Open Collections

UBC Faculty Research and Publications

The diagnostic performance of patient symptoms in screening for COPD Johnson, Kate M; Tan, Wan C; Bourbeau, Jean; Sin, Don D; Sadatsafavi, Mohsen Aug 3, 2018

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Item Metadata


52383-12931_2018_Article_853.pdf [ 630.85kB ]
JSON: 52383-1.0370938.json
JSON-LD: 52383-1.0370938-ld.json
RDF/XML (Pretty): 52383-1.0370938-rdf.xml
RDF/JSON: 52383-1.0370938-rdf.json
Turtle: 52383-1.0370938-turtle.txt
N-Triples: 52383-1.0370938-rdf-ntriples.txt
Original Record: 52383-1.0370938-source.json
Full Text

Full Text

LETTER TO THE EDITOR Open AccessThe diagnostic performance of patientsymptoms in screening for COPDKate M. Johnson1 , Wan C. Tan2, Jean Bourbeau3, Don D. Sin2,5, Mohsen Sadatsafavi1,4,5*, for the Canadian Cohortof Obstructive Lung Disease (CanCOLD) study and the Canadian Respiratory Research NetworkAbstractIt is recommended that screening for COPD be restricted to symptomatic individuals, but supporting evidence islacking. We determined the performance of wheeze, cough, phlegm, and dyspnea in discriminating COPD versusnon-COPD in a population-based sample of 1332 adults. Area Under the Receiver Operating Curves (AUC) indicatedthat symptoms had modest performance whether assessed individually (AUCs 0.55–0.62), or in combination (AUC fornumber of symptoms as the predictor 0.64). AUC improved with the inclusion of multiple other factors (AUC 0.71).Restricting screening to symptomatic individuals is unlikely to substantially improve the yield of general populationscreening for undiagnosed COPD.Keywords: Screening test, Population, Respiratory symptoms, Chronic obstructive pulmonary diseaseIntroductionChronic Obstructive Pulmonary Disease (COPD) is a com-mon inflammatory lung condition that is characterized bysymptoms of shortness of breath, cough, and sputum pro-duction [1]. Although COPD is under-diagnosed in thecommunity [2], several major guidelines, including fromthe influential US Preventive Services Task Force, have rec-ommended against the use of spirometry to screen forCOPD in asymptomatic individuals in the general popula-tion because the number-needed-to-screen (NNS) to pre-vent adverse disease outcomes is prohibitively large [3, 4].Some have advocated for case finding strategies to im-prove the diagnosis rate in the community in a morecost-effective manner, for example by targeting spirom-etry only among symptomatic individuals [5]. However,many patients with undiagnosed COPD have mild diseaseand may have few (if any) respiratory symptoms [6], andindividuals without COPD can experience symptomssimilar to those of COPD patients [7]. In addition, symp-tomatic COPD patients tend to be diagnosed earlier [6]and are therefore removed from the pool of cases thatwould be detected through a screening program. Wedetermined the diagnostic performance of patient symp-toms for screening in the general population to assesswhether the yield of screening could be improved byrestricting it to the symptomatic population.MethodsWe used data from the Canadian Cohort of ObstructiveLung Disease (CanCOLD) Study. CanCOLD was a pro-spective cohort study of 1332 adults ≥40 years who weresampled from the general Canadian population with multi-level sampling to ensure representativeness. Participantswere followed for a maximum of 3 years with visits at18-month intervals [8]. They reported their demographicinformation, smoking status and history, comorbidities,and respiratory symptoms at each visit using validatedquestionnaires. Diagnostic spirometry was performed ateach visit and persistent airflow