Original Research

COPD

Scoring System and Clinical Application of COPD Diagnostic Questionnaires*

David B. Price, MD; David G. Tinkelman, MD; Robert J. Nordyke, PhD; Sharon Isonaka, MD, MS; and R. J. Halbert, MD, MPH; for the COPD Questionnaire Study Group

Objectives: In most primary care settings, spirometric screening of all patients at risk is not practical. In prior work, we developed questionnaires to help identify COPD in two risk groups: (1) persons with a positive smoking history but no history of obstructive lung disease (case finding), and (2) patients with prior evidence of obstructive lung disease (differential diagnosis). For these questionnaires, we now present a scoring system for use in primary care. Methods: Scores for individual questions were based on the regression coefficients from logistic regression models using a spirometry-based diagnosis of obstruction as the reference outcome. Receiver operator characteristic analysis was used to determine performance characteristics for each questionnaire. Several simplified scoring systems were developed and tested. Results: For both scenarios, we created a scoring system with two cut points intended to place subjects within one of three zones: persons with a high likelihood of having obstruction (high predictive value of a positive test result); persons with a low likelihood of obstruction (high predictive value of a negative test result); and an intermediate zone. Using these scoring systems, we achieved sensitivities of 54 to 82%, specificities of 58 to 88%, positive predictive values of 30 to 78%, and negative predictive values of 71 to 93%. Conclusions: These questionnaires can be used to help identify persons likely to have COPD among specific risk groups. The use of a simplified scoring system makes these tools beneficial in the primary care setting. Used in conjunction with spirometry, these tools can help improve the efficiency and accuracy of COPD diagnosis in primary care. (CHEST 2006; 129:15311539)
Key words: diagnostic techniques; obstructive lung diseases; primary care; questionnaires; sensitivity and specificity Abbreviations: NPV negative predictive value; PPV positive predictive value; ROC receiver operator characteristic

of COPD is a widespread probU nderdiagnosis lem. Diagnostic confusion between asthma and
1

COPD, while not as widespread, appears to be an important clinical problem in some patients.2 The definitive diagnostic maneuver for COPD is spirometry1; however, despite frequent advocacy for spirometric screening,3,4 spirometry is underused.5 This is especially true in the primary care setting, the usual site of initial presentation.5,6 There is a perception that spirometric screening of all at-risk persons is impractical in primary care. This has led to efforts to identify a subset of patients for whom such screening is likely to be cost-effective.7 In contrast to general population screening programs,8 efforts to locate persons within a primary care practice are more accurately referred to as
www.chestjournal.org

case-finding programs.7 These types of programs should be directed toward groups known to have an increased prevalence of the condition to be identified. COPD prevalence is known to be increased in adults 40 years old and in persons exposed to noxious smoke or fumes, especially cigarette smoke.1 Persons with prior evidence of respiratory problems but in whom a diagnosis has not been definitively established represent another group likely to benefit from closer scrutiny. Previous work9 has shown that relatively simple questionnaires can help identify persons with an increased likelihood of fixed obstruction. Most recently, this has been demonstrated in two risk groups: (1) current and former smokers 40 years old with no prior evidence of obstructive lung disCHEST / 129 / 6 / JUNE, 2006

