BMJ Open Is Committed to Open Peer Review. As Part of This Commitment We Make the Peer Review History of Every Article We Publish Publicly Available

BMJ Open: first published as 10.1136/bmjopen-2019-035678 on 31 October 2020. Downloaded from

BMJ Open is committed to open peer review. As part of this commitment we make the peer review history of every article we publish publicly available.

When an article is published we post the peer reviewers’ comments and the authors’ responses online. We also post the versions of the paper that were used during peer review. These are the versions that the peer review comments apply to.

The versions of the paper that follow are the versions that were submitted during the peer review process. They are not the versions of record or the final published versions. They should not be cited or distributed as the published version of this manuscript.

BMJ Open is an open access journal and the full, final, typeset and author-corrected version of record of the manuscript is available on our site with no access controls, subscription charges or pay-per-view fees (http://bmjopen.bmj.com).

If you have any questions on BMJ Open’s open peer review process please email [email protected] http://bmjopen.bmj.com/ on September 28, 2021 by guest. Protected copyright. BMJ Open BMJ Open: first published as 10.1136/bmjopen-2019-035678 on 31 October 2020. Downloaded from

Primary care practitioner diagnostic action when the patient may have cancer: a vignette survey in 20 European countries ForJournal: peerBMJ Open review only Manuscript ID bmjopen-2019-035678

Article Type: Original research

Date Submitted by the 20-Nov-2019 Author:

Complete List of Authors: Harris, Michael; University of Bath, Department for Health; Universität Bern, Berner Institut für Hausarztmedizin (BIHAM) Brekke, Mette; University of Oslo, Department of General Practice and General Practice Research Unit Dinant, Geert-Jan; Maastricht University, Dept of General Practice Esteva, Magdalena; Gerencia Atencio Primaria de Mallorca, Unidad de Investigación Hoffman, Robert; Tel Aviv University, Department of Family Medicine Marzo, Mercè; Unitat de Suport a la Recerca Costa de Ponent, IDIAP Jordi Gol, Murchie, Peter; University of Aberdeen, Division of Applied Health Science

Neves, Ana Luísa; Imperial College London, Centre for Health Policy; http://bmjopen.bmj.com/ University of Porto, Also CINTESIS (Centre for Health Technology and Services Research) and MEDCIDS (Department of Community Medicine, Information and Health Decision Sciences) Smyrnakis, Emmanouil; Aristotle University of Thessaloniki , Laboratory of Primary Health Care, General Practice and Health Services Research Vedsted, Peter; Research Unit for General Practice, University of Aarhus Aubin-Auger, Isabelle; Universite Paris Diderot UFR de Medecine, General Practice

Azuri, Joseph; Tel Aviv University, Sackler Faculty of Medicine on September 28, 2021 by guest. Protected copyright. Buczkowski, Krzysztof; Nicolaus Copernicus University, Department of Family Medicine Buono, Nicola; SNAMID, Family medicine Foreva, Gergana; Medical Center BROD Babić, Svjetlana ; Croatian Health Insurance Fund Jakob, Eva; Centro de Saúde Sarria, Primary Health Centre Koskela, Tuomas; Tampere University, Faculty of Medicine and Health Technology Petek, Davorina; Univerza v Ljubljani, Department of Family Medicine Ster, Marija Petek; Univ Ljubljana, Departmento of Family medicine Puia, Aida; Iuliu Hagieganu University of Medicine and Pharmacy Faculty of Medicine, Family Medicine Department Sawicka-Powierza, Jolanta; Uniwersytet Medyczny w Bialymstoku, Department of Family Medicine Streit, Sven; Institute of Primary Health Care BIHAM, Thulesius, Hans; Lund University, Department of Research and Development

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 1 of 47 BMJ Open BMJ Open: first published as 10.1136/bmjopen-2019-035678 on 31 October 2020. Downloaded from

1 2 3 Weltermann, Birgitta; University of Bonn, Institut für Hausarztmedizin 4 Taylor, Gordon; University of Exeter, College of Medicine and Health 5 6 International health services < HEALTH SERVICES ADMINISTRATION & Keywords: 7 MANAGEMENT, Adult oncology < ONCOLOGY, PRIMARY CARE 8 9 10 11 12 13 14 15 16 For peer review only 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 http://bmjopen.bmj.com/ 34 35 36 37 38 39 40 41 on September 28, 2021 by guest. Protected copyright. 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 2 of 47

1 2

3 BMJ Open: first published as 10.1136/bmjopen-2019-035678 on 31 October 2020. Downloaded from 4 5 6 7 8 9 I, the Submitting Author has the right to grant and does grant on behalf of all authors of the Work (as defined 10 in the below author licence), an exclusive licence and/or a non-exclusive licence for contributions from authors 11 who are: i) UK Crown employees; ii) where BMJ has agreed a CC-BY licence shall apply, and/or iii) in accordance 12 with the terms applicable for US Federal Government officers or employees acting as part of their official 13 duties; on a worldwide, perpetual, irrevocable, royalty-free basis to BMJ Publishing Group Ltd (“BMJ”) its 14 licensees and where the relevant Journal is co-owned by BMJ to the co-owners of the Journal, to publish the 15 Work in this journal and any other BMJ products and to exploit all rights, as set out in our licence. 16 17 The Submitting Author accepts and understands that any supply made under these terms is made by BMJ to 18 the Submitting Author Forunless you peer are acting as review an employee on behalf only of your employer or a postgraduate 19 student of an affiliated institution which is paying any applicable article publishing charge (“APC”) for Open 20 Access articles. Where the Submitting Author wishes to make the Work available on an Open Access basis (and 21 intends to pay the relevant APC), the terms of reuse of such Open Access shall be governed by a Creative 22 Commons licence – details of these licences and which Creative Commons licence will apply to this Work are set 23 out in our licence referred to above. 24 25 Other than as permitted in any relevant BMJ Author’s Self Archiving Policies, I confirm this Work has not been 26 accepted for publication elsewhere, is not being considered for publication elsewhere and does not duplicate 27 material already published. I confirm all authors consent to publication of this Work and authorise the granting 28 of this licence. 29 30 31 32 33 34 35 36

37 http://bmjopen.bmj.com/ 38 39 40 41 42 43 44

45 on September 28, 2021 by guest. Protected copyright. 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 3 of 47 BMJ Open

1 2 1 3 Primary care practitioner diagnostic action when the patient may have BMJ Open: first published as 10.1136/bmjopen-2019-035678 on 31 October 2020. Downloaded from 4 5 6 2 cancer: a vignette survey in 20 European countries 7 8 9 10 3 Authors 11 12 13 4 Michael Harris (corresponding author), Department for Health, University of Bath, Claverton 14 15 16 5 Down, Bath BA2 7AY, UK; telephone: +44 1761 241366; fax: none. Also: Institute of Primary 17 18 6 Health Care Bern (BIHAM),For Universitypeer of review Bern, Bern, Switzerland only 19 20 21 7 [email protected] 22 23 8 Mette Brekke, Department of General Practice and General Practice Research Unit, University 24 25 26 9 of Oslo, Oslo, Norway. [email protected] 27 28 29 10 Geert-Jan Dinant, Department of General Practice, Maastricht University, Maastricht, 30 31 11 Netherlands. [email protected] 32 33 34 12 Magdalena Esteva, Research Unit, Majorca Primary Health Care Department, Balearic Islands 35 36 13 Health Research Institute (IdISBa), Preventive Activities and Health Promotion Network,

37 http://bmjopen.bmj.com/ 38 14 Carlos III Institute of Health (RedIAPP-RETICS), Palma, Spain. [email protected] 39 40 41 15 Robert D. Hoffman, Department of Family Medicine, Tel Aviv University, Tel Aviv, Israel. 42 43 44 16 [email protected]

45 on September 28, 2021 by guest. Protected copyright. 46 17 Mercè Marzo-Castillejo, Unitat de Suport a la Recerca, IDIAP Jordi Gol, Institut Català de la 47 48 49 18 Salut, Barcelona, Spain. [email protected] 50 51 52 19 Peter Murchie, Division of Applied Health Sciences - Academic Primary Care, University of 53 54 20 Aberdeen, Aberdeen, UK. [email protected] 55 56 57 21 Ana Luísa Neves, Centre for Health Policy, Imperial College, London, UK. Also CINTESIS 58 59 22 (Centre for Health Technology and Services Research) and MEDCIDS (Department of 60

1 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 4 of 47

1 2 23 Community Medicine, Information and Health Decision Sciences), Faculty of Medicine, 3 BMJ Open: first published as 10.1136/bmjopen-2019-035678 on 31 October 2020. Downloaded from 4 24 University of Porto, Porto, Portugal [email protected] 5 6 7 25 Emmanouil Smyrnakis, Laboratory of Primary Health Care, General Practice and Health 8 9 26 Services Research, Aristotle University of Thessaloniki, Thessaloniki, Greece. 10 11 27 [email protected] 12 13 14 28 Peter Vedsted, The Research Unit for General Practice, Aarhus University, Aarhus, Denmark. 15 16 17 29 [email protected] 18 For peer review only 19 20 30 Isabelle Aubin-Auger, Department of General Practice, Université Paris Diderot, Paris, France. 21 22 31 [email protected] 23 24 25 32 Joseph Azuri, Sackler Faculty of Medicine, Tel Aviv University, Tel Aviv, Israel. 26 27 33 [email protected] 28 29 30 34 Krzysztof Buczkowski, Department of Family Medicine, Nicolaus Copernicus University, 31 32 35 Toruń, Poland. [email protected] 33 34 35 36 Nicola Buono, Department of General Practice, National Society of Medical Education in 36 37 37 General Practice (SNaMID), Caserta, Italy. [email protected] http://bmjopen.bmj.com/ 38 39 40 38 Gergana Foreva, Medical Center BROD, Plovdiv, Bulgaria. [email protected] 41 42 43 39 Svjetlana Gašparović Babić, Croatian Health Insurance Fund, Rijeka, Croatia. 44 45 40 [email protected] on September 28, 2021 by guest. Protected copyright. 46 47 48 41 Eva Jakob, Primary Health Centre, Centro de Saúde Sarria, Sarria, Lugo, Spain. 49 50 51 42 [email protected] 52 53 54 43 Tuomas Koskela, Faculty of Medicine and Health Technology, Tampere University, Tampere, 55 56 44 Finland. [email protected] 57 58 59 60