limitation was defined aspost-bronchodilator FEV1/FVC < lower limit of normal.Participants were deemed to have undiagnosed COPD ifthey had persistent airflow limitation but did not reportprevious physician-diagnosed COPD, emphysema, orchronic bronchitis. Subjects with a previous diagnosisof COPD were excluded. Information was collected onthe frequency or severity of cough, phlegm, and wheezeusing three questions for each symptom. The responseswere coded as a variable ranging from 0 to 3 for eachsymptom. Breathlessness was measured using the Medical* Correspondence: msafavi@mail.ubc.ca1Respiratory Evaluation Sciences Program, Collaboration for OutcomesResearch and Evaluation, Faculty of Pharmaceutical Sciences, University ofBritish Columbia, Vancouver, Canada4Centre for Clinical Epidemiology and Evaluation, Vancouver Coastal HealthInstitute, Vancouver, Canada© The Author(s). 2018 Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0International License (, which permits unrestricted use, distribution, andreproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link tothe Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver( applies to the data made available in this article, unless otherwise stated.Johnson et al. Respiratory Research  (2018) 19:147 Council dyspnea scale. We also assessed thetotal number of symptoms experienced by each partici-pant (0–4 range).First, we determined the independent associations be-tween individual symptoms and the presence of undiag-nosed COPD (v. no COPD) using a logistic regressionmodel with symptoms as separate independent variablesand adjusting for participant demographics, comorbidities,smoking status, and pack-years. Second, we assessed thediagnostic performance of symptoms when used individu-ally to distinguish patients with undiagnosed COPD fromnon-COPD subjects. We evaluated the sensitivity and spe-cificity of each symptom at different thresholds (i.e., 0, 1, 2,or 3) for defining a patient as symptomatic. We fittedReceiver Operating Characteristic (ROC) curves to deter-mine the Area Under the Curve (AUC) for each symptomindividually, as well as their combined performance usingthe total number of symptoms. Finally, we used the AUCof the above-mentioned logistic regression model to assessthe performance of all individual symptoms and covariatestogether. An AUC of 0.5 indicates the model has no dis-criminatory ability. Generalized estimating equations wereused in all models to account for clustering of observa-tions within individuals.Ethics approval for CanCOLD was obtained from therelevant institutional review board at each study site.Written informed consent was obtained from all partic-ipants prior to study entry.ResultsThe mean age of the sample was 67.4 years [SD 9.7],44% were females, and 40% of the participants had threestudy visits. The overall prevalence of undiagnosed COPDwas 26%; 95% had mild to moderate disease based on theGOLD spirometric grading system [1]. The regressionmodel indicated that reporting wheeze, dyspnea, and coughon most days were independently related to the presence ofundiagnosed COPD (Table 1). However, symptoms alonehad poor performance in identifying patients with undiag-nosed COPD. Almost all symptoms, regardless of the sever-ity, had sensitivities and positive predictive values less than50% (Table 2).If screening was applied to the general population (a“blind” screening approach), the NNS to detect oneCOPD case would be 3.8. If screening was restricted toindividuals who reported symptoms, the NNS would be3.0, but compared with the “blind” approach, an additional17% of individuals with persistent