic area28. Limitations include the changes in national recommendations using FEV1/FVC instead of FEV1/maximum VC to confirm the diagnosis of COPD during the study period and that post-bronchodilator values were not present in all patients. Further, spirometry data was not always interpretable, which means a loss of patient data. However, the number of non-assessable spirometries was very low, and an attrition analysis showed that patients included in the analysis of spirometry interpretation and those who were excluded due to non-assessable spirometry data did not differ between any of the variables presented in Table 1 (data not shown). When completing a questionnaire there could be a risk of recall bias.

The second PRAXIS cohort included more primary health care centres (PHCCs) than the first PRAXIS cohort. However, a random selection was performed in both cohorts. Additional analysis where the extra PHCCs were excluded showed substantially unchanged results (data not shown).

The use of spirometry to confirm COPD diagnosis has increased over time, indicating improved implementation of COPD guidelines. At risk of not undergoing a diagnostic spirometry were current smokers, patients with low education and those managed in primary care. There is still a need for continuous medical educational activities to increase diagnostic accuracy.