2 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 5 of 47 BMJ Open

1 2 45 Davorina Petek, Department of Family Medicine, University of Ljubljana, Ljubljana, Slovenia. 3 BMJ Open: first published as 10.1136/bmjopen-2019-035678 on 31 October 2020. Downloaded from 4 46 [email protected] 5 6 7 47 Marija Petek Ster, Department of Family Medicine, University of Ljubljana, Ljubljana, Slovenia. 8 9 48 [email protected] 10 11 12 49 Aida Puia, Teaching Assistant, Family Medicine Department, Iuliu Hatieganu University of 13 14 50 Medicine and Pharmacy, Cluj-Napoca, Romania. [email protected] 15 16 17 51 Jolanta Sawicka-Powierza, Department of Family Medicine, Medical University of Bialystok, 18 For peer review only 19 20 52 Bialystok, Poland. [email protected] 21 22 53 Sven Streit, Institute of Primary Health Care Bern (BIHAM), University of Bern, Bern, 23 24 25 54 Switzerland. [email protected] 26 27 28 55 Hans Thulesius Department of Clinical Sciences, Lund University, Malmö, Sweden and 29 30 56 Department of Research and Development, Region Kronoberg, Sweden 31 32 57 [email protected] 33 34 35 58 Birgitta Weltermann, Institut für Hausarztmedizin, University of Bonn, Bonn, Germany. 36 37 59 [email protected] http://bmjopen.bmj.com/ 38 39 40 60 Gordon Taylor, College of Medicine and Health, University of Exeter, Exeter, UK. 41 42 43 61 [email protected] 44

45 on September 28, 2021 by guest. Protected copyright. 46 62 47 48 49 50 51 52 53 54 55 56 57 58 59 60

3 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 6 of 47

1 2 63 Abstract 3 BMJ Open: first published as 10.1136/bmjopen-2019-035678 on 31 October 2020. Downloaded from 4 5 64 Objectives 6 7 8 65 Cancer survival rates vary widely between European countries, with differences in timeliness 9 10 11 66 of diagnosis thought to be one key reason. There is little evidence on the way in which 12 13 67 different healthcare systems influence Primary Care Practitioners’ (PCPs’) referral decisions 14 15 68 in patients that could have cancer, and how this links with cancer survival. 16 17 18 69 This study aimed to Forexplore PCPs’peer diagnostic review actions in patients only with symptoms that could be 19 20 70 due to cancer, how they vary across European countries, and how they relate to cancer 21 22 23 71 survival rates. 24 25 26 72 Design 27 28 29 73 A primary care survey. PCPs were given vignettes describing patients with symptoms that 30 31 74 could indicate cancer and asked how they would manage these patients. Correlations between 32 33 34 75 the likelihood of taking immediate diagnostic action (a diagnostic test and/or referral) and 35 36 76 physician characteristics were calculated. The likelihood of taking immediate diagnostic

37 http://bmjopen.bmj.com/ 38 77 action in the different participating countries, and the correlation with national 1-year 39 40 41 78 relative cancer survival rates, were analysed. 42 43 44 79 Setting

45 on September 28, 2021 by guest. Protected copyright. 46 47 80 Centres in twenty European countries with widely varying cancer survival rates. 48 49 50 81 Participants 51 52 53 82 A total of 2,086 PCPs answered the survey question, with a median of 72 PCPs per country. 54 55 56 57 58 59 60

4 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 7 of 47 BMJ Open

1 2 83 Results 3 BMJ Open: first published as 10.1136/bmjopen-2019-035678 on 31 October 2020. Downloaded from 4 5 84 PCPs’ likelihood of immediate diagnostic action at the first consultation varied from 50 to 6 7 85 82% between countries. PCPs who were more experienced, were working in more remote 8 9 86 areas, or worked alone or in smaller practices, were more likely to take immediate diagnostic 10 11 12 87 action than their peers. There was a significant negative correlation between national 13 14 88 healthcare expenditure levels and likelihood of immediate diagnostic action (r=–0.55, 15 16 89 P=0.012). However, there was no significant correlation between the likelihood of taking 17 18 For peer review only 19 90 immediate diagnostic action and cancer survival (r=–0.27, P=0.278). 20 21 22 91 Conclusions 23 24 25 92 Europe shows large between-country variations in PCPs’ diagnostic action rates for patients 26 27 93 who could have cancer. These are linked with differences in healthcare organisation and 28 29 94 levels of healthcare investment. 30 31 32 33 95 Keywords 34 35 36 96 Delivery of Health Care; Primary Health Care; Cancer; Decision Making; Survival; Europe

37 http://bmjopen.bmj.com/ 38 39 97 Word count 40 41 42 98 3905 43 44 99 45 on September 28, 2021 by guest. Protected copyright. 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

5 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 8 of 47

1 2 100 Article Summary 3 BMJ Open: first published as 10.1136/bmjopen-2019-035678 on 31 October 2020. Downloaded from 4 101 Strengths and limitations of this study 5 6 7 102 • Recruitment of PCPs from 20 European countries, four countries from each of the 8 9 103 Northern, Southern, Eastern, Western and Central European geographical areas, provided 10 11 104 variation in geography, health systems and levels of healthcare spending. 12 13 14 105 • The questionnaire was carefully developed and piloted by GPs and other PCPs, and 15 16 106 therefore grounded in their clinical experience. 17 18 For peer review only 19 107 • While the response rate was low, it was comparable to that of other equivalent surveys of 20 21 108 primary care doctors. 22 23 24 109 • Measuring the possible link between surveyed PCPs’ stated actions and national cancer 25 26 110 survival creates the risk of an ecological fallacy, as any identified correlation may be an 27 28 111 indicator of the effect of unmeasured system factors. 29 30 31 32 33 34 35 36

37 http://bmjopen.bmj.com/ 38 39 40 41 42 43 44

45 on September 28, 2021 by guest. Protected copyright. 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

6 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 9 of 47 BMJ Open

1 2 112 Background 3 BMJ Open: first published as 10.1136/bmjopen-2019-035678 on 31 October 2020. Downloaded from 4 5 113 Cancer survival varies vary widely across Europe [1]. The fifth cycle of the European Cancer 6 7 114 Registry (EUROCARE)-based Study on Survival and Care of Cancer Patients shows that 8 9 10 115 national 1-year relative survival rates for all cancer sites vary from 58.2% to 81.1% [2]. 11 12 116 Comparison of European 1-year and 5-year relative cancer survival rates [2–5] shows that 13 14 117 some countries have higher survival from most cancers (including Belgium, France, Sweden 15 16 17 118 and Switzerland), while others have consistently lower survival (including Bulgaria, Croatia, 18 For peer review only 19 119 Poland and Scotland). This suggests that an improvement in cancer awareness and early 20 21 120 detection in relatively poorly performing countries could reduce the survival gap [4]. While 22 23 24 121 recent cancer survival rates show improvement in most countries [5], the between-country 25 26 122 differences remain [6]. However, this is not inevitable: Denmark’s considerable efforts to 27 28 123 improve early detection rates [7] have resulted in a narrowing of the gap between its own 29 30 31 124 relatively poor cancer survival rates and those of its better performing Nordic neighbours [8]. 32 33 125 There has been a call for studies which compare cancer diagnostic pathways between well 34 35 126 and poorly performing countries, to help gain an understanding of how these disparities may 36

37 http://bmjopen.bmj.com/ 38 127 be remedied [5]. 39 40 41 128 Although 1-year and 5-year relative survival can be affected by overdiagnosis and lead-time 42 43 129 biases [9,10], poorer 1-year survival in some countries is thought to be rooted in diagnostic 44 45 130 delay [11,12] and more advanced disease at diagnosis [13,14]. The more advanced a cancer is, on September 28, 2021 by guest. Protected copyright. 46 47 48 131 the more difficult it is to treat it successfully [15] and, for many but not all cancers, disease 49 50 132 stage at diagnosis is associated with survival [16,17]. There is considerable evidence that 51 52 133 longer time to diagnosis and treatment increases cancer mortality [18,19]. Timely diagnosis of 53 54 55 134 cancer is, therefore, a cornerstone of health policy throughout Europe [20]. However, there is 56 57 135 a substantial challenge in deciding where and how to achieve this [21], as it is uncertain 58 59 136 whether late diagnosis is due to cancer patients presenting later, not being referred quickly 60

7 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 10 of 47

1 2 137 enough from primary care, or whether they are inefficiently investigated, diagnosed and 3 BMJ Open: first published as 10.1136/bmjopen-2019-035678 on 31 October 2020. Downloaded from 4 138 treated in secondary care [15]. This may be a particular issue where cancer patients present 5 6 139 without ‘red-flag’ symptoms, as how the Primary Care Practitioner (PCP) acts will depend to a 7 8 9 140 large extent on local health service organisation [22]. Balanced with a desire for more timely 10 11 141 diagnosis of cancer is the need to avoid increased healthcare costs, as well as to minimise 12 13 142 overdiagnosis and overtreatment, though these latter concerns are particularly related to 14 15 16 143 cancer screening [23–25]. 17 18 For peer review only 19 144 There is little evidence on how different healthcare systems influence PCPs’ referral decisions 20 21 145 [21]. However, a large variety of non-clinical factors affect these referral decisions [22]. These 22 23 146 include the extent of gatekeeping, funding systems, access to special investigations, concerns 24 25 26 147 over litigation, and barriers to accessing specialist advice, as well as the availability of fast- 27 28 148 track programmes for suspected cancer. Whether the responsibility for early detection of 29 30 149 cancer is principally in primary or secondary care varies between countries. The International 31 32 33 150 Cancer Benchmarking Partnership (ICBP) [26] examined differences in cancer awareness and 34 35 151 beliefs between six countries with comparable wealth in an attempt to explain differences in 36 37 152 cancer survival [27]. It found a positive association between national cancer survival rates http://bmjopen.bmj.com/ 38 39 40 153 and the readiness of PCPs in those countries to investigate potential cancer symptoms [28]. 41 42 154 However, there has not yet been an investigation of how PCPs’ diagnostic actions with respect 43 44 155 to potential cancer symptoms vary across Europe, amongst countries with a wide range of 45 on September 28, 2021 by guest. Protected copyright. 46 47 156 socio-economic development, healthcare systems and healthcare spending. We therefore 48 49 157 aimed to explore the diagnostic action rates of PCPs for patients with symptoms that could be 50 51 158 due to cancer, how they compare across European countries, and how they are associated 52 53 54 159 with cancer survival. 55 56 57 58 59 60

