Symptom-Based Questionnaire for Differentiating COPD and AsthmaTinkelman D.G.a · Price D.B.b · Nordyke R.J.c, d · Halbert R.J.c, d · Isonaka S.c · Nonikov D.e · Juniper E.F.f · Freeman D.b · Hausen T.g · Levy M.L.k · Østrem A.h · van der Molen T.i · van Schayck C.P.j
aNational Jewish Medical and Research Center, Denver, Colo., USA; bUniversity of Aberdeen, Aberdeen, UK; cCerner Health Insights, Beverly Hills, Calif., and dUCLA School of Public Health, Los Angeles, Calif., USA; eWiesbaden, Germany; fMcMaster University, Hamilton, Canada; gEssen, Germany; hOslo, Norway; iUniversity of Groningen, Groningen, and jUniversity of Maastricht, Maastricht, Netherlands; kUniversity of Edinburgh, Edinburgh, UK Corresponding Author
Background: Many patients with obstructive lung disease (OLD) carry an inaccurate diagnostic label. Symptom-based questionnaires could identify persons likely to need spirometry. Objectives: We prospectively tested questions derived from a comprehensive literature review and an international Delphi panel to help identify chronic OLD (COPD) in persons with prior evidence of OLD. Methods: Subjects were recruited via random mailing to primary-care practices in Aberdeen, Scotland, and Denver, Colorado. Persons aged 40 and older reporting any prior diagnosis of OLD or any respiratory medications in the past year were enrolled. Participants answered 54 questions covering demographics and symptoms and underwent spirometry with reversibility testing. A study diagnosis of COPD was defined by fixed airway obstruction as measured by post-bronchodilator FEV1/FVC <0.70. We examined ability of individual questions in a multivariate framework to discriminate between persons with and without the study diagnosis of COPD. Results: 597 persons completed all investigations and proceeded to analysis. The list of 54 questions yielded 52 items for analyses, which was reduced to 19 items for entry into a multivariate regression model. Nine items had significant relationships with the study diagnosis of COPD, including increased age, pack-years, worsening cough, breathing-related disability or hospitalization, worsening dyspnea, phlegm quantity, cold going to the chest, and receipt of treatment for breathing. Individual items yielded odds ratios ranging from 0.33 to 20.7. This questionnaire demonstrated a sensitivity of 72.0 and a specificity of 82.7. Conclusions: A short, symptom-based questionnaire identifies persons more likely to have COPD among persons with prior evidence of OLD.
Copyright © 2006 S. Karger AG, Basel
Diagnosing obstructive lung disease (OLD), including chronic OLD (COPD) and/or asthma, in adults can be difficult in some patients . The overlap between the clinical pictures of these two disease processes can hamper a clear distinction , especially in smokers and in older adults . As medical management recommendations for these conditions have evolved, differences in therapeutic approach are becoming more distinct. As such, accurate diagnosis is increasingly important.
In most countries, patients with respiratory complaints present first to a primary care provider rather than a pulmonary specialist [4, 5]. Misdiagnosis and underdiagnosis of OLD by general practitioners is common [4, 6, 7]. In some cases, respiratory medications are prescribed without establishing a clear diagnosis. Thus, many patients with evidence of OLD carry an inaccurate diagnostic label, or none at all . This is reflected in wide variation in prescribing patterns and diagnostic labeling [9, 10].
Most respiratory guidelines agree that spirometry should be the basis for diagnosis of COPD . Lung function testing is also recognized as a mainstay of diagnosis in asthma, although the focus is more often on reversibility than on obstruction per se . In the primary care setting, access to spirometry may be limited, leading to underuse [13,14,15]. Symptom-based questionnaires could be used to help to differentiate between persons with COPD and asthma among individuals currently thought to have OLD. Used in conjunction with spirometry, these types of tools might enhance the diagnostic accuracy of OLD. A systematic literature review revealed only two questionnaires studied for this purpose. One tool was developed explicitly to differentiate asthma from COPD in the physician’s office . A second study described a computer-administered respiratory symptom questionnaire tested for its ability to discriminate among persons with no airflow limitation, chronic airflow limitation, and asthma in a controlled research setting using a small study population . Neither of these studies examined a broad array of questions to determine which might be most discriminatory, and neither tool was tested in an unselected patient population.
