Articles. Funding National Institute for Health Research.

Transcription

1 Targeted case finding for chronic obstructive pulmonary disease versus routine practice in primary care (TargetCOPD): a cluster-randomised controlled trial Rachel E Jordan, Peymané Adab, Alice Sitch, Alexandra Enocson, Deirdre Blissett, Sue Jowett, Jen Marsh, Richard D Riley, Martin R Miller, Brendan G Cooper, Alice M Turner, Kate Jolly, Jon G Ayres, Shamil Haroon, Robert Stockley, Sheila Greenfield, Stanley Siebert, Amanda J Daley, K K Cheng, David Fitzmaurice Lancet Respir Med 2016; 4: Published Online July 18, S (16) See Comment page 679 Institute of Applied Health Research, University of Birmingham, Edgbaston, Birmingham, UK (R E Jordan PhD, Prof P Adab MD, A Sitch MSc, A Enocson PhD, D Blissett MSc, S Jowett PhD, J Marsh PhD, Prof M R Miller MD, Prof K Jolly PhD, Prof J G Ayres MD, S Haroon PhD, Prof S Greenfield PhD, A J Daley PhD, Prof K K Cheng FMedSci, Prof D Fitzmaurice PhD); Research Institute for Primary Care and Health Sciences, Keele University, Keele, Staffordshire, UK (Prof R D Riley PhD); Lung Investigation Unit, University Hospitals Birmingham, NHS Foundation Trust, Birmingham, UK (B G Cooper PhD, R Stockley DSc); Queen Elizabeth Hospital Research Laboratories, Mindelsohn Way, Birmingham, UK (A M Turner PhD); and Business School, University of Birmingham, Birmingham, UK (Prof S Siebert PhD) Correspondence to: Dr Rachel E Jordan and Prof Peymané Adab, Public Health Building, Institute of Applied Health Research, University of Birmingham, Edgbaston, Birmingham B15 2TT, UK r.e.jordan@bham.ac.uk or P.adab@bham.ac.uk Summary Background Many individuals with chronic obstructive pulmonary disease (COPD) remain undiagnosed worldwide. Health-care organisations are implementing case-finding programmes without good evidence of which are the most effective and cost-effective approaches. We assessed the effectiveness and cost-effectiveness of two alternative approaches to targeted case finding for COPD compared with routine practice. Methods In this cluster-randomised controlled trial, participating general practices in the West Midlands, UK, were randomly assigned (1:1), via a computer-generated block randomisation sequence, to either a targeted case-finding group or a routine care group. Eligible patients were ever-smokers aged years without a previously recorded diagnosis of COPD. Patients in the targeted case-finding group were further randomly assigned (1:1) via their household to receive either a screening questionnaire at the general practitioner (GP) consultation (opportunistic) or a screening questionnaire at the GP consultation plus a mailed questionnaire (active). Respondents reporting relevant respiratory symptoms were invited for post-bronchodilator spirometry. Patients, clinicians, and investigators were not masked to allocation, but group allocation was concealed from the researchers who performed the spirometry assessments. Primary outcomes were the percentage of the eligible population diagnosed with COPD within 1 year (defined as post-bronchodilator forced expiratory volume in 1 s [FEV 1 ] to forced vital capacity [FVC] ratio <0 7 in patients with symptoms or a new diagnosis on their GP record) and cost per new COPD diagnosis. Multiple logistic and Poisson regression were used to estimate effect sizes. Costs were obtained from the trial. This trial is registered with ISRCTN, number ISRCTN Findings From Aug 10, 2012, to June 22, 2014, eligible patients from 54 diverse general practices were randomly assigned and completed the trial. At 1 year, 1278 (4%) cases of COPD were newly detected in eligible patients in the targeted case-finding group compared with 337 (1%) cases in patients in the routine care group (adjusted odds ratio [OR] 7 45 [95% CI ], p<0 0001). The percentage of newly detected COPD cases was higher in the active case-finding group (822 [5%] of ) than in the opportunistic case-finding group (370 [2%] of ; adjusted OR 2 34 [ ], p<0 0001; adjusted risk difference 2 9 per 100 patients [95% CI ], p<0 0001). Active case finding was more cost-effective than opportunistic case finding ( 333 vs 376 per case detected, respectively). Interpretation In this well established primary care system, routine practice identified few new cases of COPD. An active targeted approach to case finding including mailed screening questionnaires before spirometry is a cost-effective way to identify undiagnosed patients and has the potential to improve their health. Funding National Institute for Health Research. Introduction Chronic obstructive pulmonary disease (COPD) is a common long-term respiratory disorder with high health-care and societal burden, 1 with estimated annual costs of US$50 billion (2010) to the US 1 and 48 6 billion (2011) to the EU economies. 2 Worldwide, the diagnosed prevalence of COPD is estimated to be about 5% of adults older than 40 years, 3 but at least an equivalent number of people with significant symptomatic COPD remain undiagnosed and could potentially benefit from effective interventions. 4,5 There is a worldwide drive to identify these missing millions 6 9 and many health-care organisations are implementing case-finding programmes, although without good evidence of which are the most effective and cost-effective approaches. There are several reports of uncontrolled studies of case finding, with heterogeneous populations, interventions, and case definitions. 10 However, only one published randomised controlled trial, within family medical practices in the USA, has assessed a systematic approach to identifying previously undiagnosed clinical COPD compared with usual care. 11 This study showed that initial opportunistic administration of a screening Vol 4 September 2016