8 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 11 of 47 BMJ Open

1 2 160 Methods 3 BMJ Open: first published as 10.1136/bmjopen-2019-035678 on 31 October 2020. Downloaded from 4 5 161 Design 6 7 8 162 We provided clinical vignettes to PCPs from twenty European countries with markedly 9 10 11 163 different socio-economic and healthcare systems. The vignettes described patients presenting 12 13 164 with symptoms that could indicate cancer. Recruitment started in November 2015 and was 14 15 165 completed at the end of 2016. 16 17 18 166 Study population For peer review only 19 20 21 167 The Örenäs Research Group (ÖRG) is a European collaborative of primary care researchers, 22 23 24 168 formed in 2013 to study the factors influencing national variations in the early diagnosis of 25 26 169 cancer in primary care. The research was conducted in 25 ÖRG centres in 20 countries across 27 28 170 Europe: Bulgaria, Croatia, Denmark, England, Finland, France, Germany, Greece, Israel, Italy, 29 30 31 171 Netherlands, Norway, Poland, Portugal, Romania, Scotland, Slovenia, Spain, Sweden and 32 33 172 Switzerland. Medical doctors were eligible for the survey if they were working mainly in 34 35 173 primary care. These doctors, here referred to collectively as ‘Primary Care Practitioners’, 36

37 http://bmjopen.bmj.com/ 38 174 included General Practitioners (GPs) and other doctors who had other specialist training but 39 40 175 worked in the community and could be accessed directly by patients without referral. 41 42 43 176 Development of the questionnaire 44

45 on September 28, 2021 by guest. Protected copyright. 46 177 Following a literature review, ÖRG investigators developed a questionnaire designed to elicit 47 48 49 178 PCPs’ diagnostic actions for patients that could have cancer. A questionnaire with five clinical 50 51 179 vignettes (three new ones, and two designed and validated by the ICBP [29] and used with 52 53 180 permission) was piloted by the ÖRG local leads in January 2015 to check validity. More 54 55 56 181 information about this process is given elsewhere [30]. One of the vignettes was found to be 57 58 182 invalid and was removed. The next version of the questionnaire, in English, was then piloted 59 60

9 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 12 of 47

1 2 183 by 49 PCPs in 16 ÖRG member countries in July 2015. No changes to the vignettes were made 3 BMJ Open: first published as 10.1136/bmjopen-2019-035678 on 31 October 2020. Downloaded from 4 184 following this second pilot. 5 6 7 185 ÖRG leads arranged for translations of the questionnaire into their local languages where 8 9 186 these were not English, a total of 19 translations from the original English. Translation and 10 11 187 validation by backtranslation were done in a standardised way [31] and are described 12 13 14 188 elsewhere [32]. 15 16 17 189 Description of the questionnaire 18 For peer review only 19 20 190 The questionnaire consisted of 47 items and was divided into four sections: (a) demographic 21 22 191 questions (five questions about years since graduation, gender, type and rural/urban location 23 24 25 192 of practice and number of doctors working in the practice); (b) referral availability questions 26 27 193 (two questions about tests and specialist opinions that were either directly or indirectly 28 29 194 available to the respondent); (c) four fictitious clinical vignettes and (d) 20 health system 30 31 32 195 factor questions. Each of the vignettes provided information on the patient’s presenting 33 34 196 symptoms, previous medical history, medication, clinical findings and other relevant 35 36 197 information (Appendix 1). The vignettes were: 37 http://bmjopen.bmj.com/ 38 39 198 1. A 62-year-old male smoker with chronic obstructive pulmonary disease and now a two- 40 41 42 199 week history of a productive cough; positive predictive value (PPV) for lung cancer: 3.6% 43 44 200 [33];

45 on September 28, 2021 by guest. Protected copyright. 46 47 201 2. A 53-year-old woman with lower abdominal pain and abdominal distension; PPV for 48 49 202 ovarian cancer: 3.1% [34]; 50 51 52 203 3. A 35-year-old breastfeeding woman with an abnormal nipple discharge and eczematous 53 54 204 changes around the nipple; PPV for breast cancer: 1.2% [35]; 55 56 57 205 4. A 22-year-old man with coeliac disease who now has abdominal pain, rectal bleeding and 58 59 206 diarrhoea; PPV for colorectal cancer: 3.4% [33]. 60

10 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 13 of 47 BMJ Open

1 2 207 For each patient a range of five possible management decisions was given, with an invitation 3 BMJ Open: first published as 10.1136/bmjopen-2019-035678 on 31 October 2020. Downloaded from 4 208 to choose as many as needed: 5 6 7 209  ‘I would write an appropriate prescription for the patient.’ 8 9 210  ‘I would arrange to see the patient again for follow-up and reassessment.’ 10 11 12 211  ‘I would not arrange formal follow-up, but would tell the patient under what 13 14 212 circumstances he should see me again.’ 15 16 213  ‘I would organise an investigation at this consultation.’ 17 18 For peer review only 19 214  ‘I would refer the patient to a specialist at this consultation.’ 20 21 22 215 Those that chose to investigate the patient were able to select from a range of possible 23 24 216 diagnostic tests. The response of primary interest was a PCPs’ management choice that would 25 26 217 be likely to identify a cancer as a cause of the patients’ symptoms, by either opting to request 27 28 29 218 a significant diagnostic test or by referring to a specialist. The tests used in the analysis were: 30 31 219 a plain chest X-ray or lung computerised tomography (CT) for the lung vignette; a tumour 32 33 220 marker, diagnostic ultrasound or CT for the ovarian vignette; an ultrasound of the breast or 34 35 36 221 mammography for the breast vignette; and diagnostic ultrasound, sigmoidoscopy,

37 http://bmjopen.bmj.com/ 38 222 colonoscopy or CT colonography for the colorectal vignette. A factor analysis of the results of 39 40 223 the survey section on the effect of health system factors is reported separately [36]. 41 42 43 44 224 Sample size

45 on September 28, 2021 by guest. Protected copyright. 46 47 225 We aimed for a total sample size of at least 1000 PCPs, with at least 50 responses from each of 48 49 226 the participating countries. 50 51 52 227 Recruitment of participants 53 54 55 228 Each ÖRG local lead was asked to email an invitation to take part in the survey to the PCPs in 56 57 229 their local health district or jurisdiction, and to recruit at least 50 participants. In six countries 58 59 60 230 (Denmark, Norway, Portugal, Romania, Slovenia, Sweden), the invitation was distributed to a

11 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 14 of 47

1 2 231 national sample. The recruitment email stated that the research aimed to identify which 3 BMJ Open: first published as 10.1136/bmjopen-2019-035678 on 31 October 2020. Downloaded from 4 232 health system factors affect PCPs’ decisions to refer patients for further investigation. The 5 6 233 possibility of cancer as being a cause of the vignette symptoms was not mentioned in either 7 8 9 234 the recruitment email or the survey. As low survey response rates are common in primary 10 11 235 care [37] and can vary between jurisdictions, any local leads who had difficulty in achieving 12 13 236 the required sample sizes were asked to increase the number of responses by using 14 15 16 237 snowballing, a recognised technique for recruiting hard-to-reach populations in health studies 17 18 238 [38,39]. For peer review only 19 20 21 239 Distribution of the questionnaire 22 23 24 240 The questionnaire was designed using SurveyMonkey (SurveyMonkey, California, USA). 25 26 241 Because of the study’s wide geographical coverage, on-line delivery of the questionnaire was 27 28 29 242 used; this methodology has previously been successfully used in research involving cancer 30 31 243 care professionals [40]. 32 33 34 244 Statistical analysis 35 36 37 245 Demographic questions and those relating to decisions to arrange a diagnostic test and to http://bmjopen.bmj.com/ 38 39 40 246 refer to a specialist were analysed using descriptive statistics. As it was considered that some 41 42 247 PCPs would not organise a diagnostic test because they were referring to a specialist, and 43 44 248 conversely some PCPs would not refer to a specialist because they were organising a 45 on September 28, 2021 by guest. Protected copyright. 46 47 249 diagnostic test, we also used a composite measure of a decision to arrange a diagnostic test 48 49 250 and/or refer to a specialist, i.e. the likelihood of taking immediate diagnostic action for cancer. 50 51 251 For each individual PCP, mean immediate diagnostic action rates were calculated from the 52 53 54 252 four individual vignette responses. From those, mean immediate diagnostic action rates were 55 56 253 calculated for each country. For comparisons between countries, medians and ranges were 57 58 254 calculated. 59 60

12 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 15 of 47 BMJ Open

1 2 255 For each country, the mean 1-year and 5-year relative cancer survival rates for the four 3 BMJ Open: first published as 10.1136/bmjopen-2019-035678 on 31 October 2020. Downloaded from 4 256 cancers of interest (lung, ovary, breast and colorectal) were calculated from EUROCARE-5 5 6 257 data [2], using International Cancer Survival Standards (ICSS) which standardise survival by 7 8 9 258 age according to cancer site [41]. These survival rates are shown in Table 1. While long-term 10 11 259 relative cancer survival would be an outcome that eliminates lead-time and length-time bias, 12 13 260 there are no standardised data for this for the 20 European countries that we studied. Linear 14 15 16 261 correlations with their 95% confidence intervals (CIs) and 2-tailed significance levels were 17 18 262 estimated for the proportionsFor peer of PCPs opting review to take diagnostic only action and national 1-year 19 20 263 relative cancer survival rates, and between PCPs’ likelihood of organising a diagnostic test and 21 22 23 264 their likelihood of referral to a specialist. As 1-year relative survival rates for cancer can be 24 25 265 affected by lead-time bias, we made a sensitivity analysis using 5-year survival. For countries 26 27 266 with low or unknown response rates, we were unsure how representative their respondents’ 28 29 30 267 answers were of their wider PCP population, so we also made a sensitivity analysis that 31 32 268 excluded the three countries in which the response rate was either unknown or was <10%. 33 34 269 In all the analyses the countries are weighted equally, so that differing numbers of 35 36