1531

Downloaded From: http://journal.publications.chestnet.org/ on 11/22/2012

ease (case-finding scenario)10; and (2) persons 40 years old with prior evidence of obstructive lung disease (differential diagnosis scenario).11 We now describe the development of a scoring system for these questionnaires suitable for use in a primary care setting. Materials and Methods
Details of the development of the questionnaires have been described elsewhere.10,11 In brief, two study sites (Aberdeen, Scotland and Denver, CO) were chosen for the evaluation. Subjects 40 years old were randomly selected from primary care practice rosters in these sites and invited by mail to participate in the study. Eligible respondents were enrolled after providing informed consent. Respondents were eligible if they reported the following: (1) a positive smoking history (current or former smokers), with no prior evidence of respiratory diagnosis (eg, no prior respiratory diagnosis and no respiratory medications within the past year); or (2) prior evidence of respiratory diagnosis (eg, any prior respiratory diagnosis or any respiratory medications within the past year), regardless of smoking status. Participants completed a questionnaire covering demographics and symptoms and then underwent spirometry with reversibility testing. Study diagnoses were based on guidelines developed by the Global Initiative for Chronic Obstructive Lung Disease1 and the Global Initiative for Asthma.12 A study diagnosis of COPD was assigned to persons with postbronchodilator FEV1/FVC ratio 0.70. For the differential diagnosis analysis, a diagnosis of asthma was assigned to persons with postbronchodilator FEV1/ FVC ratio 0.70 and FEV1 reversibility 200mL and 12% of baseline. Persons with no reversibility received a diagnosis of probable asthma if they had a prior diagnosis of asthma or were receiving long-term corticosteroids. The study was approved by *From the University of Aberdeen (Dr. Price), Aberdeen, Scotland; National Jewish Medical and Research Center (Dr. Tinkelman), Denver, CO; and Cerner Health Insights (Drs. Nordyke, Isonaka, and Halbert), Beverly Hills, CA. A list of COPD Questionnaire Study Group members is presented in the Appendix. This project was funded by Boehringer Ingelheim International GmbH and Pfizer Inc. Portions of this article were presented at meetings of the International Primary Care Respiratory Group (International Congress: Melbourne, Australia; February 19 to 22, 2004) and the Asian Pacific Society of Respirology (Ninth Congress: Hong Kong, December 10 to 13, 2004). Dr. Price has received honoraria for speaking at sponsored meetings and serving on advisory panels for the following companies marketing COPD products: AstraZeneca, Boehringer Ingelheim, GlaxoSmithKline, Novartis, and Pfizer. He or his research team have received funding for research projects from the same companies. Dr. Tinkelman is an employee of National Jewish Medical and Research Center, which received funding from Boehringer Ingelheim to participate in this study. At the time this work was performed, Drs. Nordyke, Isonaka and Halbert were employees of Cerner Health Insights, which provides consulting services to the pharmaceutical industry, including the sponsors of this project. Manuscript received October 3, 2005; revision accepted December 3, 2005. Reproduction of this article is prohibited without written permission from the American College of Chest Physicians (www.chestjournal. org/misc/reprints.shtml). Correspondence to: R. J. Halbert, MD, MPH, Cerner Health Insights, 9100 Wilshire Blvd, Suite 655E, Beverly Hills, CA 90212; e-mail: rhalbert@cerner.com
1532

ethics committees at the two sites. A description of the study populations is provided in Table 1. In previous reports,10,11 we described the development and initial performance evaluation of the questionnaires. Prior to analysis, each of the study samples were randomly assigned into two subsamples in order to reduce potential biases introduced by same-sample predictions.13 For each study sample, a development subsample (70%) was used to create the questionnaires, while a performance subsample (30%) was used to evaluate the performance characteristics of the questionnaires. Using the development subsamples, item reduction was carried out. Based on the reduced item sets, multivariate logistic regression models were constructed to identify the best performing questions to discriminate between persons with and without COPD in each risk group, again using the development subsamples. All items showing statistical significance at p 0.05 were retained for the final questionnaires, which are provided in Table 2. Using the performance subsamples, receiver operator characteristic (ROC) curves were constructed. Several performance parameters were

Table 1Sample Description*

Differential Case Finding Diagnosis (Full Sample, (Full Sample, n 818) n 597) 58.2 11.2 26.0 31.2 24.2 18.6 49.3 28.3 5.7 33.5 33.3 33.3 44.5 55.5 25.6 24.3 34.8 20.2 33.0 12.0 87.0 6.8 0.2 0.5 5.4 94.4 17.0 95.8 15.5 98.2 9.7 155 18.9 58.7 11.4 26.3 28.3 26.1 19.3 38.3 29.0 6.7 33.6 33.0 33.4 24.2 53.6 22.2 19.1 23.9 55.1 12.7 19.4 12.7 90.8 4.4 0.2 1.0 3.7 83.6 23.8 91.4 18.9 89.9 16.3 235 39.4