2 Research in context Evidence before this study Chronic obstructive pulmonary disease (COPD) is substantially underdiagnosed worldwide. Mass population screening with spirometry is not recommended, although early identification of undiagnosed patients with clinical disease is generally supported because COPD has substantial patient, health-care, and societal burden, which could potentially be lessened through timely diagnosis. We undertook a comprehensive systematic review of the effectiveness of case-finding studies, published in 2015, using synonyms for the search terms COPD, case-finding screening tests in MEDLINE and Embase with no language restrictions. We found 39 published studies, but few included a comparator group. Only one randomised controlled trial has compared a case-finding approach with routine care to identify clinically significant COPD, identifying that opportunistic administration of a screening questionnaire at practice visits was twice as effective as routine care. Non-comparative studies suggest that high yields could be achieved from active case finding with mailed questionnaires to high-risk patients. No previous trial has compared the clinical and cost-effectiveness of alternative approaches to case finding in a real-life primary care setting. Added value of this study To our knowledge, TargetCOPD is the first randomised controlled trial to assess the effectiveness and cost-effectiveness of active case finding for COPD compared with opportunistic case finding and to compare these approaches with routine care. In 54 general practices, we found that when targeting eversmokers aged years, active case finding was more than twice as effective as opportunistic case finding, and that together, these targeted approaches were more than seven times as effective as current routine care. Active case finding was also more cost-effective than opportunistic case finding. Implications of all the available evidence In a real-life setting, active case finding comprising an initial mailed questionnaire followed by spirometry should be recommended for identifying undiagnosed COPD in primary care. An important proportion of these previously undiagnosed patients have significant breathlessness, and earlier identification and management with effective treatments including inhalers and pulmonary rehabilitation has the potential to improve their health. This trial provides the evidence for guidelines, which was previously lacking. questionnaire was twice as effective as usual care in identifying new cases of COPD. Other uncontrolled studies have suggested that sending a screening questionnaire by post before spirometry assessment is a promising alternative method, 10 but the comparative effectiveness and cost-effectiveness of the two approaches have not been tested. Modelling of observational data has suggested that an active approach (including both mailed questionnaires and opportunistic administration) could be twice as effective as an opportunistic-only approach, and identify patients with significant potential to benefit from treatment. 4 This large, pragmatic, cluster-randomised controlled trial investigated the effectiveness and costeffectiveness of these two methods of identifying undiagnosed COPD in a UK primary care setting. Methods Study design and participants TargetCOPD was a pragmatic cluster-randomised controlled trial comparing an active and an opportunistic targeted approach to case finding for undiagnosed COPD versus routine practice in primary care. Details of our study design have been published elsewhere. 12 UK ethical and research governance approvals were obtained. All 354 general practices (family medical practices) in the Birmingham and Black Country region of the West Midlands, UK, were invited to take part. Automated computer searches of electronic health records (appendix) identified ever-smokers aged years without a previous diagnosis of COPD. Physicians then excluded at their discretion individuals they considered unsuitable (eg, individuals who were unable to give informed consent, have terminal illness, or were pregnant). The remaining patients were classed as approved to receive a screening questionnaire. Randomisation and masking Participating practices were randomly assigned in three blocks, with variable block size, to targeted case finding or routine care groups using a computer-generated randomisation sequence from a published algorithm for cluster-randomised trials, which balanced key practice characteristics (deprivation [Index of Multiple Deprivation score of the practice], ethnic origin [percentage of white patients], practice list size, age [percentage aged 45 years], and proportion of patients on the COPD register 13 ) with stratification for linked practices. 12,14 Allocation was done by the trial statistician. We chose a cluster-randomised design so that healthcare staff in the routine care group did not change their usual practice for control patients because of raised awareness about case finding, and because we wanted to assess a whole practice organisational approach that could be undertaken in a real-life setting. Practices were unaware of their allocation until baseline patient data had been provided and they had agreed to commence the study. Within the targeted case-finding group, individual households were further randomly assigned, via block randomisation (block size 20) using a random See Online for appendix Vol 4 September

3 For more on the EQ-5D see html number generator (automatically applied when participants from each practice were uploaded to the trial database), to active or opportunistic case finding. This approach ensured that patients from the same household were allocated the same intervention to avoid contamination. The allocation ratio was 1:1 for randomisation of both general practices and households. Data processing was computerised and research assistants who performed the spirometry assessments were unaware of group allocation. Procedures The case-finding intervention was applied for 12 months in each practice, with a staggered start (Aug 10, 2012, to June 22, 2013). 12 practices (six in the targeted casefinding group, six in the routine care group) formed an internal pilot phase to test procedures. In the targeted case-finding group, eligible patients were allocated to receive a brief screening questionnaire (appendix) in either of two ways: (1) patients electronic health records were flagged to prompt the general practitioner (GP) or general practice staff to hand out the questionnaire during any visit to their general medical practice (opportunistic case finding); and (2) in addition to opportunistic distribution, the screening questionnaire was mailed to patients homes (with prepaid return envelope), with reminders sent after 4 and 8 weeks (active case finding). Electronic health record prompts were removed after receipt of the questionnaire. All participants in the targeted case-finding group were provided with patient information leaflets and a standard letter of invitation from their GP and respondents were considered to have implied consent by returning their questionnaire with their personal details. The questionnaire was piloted by our patient advisory group. In the routine care group, GPs were expected to follow UK guidance, which recommends that patients older than 35 years should be investigated for COPD with spirometry if they present with chronic cough or phlegm, exertional breathlessness, or wheeze. 