37 270 respondents do not bias the results. http://bmjopen.bmj.com/ 38 39 40 271 Patient and public involvement 41 42 272 There was no patient or public involvement in this study. 43 44

45 on September 28, 2021 by guest. Protected copyright. 46 273 Results 47 48 49 50 274 A total of 2,086 PCPs completed the questionnaire. There was a median of 72 respondents per 51 52 275 country, range 59-446 (Table 1). 53 54 55 276 (Place Table 1 here) 56 57 58 59 60

13 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 16 of 47

1 2 277 The response rate for two countries was unknown. For the other 18 countries, the median 3 BMJ Open: first published as 10.1136/bmjopen-2019-035678 on 31 October 2020. Downloaded from 4 278 response rate was 24.8% (range 7.1% to 65.6%). Participants’ demographic distributions are 5 6 279 shown in Table 2. 7 8 9 280 (Place Table 2 here) 10 11 12 281 Organising a diagnostic test 13 14 15 282 The range of PCPs who stated that they would organise a diagnostic test at this first 16 17 18 283 consultation varied Forfrom 35.6%-80.1%, peer medianreview 53.0% (Figure only 1). 19 20 284 (Place Figure 1 here) 21 22 23 24 285 Referring the patient to a specialist 25 26 27 286 Across the participating countries, a median of 34.2%, range 12.3%-64.7%, of PCPs decided to 28 29 287 refer the patient to a specialist at the first consultation (Figure 2). 30 31 32 288 (Place Figure 2 here) 33 34 35 289 Arranging a diagnostic test and/or referring the patient to a specialist 36

37 http://bmjopen.bmj.com/ 38 290 There was a strong correlation between PCPs’ likelihood of taking immediate diagnostic 39 40 291 action (organising a diagnostic test and/or referring to a specialist): r=0.77, P=<0.001. Across 41 42 43 292 the surveyed countries, the proportion of PCPs who would take diagnostic action at this first 44

45 293 consultation varied from 50.0% to 82.1%, median 59.9% (Figure 3). on September 28, 2021 by guest. Protected copyright. 46 47 48 294 (Place Figure 3 here) 49 50 51 295 Correlation between decision to take diagnostic action and national cancer survival rates 52 53 54 296 Overall, there was no significant correlation between the likelihood of immediate diagnostic 55 56 297 action (diagnostic test and/or referral) and national 1-year relative cancer survival rates, 57 58 59 298 r=-0.27, 95% CI –0.65 to 0.23, P=0.278 (Figure 4). In the sensitivity analysis based on national 60

14 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 17 of 47 BMJ Open

1 2 299 5-year relative cancer survival rates, the values were similar, r=–0.28, 95% CI –0.66 to 0.22, 3 BMJ Open: first published as 10.1136/bmjopen-2019-035678 on 31 October 2020. Downloaded from 4 300 P=0.267. In the sensitivity analysis that excluded the three countries in which the response 5 6 301 rate was either unknown or was <10%, there was also no significant correlation between the 7 8 9 302 likelihood of immediate diagnostic action and national 1-year relative cancer survival rates 10 11 303 (r=–0.28, 95% CI -0.25 to 0.68, P=0.301). The degree of correlation between the likelihood of 12 13 304 immediate diagnostic action and national 1-year relative cancer survival varied between the 14 15 16 305 vignettes: possible lung cancer vignette, r=–0.28, 95% CI -0.66 to 0.22, P=0.261; possible 17 18 306 ovarian cancer vignette,For r=–0.55, peer 95% CI review –0.81 to –0.11, Ponly=0.019; possible breast cancer 19 20 307 vignette, r=–0.17, 95% CI –0.59 to 0.33, P=0.514; possible colorectal cancer vignette, r=0.02, 21 22 23 308 95% CI –0.45 to 0.48, P=0.146. 24 25 26 309 (Place Figure 4 here) 27 28 29 310 Correlation of diagnostic action with PCP demographics 30 31 32 311 There was no significant gender difference for the likelihood of taking immediate diagnostic 33 34 312 action: for female PCPs, the likelihood was 62.7%, for male GPs it was 61.3%, P=0.346. There 35 36 313 was, however, a link between the number of years since graduation and the likelihood of 37 http://bmjopen.bmj.com/ 38 39 314 taking immediate diagnostic action (Table 3): a mean of 55.5% of PCPs who had graduated 40 41 315 less than 10 years ago were likely to take diagnostic action, compared with a mean of 63.4% 42 43 316 for those who had graduated 10 or more years ago, P<0.001. 44

45 on September 28, 2021 by guest. Protected copyright. 46 317 (Place Table 3 here) 47 48 49 318 In the twelve countries with respondents who self-identified as working in remote or island 50 51 319 practices, those PCPs were significantly more likely to take immediate diagnostic action than 52 53 54 320 their colleagues (71.4% vs. 60.7%, P=0.021). There was a trend towards lower likelihood of 55 56 321 diagnostic testing and/or referral with larger practices, Table 4. 57 58 59 322 (Place Table 4 here) 60

15 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 18 of 47

1 2 323 Correlation of immediate diagnostic action with national healthcare expenditures 3 BMJ Open: first published as 10.1136/bmjopen-2019-035678 on 31 October 2020. Downloaded from 4 5 324 National healthcare expenditures, measured as Purchasing Power Parity (PPP) Current 6 7 325 International Dollars per capita [42], are given in Table 1. There was a significant negative 8 9 326 correlation between national per capita healthcare expenditure and PCPs’ likelihood of taking 10 11 12 327 immediate diagnostic action (r=–0.55, P=0.012), Figure 5. 13 14 328 (Place Figure 5 here) 15 16 17 18 329 For peer review only 19 Discussion 20 21 22 330 Principal findings 23 24 25 331 When faced with vignettes of patients with symptoms that could be due to cancer, there was a 26 27 332 marked variation between different European countries in PCPs’ stated actions. In all the 28 29 30 333 participating countries at least half of PCPs would have taken immediate diagnostic action (i.e. 31 32 334 either organised a diagnostic test or referred the patients to a specialist, or both). There was a 33 34 335 significant negative correlation between national healthcare expenditure levels and likelihood 35 36

37 336 of PCPs’ immediate diagnostic action, but no significant correlation between the likelihood of http://bmjopen.bmj.com/ 38 39 337 taking diagnostic action and cancer survival. PCPs who were more likely to arrange a 40 41 338 diagnostic test were also more likely to refer their patients to a specialist at the same time. 42 43 44 339 PCPs who had graduated more recently were less likely to take diagnostic action in these 45 on September 28, 2021 by guest. Protected copyright. 46 47 340 vignettes than their more experienced peers, but PCPs working in more remote locations 48 49 341 were more likely to take diagnostic action than their colleagues in other localities. PCPs 50 51 342 working alone or in smaller practices were more likely to take diagnostic action than those in 52 53 54 343 larger practices. 55 56 57 58 59 60

16 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 19 of 47 BMJ Open

1 2 344 Strengths and limitations of this study 3 BMJ Open: first published as 10.1136/bmjopen-2019-035678 on 31 October 2020. Downloaded from 4 5 345 One of the strengths of our study is the wide spectrum of participating centres, with four 6 7 346 countries from each of the Central, Eastern, Northern, Southern and Western European 8 9 347 geographical areas, providing variation in geography, socioeconomic and health systems, and 10 11 12 348 levels of healthcare spending. It included the views of PCPs who are not usually involved in 13 14 349 research. The questionnaire was carefully developed and piloted by GPs and other PCPs, and 15 16 350 therefore grounded in their clinical experience. While low survey response rates are common 17 18 For peer review only 19 351 in primary care [37] and are known to vary between countries, those in our study compared 20 21 352 favourably with those of a recent ICBP survey, in which response rates varied from 5.5% to 22 23 353 45.6% [28]. There is evidence that responses to vignettes in surveys correspond well to 24 25 26 354 clinical practice [43], and such surveys have previously been used to study primary care 27 28 355 investigation preferences in patients who could have cancer [28,44]. 29 30 31 356 Our study examined groups of PCPs as the unit of observation. Measuring the possible link 32 33 357 between these groups’ stated actions and national cancer survival creates the risk of an 34 35 36 358 ecological fallacy [45], as any identified correlation may be an indicator of the effect of

37 http://bmjopen.bmj.com/ 38 359 unmeasured system factors. While the demographic data that we collected included the 39 40 360 gender of participants and the number of years that they had been in practice, we have found 41 42 43 361 no equivalent data on national PCP populations that would allow us to assess how 44

45 362 representative our samples were. on September 28, 2021 by guest. Protected copyright. 46 47 48 363 Most samples were taken from each local lead’s own locality, and these may not have been 49 50 364 representative of their nations as a whole [46]. The recruitment method used in this study 51 52 365 resulted in variable response rates, leading to a risk of non-response bias [37]. However, the 53 54 55 366 goal of 50 survey participants per country and more than 1000 respondents in total was 56 57 367 achieved. We have no data on non-responders as the survey was anonymous, and the low 58 59 368 response rates in some countries means that we are unable to be sure how representative 60

17 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 20 of 47

1 2 369 their results are of their wider PCP population. However, the respondent anonymity might 3 BMJ Open: first published as 10.1136/bmjopen-2019-035678 on 31 October 2020. Downloaded from 4 370 have reduced the risk of social desirability bias. It is possible that the PCPs with the most 5 6 371 interest in diagnostic decision-making were the most likely to respond. Our respondents were 7 8 9 372 only able to select their management decisions from the options that we gave them; this 10 11 373 means they might have selected an option that they would not have thought of without 12 13 374 prompting, which may have affected the diagnostic testing and referral rates. 14 15 16 17 375 Interpretation of the results 18 For peer review only 19 376 Diagnostic testing and referral rates for these vignettes differed widely between participating 20 21 22 377 countries. Whereas we might expect that referring a patient would result in PCPs being less 23 24 378 likely to organise a diagnostic test on their patient, and vice versa, this was not confirmed in 25 26 379 the survey: the more likely PCPs were to refer a patient, the more likely they were to organise 27 28 29 380 a diagnostic test at the same time. The reasons for this are unclear, but it may be that PCPs 30 31 381 who are more worried about patients with unexplained symptoms are more likely to take all 32 33 34 382 the measures available to help them make a diagnosis. 35 36 383 While we found that PCPs working in remote locations were more likely to take immediate 37 http://bmjopen.bmj.com/ 38 39 384 diagnostic action, this may be due to confounding, as doctors with young families have been 40 41 385 found to be less likely to work rurally [47], and we found that more experienced PCPs were 42 43 386 more likely to take diagnostic action, possibly because they are more likely to have previously 44