Characteristics Demographics Age, yr Age range, % 4049 5059 6069 70 Male gender Body mass index Body mass index categories Low tertile Middle tertile High tertile Smoking status Current smoker Former smoker Never-smoker Pack-years Pack-year categories, % 014 1524 2549 50 Race/ethnicity, % Non-Hispanic white Non-Hispanic black Hispanic Asian Other Pulmonary function, % of predicted Postbronchodilator FEV1 Postbronchodilator FVC Postbronchodilator FEV1/FVC Study diagnosis of COPD Postbronchodilator FEV1/FVC 0.70

*Data are presented as % or mean SD. Never-smokers were not eligible for this study arm.
Original Research

Downloaded From: http://journal.publications.chestnet.org/ on 11/22/2012

Table 2COPD Diagnostic Questionnaires

Questionnaire Items Case-finding questionnaire Age group, yr Body mass index, tertiles Smoking intensity, pack-years Does the weather affect your cough? Do you ever cough up phlegm (sputum) from your chest when you dont have a cold? Do you usually cough up phlegm (sputum) from your chest first thing in the morning? How frequently do you wheeze? Do you have or have you had any allergies? Differential diagnosis questionnaire Age group, yr Smoking intensity, pack-years Have you coughed more in the past few years? During the past 3 yr have you had any breathing problems that have kept you off work, indoors, at home, or in bed? Have you ever been admitted to hospital with breathing problems? Have you been short of breath more often in the past few years? On average, how much phlegm (sputum) do you cough up most days? If you get a cold, does it usually go to your chest? Are you taking any treatment to help your breathing? Response Categories 4049/5059/6069/70 25.4/25.429.7/ 29.7 014/1524/2549/50 Yes/no Yes/no Yes/no Occasionally or more often/never Yes/no 4049/5059/6069/70 014/1524/2549/50 Yes/no Yes/no Yes/no Yes/no None, or 15 mL/d/ 15 mL/d Yes/no Yes/no

examined, including sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV). These parameters were determined at several points on the ROC curve. Several alternate combinations were then tested in comparison to these baseline models in order to identify the best-performing question sets. In this article, we report the creation of a scoring system suitable for use in the primary care setting for each questionnaire. Using the full samples, logistic regression models with the final question sets were used to refine (or finalize) the confidence intervals for the estimated coefficients for each question item. Since regression coefficients are linear (as opposed to odds ratios, which are nonlinear), these were used as a basis for scoring. In

order to facilitate ease of use in the primary care setting, several simplified scoring algorithms were tested against the performance obtained using the raw coefficients. At each stage of the analysis, clinical input was used to evaluate the face validity of the analytic process. All analyses were performed using statistical software (STATA; StataCorp; College Station, TX).

Results The ROC curves are presented for the casefinding (Fig 1) and differential diagnosis (Fig 2)

Figure 1. ROC curve, case-finding questionnaire (performance subsample, n 246).

www.chestjournal.org CHEST / 129 / 6 / JUNE, 2006

1533

Downloaded From: http://journal.publications.chestnet.org/ on 11/22/2012

Figure 2. ROC curve, differential diagnosis questionnaire (performance subsample, n 180).

questionnaires, along with selected performance characteristics (Table 3). For each ROC curve, two cut points were determined by assessing the performance characteristics of the questions to choose those points representing the optimal combination of PPV, NPV, and the distribution of subjects between cut points. This system places subjects within one of three zones, depending on their likelihood of having obstruction: (1) increased likelihood of obstruction (optimum PPV); (2) intermediate likelihood of ob-

struction; or (3) decreased likelihood of obstruction (optimum NPV). Expressed another way, these zones represent, respectively, a probability of COPD that is (1) increased, (2) unchanged, or (3) reduced, compared to the risk group as a whole. Since PPV and NPV vary by prevalence, the scoring process was in part dependent on the baseline prevalence of obstruction within each group: 18.7% in the casefinding sample and 43.3% in the differential diagnosis sample.