airflow limitation wouldbe missed.The ROC curves indicated that wheeze had the bestperformance among all symptoms (AUC = 0.62), followedby cough and dyspnea (each AUC = 0.57), and phlegm(AUC= 0.55). The total number of symptoms performedmarginally better than any symptom alone (AUC= 0.64).AUC improved by ~ 0.06 when each of the symptoms wascombined with smoking history (measured as pack-years),resulting in an AUC of 0.67 for a model that includedwheeze and pack-years. The model that included all indi-vidual symptoms and covariates improved the AUC to0.71 (Fig. 1).DiscussionOur results indicate that symptoms are relatively poor atdiscriminating undiagnosed COPD patients from non-COPD subjects in a general population. The use of symp-toms for screening is unlikely to significantly improve thediagnostic yield compared with “blind” screening in thegeneral population. These data highlight the apparentparadox in finding strong associations between symptomsTable 1 Association between symptoms and other patientcharacteristics with the odds of having undiagnosed COPD(v. no COPD)OR 95% CI p-valueCough1 (vs.0) 1.08 0.86–1.36 0.492 (vs.0) 1.38 0.93–2.04 0.113 (vs.0) 1.35 1.09–1.68 0.01Wheeze1 (vs.0) 1.34 1.03–1.73 0.032 (vs.0) 1.74 1.38–2.20 < 0.013 (vs.0) 1.77 0.93–3.38 0.08Phlegm1 (vs.0) 1.21 0.89–1.66 0.232 (vs.0) 1.35 0.77–2.36 0.303 (vs.0) 1.24 0.93–1.65 0.14Dyspnea2 (vs.1) 1.26 1.05–1.50 0.013 (vs.1) 2.05 1.30–3.22 < 0.014 (vs.1) 1.52 0.76–3.03 0.245 (vs.1) 3.35 1.88–5.97 < 0.01Agea 0.83 0.74-0.94 < 0.01Female (vs. male) 1.03 0.81–1.30 0.82BMIa 0.78 0.68-0.89 < 0.01Caucasian (vs. non-Caucasian) 2.39 1.28–4.48 0.01Comorbidities1 comorbidity (vs. 0) 0.93 0.76–1.14 0.512 comorbidities (vs. 0) 0.75 0.48–1.18 0.21Smoking between visits (vs. no) 1.16 0.96–1.41 0.13Smoking Pack-Years20–40 (vs. < 20) 2.09 1.53–2.86 < 0.01> 40 (vs. < 20) 3.09 2.22–4.32 < 0.01BMI body mass index, CI confidence interval, OR odds ratioaVariables were converted to z-scores in the regression modelJohnson et al. Respiratory Research  (2018) 19:147 Page 2 of 5Table 2 Prevalence of each symptom severity category in the whole population (‘Prev’) across all study visits, and the prevalence ofundiagnosed COPD (‘COPD+’) within that symptom severity category. Sensitivity (‘TP’), specificity (‘TN’), positive predictive value(‘PPV’), and negative predictive value (‘NPV’) of each symptom when used alone to classify undiagnosed COPD (v. no COPD) usingdifferent severity thresholdsSymptom severity Cougha Wheezeb Phlegmc Dyspnead Total SymptomsePrev, COPD+ Prev, COPD+ Prev, COPD+ Prev, COPD+ Prev, COPD+0 72, 23% 78, 21% 83, 24% 70, 23% 45, 17%1 12, 29% 10, 42% 6, 30% 26, 33% 29, 28%2 3, 35% 10, 49% 1, 47% 3, 56% 16, 33%3 13, 42% 2, 51% 9, 39% 1, 30% 7, 50%4 < 1, 67% 4, 52%TP, TNPPV, NPVTP, TNPPV, NPVTP, TNPPV, NPVTP, TNPPV, NPVTP, TNPPV, NPV0 vs. > 0 37, 76%36, 77%39, 84%46, 79%24, 85%37, 76%40, 74%35, 77%71, 50%34, 83%≤1 vs. > 1 24, 88%41, 76%23, 91%49, 77%16, 91%40, 75%7, 98%,52, 75%41, 78%40, 79%≤2 vs. > 2 21, 90%42, 76%4, 99%51, 74%14, 92%39, 75%1, 99%,38, 74%21, 93%51, 77%≤3 vs. > 3 < 1, > 99%67, 74%7, 98%52, 75%Prev prevalence, COPD+ undiagnosed COPD, TP true positive (sensitivity), TN true negative (specificity), PPV positive predictive value, NPV negative predictive valuePatients were asked, since your last visit:a1) Do you usually cough when you don’t have a cold? 1a) Are there months you cough most days? 1b) Do you cough most days for as much as 3 months?b2) Have you had any wheezing or whistling in your chest? 