15 National Quality and Outcomes Framework (QOF) data show that more than 80% of patients newly diagnosed with COPD receive confirmatory spirometry within 12 months. 13 Consent was not required from patients in the routine care group because we did not obtain patient-identifiable data. Patients in the targeted case-finding group who responded to the screening questionnaire and reported any of the following respiratory symptoms (chronic cough or phlegm for 3 months for at least 2 years, wheeze in the past 12 months, or dyspnoea of MRC grade 2 or more) were invited to attend their general practice for confirmatory spirometry. Attendees provided signed informed consent at the start of the assessment. Post-bronchodilator spirometry was done according to American Thoracic Society (ATS) and European Respiratory Society (ERS) 2005 guidelines 16 using an ultrasonic flow head (Spiroson-AS, ndd Medical Technologies, Zurich, Switzerland) with bespoke software (designed by MRM) by blinded, trained research assistants with immediate visual quality control monitoring and feedback. Every trace was over-read and quality of blows graded according to standard criteria. 17 Patients also completed a short questionnaire to ascertain out-of-pocket expenses for attendance (appendix) and health status (EuroQol Group 5-Dimension Self-Report Questionnaire [EQ-5D]). Patients height was measured to the nearest cm with a portable stadiometer (or estimated using arm-span if necessary) to compute predicted FEV 1 values. Outcomes Primary outcomes were the percentage of the eligible population diagnosed with COPD within 1 year, and cost per additional case identified in each group, comparing (1) targeted case finding versus routine care, and (2) active versus opportunistic case finding. Secondary outcomes were feasibility (process measures including uptake and resource needs) and efficiency (number needed to target to identify one person likely to benefit). New cases in the targeted case-finding group at the end of the 12-month period were identified either through the trial spirometry assessment visit (airflow obstruction defined as post-bronchodilator forced expiratory volume in 1 s [FEV 1 ] to forced vital capacity [FVC] ratio <0 7 in the presence of respiratory symptoms, in line with recommended UK guidelines 15 ) or through the electronic health records by automated searches with clinical (Read) codes (appendix), to ensure fair comparison with the routine care group. New cases in the routine care group were identified from the electronic health record only because patients in this group did not receive a trial spirometry assessment. In sensitivity analyses, we used the lower limit of normal definition (Global Lung Function Initiative [GLI] 2012 equations 18 ) to define airflow limitation. Additional data from the electronic health records were collected at baseline and 12 months to ascertain patient characteristics, medical conditions, and study outcomes. To estimate opportunistic questionnaire distribution in the targeted case-finding group, 30 patients from each of the opportunistic and active case-finding groups in each practice were randomly sampled to ascertain whether electronic health record prompts had been removed. Statistical analysis The significance level for multiple testing was adjusted to provide a total of 5% significance level across the two primary effectiveness outcomes: 0 25% for the opportunistic versus active case-finding comparison, and 4 75% for the targeted case-finding versus routine care comparison. The sample size calculation was computed from estimates presented in our published model of case finding, 4 which used data from the Health Survey for England and published literature to estimate Vol 4 September 2016

4 values for different stages of the process (provided in detail in our protocol 12 ). For the opportunistic versus active case-finding comparison, we assumed that 50% of patients allocated to the active case-finding group would respond; of the remaining patients, 91% would visit their GP at least once in 12 months, 50% would be offered the questionnaire, of whom 90% would fill it out. In the opportunistic case-finding group, we assumed that 50% of patients would be offered the questionnaire, of whom 90% would complete it. Of all respondents to the questionnaire in both case-finding groups, we assumed 48% would report symptoms and be invited to spirometry, of whom 70% would attend, and 17% of the attendees would have COPD. This finding would lead to yields of 2 3% in the opportunistic and 4 0% in the active case-finding groups. At a 0 25% significance level, 3904 patients per case-finding group were required to detect this difference with 90% power. 4,12 For the targeted case-finding versus routine care comparison, the proportion of new COPD cases detected in the targeted case-finding group (averaged across both active and opportunistic subgroups) was assumed to be 3 15%. The proportion of new COPD cases detected in the routine care group was assumed to be 0 75%. At the 4 75% significance level, with 80% power, these assumptions led to an unadjusted sample size required of 545 per group, to detect a difference of 2 4% between targeted case-finding and routine care groups. We expected about 40% of a practice population to be aged years with 57% of this age group being ever-smokers without a previous diagnosis of COPD. Assuming, therefore, a conservative 1000 eligible patients per practice of average list size 6000, and adjusting for clustering of patients within practices, assuming a conservative intra-cluster correlation coefficient (ICC) of 0 05, the sample size required was per group, equivalent to 28 practices per group. All analyses were done with Stata 13. The process measures such as questionnaire response rates are presented using simple descriptive statistics. For the primary outcomes, all models used outcomes for individual participants and were adjusted for practicelevel deprivation (IMD), patient ethnic origin, and age, as predefined in the protocol. 