Table 3Performance Characteristics, COPD Diagnostic Questionnaires

Questionnaires Case-finding (baseline COPD probability 18.7%) Cut point performance Score PPV, % Specificity, % NPV, % Sensitivity, % Zone performance Proportion of population in zone, % Probability of obstruction, % Differential diagnosis (baseline probability 43.3%) Cut point performance Score PPV Specificity NPV Sensitivity Zone performance Proportion of population in zone Probability of obstruction Cut Point A Cut Point B High Likelihood Intermediate Likelihood Low Likelihood

16.5 37.0 77.0 89.0 58.7

19.5 30.3 57.5 92.7 80.4 29.7 37.0 19.9 20.4 50.4 7.3

18.5 77.8 88.2 71.4 53.8

24.5 63.4 63.7 82.3 82.1 30.0 77.8 26.1 46.8 43.9 17.1

1534

Original Research

Downloaded From: http://journal.publications.chestnet.org/ on 11/22/2012

Table 4 Final Scoring, Case-Finding Questionnaire (Full Sample, n 798)

Question Items Age group, yr 4049 (reference) 5059 6069 70 Body mass index, kg/m2 24 (reference) 25.429.7 29.7 Smoking intensity, pack-yr 014 (reference) 1524 2549 50 Symptoms/history Weather affects cough Phlegm without a cold Phlegm in the morning Wheeze frequency (any) Have or had any allergies *See text for derivation of score. Odds Ratio 1.00 2.20 4.73 7.77 1.00 0.44 0.35 1.00 1.63 1.99 4.05 1.68 1.81 0.54 2.08 0.52 p Value Coefficient 0.00 0.79 1.55 2.05 0.00 0.83 1.05 0.00 0.49 0.69 1.40 0.521 0.60 0.62 0.73 0.65 Score* 0 4 8 10 5 1 0 0 2 3 7 3 3 3 (no) 4 3 (no)

0.022 0.001 0.001

0.002 0.001

0.112 0.010 0.001 0.089 0.013 0.022 0.001 0.005

Odds ratios and regression coefficients are presented for items in the case-finding (Table 4) and differential diagnosis (Table 5) questionnaires, along with the final score for each item. In the final scoring system, coefficients were multiplied by five, rounded to the nearest integer, and scoring reversed for question items with negative scores. These transformations were done to facilitate easy calculation of the total score, derived by summing the values for each response. Assignment of patients to the three

likelihood zones using this scoring system did not differ significantly from the assignment using raw regression coefficients. Since the selection criteria for question items to be included in the final questionnaires were based on the development subsamples, there were some parameter changes when evaluating these question sets using the full samples, including both increases and decreases in signal strength. Notably, three questions that demonstrated statistically significant contribu-

Table 5Final Scoring, Differential Diagnosis Questionnaire (Full Sample, n 571)

Question Items Age group, yr 4049 (reference) 5059 6069 70 Smoking intensity, pack-yr 014 (reference) 1524 2549 50 Symptoms/history Coughed more in recent years Breathing-related disability Breathing-related hospitalization Dyspnea more in recent years Phlegm quantity 15 mL/d Cold goes to chest Treatment for breathing *See text for derivation of score.
www.chestjournal.org CHEST / 129 / 6 / JUNE, 2006

Odds Ratio 1.00 2.91 5.64 9.25 1.00 1.77 4.04 5.51 0.76 0.55 3.35 1.18 2.18 2.23 2.81

p Value

Coefficient 0.00 1.07 1.73 2.22 0.00 0.57 1.40 1.71 0.28 0.60 1.21 0.16 0.78 0.80 1.03