We prospectively tested a group of questions to help identify COPD among persons aged 40 and over with prior evidence of OLD drawn from primary care practices in two countries.
Except as noted, methods were similar to those described in the accompanying report . That report includes a CONSORT diagram for the entire study, including this study arm.
Respondents were enrolled if they reported prior evidence of obstructive airways disease (prior diagnosis of asthma or COPD, or any respiratory medications within the past year) and provided informed consent. Subjects taking bronchodilators were asked to withhold these prior to visiting the clinic. A separate consent was provided for bronchodilator withholding.
Questionnaire administration and spirometry procedures were identical to those previously described .
Subjects were placed into one of five mutually exclusive diagnostic categories based on spirometric criteria (table 1). In addition to study diagnoses based upon fixed obstruction (postbronchodilator FEV1/FVC <0.70) and bronchodilator reversibility (change in FEV1 = 200 ml and 12% of baseline), persons with no objective evidence of OLD were subdivided into study diagnoses of ‘Probable asthma’ and ‘Probable normal’ based upon the presence or absence of a prior diagnosis of asthma or chronic administration of corticosteroids.
|Table 1. Study diagnoses|
The study sample (n = 597) was split into two subsamples for the development (development sample, 70%, n = 417) and evaluation (performance sample, 30%, n = 180) of the questionnaire. All other steps in the analysis were conducted in the manner previously described .
In this arm of the study, 634 persons completed all investigations. Spirometric results were unsuitable for analysis in 5.8% of these, leaving 597 for analysis. Table 2 displays selected characteristics of the study population. Although the distribution of characteristics was generally similar between sites, respondents from Aberdeen were significantly more likely to be older, male, and white. As seen in table 3, Aberdeen subjects were more likely to have a study diagnosis of COPD (45 vs. 35.6% in Denver, p = 0.021), and were less likely to receive a study diagnosis of ‘probable normal’ (p < 0.001).
|Table 2. Sample description|
|Table 3. Study diagnoses, by smoking status and study site|
Bivariate results from the development sample are shown in table 4, which presents the proportion of persons with COPD and the proportion of persons without COPD who gave a positive response for each item. Of 52 items, almost half showed statistical differences between these two groups at the 0.05 level.
|Table 4. Bivariate relationships between candidate variables and study diagnosis (development sample, n = 417)|
After item reduction, the final regression model (table 5) included 19 items, of which 11 showed some significance at the 0.05 level. Two of these items – BMI and wheeze frequency – showed significance in only one response category, and were dropped from the final questionnaire. The final questionnaire contained 9 question items and is presented in table 6. Figure 1 presents the ROC curve from this questionnaire using the performance sample. At the inflection point of the ROC curve, this questionnaire demonstrated a sensitivity of 72.0 and a specificity of 82.7.
|Table 5. Predictive ability of reduced item set (development sample, n = 375)|
|Table 6. Final differential diagnosis questionnaire|
|Fig. 1. ROC curve for final questionnaire (performance sample, n = 180).|
This is the first prospective analysis of the utility of patient-reported information for identifying persons in a primary care setting likely to have COPD among those with prior history of OLD. Prior studies have been small or have validated short questionnaires against preselected patients with known prior diagnosis. We focused on persons aged 40 and older with a prior respiratory history or medications in order to represent the patient population likely to be the focus of efforts in actual clinical practice, i.e. those with suspected OLD.