12 To compare opportunistic and active case-finding groups, logistic regression was used including a fixed effect for each practice (adjustment for household made little difference to the results and is therefore not presented). Adjusted relative risks and risk differences were also estimated using Poisson regression with robust standard errors, 19 and from these, numbers needed to target were computed. The analysed population included only those patients approved by their GP to receive a screening questionnaire. To compare targeted case-finding and routine care groups, the same approach was taken but multilevel regression models were used (logistic and Poisson with Assumption Source Cost per patient ( ) Searches of patient records, screening, and flagging notes with prompts Running the search 20 min practice manager time per practice Estimate from trial 0 01 Screening patient list 1 min GP time per patient selected 3 20 report 20 Attaching prompts 15 s administration time per patient record Estimate from trial 0 94 Total 4 15 Invitation letters and questionnaires Drafting and approving letter Administering mailmerge ( 3) Administering postal questionnaires Stationery cost for postal questionnaire Total per postal questionnaire/reminder Completing the questionnaire in clinic Processing questionnaires and allocating appointment times Appointment booking Book appointments Proportion of patients that cancelled or rebooked 1 h practice manager time per practice per mail-out 2 h administration time per practice per mail-out 1 8 min administration time per letter sent, plus stationery cost Six sheets of headed note paper and letter head, stamp, envelope, and prepaid return envelope robust standard errors) with a separate random effect for targeted and routine practices, to allow for clustering and between-practice heterogeneity in the underlying log odds of COPD case detection for each group. In this case, the analysed population included all eligible patients meeting the entry criteria. An estimate of the ICC was obtained by fitting a multilevel linear null model. This trial is registered with ISRCTN, number ISRCTN report report report report Six sheets of headed notepaper and 1 min of GP time per questionnaire completed in clinic 30 min administration time per questionnaire completed 10 min plus stamp, letter, and envelope for 100% of appointments and text message for 49% of appointments All booked appointments divided by appointment attended report, 20 estimate from trial report, 20 estimate from trial report, 20 estimate from trial Total per appointment 8 25 Spirometry Staff costs 50 min clinical support worker time, 3 min Estimate from trial reception time Training 4 weeks training, plus 2-day workshop, 1-day Estimate from trial 4 33 refresher course; annual cost, assuming 372 tests per year and 3-yearly reassessment Room costs 50 min per appointment (allowing for Estimate from trial patients who did not attend) Travel costs Average 2 85 miles per attendance, 0 40 Estimate from trial 1 14 per mile Equipment Use of spirometer and laptop; single use of Estimate from trial mouthpiece, spacer, and salbutamol Total GP=general practitioner. NIHR=National Institute for Health Research. Table 1: Screening process assumptions and costs Vol 4 September

5 354 general practices invited to take part 298 did not participate 56 agreed to participate 2 withdrew 54 randomly assigned Cluster randomisation Targeted case finding 27 general practices patients* Routine practice 27 general practices patients randomised by household patients assigned to opportunistic case finding patients assigned to active case finding 6 withdrew data 16 withdrew data completed trial completed trial (89%) attended GP surgery 9570 did not respond 49% (estimated) GP/nurse asked questions 1973 (13%) responded to questionnaire 5042 (33%) responded to mailed questionnaire 748 (5%) responded to opportunistic questionnaire 18 (<1%) responded by unknown means Figure: Trial profile Note that search terms were applied to electronic health records to identify ever-smokers, but any patients found to be never-smokers were included in the primary analysis to maintain the pragmatic approach. GP=general practitioner. *Of patients, were approved to be allocated a screening questionnaire by GPs. Percentage of all approved patients randomised in the nested trial. Percentage of all eligible in cluster comparison. See appendix for a more detailed trial profile (55%) positive respiratory symptoms 543 (50%) attended for spirometry 189 diagnosed by trial 370 (2% ) diagnosed with COPD 181 diagnosed by GP 662 diagnosed by trial 160 diagnosed by GP 1278 (4% ) patients diagnosed in targeted case-finding group 3264 (56%) positive respiratory symptoms 2065 (63%) attended for spirometry 822 (5% ) diagnosed with COPD 337 (1% ) patients diagnosed by GP Vol 4 September 2016

6 Economic analyses The base-case analysis estimated the cost per additional case detected with active and opportunistic case finding compared with routine care, taking a health-care perspective and using multilevel modelling in line with the main trial analysis. 12 We calculated incremental costeffectiveness ratios by dividing the difference in mean per patient costs by the difference in mean patient Targeted case finding Routine care Practice characteristics Total practices Practice list size 5762 (3482) 5811 (3451) IMD score 35 2 ( ) 36 0 ( ) White patients (%) 80 5% (20 5) 79 1% (21 9) Age 45 years (%) 38 5% ( ) Diagnosed prevalence of COPD (%) Patient characteristics 38 2% ( ) 1 6% (0 6) 1 7% (0 7) Total patients * Sex Male (54%) (52%) Female (46%) (48%) Age (years) 55 3 ( ) Ethnic origin 55 6 ( ) White (55%) (57%) Mixed 265 (1%) 223 (1%) Asian 1742 (5%) 2044 (5%) African Caribbean 1220 (4%) 1025 (2%) Other 591 (2%) 353 (1%) Missing (33%) (34%) Smoking status Never-smoker 6580 (20%) (26%) Ex-smoker (42%) (40%) Current smoker (36%) (34%) Missing 428 (1%) 173 (<1%) Comorbidities Asthma 3117 (11%) 3834 (9%) Ischaemic heart disease 1881 (7%) 2341 (6%) Chest infection in previous 3 years 2694 (10%) 4377 (11%) Number of patients sharing household (68%) (58%) (29%) (38%) 3 or more 722 (2%) 1664 (4%) Data are mean (SD), median (IQR), or n (%), unless otherwise