Score* 0 5 9 11 0 3 7 9 1 (no) 3 (no) 6 1 4 4 5

0.001 0.001 0.001

0.075 0.001 0.001 0.235 0.015 0.001 0.535 0.002 0.002 0.001

1535

Downloaded From: http://journal.publications.chestnet.org/ on 11/22/2012

tions to the final questionnaire models fell below the threshold of statistical significance in the final performance evaluation. One item in the case-finding instrument (weather affects cough) and two items in the differential diagnosis instrument (cough or dyspnea more in recent years) changed significance in this way. Omitting these items produced no significant change in questionnaire performance; however, due to the assumptions of the split-sample technique, these questions were retained in the final models. Discussion Prior work10,11 has demonstrated that questionnaires based on patient-reported information can be used to identify persons likely to have COPD among specific risk groups. This study adds a simple scoring system to make these instruments practical for use in the primary care setting. Using this scoring system, these instruments demonstrated ROC performance comparable to that of other respiratory screening questionnaires, as illustrated in Table 6. In COPD, for example, retrospective analyses using data from the Third National Health and Nutrition Examination Survey8,9 achieved sensitivities ranging from 54 to 86% and specificities from 40 to 71%. Symptombased screening tools for asthma have been reported to achieve sensitivities from 38 to 80% in adults and 23 to 86% in children, with specificities ranging from 64 to 99% in adults and 55 to 100% in children.14 16 These values compare well with airway hyperresponsiveness, which showed a sensitivity of 39%, specificity of 90%, PPV of 62%, and NPV of 78% when

compared with physician-diagnosed asthma.17 Similar results have been described with recommended screening examinations for diseases such as breast18 and colorectal cancer.19 Our scoring system was designed to address both clinical and policy goals. For the clinical goals, we conformed to the logic of the clinical decisionmaking process in order to make the system easy to use and interpret by the busy general practitioner. We focused on two key clinical decisions. For the first decision, whom should I refer for spirometry? the goal was to identify persons at a high likelihood of obstruction with a reasonable degree of certainty. In practice, we chose PPV as the most relevant parameter for this decision, and defined reasonable as approximately 50% certainty. For the second decision, for which patients may spirometry be safely deferred? the goal was to identify persons at low likelihood of obstruction with a high degree of certainty. We chose NPV as the most relevant parameter for this decision, and defined high as approximately 90%. Thus, these two clinical decisions define three likelihood zones: (1) the zone of increased likelihood allows physicians to refer patients for spirometry with a reasonable (approximately 50%) certainty that they do indeed have fixed obstruction; (2) the intermediate zone includes patients for whom the prior probability of obstruction is not definitely higher or lower than the baseline for the risk group as a whole; (3) the zone of low likelihood allows physicians to defer spirometry with a high (approximately 90%) certainty that there is a no fixed obstruction in these patients. For policy purposes, we anticipated that the ques-

Table 6 Predictive Ability of Screening Methods for COPD and Other Conditions*
Diseases and Tests COPD Current study: case finding Current study: differential diagnosis COPD case finding9 COPD differential diagnosis9 COPD population screener8 Asthma Adult questionnaire15 Childhood questionnaire16 Childhood questionnaire17 Airway hyperresponsiveness18 Breast cancer19 Mammography Clinical breast examination Colorectal cancer20 Fecal occult blood test Sigmoidoscopy Barium enema *N/A not available. 1536
Original Research

Sensitivity, % 5980 5482 5471 6375 7786 3880 2370 5586 39 7196 4069 3070 6878 3253

Specificity, % 5877 6488 6371 6370 4063 6499 9199 55100 90 9497 8899 9098 N/A 8285