Among patients with suspected OLD, many general practitioners do not consider a diagnosis of COPD unless patients have a ‘classic’ presentation . These patients tend to be older, have more clinical disability, and are generally sicker in appearance. However, a substantial group of individuals with COPD do not fit this description and are often misdiagnosed as having asthma. Correct diagnosis only matters when it alters treatment recommendations. The current recommended approach to asthma focuses on the early use of anti-inflammatory ‘controllers’, principally corticosteroids, and the measurement of day-to-day variability in respiratory function using peak flow measurements . For COPD, symptomatic relief is paramount, with corticosteroids playing a more limited role . However, the primary reason for early, correct identification of these patients is to actively begin a program of breaking the addictive behavior pattern associated with cigarette smoking. Only smoking cessation can lead to substantive changes in the progressive decrement of lung function characteristic of COPD, although smoking reduction may be a promising approach for recalcitrant patients . The earlier these efforts are begun, the better the chances for patients to enjoy better long-term lung function, additional life years, and improved quality of life. Patients identified earlier in the course of their disease have a greater incentive to quit, while their efforts can still make a difference .
In many primary care practices, there is a large group of patients who enter the practice with a prior respiratory diagnosis, or who have received respiratory medications without a clearly established diagnosis. In order to efficiently identify which of these patients need further evaluation with spirometry, the general practitioner needs help in identifying those patients who are most likely to have fixed obstruction. A similar thought process is needed when patients’ symptoms and clinical findings do not lead to a clear diagnosis. There is also apparent confusion regarding appropriate therapeutic management in each of these diseases [23, 24].
It can be difficult to definitively establish a diagnosis of asthma during a single clinical encounter. Furthermore, there is currently no ‘gold standard’ for asthma diagnosis using objective (i.e., spirometric) measures . Since our goal was to facilitate spirometry as the primary diagnostic maneuver, we focused on identifying persons with fixed (postbronchodilator) obstruction. The strongest discriminators for fixed obstruction were increasing age and pack-years of smoking, dyspnea that worsened in recent years, breathing-related hospitalization, phlegm in quantities more than 15 ml per day, and a tendency for colds to ‘go to the chest’. Coughing more in recent years, and breathing-related disability (‘off work, indoors, at home, or in bed’) suggested a decreased likelihood of COPD.
In addition to the well-known relationships between COPD, age and smoking, BMI has shown associations with COPD [26, 27] and with bronchial hyperresponsiveness . In this population of patients with prior evidence of pulmonary disease, the relationship between BMI and fixed obstruction was relatively weak and only held for persons with a BMI over 30.7. BMI was dropped from the final questionnaire.
Symptoms were quite commonly reported in our sample, as expected in a population with prior respiratory diagnosis or medications. Many of the symptom items tested showed significant discriminatory ability in the unadjusted analysis. Using cough as an example, 12 items probed different aspects of cough and 6 of these were significant at the 0.05 level. Several items often considered important in clinical practice – such as morning cough, coughing in the winter, and cough affected by the weather or cold air – showed little discriminatory power. Only one cough item, ‘Have you coughed more in the past few years?’ remained significant after adjustment for other variables, and this was negatively associated with COPD. Using the same question item, van Schayck et al.  found that smokers answering positively were 2.5 times more likely to have an FEV1 <80% predicted than were those answering negatively. That study excluded persons with respiratory medications, however, making direct comparisons difficult.
Similarly, dyspnea and phlegm were commonly reported among persons with and without COPD, and many of these items were discriminatory in unadjusted analysis. The MRC Dyspnoea Scale, which showed excellent discrimination in bivariate analysis, showed almost no discriminatory power after adjustment for other factors. The MRC definition of chronic phlegm (phlegm on most days for three consecutive months for two years or more), was reported by less than 1% of persons, and did not enter the final model. Quantity of phlegm, an issue familiar to many physicians, had never been previously tested. Indeed, we found no standardized questions in the literature or even standardized response categories for sputum volume, and developed a new item for this study. This item entered the final regression model and remained a significant discriminator even after adjustment for other variables.