45 on September 28, 2021 by guest. Protected copyright. 46 387 seen patients with those symptoms who were subsequently found to have a serious diagnosis. 47 48 49 388 The absence of significant correlation between the likelihood of taking immediate diagnostic 50 51 389 action and cancer survival could be related to a range of factors not assessed in the study, 52 53 390 including variations in time between first symptoms and presentation to a doctor [48], and, in 54 55 56 391 some countries, a greater reliance on poorly funded secondary care services. 57 58 59 60

18 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 21 of 47 BMJ Open

1 2 392 Comparison with existing literature 3 BMJ Open: first published as 10.1136/bmjopen-2019-035678 on 31 October 2020. Downloaded from 4 5 393 In contrast to our study, an ICBP study showed a positive correlation between jurisdictional 6 7 394 cancer survival rates and readiness to investigate or refer to secondary care in 4 out of 5 of its 8 9 395 vignettes [28]. However, it only studied six, relatively wealthy, countries (Australia, Canada, 10 11 12 396 Denmark, Norway, Sweden and the United Kingdom). While the ICBP research found no 13 14 397 health system characteristics that explained their findings, our study shows a link between 15 16 398 PCP diagnostic actions and practice location, size of practice and PCP experience. Our finding 17 18 For peer review only 19 399 of a lower likelihood of diagnostic action for PCPs working in larger practices corresponds 20 21 400 with other study findings [49]. Our evidence that PCPs working in more remote locations 22 23 401 were more likely to take diagnostic action for these vignettes links across to evidence that 24 25 26 402 such remote living is associated with more rapid cancer diagnosis and treatment following GP 27 28 403 referral [50]. While our data show higher diagnostic action rates in PCPs with more years 29 30 404 since graduation, the opposite was found in a Finnish study [51], and no difference was found 31 32 33 405 in a United Kingdom study [52]. However, those two studies did not specifically study 34 35 406 referrals for suspected cancer: it may be that experienced PCPs are more likely to recognise 36 37 407 symptoms that suggest a possibility of cancer, even in the absence of ‘red-flag’ symptoms. http://bmjopen.bmj.com/ 38 39 40 408 The extent to which respondents were gatekeepers, and needed to authorise their patients’ 41 42 43 409 access to specialist care and diagnostic tests [53], may have been a factor in their diagnostic 44

45 410 actions. There has been a suggestion that stronger gatekeeper systems are linked with lower on September 28, 2021 by guest. Protected copyright. 46 47 411 1-year relative cancer survival than non-gatekeeper systems [54], possibly because 48 49 50 412 gatekeeping can result in cost and resource decisions which reduce the likelihood of early 51 52 413 referral [55]. However, there are important variations in the level of gatekeeping between 53 54 414 countries, with no simple binary model as to whether or not a country has a ‘GP-as-gate- 55 56 57 415 keeper’ system, and a European study found no association between cancer survival and a 58 59 416 probability of presentation to a GP [30]. 60

19 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 22 of 47

1 2 417 Implications for research and practice 3 BMJ Open: first published as 10.1136/bmjopen-2019-035678 on 31 October 2020. Downloaded from 4 5 418 We have found no evidence that the poorer survival in some of our surveyed countries is due 6 7 419 to delayed referral from primary care. It may be that in these countries poorer survival is 8 9 420 related to cancer patients presenting later, or being less efficiently investigated, diagnosed 10 11 12 421 and treated in secondary care. It may also be that in those countries a stronger gate-keeper 13 14 422 role for primary care, allowing secondary care to focus on patients that have a higher risk of 15 16 423 cancer, would be more effective. 17 18 For peer review only 19 424 While it might be expected that PCPs who are more likely to arrange a diagnostic test would 20 21 425 be less likely to refer their patient in the same consultation, we found the opposite was the 22 23 24 426 case, and research is needed to explain this. We cannot explain our finding that PCPs in 25 26 427 countries with a higher per capita healthcare expenditure were less likely to take immediate 27 28 428 diagnostic action, and the reasons for this need investigation. 29 30 31 32 33 429 Conclusions 34 35 36 430 When given vignettes of patients with a low but significant possibility of cancer, more than

37 http://bmjopen.bmj.com/ 38 431 half of PCPs across Europe would take diagnostic action, most often by ordering diagnostic 39 40 432 tests. However, there are substantial between-country differences, which are linked with the 41 42 43 433 healthcare organisation in those countries, as well as their levels of investment in healthcare. 44

45 434 In countries with higher healthcare expenditure, immediate diagnostic action is less likely. on September 28, 2021 by guest. Protected copyright. 46 47 48 49 435 List of abbreviations 50 51 52 436 CI Confidence Interval 53 54 55 437 CT Computerised Tomography 56 57 58 438 GNP Gross National Product 59 60

20 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 23 of 47 BMJ Open

1 2 439 GP General practitioner 3 BMJ Open: first published as 10.1136/bmjopen-2019-035678 on 31 October 2020. Downloaded from 4 440 ICSS International Cancer Survival Standards 5 6 7 441 ICBP International Cancer Benchmarking Partnership 8 9 10 442 ÖRG Örenäs Research Group 11 12 13 443 PCP Primary Care Physician 14 15 16 444 PPP Purchasing Power Parity 17 18 For peer review only 19 445 PPV Positive Predictive Value 20 21 22 23 446 Declarations 24 25 26 447 Ethics and consent to participate 27 28 29 448 Ethical approval for the study was given by the University of Bath Research Ethics Approval 30 31 449 Committee for Health (approval date: 24th November 2014; REACH reference number: EP 32 33 34 450 14/15 66). Other countries’ study leads either achieved local ethical approval or gave 35 36 451 statements that formal ethical approval was not needed in their jurisdictions (see 37 http://bmjopen.bmj.com/ 38 39 452 supplementary file). Consent was implied by agreeing to take part in the survey. 40 41 42 453 Availability of data 43 44 45 454 To avoid the risk of identification of individual participants, the datasets generated and on September 28, 2021 by guest. Protected copyright. 46 47 48 455 analysed during the current study are not publicly available. However, they are available from 49 50 456 the corresponding author on reasonable request. 51 52 53 457 Competing interests 54 55 56 458 None declared. 57 58 59 60

21 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 24 of 47

1 2 459 Patient consent for publication 3 BMJ Open: first published as 10.1136/bmjopen-2019-035678 on 31 October 2020. Downloaded from 4 5 460 Not required. 6 7 8 461 Funding 9 10 11 462 This study received no specific grant from any funding agency in the public, commercial or 12 13 463 not-for-profit sectors. ALN’s time is supported by the National Institute for Health Research 14 15 16 464 (NIHR) Imperial Patient Safety Translation Research Centre, with her infrastructure support 17 18 465 provided by the NIHRFor Imperial peer Biomedical review Research Centre only (BRC). 19 20 21 466 Author contributions 22 23 24 467 IA-A, JA, KB, MB, NB, G-JD, ME, GF, SGB, MH, RH, EJ, TK, MM-C, PM, ALN, AP, DP, MPS, JS-P, ES, 25 26 27 468 SS, GT, HT, PV and BW participated in the study design. All authors except GT were involved in 28 29 469 the data collection. All authors contributed to the manuscript and approved the final version. 30 31 470 MH had overall responsibility for the study design, recruitment of local leads, analysis of data 32 33 34 471 and interpretation of results. GT advised on the study design and the statistical analysis. 35 36

37 472 Acknowledgements http://bmjopen.bmj.com/ 38 39 40 473 We would like to thank all the PCPs who piloted the questionnaire and those who completed 41 42 474 the survey. We would also like to thank the European GP Research Network for its support. 43 44 475 We are grateful to Prof. Barbara Silverman and Prof. Lital Keinan for the data on cancer 45 on September 28, 2021 by guest. Protected copyright. 46 47 476 survival rates in Israel, and to Dr Yochai Schonmann for his work on those data. Two of the 48 49 477 vignettes were used by kind permission of the ICBP; we also thank Dr Peter Murchie and Dr 50 51 478 Rhona Auckland, who generously provided the other two vignettes. Prof. Antonius Schneider 52 53 54 479 kindly organised the Technical University of Munich’s data collection. 55 56 57 58 59 60

22 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 25 of 47 BMJ Open

1 2 480 References 3 BMJ Open: first published as 10.1136/bmjopen-2019-035678 on 31 October 2020. Downloaded from 4 5 481 1 Møller H, Linklater KM, Robinson D. A visual summary of the EUROCARE-4 results: a UK 6 7 482 perspective. Br J Cancer 2009;101:S110-4. 8 9 10 11 483 2 EUROCARE. EUROCARE-5. Istituto Nazionale Tumori (Milan) and Istituto Superiore di 12 13 484 Sanità (Rome) 2014. 14 15 16 485 3 EUROCARE. EUROCARE-4. Istituto Nazionale Tumori (Milan) and Istituto Superiore di 17 18 For peer review only 19 486 Sanità (Rome) 2011. 20 21 22 487 4 Thomson CS, Forman D. Cancer survival in England and the influence of early diagnosis: 23 24 488 what can we learn from recent EUROCARE results? Br J Cancer 2009;101 Suppl 2:S102– 25 26 27 489 S109. doi:10.1038/sj.bjc.6605399 28 29 30 490 5 De Angelis R, Sant M, Coleman MP, et al. Cancer survival in Europe 1999-2007 by country 31 32 491 and age: results of EUROCARE-5 - a population-based study. Lancet Oncol 2014;15:23–34. 33 34 35 36 492 6 Coleman MP, Forman D, Bryant H, et al. Cancer survival in Australia, Canada, Denmark,

37 http://bmjopen.bmj.com/ 38 493 Norway, Sweden, and the UK, 1995-2007 (the International Cancer Benchmarking 39 40 494 Partnership): an analysis of population-based cancer registry data. Lancet 2011;377:127– 41 42 43 495 38. 44