PPV, % 3037 6378 2025 6869 3342 542 6396 1066 62 222 450 238 N/A N/A

NPV, % 8993 7182 9194 6470 8694 99 N/A 8499 78 N/A N/A N/A N/A N/A

Downloaded From: http://journal.publications.chestnet.org/ on 11/22/2012

tionnaires would be used in diverse health-care contexts. We therefore tried to make the system adaptable to different regional needs. Taking into account the performance characteristics of the instruments, the three-zone approach was chosen to allow for some degree of flexibility in policy recommendation between different health systems, with different budgetary constraints and different policy priorities. These decisions (and subsequent policy recommendations) will also be affected by the clinical perception of the background prevalence of COPD in the group to whom the questionnaires are administered (ie, the prior probability of fixed obstruction). We anticipate that persons assigned to the highlikelihood zones will be recommended for spirometry in most regions, and that persons in the lowlikelihood zones will be deferred in most regions. Recommendations for the intermediate zones will depend on local conditions. In some regions, for example, persons falling within the intermediate zones will be recommended for spirometry with the intention of minimizing the number of cases missed. In regions with fewer health resources (or different health-spending priorities), persons assigned to this zone might be followed up clinically with spirometry deferred until a later date, so as to minimize the number of unnecessary procedures. By incorporating the scoring system described here, the questionnaires become powerful tools for helping the primary care clinician efficiently identify persons with obstructive lung disease. Most physicians will note that they are likely to ask at least some of the questions included in our instruments as part of their routine workup. Placing these questions within the framework described here provides added value in at least two ways. First, because of complex interactions between question items, combining the questions into a single instrument changes their individual significance. The process described here helps identify the relationships between several key questions, providing an added value above the diverse questions asked by clinicians. Second, creating questionnaires with defined scoring and performance characteristics provides quantitative estimates of the risk of obstruction for the clinician. Case-Finding Application A key goal in seeking out previously undiagnosed cases of a disease among persons at known increased risk is to identify these new cases at the lowest incremental cost. The case-finding questionnaire could be very useful to enhance the efficiency of such current screening efforts using spirometry
www.chestjournal.org

alone. We see at least two potential applications: (1) improving the efficiency of COPD diagnosis in smokers who present with respiratory symptoms (but no prior diagnosis), and (2) assisting in the early detection of COPD in smokers outside clinical settings. In the first application, clinicians who feel that spirometric evaluation of all current and former smokers in their practice is impractical may use the questionnaire to prescreen those smokers at highest likelihood. This allows more efficient use of office and referral resources, minimizing the administrative burden on the practice. Since the questionnaire may be given unsupervised to patients in the waiting area, the only time required of the office staff is that required to calculate and interpret the results. The simplified scoring system presented here reduces this administrative burden considerably: a computerized version of the questionnaires could reduce it even further. In the second application, the questionnaire could be used as part of an outreach program to screen smokers for potential lung health problems in settings outside the clinic, such as a worksite or general population screening program. These types of programs, often associated with a media awareness campaign or similar marketing outreach, can be very helpful in stimulating persons to consider the potential for lung damage due to smoking.20 In both these applications, the prevalence of COPD in the intermediate zone is expected to approximate 20%. Thus, the administrator of the program (a clinician in many cases) should determine in advance whether this prevalence is high enough to warrant follow-on evaluation with spirometry, and implement accordingly. Differential Diagnosis Application The differential diagnosis questionnaire could be useful as part of the evaluation process whenever there is lack of diagnostic clarity among patients suspected of having obstructive lung disease. This might include patients who have previously received respiratory medications without assignment of a specific diagnosis, or patients in whom the previously assigned diagnosis is in doubt. As with the casefinding application, more efficient use of spirometry is likely to result. However, enhanced diagnostic accuracy may provide additional benefit by improving the overall management of patients with these diseases. The expected probability of fixed obstruction in this group is approximately 50%, which should generate a spirometry referral in most developed countries.
CHEST / 129 / 6 / JUNE, 2006