2a) Do you only have wheezing or whistling when you have a cold? 2b) Have you had anattack of wheezing or whistling that made you short of breath?c3) Do you usually have phlegm in your chest when you don’t have a cold? 3a) Are there months you have phlegm most days? 3b) Do you hav ephlegm most days for as many as 3 months?dScores on the Medical Research Council (MRC) Dyspnea scale are subtracted by 1eThe sum of the number of individual symptoms that participants reportedFig. 1 Receiver operating characteristic (ROC) curves for a model with all of the symptoms and covariates included (‘All variables’), as well as foreach of the symptoms individually and the total number of symptoms reported by study participants (‘Total symptoms’)Johnson et al. Respiratory Research  (2018) 19:147 Page 3 of 5and undiagnosed COPD, yet poor diagnostic performancewhen symptoms are used to diagnose these “hidden”COPD cases. This observation is consistent with thewell-established notion that a predictor can be strongly as-sociated with an outcome while still being a poor classifierof that outcome [9]. Association models are useful forevaluating relationships at the population level, butclassification models are more relevant to decisions at anindividual level, specifically whether a test (e.g., the presenceof a symptom) can detect the underlying disease state(e.g., undiagnosed COPD).Previous studies have evaluated the merits of opportun-istic case detection based on patient characteristics at thepoint of care [10, 11]. They generally found that patientcharacteristics and symptoms have modest capacity in de-tecting undiagnosed COPD [10–12]. A unique feature ofour study is its population-based sample, which providesnew evidence on whether the yield of population-basedscreening can be improved if symptoms are considered inthe inclusion criterion (e.g., through advertisements forreferral of symptomatic individuals for lung functiontesting). Although the costs and benefits of early interven-tion between symptomatic and asymptomatic patients werenot considered here, our results do not support the use ofsymptoms in “case finding” for COPD. Symptoms shouldbe used in conjunction with other characteristics such aspack-years of smoking to improve the diagnostic perform-ance. Future studies should evaluate the cost-effectivenessof this approach considering the long-term outcomes asso-ciated with earlier diagnosis of COPD.AbbreviationsAUC: Area Under the Curve; CanCOLD: Canadian Cohort of Obstructive LungDisease; COPD: Chronic Obstructive Pulmonary Disease; FEV1: ForcedExpiratory Volume in 1 s; FVC: Forced Vital Capacity; GOLD: Global Initiativefor chronic Obstructive Lung Disease; NNS: Number Needed to Screen;ROC: Receiver Operating Characteristic curves; SD: Standard DeviationAcknowledgementsThe authors thank the men and women who participated in the study andindividuals in the CanCOLD Collaborative Research Group.Members of the CanCOLD Collaborative Research Group are as follows. ExecutiveCommittee: Jean Bourbeau (McGill University, Montreal, Canada); Wan C. Tan, J.Mark FitzGerald; Don Sin (UBC, Vancouver, Canada); Darcy Marciniuk (Universityof Saskatoon, Saskatoon, Canada); Dennis E. O’Donnell (Queen’s University,Kingston, Canada); Paul Hernandez (Dalhousie University, Halifax, Canada);Kenneth R. Chapman (University of Toronto, Toronto, Canada); Robert Cowie(University of Calgary, Calgary, Canada); Shawn Aaron (University of Ottawa,Ottawa, Canada); F. Maltais (University of Laval, Quebec City, Canada). InternationalAdvisory Board: Jonathon Samet (Keck School of Medicine of USC, Los Angeles,CA); Milo Puhan (John Hopkins School of Public Health, Baltimore, MD); QutaybaHamid (McGill University, Montreal, Canada); James