Wheeze is particularly noteworthy, as it is often considered more typical of asthma than of COPD. A large proportion of our respondents reported wheezing, whether or not they had COPD. Items addressing the presence, timing, and chronicity of wheeze were not discriminatory. Only wheeze frequency entered the regression model, where it was discriminatory in one response category, wheezing every day. As a result, this item was not included in the final questionnaire.
It is interesting to compare these results with symptom patterns commonly offered for differentiating COPD and asthma. Our results support the importance of the progressive nature of dyspnea in identifying persons with COPD, but a similarly worded question on the progression of cough was negatively associated with obstruction. Other items – such as dyspnea during exercise , chronic productive cough , night time waking with dyspnea or wheeze , and day-to-day variation of symptoms  – were common in persons with and without fixed obstruction, and added little discriminatory ability in this study population. A planned further analysis of the relationships between symptom items may shed more light on these findings.
Other items, such as the tendency for a cold to ‘go to your chest’ and taking ‘any treatment to help your breathing’ were significantly associated with COPD. The former item, taken from the American Thoracic Society – Division of Lung Diseases questionnaire , addresses the frequency of lower respiratory complications following viral infections. While nondiscriminatory in the unadjusted analysis, this was significant after adjusting for other variables. The latter item, from the 1987 revision of the European Community for Coal and Steel questionnaire , was discriminatory in both unadjusted and adjusted analyses.
Several questions on exposures thought to be useful in discriminating between asthma and COPD did not prove so in our analysis. For example, several items related to allergy history and response to noxious or allergic triggers showed only marginal discriminatory abilities in unadjusted analysis, and did not enter the final model.
In addition to the limitations previously described , there may have been between-site differences in response rate that could have biased our results. Thus, the observed differences in age and sex distribution and in study diagnoses could have resulted from differential responses to the mailed invitation, or from actual differences in respiratory diagnostic labeling or prescribing at the two sites. Such differences, however, are less likely to have differentially affected answers to question items in persons with similar spirometric results.
Confusion between asthma and COPD, combined with underuse of spirometry in the primary care setting, contributes to misdiagnosis of OLD. Our results suggest that a simple self-administered questionnaire can facilitate differentiation between these diseases in order to ensure appropriate use of spirometry. This questionnaire could be useful as part of the evaluation process whenever there is lack of diagnostic clarity in patients suspected of having OLD, including patients who have previously received respiratory medications without assignment of a specific diagnosis. More efficient use of spirometry and enhanced diagnostic accuracy may serve to improve overall management of patients with these diseases.
This project was supported by a grant by Boehringer Ingelheim International GmbH and Pfizer Inc. The authors would like to thank the following: the subjects who participated in the study; the staff who collected the data; study coordinators Jan Caldow in Aberdeen and Melanie Gleason in Denver; Dr. John L. Adams, who provided statistical advice; Dr. Michael Levine, who served as a spirometry reviewer; and Drs. Claus Justus of Boehringer Ingelheim and William Erhardt of Pfizer, for support and encouragement throughout.
R.J. Halbert, MD, MPH
Cerner Health Insights
9100 Wilshire Blvd., Suite 655E
Beverly Hills, CA 90212 (USA)
Tel. +1 310 598 4462, Fax +1 816 936 1862, E-Mail email@example.com
Received: March 31, 2005
Accepted after revision: September 6, 2005
Published online: December 5, 2005
Number of Print Pages : 10
Number of Figures : 1, Number of Tables : 6, Number of References : 31
Respiration (International Journal of Thoracic Medicine)
Vol. 73, No. 3, Year 2006 (Cover Date: May 2006)
Journal Editor: Bolliger, C.T. (Cape Town)
ISSN: 0025–7931 (print), 1423–0356 (Online)
For additional information: http://www.karger.com/RES