1537

Downloaded From: http://journal.publications.chestnet.org/ on 11/22/2012

Limitations Our response rates were fairly low (24.2% in Aberdeen and 6.1% in Denver). Since we mailed to a random sample of all persons 40 years old irrespective of smoking or lung disease history, we cannot determine exactly how many of the persons who failed to respond were actually ineligible for the study. It is difficult to know how nonresponse bias might have affected our primary study question of which questions are associated with obstruction. At any rate, our subjects volunteered to participate and do not necessarily reflect the general nature of the primary care practice population, though they may represent persons likely to come forward for screening in primary care. Our questionnaires were tested and validated using different subsets drawn from the same population. Despite the splitting of the samples to reduce same-sample bias, other potential sources of bias could have affected the performance of the final instruments. Systematic differences in diagnostic or prescribing patterns, for example, would affect the assignment of patients into the case-finding or differential diagnosis sample. Our sample size was quite large relative to most previous studies of respiratory diagnostic questionnaires; nevertheless, we found some instability in the regression coefficients and variance estimates between development and performance subsamples. This may represent idiosyncratic differences between these two subsamples or unmeasured systematic differences. Resolution of these questions requires validation in an independent population sample. We made our study diagnoses based on spirometry performed during a single visit. This is an epidemiologic case-definition approach and thus represents an oversimplification of clinical diagnosis. It is possible that some proportion of patients might have received a different study diagnosis had they been subjected to a full diagnostic workup, including a course of steroid therapy. Although most of the question items included in the final questionnaires were derived from previously validated instruments, test-retest reliability of the final instruments should be performed. Finally, different clinical populations are likely to have different baseline prevalences of airway obstruction. Since the questionnaire scoring system is based on positive and negative predictive values (which are sensitive to prevalence), questionnaire performance is likely to be different in these populations. The PPV increases as prevalence increases while the NPV decreases with increasing prevalence, although most of the changes in performance parameters are generally small within the ranges likely to
1538

be seen in clinical practice. For example, if we administered the case-finding questionnaire to persons with a baseline obstruction prevalence of 30% (50% higher than observed in our results), we would expect the questionnaire to have a higher PPV (52%) at cut point A and a slightly lower NPV (87%) at cut point B. This translates into an improved confidence in referring for spirometry and slightly decreased confidence in deferring this test. If we administered the differential diagnosis questionnaire to a population with the same prevalence of 30%, we would expect a lower PPV (66%) at cut point A leading to decreased confidence in spirometry referral, and a slightly higher NPV (89%) at cut point B leading to increased confidence in deferring spirometry. For these reasons, administering these questionnaires using modified selection criteria (for example, using the questionnaires in an unselected general population, or using different age thresholds) is not supported by these results.

Conclusions The questionnaires described here can be used to identify persons likely to have COPD among specific risk groups. This can be done with acceptable performance characteristics. The use of a simple scoring system makes these tools beneficial in the primary care setting. Used in conjunction with spirometry, these tools can help improve the efficiency and accuracy of COPD diagnosis in primary care. Future work should address the reliability of the instruments, validity in different populations and practice environments, and the cost impact of implementing these tools.
ACKNOWLEDGMENT: The authors thank 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; and Dr. Michael Levine, who served as a spirometry reviewer.

Appendix
Members of the COPD Questionnaire Study Group include William Erhardt, MD, Pfizer Inc, New York, NY; Daryl Freeman, MD, University of Aberdeen, Aberdeen, Scotland; R. J. Halbert, MD, MPH, Cerner Health Insights, Beverly Hills, CA; Thomas Hausen, MD, Essen, Germany; Sharon Isonaka, MD, MS, Cerner Health Insights, Beverly Hills, CA; Elizabeth F. Juniper, MSc, McMaster University, Hamilton, ON, Canada; Claus Justus, DVM, Boehringer Ingelheim International GmbH, Ingelheim, Germany; Mark L. Levy, MD, University of Edinburgh, Edinburgh, Scotland; Dmitry Nonikov, MD, Wiesbaden, Germany; Robert J. Nordyke, PhD, Cerner Health Insights, Beverly Hills, CA; Anders strem, MD, Oslo, Norway; David B. Price, MD, University of Aberdeen, Aberdeen, Scotland; David G. Tinkelman, MD, National Jewish Medical and Research
Original Research

Downloaded From: http://journal.publications.chestnet.org/ on 11/22/2012

Center, Denver, CO; Thys van der Molen, MD, University of Groningen, Groningen, the Netherlands; and Constant P. van Schayck, PhD, University of Maastricht, Maastricht, the Netherlands.