indicated. Practice characteristics were obtained from ward and GP practice-level statistics. IMD=Index of Multiple Deprivation. *Results exclude 22 randomised patients who withdrew their data. From 22 targeted case-finding practices ( patients) and 26 routine care practices (41 100) with available data. Patients reside in the same household. Table 2: Baseline characteristics of randomised general practices outcomes (cases detected). We used a bottom-up approach to cost the case-finding strategies, using data collected within the trial (table 1). Standard National Health Service (NHS) unit costs and trial-specific costs were applied to calculate the costs of each process (appendix) using 2013 prices. Spirometry costs were calculated assuming that the assessment would be provided by an NHS outreach service. Reusable equipment was amortised over 5 years, and training costs spread over 3 years, using a discount rate of 3 5%. Costs incurred by patients and their families to attend spirometry assessments (including travel, care, and cost of time off work) were included in a sensitivity analysis and obtained by patient self-report while attending the assessment. Time spent away from paid work was valued at the average national wage rate. Costs incurred after COPD was confirmed were not considered in this analysis. Sensitivity analyses were also done to estimate the cost-effectiveness of different models of care and alternative case-finding scenarios. We considered three alternative models of care, including a GP-led model, a community-led model, and a secondary care model (appendix). Role of the funding source The funder of the study had no role in study design, data collection, data interpretation, or writing of the report. The corresponding authors had full access to all the data in the study and had final responsibility for the decision to submit for publication. Targeted case finding* Routine care Total eligible patients Deceased 255 (1%) 527 (1%) Left practice 1972 (6%) 2492 (6%) Primary outcome (NICE criteria) 15 COPD identified by trial 851 (3%) COPD additionally identified 427 (1%) 337 (1%) by GP Total COPD newly diagnosed 1278 (4%) 337 (1%) Crude risk difference (95% CI) ( ) Crude rate ratio (95% CI) 4 86 ( ) Adjusted odds ratio (95% CI) ; p value 7 45 ( ); p< Adjusted risk difference (95% CI) ( ) Data are n (%) unless otherwise indicated. COPD=chronic obstructive pulmonary disease. GP=general practitioner. NICE=National Institute for Health and Care Excellence. *Results exclude 22 randomised patients who withdrew their data. COPD defined as post-bronchodilator forced expiratory volume in 1 s (FEV 1 ) to forced vital capacity (FVC) ratio less than 0 7. Adjusted for clustering of general practices, Index of Multiple Deprivation score of practice, patient age, and patient ethnic origin using multilevel logistic regression. Obtained using adjusted estimate of relative risk (7 23 [95% CI ]) from multilevel Poisson regression model with random effects and robust standard errors, together with the baseline risk estimate of 0 8% from routine care group. Table 3: Effectiveness of targeted case finding compared with routine care Vol 4 September

7 Results 56 general practices agreed to participate (appendix), but two withdrew before randomisation (figure). Practices took part between Aug 10, 2012, and June 22, No practices were lost to follow-up by 1 year. Practice and participant characteristics were well balanced between targeted case finding and routine care groups (table 2), although the routine care group ( patients) was larger than the targeted case-finding group ( patients) with some slight differences in distribution of smoking status and participants living in shared households. Practice characteristics were broadly representative of those in the Birmingham and Black Country region with a wide range of socioeconomic deprivation, practice size, and populations served. For the nested trial within the targeted case-finding group, which compared opportunistic case finding with active case finding, of a total of patients, were approved to be allocated a screening questionnaire by GPs (figure) were allocated to the opportunistic Active case finding Opportunistic case finding Total Total patients approved by GP for contact Deceased by 12 months 93 (1%) 92 (1%) 185 (1%) Left practice by 12 months 870 (6%) 897 (6%) 1767 (6%) Primary outcome (NICE criteria) 15 COPD identified by trial* 662 (4%) 189 (1%) 851 (3%) COPD additionally identified by GP 160 (1%) 181 (1%) 341 (1%) Total diagnosed with COPD 822 (5%) 370 (2%) 1192 (4%) Crude risk difference (95% CI) ( ) Crude rate ratio (95% CI) 2 22 ( ) Adjusted odds ratio (95% CI) ; p value 2 34 ( ); p< Adjusted risk difference (95% CI) ( ) Characteristics of patients newly diagnosed through trial Total number Severity of airflow obstruction Mild (FEV 1 80% predicted) 352 (53%) 107 (57%) 459 (54%) Moderate (FEV 1 <80% and 50% 288 (44%) 72 (38%) 360 (42%) predicted) Severe (FEV 1 <50% and 30% predicted) 19 (3%) 10 (5%) 29 (3%) Very severe (FEV 1 <30% predicted) 3 (<1%) 0 3 (<1%) Breathlessness (MRC dyspnoea score 172 (26%) 59 (31%) 231 (27%) grade 3 or more) Current smoker 244 (37%) 56 (30%) 300 (35%) Previous self-report of COPD, chronic bronchitis, or emphysema 74 (11%) 18 (10%) 92 (11%) Data are n (%) unless otherwise indicated. Results exclude 22 randomised patients who withdrew their data. COPD=chronic obstructive pulmonary disease. FEV 1 =forced expiratory volume in 1 s. GP=general practitioner. NICE=National Institute for Health and Care Excellence. *COPD defined as post-bronchodilator FEV 1 to forced vital capacity [FVC] ratio less than 0 7. Adjusted for general practice, Index of Multiple Deprivation score of practice, patient age, and patient ethnic origin using logistic regression. Obtained using estimate of rate ratio of 2 30 (95% CI ) from multilevel Poisson regression model with random effects and robust standard errors, together with the 2 4% baseline risk estimate from routine care group. Table 4: Effect iveness of active compared with opportunistic case finding group, and to the active group (appendix), and 22 formally withdrew from the trial. Patient characteristics were similar between groups (appendix). Our analyses exclude the 22 patients who withdrew their data. In the opportunistic case-finding group, (89%) of patients consulted their GP or primary