C. Hogg (UBC James HoggResearch Center, Vancouver, Canada). Operations Center: Jean Bourbeau (PI), CaroleJabet, Palmina Mancino, (McGill University, Montreal, Canada); Wan C. Tan (co-PI),Don Sin, Sheena Tam, Jeremy Road, Joe Comeau, Adrian Png, Harvey Coxson,Jonathon Leipsic, Cameron Hague (University of British Columbia James HoggResearch Center, Vancouver, Canada). Economic Core: Mohsen Sadatsafavi(University of British Columbia, Vancouver, Canada). Public Health Core: Teresa To,Andrea Gershon (University of Toronto, Toronto, Canada). Data Management andQuality Control: Wan C. Tan, Harvey Coxson (UBC, Vancouver, Canada); JeanBourbeau, Pei Zhi Li, Zhi Song, Yvan Fortier, Andrea Benedetti, Dennis Jensen(McGill University, Montreal, Canada). Field Centers: Wan C. Tan (Vancouver PI),Christine Lo, Sarah Cheng, Cindy Fung, Nancy Haynes, Junior Chuang, Licong Li,Selva Bayat, Amanda Wong, Zoe Alavi, Catherine Peng, Bin Zhao, NathalieScott-Hsiung, Tasha Nadirshaw (UBC James Hogg Research Center, Vancouver,Canada); Jean Bourbeau (Montreal PI), Palmina Mancino, David Latreille,Jacinthe Baril, Laura Labonté (McGill University, Montreal, Canada); KennethChapman (Toronto PI), Patricia McClean, Nadeen Audisho (University ofToronto, Toronto, Canada); R. Cowie and B. Walter (Calgary PI), Ann Cowie,Curtis Dumonceaux, Lisette Machado (University of Calgary, Calgary,Canada); Paul Hernandez (Halifax PI), Scott Fulton, Kristen Osterling(Dalhousie University, Halifax, Canada); Shawn Aaron (Ottawa PI), KathyVandemheen, Gay Pratt, Amanda Bergeron (University of Ottawa, Ottawa,Canada); Denis O’Donnell (Kingston PI), Matthew McNeil, Kate Whelan(Queen’s University, Kingston, Canada); François Maltais (Quebec PI),Cynthia Brouillard (Université Laval, Quebec City, Canada); Darcy Marciniuk(Saskatoon PI), Ron Clemens, Janet Baran (University of Saskatoon,Saskatoon, Canada).FundingThe current study was funded by a Canadian Lung Association Breathing asOne Studentship Award and the Canadian Institutes of Health Research(application number 142238). The Canadian Cohort Obstructive Lung Disease(CanCOLD) study is currently funded by the Canadian Respiratory ResearchNetwork (CRRN); industry partners: Astra Zeneca Canada Ltd.; BoehringerIngelheim Canada Ltd.; GlaxoSmithKline Canada Ltd.; and Novartis. Researchersat RI-MUHC Montreal and Icapture Centre Vancouver lead the project. Previousfunding partners are the CIHR (CIHR/Rx&D Collaborative Research ProgramOperating Grants 93326); the Respiratory Health Network of the Fonds dela recherche en santé du Québec (FRSQ); industry partners: Almirall; MerckNycomed; Pfizer Canada Ltd.; and Theratechnologies. The funders had no rolein study design, data collection and analysis, or preparation of the manuscript.Availability of data and materialsThe data analysed in the current study are not publicly available but may bemade available from the CanCOLD Research Group upon reasonable request.Authors’ contributionsWT and JB are co-Principal Investigators of the CanCOLD study. MS and KJformulated the current study idea. KJ performed all data analyses and wrotethe first draft of the manuscript. All authors contributed to interpretation offindings, critically commented on the manuscript and approved the finalversion. MS is the guarantor of the manuscript.Ethics approval and consent to participateEthics approval for CanCOLD was obtained from the relevant institutionalreview board at each study site. Written informed consent was obtainedfrom all participants prior to study entry.Consent for publicationNot applicable.Competing interestsThe authors declare that they have no competing interests.Publisher’s NoteSpringer Nature remains neutral with regard to jurisdictional claims inpublished maps and institutional affiliations.Author details1Respiratory Evaluation Sciences Program, Collaboration for OutcomesResearch and Evaluation, Faculty of Pharmaceutical Sciences, University ofBritish Columbia, Vancouver, Canada. 