45 on September 28, 2021 by guest. Protected copyright. 46 496 7 Storm HH, Gislum M, Engholm G. [Cancer survival before and after initiating the Danish 47 48 497 Cancer Control plan]. Ugeskr Laeger 2008;170:3065–9. 49 50 51 52 498 8 Allemani C, Matsuda T, Carlo VD, et al. Global surveillance of trends in cancer survival 53 54 499 2000–14 (CONCORD-3): analysis of individual records for 37 513 025 patients diagnosed 55 56 57 500 with one of 18 cancers from 322 population-based registries in 71 countries. Lancet 58 59 501 2018;391:1023–75. doi:10.1016/S0140-6736(17)33326-3 60

23 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 26 of 47

1 2 502 9 Zahl PH, Jorgensen KJ, Gotzsche PC. Overestimated lead times in cancer screening has led 3 BMJ Open: first published as 10.1136/bmjopen-2019-035678 on 31 October 2020. Downloaded from 4 503 to substantial underestimation of overdiagnosis. Br J Cancer 2013;109:2014-9. doi: 5 6 504 10.1038/bjc.2013.427. Epub 2013 Aug 20. 7 8 9 10 505 10 Carter JL, Coletti RJ, Harris RP. Quantifying and monitoring overdiagnosis in cancer 11 12 506 screening: a systematic review of methods. BMJ 2015;350:g7773.:10.1136/bmj.g7773. 13 14 15 507 11 Maringe C, Walters S, Rachet B, et al. Stage at diagnosis and colorectal cancer survival in 16 17 18 508 six high-income Forcountries: peer a population-based review study onlyof patients diagnosed during 2000- 19 20 509 2007. Acta Oncol 2013;52:919-32. doi: 10.3109/0284186X.2013.764008. Epub 2013 Apr 21 22 510 15. 23 24 25 26 511 12 Walters S, Maringe C, Coleman MP, et al. Lung cancer survival and stage at diagnosis in 27 28 512 Australia, Canada, Denmark, Norway, Sweden and the UK: a population-based study, 29 30 513 2004-2007. Thorax 2013;68:551–64. doi:10.1136/thoraxjnl-2012-202297 31 32 33 34 514 13 Richards MA. The size of the prize for earlier diagnosis of cancer in England. Br J Cancer 35 36 515 2009;101 Suppl 2:S125–S129. doi:10.1038/sj.bjc.6605402 www.bjcancer.com

37 http://bmjopen.bmj.com/ 38 39 516 14 Woods LM, Coleman MP, Lawrence G, et al. Evidence against the proposition that “UK 40 41 42 517 cancer survival statistics are misleading”: simulation study with National Cancer Registry 43 44 518 data. BMJ 2011;342:d3399.:10.1136/bmj.d3399.

45 on September 28, 2021 by guest. Protected copyright. 46 47 519 15 Butler J, Foot C, Bomb M, et al. The International Cancer Benchmarking Partnership: An 48 49 50 520 international collaboration to inform cancer policy in Australia, Canada, Denmark, 51 52 521 Norway, Sweden and the United Kingdom. Health Policy 2013;112:148–55. 53 54 55 522 doi:https://doi.org/10.1016/j.healthpol.2013.03.021 56 57 58 59 60

24 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 27 of 47 BMJ Open

1 2 523 16 Green T, Atkin K, Macleod U. Cancer detection in primary care: insights from general 3 BMJ Open: first published as 10.1136/bmjopen-2019-035678 on 31 October 2020. Downloaded from 4 524 practitioners. Br J Cancer 2015;112:S41. doi:10.1038/bjc.2015.41 5 6 525 https://www.nature.com/articles/bjc201541#supplementary-information 7 8 9 10 526 17 Jacobsen MM, Silverstein SC, Quinn M, et al. Timeliness of access to lung cancer diagnosis 11 12 527 and treatment: A scoping literature review. Lung Cancer 2017;112:156–64. 13 14 528 doi:10.1016/j.lungcan.2017.08.011 15 16 17 18 529 18 Tørring ML, FrydenbergFor M,peer Hansen RP,review et al. Evidence only of increasing mortality with longer 19 20 530 diagnostic intervals for five common cancers: a cohort study in primary care. Eur J Cancer 21 22 531 2013;49:2187–98. doi:10.1016/j.ejca.2013.01.025 23 24 25 26 532 19 Neal RD, Tharmanathan P, France B, et al. Is increased time to diagnosis and treatment in 27 28 533 symptomatic cancer associated with poorer outcomes? Systematic review. Br J Cancer 29 30 534 2015;112 Suppl 1:S92–107. doi:10.1038/bjc.2015.48 31 32 33 34 535 20 Malmström M, Rasmussen BH, Bernhardson B-M, et al. “It is important that the process 35 36 536 goes quickly, isn’t it?” A qualitative multi-country study of colorectal or lung cancer

37 http://bmjopen.bmj.com/ 38 537 patients’ narratives of the timeliness of diagnosis and quality of care. Eur J Oncol Nurs 39 40 41 538 2018;34:82–8. doi:10.1016/j.ejon.2018.04.002 42 43 44 539 21 Foot C, Harrison T. How to improve cancer survival: explaining England’s relatively poor

45 on September 28, 2021 by guest. Protected copyright. 46 540 rates. 2011. 47 48 49 50 541 22 Harris M, Frey P, Esteva M, et al. How health system factors influence referral decisions in 51 52 542 patients that may have cancer: European symposium report. J Cancer Res Ther 53 54 55 543 2016;4(1):7–10. doi:10.14312/2052-4994.2016-2 56 57 58 59 60

25 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 28 of 47

1 2 544 23 Davies L, Petitti DB, Martin L, et al. Defining, Estimating, and Communicating 3 BMJ Open: first published as 10.1136/bmjopen-2019-035678 on 31 October 2020. Downloaded from 4 545 Overdiagnosis in Cancer Screening. Ann Intern Med 2018;169:36–43. doi:10.7326/M18- 5 6 546 0694 7 8 9 10 547 24 Jung M. Breast, prostate, and thyroid cancer screening tests and overdiagnosis. Curr Probl 11 12 548 Cancer 2017;41:71–9. doi:10.1016/j.currproblcancer.2016.11.006 13 14 15 549 25 Dubben H-H. [Early detection of prostate cancer: harm verified, benefit not verifiable]. 16 17 18 550 BundesgesundheitsblattFor Gesundheitsforschungpeer review Gesundheitsschutz only 2014;57:318–26. 19 20 551 doi:10.1007/s00103-013-1904-1 21 22 23 552 26 International Cancer Benchmarking Partnership. ICBP Newsletter. 2011. 24 25 26 27 553 27 Forbes LJL, Simon AE, Warburton F, et al. Differences in cancer awareness and beliefs 28 29 554 between Australia, Canada, Denmark, Norway, Sweden and the UK (the International 30 31 555 Cancer Benchmarking Partnership): do they contribute to differences in cancer survival? 32 33 34 556 Br J Cancer 2013;108:292–300. 35 36

37 557 28 Rose PW, Rubin G, Perera-Salazar R, et al. Explaining variation in cancer survival between http://bmjopen.bmj.com/ 38 39 558 11 jurisdictions in the International Cancer Benchmarking Partnership: a primary care 40 41 42 559 vignette survey. BMJ Open 2015;5:e007212. doi: 10.1136/bmjopen-2014-007212. 43 44

45 560 29 Rose PW, Hamilton W, Aldersey K, et al. Development of a survey instrument to on September 28, 2021 by guest. Protected copyright. 46 47 561 investigate the primary care factors related to differences in cancer diagnosis between 48 49 50 562 international jurisdictions. BMC Fam Pr 2014;15:122.:10.1186/1471-2296-15-122. 51 52 53 563 30 Harris M, Frey P, Esteva M, et al. How the probability of presentation to a primary care 54 55 56 564 clinician correlates with cancer survival rates: a European survey using vignettes. Scand J 57 58 565 Prim Health Care 2017;:27–34. doi:10.1080/02813432.2017.1288692 59 60

26 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 29 of 47 BMJ Open

1 2 566 31 Brett J, Staniszewska S, Mockford C, et al. Mapping the impact of patient and public 3 BMJ Open: first published as 10.1136/bmjopen-2019-035678 on 31 October 2020. Downloaded from 4 567 involvement on health and social care research: a systematic review. Health Expect 5 6 568 2014;17:637–50. doi:10.1111/j.1369-7625.2012.00795.x 7 8 9 10 569 32 Harris M, Taylor G, Örenäs Research Group. How health system factors affect primary care 11 12 570 practitioners’ decisions to refer patients for further investigation: protocol for a pan- 13 14 571 European ecological study. BMC Health Serv Res 2018;18:338. doi:10.1186/s12913-018- 15 16 17 572 3170-2 18 For peer review only 19 20 573 33 Hamilton W. The CAPER studies: five case-control studies aimed at identifying and 21 22 574 quantifying the risk of cancer in symptomatic primary care patients. Br J Cancer 2009;101 23 24 25 575 Suppl 2:S80-6. doi:10.1038/sj.bjc.6605396 26 27 28 576 34 Hamilton W, Peters TJ, Bankhead C, et al. Risk of ovarian cancer in women with symptoms 29 30 577 in primary care: population based case-control study. BMJ 2009;339:b2998. 31 32 33 578 doi:10.1136/bmj.b2998 34 35 36 579 35 Walker S, Hyde C, Hamilton W. Risk of breast cancer in symptomatic women in primary

37 http://bmjopen.bmj.com/ 38 580 care: a case-control study using electronic records. Br J Gen Pract 2014;64:e788-93. 39 40 41 581 doi:10.3399/bjgp14X682873 42 43 44 582 36 Harris M, Vedsted P, Esteva M, et al. Identifying important health system factors that

45 on September 28, 2021 by guest. Protected copyright. 46 583 influence primary care practitioners’ referrals for cancer suspicion: a European cross- 47 48 49 584 sectional survey. BMJ Open 2018;8. doi:10.1136/bmjopen-2018-022904 50 51 52 585 37 Pit SW, Vo T, Pyakurel S. The effectiveness of recruitment strategies on general 53 54 55 586 practitioner’s survey response rates - a systematic review. BMC Med Res Methodol 56 57 587 2014;14:76. doi:10.1186/1471-2288-14-76 58 59 60