care professional within the 12-month period (figure). Sampling of record flags across practices suggested that screening questionnaires were distributed to 48 5% (95% CI ) of eligible patients in the opportunistic case-finding group (range 0 100%). All eligible patients in the active case-finding group received a questionnaire. In the opportunistic case-finding group (appendix), 1973 (13%) of eligible patients returned their questionnaire (range %). In the active casefinding group, 5808 (38%) of returned the screening questionnaires, most (5042 [87%]) after receiving it through the post rather than at the medical practice (appendix). Responses improved following each reminder (2312 [15%] of after the initial invite, 1624 [11%] after the first reminder, 1086 [7%] after the second reminder, 20 [<1%] unknown), and varied by general practice (range %; appendix) (56%) of 7781 respondents to the screening questionnaire reported symptoms triggering an invitation for spirometry assessment (3264 [56%] of 5808 in the opportunistic case-finding group vs 1077 [55%] of 1973 in the active case-finding group) (40%) of 7781 respondents reported dyspnoea of MRC grade 2 or worse, 2989 (38%) reported wheeze in the past 12 months, 926 (12%) reported chronic cough, and 727 (9%) reported chronic phlegm (appendix). Chronic cough (752 [13%] of 5808 in the active case-finding group vs 174 [9%] of 1973 in the opportunistic case-finding group) and chronic phlegm (589 [10%] of 5808 vs 138 [7%] of 1973) were more commonly reported in the active case-finding group, although the prevalence of dyspnoea and wheeze was similar in both groups. Non-responders were more likely to be younger, current smokers, and of non-white British ethnicity than respondents. 543 (50%) of 1077 participants with respiratory symptoms in the opportunistic case-finding group, and 2065 (63%) of 3264 in the active case-finding group attended a spirometry assessment (appendix) (93%) of 2658 blows met grade A C quality criteria (33%) of 2608 assessed patients had airflow obstruction defined as FEV 1 to FVC ratio less than 0 7, and 529 (20%) of 2608 had airflow obstruction when defined by the lower limit of normal criteria (GLI equations). In the routine care group, 337 (1%) new cases of COPD were diagnosed in patients after 1 year, compared with 1278 (4%) in eligible patients (including those not approved by the GP for receiving a trial screening questionnaire) in the targeted case-finding group (opportunistic plus active; table 3). After adjusting for age, IMD score, ethnic origin, and clustering of practices Vol 4 September 2016

8 (ICC 0 02), the likelihood of detecting undiagnosed COPD was much greater in the targeted case-finding group (adjusted odds ratio [OR] 7 45 [95% CI ]; p<0 0001). With a baseline risk of 0 8%, the adjusted risk difference was 4 9 per 100 patients (95% CI ) and therefore 21 (95% CI ) patients would require targeting with a screening questionnaire (number needed to screen) to identify one new case. Sensitivity analyses with COPD defined by the lower limit of normal criteria (GLI equations) 18 attenuated the effect in the targeted case-finding group (996 of new cases; 3% yield), although the approach remained significantly better than for the routine care group (adjusted OR 5 41 [95% CI ]; p<0 0001). An analysis including only confirmed ever-smokers also slightly attenuated the effect (OR 6 33 [ ]; p<0 0001). After 1 year, the percentage of newly detected COPD cases was higher in the active case-finding group (822 [5%] of ) than in the opportunistic case-finding group (370 [2%] of ), giving an adjusted OR of 2 34 (95% CI ; p<0 0001) and, assuming a baseline risk of 2 4%, an adjusted risk difference of 2 9 per 100 patients (95% CI ; table 4). This effect varied across practices (appendix), although all but one had an OR estimate in favour of the active case-finding group. 35 (95% CI ) screening questionnaires would need to be mailed to identify one extra COPD patient. Among newly diagnosed patients, severity of airflow obstruction was similar in both case-finding groups (table 4), with 819 (96%) of 851 classified as either mild (459 [54%]) or moderate (360 [42%]). 231 (27%) of 851 newly diagnosed patients reported dyspnoea of MRC grade 3 or more, and 300 (35%) were current smokers. In sensitivity analyses with COPD defined by the lower limit of normal criteria (GLI equations), 607 (4%) of patients in the active case-finding group and 303 (2%) of in the opportunistic case-finding group had undiagnosed COPD identified (adjusted OR 2 07 [95% CI ]; p<0 0001). Compared with routine care, active case finding was more cost-effective ( 333 [US$476] per additional case detected) than opportunistic case finding ( 376 [$538]; table 5). The incremental cost-effectiveness ratio (ICER) for active case finding was 573 ($889) per additional case detected compared with opportunistic case finding. Inclusion of patient-incurred costs made little difference to the results although alternative GP-led or tariff models Active case finding versus routine care New cases identified in the active group* Difference in mean per-patient costs ( ) Adjusted difference in mean cases detected Cost per additional case detected ( ) Opportunistic case finding versus routine care New cases identified in the opportunistic group* Difference in mean per-patient costs ( ) Adjusted difference in mean cases detected Base case 822 (100%) (100%) Including patient costs GP-led model Community model Tariff model Altering target groups Ever-smokers only 707 (86%) (87%) Current smokers only 317 (39%) (44%) Aged 65 years 378 (46%) (55%) Aged 60 years 534 (65%) (71%) Aged 55 years 646 (79%) (82%) Aged 50 years 733 (89%) (91%) Aged 45 years 786 (96%) (96%) Altering triggers for spirometry invitation Cough and phlegm only 383 (47%) (63%) Dyspnoea only 655 (80%) (85%) Excluding wheeze 713 (87%) (91%) Altering processes 3-month period for 786 (96%) (84%) recruitment One reminder 740 (90%) (100%) No reminders 589 (72%) (100%) Cost per additional case detected ( ) GP=general practitioner. *Raw numbers only, given as n (% of base cases identified). Details provided in the appendix. Alternative scenarios assume that additional patients are not diagnosed outside of the trial processes instead. Table 5: Results of the cost-effectiveness analysis with sensitivity analyses Vol 4 September