10

11

References
1 Global Initiative for Chronic Obstructive Lung Disease. Global strategy for the diagnosis, management, and prevention of chronic obstructive pulmonary disease, updated 2004. Available at: www.goldcopd.com. Accessed November 8, 2005 2 van Schayck CP. Diagnosis of asthma and chronic obstructive pulmonary disease in general practice. Br J Gen Pract 1996; 46:193197 3 National Lung Health Education Program. Strategies in preserving lung health and preventing COPD and associated diseases. Chest 1998; 113(2 Suppl):123S163S 4 Ferguson GT, Enright PL, Buist AS, et al. Office spirometry for lung health assessment in adults: a consensus statement from the National Lung Health Education Program. Chest 2000; 117:1146 1161 5 van Weel C. Underdiagnosis of asthma and COPD: is the general practitioner to blame? Monaldi Arch Chest Dis 2002; 57:65 68 6 Korsten AMMH, van Schayck CP. New international primary care airways guidelines (IPAG). Revue Franc aise dAllergologie et dImmunologie Clinique 2003; 43:246 248 7 van Schayck CP, Loozen JM, Wagena E, et al. Detecting patients at a high risk of developing chronic obstructive pulmonary disease in general practice: cross sectional case finding study. BMJ 2002; 324:1370 8 Calverley PMA, Nordyke RJ, Halbert RJ, et al. Development of a population-based screening questionnaire for COPD. J COPD 2005; 2:225232 9 van Schayck CP, Halbert RJ, Nordyke RJ, et al. Comparison of existing symptom-based questionnaires for identifying 12

13 14

15

16 17 18

19

20

COPD in the general practice setting. Respirology 2005; 10:323333 Price DB, Tinkelman DG, Halbert RJ, et al. Symptom-based questionnaire for identifying COPD in smokers. Respiration 2006; 73:285295 Tinkelman DG, Price DB, Nordyke RJ, et al. Symptom-based questionnaire for differentiating COPD and asthma. Respiration 2006; 73:296 305 Global Initiative for Asthma. Global strategy for asthma management and prevention, updated 2004. Available at: www.ginasthma.com. Accessed November 8, 2005 Cox DR. A note on data-splitting for the evaluation of significance levels. Biometrika 1975; 62:441 444 Sistek D, Tschopp JM, Schindler C, et al. Clinical diagnosis of current asthma: predictive value of respiratory symptoms in the SAPALDIA study. Eur Respir J 2001; 17:214 219 Frank TL, Frank PI, McNamee R, et al. Assessment of a simple scoring system applied to a screening questionnaire of asthma in children aged 515 yrs. Eur Respir J 1999; 14:1190 1197 Hall CB, Wakefield D, Rowe TM, et al. Diagnosing pediatric asthma: validating the Easy Breathing Survey. J Pediatr 2001; 139:227233 Peat JK, Toelle BG, Marks GB, et al. Continuing the debate about measuring asthma in population studies. Thorax 2001; 56:406 411 Humphrey LL, Helfand M, Chan BK, et al. Breast cancer screening: a summary of the evidence for the U.S. Preventive Services Task Force. Ann Intern Med 2002; 137(5 Part 1):347360 Pignone M, Rich M, Teutsch SM, et al. Screening for colorectal cancer in adults at average risk: a summary of the evidence for the U.S. Preventive Services Task Force. Ann Intern Med 2002; 137:132141 Zielinski J, Bednarek M. Early detection of COPD in a high-risk population using spirometric screening. Chest 2001; 119:731736

www.chestjournal.org

CHEST / 129 / 6 / JUNE, 2006

1539

Downloaded From: http://journal.publications.chestnet.org/ on 11/22/2012