27 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 30 of 47

1 2 588 38 Shaghaghi A, Bhopal RS, Sheikh A. Approaches to Recruiting ‘Hard-To-Reach’ Populations 3 BMJ Open: first published as 10.1136/bmjopen-2019-035678 on 31 October 2020. Downloaded from 4 589 into Re-search: A Review of the Literature. Health Promot Perspect 2011;1:86–94. 5 6 590 doi:10.5681/hpp.2011.009 7 8 9 10 591 39 Atkinson R, Flint J. Accessing Hidden and Hard-to-Reach Populations: Snowball Research 11 12 592 Strategies. 2001. 13 14 15 593 40 Dobrow MJ, Orchard MC, Golden B, et al. Response audit of an Internet survey of health 16 17 18 594 care providers andFor administrators: peer implicationsreview for determinationonly of response rates. J Med 19 20 595 Internet Res 2008;10:e30. doi:10.2196/jmir.1090 21 22 23 596 41 Corazziari I, Quinn M, Capocaccia R. Standard cancer patient population for age 24 25 26 597 standardising survival ratios. Eur J Cancer Oxf Engl 1990 2004;40:2307–16. 27 28 598 doi:10.1016/j.ejca.2004.07.002 29 30 31 599 42 The World Bank. Current health expenditure per capita, PPP (current international $). 32 33 34 600 2018.https://data.worldbank.org/indicator/SH.XPD.CHEX.PP.CD 35 36

37 601 43 Peabody JW, Luck J, Glassman P, et al. Measuring the quality of physician practice by using http://bmjopen.bmj.com/ 38 39 602 clinical vignettes: a prospective validation study. Ann Intern Med 2004;141:771–80. 40 41 42 43 603 44 Banks J, Hollinghurst S, Bigwood L, et al. Preferences for cancer investigation: a vignette- 44

45 604 based study of primary-care attendees. Lancet Oncol 2014;15:232–40. on September 28, 2021 by guest. Protected copyright. 46 47 605 doi:10.1016/S1470-2045(13)70588-6 48 49 50 51 606 45 Freedman D. Ecological Inference and the Ecological Fallacy. 1999. 52 53 54 607 46 Greenacre Z. The Importance of Selection Bias in Internet Surveys. Open J Stat 55 56 57 608 2016;6:397–404. doi:http://dx.doi.org/10.4236/ojs.2016.63035 58 59 60

28 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 31 of 47 BMJ Open

1 2 609 47 McGrail MR, Russell DJ, O’Sullivan BG. Family effects on the rurality of GP’s work location: 3 BMJ Open: first published as 10.1136/bmjopen-2019-035678 on 31 October 2020. Downloaded from 4 610 a longitudinal panel study. Hum Resour Health 2017;15. doi:10.1186/s12960-017-0250-z 5 6 7 611 48 Esteva M, Leiva A, Ramos M, et al. Factors related with symptom duration until diagnosis 8 9 10 612 and treatment of symptomatic colorectal cancer. BMC Cancer 2013;13:87. 11 12 13 613 49 Verstappen WH, ter Riet G, Dubois WI, et al. Variation in test ordering behaviour of GPs: 14 15 614 professional or context-related factors? Fam Pract 2004;21:387–95. 16 17 18 615 doi:10.1093/fampra/cmh408For peer review only 19 20 21 616 50 Turner M, Fielding S, Ong Y, et al. A cancer geography paradox? Poorer cancer outcomes 22 23 617 with longer travelling times to healthcare facilities despite prompter diagnosis and 24 25 26 618 treatment: a data-linkage study. Br J Cancer 2017;117:439–49. doi:10.1038/bjc.2017.180 27 28 29 619 51 Vehvilainen AT, Kumpusalo EA, Voutilainen SO, et al. Does the doctors’ professional 30 31 620 experience reduce referral rates? Evidence from the Finnish referral study. Scand J Prim 32 33 34 621 Health Care 1996;14:13–20. 35 36

37 622 52 Wilkin D, Smith AG. Variation in general practitioners’ referral rates to consultants. J R Coll http://bmjopen.bmj.com/ 38 39 623 Gen Pr 1987;37:350–3. 40 41 42 43 624 53 Franks P, Clancy CM, Nutting PA. Gatekeeping revisited--protecting patients from 44

45 625 overtreatment. N Engl J Med 1992;327:424–9. doi:10.1056/nejm199208063270613 on September 28, 2021 by guest. Protected copyright. 46 47 48 626 54 Vedsted P, Olesen F. Are the serious problems in cancer survival partly rooted in 49 50 51 627 gatekeeper principles? Br J Gen Pr 2011;61:512–3. 52 53 54 628 55 Neal RD. Commentary. Cancer diagnosis - the role of urgent referral guidelines. Br J Gen 55 56 57 629 Pract 2010;60:127. doi:10.3399/bjgp10X483427 58 59 60 630

29 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open: first published as 10.1136/bmjopen-2019-035678 on 31 October 2020. Downloaded from

BMJ Open Page 32 of 47

1 2 Table 1. Number of respondents per country, response rates, mean national cancer survival rates for the four cancers of interest, and 3 healthcare expenditures (given as PPP Current International Dollars per capita). 4 5 6 7 8 Number of Number of Response 1-year 5-year Healthcare 9 10 respondents (% of PCPs rate (%) relative relative expenditure 11 12 all respondents)For peerinvited reviewcancer onlycancer per capita, 13 14 * * 15 survival (%) survival PPP$ 16 http://bmjopen.bmj.com/ 17 (%) 18 19 20 Respondents Bulgaria 59 (2.8) 90 65.5 59.6 38.4 1399 21 22 per country 23 Croatia 67 (3.2) 292 22.9 63.7 44.7 1652 24 on September 28, 2021 by guest. Protected copyright. 25 (in 26 Denmark 107 (5.1) 400 26.8 69.0 45.4 4782 27 alphabetical 28 29 order) England 65 (3.1) 300 21.7 65.2 42.7 3377 30 31 32 Finland 65 (3.1) 178 36.5 73.2 50.3 3701 33 34 35 France 59 (2.8) 550 10.7 74.9 49.8 4508 36 37 38 Germany 103 (4.9) 242 42.6 73.5 50.3 5182 39 40 41 42 30 43 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 BMJ Open: first published as 10.1136/bmjopen-2019-035678 on 31 October 2020. Downloaded from

Page 33 of 47 BMJ Open

1 2 Greece 68 (3.3) 318 21.4 Data not available 2098 3 4 5 Israel 75 (3.6) 339 22.1 79.2** 58.3* 2599 6 7 8 Italy 63 (3.0) 200 31.5 72.9 49.4 3239 9 10 11 Netherlands 113 (5.4) 1601 7.1 72.0 49.1 5202 12 For peer review only 13 14 Norway 90 (4.3) 500 18.0 72.8 49.9 6347 15 16 http://bmjopen.bmj.com/ 17 Poland 152 (7.3) 422 36.0 65.8 41.5 1570 18 19 20 Portugal 65 (3.1) 227 28.6 71.0 48.2 2690 21 22 23 Romania 177 (8.5) Not known Data not available 1079 24 on September 28, 2021 by guest. Protected copyright. 25 26 Scotland 65 (3.1) 350 18.6 66.5 43.7 3377 27 28 29 Slovenia 104 (5.0) 352 29.5 69.5 44.8 2698 30 31 32 Spain 446 (21.4) Not known 70.3 46.8 2966 33 34 35 Sweden 79 (3.8) 400 19.8 75.9 51.5 5219 36 37 38 Switzerland 64 (3.1) 100 64.0 75.7 50.2 6468 39 40 41 42 31 43 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 BMJ Open: first published as 10.1136/bmjopen-2019-035678 on 31 October 2020. Downloaded from

BMJ Open Page 34 of 47

1 2 Total 2086 (100) 3 4 5 6 7 * Calculated using International Cancer Survival Standards (ICSS). 8 9 10 ** Calculated from data provided by B. Silverman, Israel Ministry of Health (personal communication, 7 September 2017) and Y. Schonmann, 11 12 London School of Hygiene & Tropical ForMedicine (personal peer communication, review 7 September 2018). only 13 14 15 16 http://bmjopen.bmj.com/ 17 18 19 20 21 22 23 24 on September 28, 2021 by guest. Protected copyright. 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 32 43 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 Page 35 of 47 BMJ Open

1 2 Table 2. Demographic distributions of respondents. 3 BMJ Open: first published as 10.1136/bmjopen-2019-035678 on 31 October 2020. Downloaded from 4 5 6 7 8 Number (%) 9 10 11 Gender Female 1274 (61.1) 12 13 14 Male 790 (37.9) 15 16 17 Not stated 22 (1.1) 18 For peer review only 19 20 21 22 Years since graduation <10 years 331 (15.5) 23 24 25 10-19 years 553 (26.9) 26 27 28 20-29 years 609 (29.2) 29 30 31 30-39 years 499 (23.9) 32 33 34 40 years or over 76 (3.6) 35 36

37 Not stated 18 (0.9) http://bmjopen.bmj.com/ 38 39 40 41 42 Site of practice Urban 1238 (59.3) 43 44 45 Rural 485 (23.3) on September 28, 2021 by guest. Protected copyright. 46 47 48 Remote or Island 56 (2.7) 49 50 51 Mixed 295 (14.1) 52 53 54 Not stated 12 (0.6) 55 56 57 58 59 60

33 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 36 of 47

1 2 Number of doctors in 1 286 (13.7) 3 BMJ Open: first published as 10.1136/bmjopen-2019-035678 on 31 October 2020. Downloaded from 4 practice 5 2 233 (11.2) 6 7 8 3 226 (10.8) 9 10 11 4-5 347 (16.6) 12 13 14 6-7 259 (12.4) 15 16 17 8-9 172 (8.2) 18 For peer review only 19 10 or more 542 (26.0) 20 21 22 Not stated 21 (1.0) 23 24 25 26 27 28 29 30 31 32 33 34 35 36

37 http://bmjopen.bmj.com/ 38 39 40 41 42 43 44

45 on September 28, 2021 by guest. Protected copyright. 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

34 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 37 of 47 BMJ Open

1 2 Table 3. Overall likelihood of immediate diagnostic action by years since graduation. 3 BMJ Open: first published as 10.1136/bmjopen-2019-035678 on 31 October 2020. Downloaded from 4 5 6 7 Years since Likelihood of immediate 8 9 graduation diagnostic action, % 10 11 12 <10 years 55.5 13 14 15 10-19 years 62.8 16 17 18 20-29 years For peer63.1 review only 19 20 21 30-39 years 64.5 22 23 24 40 years or over 64.1 25 26 27 28 29 30 31 32 33 34 35 36