9 (where spirometry would be done in hospital outpatient appointments) were more expensive. ICERs for both opportunistic and active case finding were lower when older age groups or current smokers only were targeted, although at the expense of identifying many fewer cases. If no reminders were sent to patients, the opportunistic approach would be more cost-effective than the active approach. However, in most alternative scenarios, active case finding identified more new cases than opportunistic case finding. Discussion The identification of undiagnosed COPD remains an important priority worldwide. Despite little available evidence, national guidelines recommend opportunistic case finding in the primary care setting. 15 In our large, generalisable, real-life randomised controlled trial, very few new cases were identified in routine care. As expected, we showed that a systematic targeted approach was markedly more effective than routine care at identifying new cases of COPD. More importantly, we found that an active case-finding approach was twice as effective as opportunistic case finding, and more costeffective. By contrast with many previous studies, we sought confirmation of clinical cases of COPD with spirometry in our targeted group, and assumed (in line with published data) that most new cases routinely diagnosed by GPs would meet appropriate spirometric data. There is only one other relevant trial, undertaken among family medical practices in the USA, which compares the effectiveness of a case-finding approach against usual care, although it did not confirm COPD with spirometry and might not be comparable to the UK or similar primary care settings. 11 Consistent with our findings, this US trial demonstrated that a structured approach to case finding by use of a screening questionnaire administered opportunistically was more effective than routine care in identifying new cases of COPD, although the effect size was smaller (OR 2 20, 95% CI ) and diagnostic yield in routine care was lower (0 49%) than in our study. 11 There have also been many uncontrolled assessments of case-finding approaches for COPD with similar yields to our case-finding group. 10 However, other than our pilot study, 21 our trial is the first to investigate the most costeffective method of administering an initial screening questionnaire, and to compare the results with routine care. The findings from our trial also substantiate the results from our published model, 4 which suggested the likely superiority of active over opportunistic case finding. In the active case-finding group, around a third of eligible participants responded to the postal screening questionnaires, which is consistent with a trial in the Netherlands that compared two methods of processing a postal screening questionnaire, 22 and with uncontrolled studies described in our systematic review. 10 The response rate in our opportunistic case-finding group was lower than we expected from our model, 4 but similar to our pilot study 21 (13% on average across the practices and a maximum of 31%). Unlike other studies, 10 our uptake rates include the whole eligible population as the denominator. It is also likely that our study reflects real life in busy UK primary care, where opportunistic administration of a questionnaire might not always be possible. Opportunistic response rates would need to be at least 40% to reach the yield observed with the active case-finding approach. Our work builds on other published investigations. A particularly relevant programme of work in the Netherlands, the Detection, Intervention and Monitoring of COPD and Asthma (DIMCA) programme, which began in 1991, sought to detect individuals with signs of COPD and asthma at an early stage by use of a two-stage process of screening and monitoring. Findings showed that 52% of adults aged years had early signs and symptoms of COPD or asthma, and 7 7% overall showed persistently decreased lung function or increased bronchial hyper-responsiveness. 23,24 Our finding that newly identified cases tended to have mild or moderate airflow obstruction was in keeping with previous case-finding studies 10 and patients newly diagnosed by case finding had generally less severe disease than patients newly diagnosed in primary care. 25 Nevertheless, in accordance with our model, most new cases could potentially benefit from inhalers, 26,27 selfmanagement support, 28 smoking cessation, 29 and vaccinations. 30 A substantial proportion (27%) also had significant breathlessness with potential to benefit from pulmonary rehabilitation. 31 However, it is important to note that the published evidence available on the effectiveness of these interventions might not directly apply to milder or case-found patients with COPD, for whom new studies are urgently needed. 32 A notable strength of our trial was the achievement of high-quality post-bronchodilator spirometry for COPD diagnosis. We have demonstrated the feasibility of undertaking excellent quality spirometry outside of a specialist setting with previously untrained staff, implementing a rigorous training programme and quality control system. However, there were several practical difficulties. The use of electronic searches of primary care records to identify ever-smokers without previous COPD diagnoses was not always accurate. We acknowledge that there were more never-smokers reported in the routine care group, which, as shown in sensitivity analyses (OR removing never-smokers 6 33 [95% CI ]), might have slightly exaggerated the effect size. Although we aimed to exclude patients with a previous diagnosis of COPD using information from electronic health records, 5% of the respondents to the screening questionnaire self-reported ever having been told that they had COPD, chronic bronchitis, or asthma. However, this finding is unlikely to have influenced the direction of the results because only Vol 4 September 2016