2Centre for Heart Lung Innovation (theJames Hogg Research Centre), St. Paul’s Hospital, Vancouver, Canada.3Respiratory Epidemiology and Clinical Research Unit, McGill University,Montreal, Canada. 4Centre for Clinical Epidemiology and Evaluation,Vancouver Coastal Health Institute, Vancouver, Canada. 5Institute for Heartand Lung Health, Department of Medicine, The University of BritishColumbia, Vancouver, Canada.Johnson et al. Respiratory Research  (2018) 19:147 Page 4 of 5Received: 3 June 2018 Accepted: 26 July 2018References1. From the Global Strategy for the Diagnosis, Management and Prevention ofCOPD. Global Initiative for Chronic Obstructive Lung Disease (GOLD)[Internet]. 2017 [cited 2017 May 1]. Available from: http://goldcopd.org2. Lamprecht B, Soriano JB, Studnicka M, Kaiser B, Vanfleteren LE, Gnatiuc L, etal. Determinants of underdiagnosis of COPD in national and internationalsurveys. Chest. 2015;148:971–85.3. Qaseem A, Wilt TJ, Weinberger SE, Hanania NA, Criner G, van der Molen T,et al. Diagnosis and management of stable chronic obstructive pulmonarydisease: a clinical practice guideline update from the American College ofPhysicians, American College of Chest Physicians, American Thoracic Society,and European Respiratory Society. Ann Intern Med. 2011;155:179–91.4. US Preventive Services Task Force (USPSTF), Siu AL, Bibbins-Domingo K,Grossman DC, Davidson KW, Epling JWJ, et al. Screening for ChronicObstructive Pulmonary Disease: US Preventive Services Task ForceRecommendation Statement. JAMA. 2016;315:1372–7.5. Kaplan A, Thomas M. Screening for COPD: the gap between logic andevidence. Eur Respir Rev. 2017;26:160113.6. Johnson KM, Bryan S, Ghanbarian S, Sin DD, Sadatsafavi M. Characterizingundiagnosed chronic obstructive pulmonary disease: a systematic reviewand meta-analysis. Respir Res. 2018;19:26.7. Tan WC, Bourbeau J, Hernandez P, Chapman KR, Cowie R, FitzGerald JM, etal. Exacerbation-like respiratory symptoms in individuals without chronicobstructive pulmonary disease: results from a population-based study. Thorax.2014;69:709–17.8. Bourbeau J, Tan WC, Benedetti A, Aaron SD, Chapman KR, Coxson HO, et al.Canadian cohort obstructive lung disease (CanCOLD): fulfilling the need forlongitudinal observational studies in COPD. COPD. 2014;11:125–32.9. Pepe MS. Evaluating technologies for classification and prediction inmedicine. Stat Med. 2005;24:3687–96.10. Stanley AJ, Hasan I, Crockett AJ, van Schayck OCP, Zwar NA. COPD diagnosticquestionnaire (CDQ) for selecting at-risk patients for spirometry: a cross-sectional study in Australian general practice. NPJ Prim Care Respir Med.2014;24:14024.11. Martinez FJ, Raczek AE, Seifer FD, Conoscenti CS, Curtice TG, D’Eletto T, et al.Development and initial validation of a self-scored COPD populationscreener questionnaire (COPD-PS). COPD J Chronic Obstr Pulm Dis.2008;5:85–95.12. Gershon AS, Hwee J, Chapman KR, Aaron SD, O’Donnell DE, Stanbrook MB,et al. Factors associated with undiagnosed and overdiagnosed COPD. EurRespir J. 2016;48:561–4.Johnson et al. Respiratory Research  (2018) 19:147 Page 5 of 5


Citation Scheme:


Citations by CSL (citeproc-js)

Usage Statistics



Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            async >
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:


Related Items