10 10% of those newly diagnosed in our targeted case-finding group reported having these disorders. Also, because of the large volume of participants, we were only able to invite patients for spirometry once, relying on reminder text messages (if possible) for those who did not attend to contact the office for further appointments. This approach led to lower attendance rates than expected, but was unlikely to make a substantial difference to the comparative effect size. Furthermore, it was difficult to know with certainty the proportion of opportunistic questionnaires administered, but it was clearly suboptimal. Nevertheless, this was a pragmatic trial and these issues reflect real life in primary care in the UK. A further point of debate is the criteria of airflow obstruction we were required to use in our assessments for them to be comparable with current UK guidance. Use of the fixed ratio (FEV 1 /FVC <0 7) tends to overestimate the prevalence of new cases of COPD compared with the lower limit of normal alternative criteria, 33 particularly among older men. Use of these criteria in the targeted case-finding group attenuated the yield by about 25%; however, it is unlikely that there would be any relative difference between the targeted case-finding and routine care groups if these criteria were applied to both. Additionally, the risk differences were estimated using the results from Poisson regression and the baseline risk estimates. This approach might not fully account for variability in the baseline risk. In the future, it would be useful to identify more efficient ways to uncover new cases. In our study, to identify one new case of COPD, screening questionnaires had to be provided to 21 extra patients. Algorithms derived from GP records might better predict new cases of COPD and allow more efficient targeting of invitations, but require further testing. 34 Our sensitivity analyses also suggested that limiting to one postal reminder and targeting those aged 50 years or older might be less expensive with minimal loss in effectiveness. A postal-only approach would be nearly as effective and avoid the need for opportunistic administration during busy consultation times. Finally, before a national case-finding programme can be recommended, the longer term effects of case finding on health outcomes should be studied. The DIMCA programme in the Netherlands suggested that long-term prognosis might not be improved by screening. 24 It would be important to establish whether earlier identification by our approach would lead to effective management and health gains that would outweigh the cost to the health service of the medications and management and the potential cost to the patient of having a label of COPD. Currently, there is insufficient evidence regarding this issue. In conclusion, in this well established primary care system, routine practice identified few new cases of COPD. An active targeted approach is a highly effective and cost-effective way to identify patients with undiagnosed clinically important COPD and has the potential to improve their health. Contributors The concept of a need for the trial was identified by REJ and PA and the design was refined and shaped in discussion with DF and KKC. AS undertook the statistical analyses with guidance and input from RDR and JM. JM undertook the randomisation. DB and SJ undertook the economic analyses. AE, MRM, BGC, and SH designed the approach to training and spirometry quality control procedures. REJ wrote the first draft of the report and led the oversight of the trial with PA, DF, and AE. All authors had input into the design and analysis of the trial, and the manuscript. Declaration of interests KKC reports grants from Pfizer China, outside of the submitted work. RS reports personal fees from Boehringer Ingelheim, personal fees from GSK, personal fees from Chiesi, personal fees from Takeda, personal fees from Novartis, personal fees from Polyphor, grants from CSL Behring, grants from Talecris, and personal fees from Dyax, outside of the submitted work. AMT reports grants from Linde REAL fund, grants from Alpha 1 Foundation, non-financial support from GSK, non-financial support from Boehringer Ingelheim, personal fees and non-financial support from Chiesi, and personal fees and non-financial support from AZ, outside of the submitted work. REJ, KJ, RDR, and SH report grants from NIHR, during the conduct of the study. All other authors declare no competing interests. Acknowledgments We thank the GPs and patients for taking part in this study, and our patient advisory group for their useful comments throughout. We thank the Clinical Research Network West Midlands at the University of Birmingham for recruiting the GPs, and finally we thank the BLISS research team for their hard work in making this trial possible. This paper summarises independent research funded by the National Institute for Health Research (NIHR) under its Programme Grants for Applied Research (grant reference number RP-PG ). KJ is part-funded by the NIHR CLAHRC West Midlands. The views expressed are those of the authors and not necessarily those of the National Health Service (NHS), the NIHR, or the Department of Health. TargetCOPD is part of the Birmingham Lung Improvement Studies (BLISS). References 1 Global Initiative for Chronic Obstructive Lung Disease (GOLD). Global strategy for the diagnosis, management and prevention of COPD (accessed May 30, 2016). 2 European Respiratory Society. European lung white book. Respiratory health and disease in Europe Chapter 2 The economic burden of lung diseases. (accessed May 30, 2016). 3 Halbert RJ, Natoli JL, Gano A, Badamgarav E, Buist AS, Mannino DM. Global burden of COPD: systematic review and meta-analysis. Eur Respir J 2006; 28: Jordan R, Lam K-B, Cheng K, et al. Case finding for chronic obstructive pulmonary disease: a model for optimizing a targeted approach. Thorax 2010; 65: Mannino DM, Gagnon RC, Petty TL, Lydick E. Obstructive lung disease and low lung function in adults in the United States: data from the National Health and Nutrition Examination Survey, Arch Intern Med 2000; 160: Buist AS, McBurnie M, Vollmer WM, et al, on behalf of the BOLD Collaborative Research Group. International variation in the prevalence of COPD (the BOLD Study): a population-based prevalence study. Lancet 2007; 370: British Lung Foundation. Invisible lives. Chronic obstructive pulmonary disease (COPD) finding the missing millions Lives-report.pdf (accessed July 4, 2016). 8 Department of Health. An outcomes strategy for people with chronic obstructive pulmonary disease (COPD) and asthma in England Publications/PublicationsPolicyAndGuidance/DH_ (accessed May 30, 2016). 9 COPD Foundation. Am I at risk for COPD? copdfoundation.org/take-action/get-tested/am-i-at-risk.aspx (accessed July 4, 2016). Vol 4 September