37 http://bmjopen.bmj.com/ 38 39 40 41 42 43 44

45 on September 28, 2021 by guest. Protected copyright. 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

35 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 38 of 47

1 2 Table 4. Overall likelihood of immediate diagnostic action by practice size. 3 BMJ Open: first published as 10.1136/bmjopen-2019-035678 on 31 October 2020. Downloaded from 4 5 6 7 Number of PCPs in Likelihood of immediate 8 9 10 respondent’s practice diagnostic action, % 11 12 13 1 72.5 14 15 2 67.4 16 17 18 3 For peer 62.5review only 19 20 21 4-5 62.9 22 23 24 6-7 58.2 25 26 27 8-9 57.1 28 29 30 10 or more 57.6 31 32 33 34 35 36

37 http://bmjopen.bmj.com/ 38 39 40 41 42 43 44

45 on September 28, 2021 by guest. Protected copyright. 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

36 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 39 of 47 BMJ Open

1 2 Figure Legends 3 BMJ Open: first published as 10.1136/bmjopen-2019-035678 on 31 October 2020. Downloaded from 4 5 Figure 1. Percentage of PCPs in each country who would organise a diagnostic test. 6 7 8 Figure 2. Percentage of PCPs in each country who would refer the patients to a specialist. 9 10 11 Figure 3. Percentage of PCPs in each country who would organise an investigation and/or 12 13 refer the patients to a specialist. 14 15 16 Figure 4. Percentage of PCPs in each country who would organise a diagnostic test. 17 18 For peer review only 19 Figure 5. Percentage of PCPs in each country who would refer the patients to a specialist. 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36

37 http://bmjopen.bmj.com/ 38 39 40 41 42 43 44

45 on September 28, 2021 by guest. Protected copyright. 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

37 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 40 of 47

1 2

3 BMJ Open: first published as 10.1136/bmjopen-2019-035678 on 31 October 2020. Downloaded from 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 For peer review only 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 http://bmjopen.bmj.com/ 38 39 40 41 42 43 44

45 on September 28, 2021 by guest. Protected copyright. 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 41 of 47 BMJ Open

1 2

37 http://bmjopen.bmj.com/ 38 39 40 41 42 43 44

45 on September 28, 2021 by guest. Protected copyright. 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 42 of 47

1 2

37 http://bmjopen.bmj.com/ 38 39 40 41 42 43 44

45 on September 28, 2021 by guest. Protected copyright. 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 43 of 47 BMJ Open

1 2

37 http://bmjopen.bmj.com/ 38 39 40 41 42 43 44

45 on September 28, 2021 by guest. Protected copyright. 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 44 of 47

1 2

37 http://bmjopen.bmj.com/ 38 39 40 41 42 43 44

45 on September 28, 2021 by guest. Protected copyright. 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 45 of 47 BMJ Open

1 Appendix 1. Vignettes used in the study. 2 Two of the vignettes were used by kind permission of the ICBP. Dr Peter Murchie and Dr

3 BMJ Open: first published as 10.1136/bmjopen-2019-035678 on 31 October 2020. Downloaded from 4 Rhona Auckland generously provided the other two vignettes. 5 6 7 First clinical case: Alexander 8 Alexander is 62 years old. He has had a respiratory tract infection for two weeks, with 9 10 increased sputum production and increased use of his salbutamol inhaler. He has recently had 11 a constant ache in his left shoulder. 12 He had chronic obstructive pulmonary disease (COPD) diagnosed by spirometry 2 years ago. 13 14 There is no other relevant past medical history. 15 He smokes 20 cigarettes a day and has done for over 40 years. 16 17 His current treatment is tiotropium, 1 dose daily, and a salbutamol inhaler for use as required. 18 There are no significant findiFor peerngs on physical examination and he is not acutely unwell. review only 19 20 21 Second clinical case: Maria 22 23 A 53 year-old woman, whose last menstrual period was 6 months ago, has had colicky right 24 lower abdominal pain for three weeks. She has noticed that her abdomen seems swollen and 25 she has urinary frequency. She has had no change in the frequency or consistency of her stools 26 (faeces). 27 28 She has had the same sexual partner for 20 years. 29 30 She is a frequent attender, often with complaints that remain undiagnosed. 31 There are no significant findings on abdominal examination and she is not acutely unwell. 32 33 Second clinical case: Victoria 34 35 Victoria is a 35 year-old who is breast-feeding. She has pain in her left breast. It is not related 36 to her menstrual cycle. She has recently noticed a small amount of abnormal discharge from 37 her nipple, with some eczema around the nipple. http://bmjopen.bmj.com/ 38 39 She has a long history of atopic dermatitis, which has only been on her elbows and knees and 40 has been treated by emollients. 41 42 She has lost weight rapidly but she is still heavier than she was before her pregnancy. 43 Her mother had a mastectomy for breast cancer 5 years ago, but she was told that this was not 44

45 familial. on September 28, 2021 by guest. Protected copyright. 46 Clinical examination confirms eczematous change around the nipple but there are no 47 abnormalities palpable in the breast. 48 49 50 Fourth clinical case: Peter 51 Peter is 22 years old. Over the past 4 weeks, he has had increasing episodes of abdominal pain 52 and rectal bleeding, often with diarrhoea. He has sometimes had rectal bleeding before, but he 53 54 has never been worried about it. 55 Two years ago, he was diagnosed as having coeliac disease, and he has had a strict gluten-free 56 diet since then. 57 58 His diet has not changed recently, but he has lost 5kg of weight. 59 60 There are no significant findings on physical examination, and he is not acutely unwell.

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 46 of 47

1 2 Supplementary file. Ethical and other approvals obtained in each Örenäs

3 BMJ Open: first published as 10.1136/bmjopen-2019-035678 on 31 October 2020. Downloaded from 4 Research Group participating jurisdiction 5 6 7 Date of Ethics Approvals obtained Reference 8 Approval 9 10 Bulgaria 29 October Medical University Plovdiv Ethical P-7820 11 2015 Commission 12 13 Croatia 16 December Nastavni Zovod Za Javno Zdravstvo 08-820-61/31-15 14 2016 15 16 Denmark 7 May 2016 Danish Data Protection Agency; 2009-41-3471 17 according to Danish law and the 18 For peerCentral Denmark Region Committees review only 19 on Health Research Ethics, approval 20 21 by the National Committee on Health 22 Research Ethics was not required as 23 no biomedical intervention was 24 performed. 25 26 Finland 16 November Academic Ethics Committee of the 16 November 2016 27 2016 Tampere Region 28 29 France N/A In France, research ethics approval 30 was not required as no biomedical 31 intervention was performed. 32 33 Germany 15 January Ethik-Kommission Universität 16-6747-BO 34 2016 Duisberg-Essen 35 36 Greece N/A In Greece, research ethics approval

37 was not required as no biomedical http://bmjopen.bmj.com/ 38 intervention was performed. 39 40 Israel N/A In Israel, research ethics approval 41 was not required as no biomedical 42 intervention was performed. 43 44 Italy N/A In Italy the approval of the ethical Decreto Legislativo n.

45 committee is not required when a 211 (24 giugno on September 28, 2021 by guest. Protected copyright. 46 study is neither an interventional 2003)˂2001/20/EC 47 nor an observational study on 48 pharmacological treatment. 49 50 Netherlands 27 June 2016 medisch- METC 16-4-113 51 ethischetoetsingscommissie (METC) 52 azM/UM Maastricht UMC+ 53 54 Norway N/A In Norway, research ethics approval 55 was not required as no biomedical 56 57 intervention was performed. 58 Poland 28 January Komisja Bioetyczena Uniwersytetu R_I_022/10/2016 59 2016 Medycznego w Bialymstoku 60

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 47 of 47 BMJ Open

Portugal N/A In Portugal, research ethics approval 1 was not required as no biomedical 2 intervention was performed.

3 BMJ Open: first published as 10.1136/bmjopen-2019-035678 on 31 October 2020. Downloaded from 4 Romania N/A In Romania, research ethics approval 5 was not required as no biomedical 6 7 intervention was performed. 8 Slovenia 8 December Komisija Republike Slovenije KME 113/08/14 9 10 2014 Medicinsko Etiko 11 Spain 25 October Comissio d’Investigacio Govern de Palma 27oct15 12 2015 les Illes Balears 13 14 23 Decmber Informe del Comite Etic P15/159 15 2015 d’Investigacio Clinica 16 17 Sweden N/A In Sweden, research ethics approval 18 For peerwas not required as no biomedical review only 19 intervention was performed. It does 20 21 not fall under the law of research on 22 human subjects to ask professionals 23 about their work and how they 24 perceive it. 25 26 Switzerland N/A Swiss law on human research 27 (Humanforschungsgesetz, HFG) does 28 not require that an ethics committee 29 approve collection and analysis of 30 non-medical and anonymous data. 31 32 United 24 November Research Ethics Approval Committee EP 14/15 66 33 Kingdom 2014 for Health, University of Bath 34 35 36

37 http://bmjopen.bmj.com/ 38 39 40 41 42 43 44

45 on September 28, 2021 by guest. Protected copyright. 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 48 of 47

1 2 Primary care practitioner diagnostic action when the patient may have cancer: a vignette survey in 20

3 European countries BMJ Open: first published as 10.1136/bmjopen-2019-035678 on 31 October 2020. Downloaded from 4 5 STROBE Statement – checklist of items that should be included in reports of cross-sectional studies 6 Item 7 No Recommendation 8 Title and abstract 1 (a) Indicate the study’s design with a commonly used term in the title or the abstract 9 10 [Within the title page 1 and Design section of the abstract page 4] 11 (b) Provide in the abstract an informative and balanced summary of what was done 12 and what was found [pages 4-5] 13 14 Introduction 15 Background/rationale 2 Explain the scientific background and rationale for the investigation being reported 16 [Pages 7-8] 17 Objectives 3 State specific objectives, including any prespecified hypotheses [Page 8] 18 For peer review only 19 Methods 20 Study design 4 Present key elements of study design early in the paper [Page 9] 21 Setting 5 Describe the setting, locations, and relevant dates, including periods of recruitment, 22 23 exposure, follow-up, and data collection [Pages 9-10] 24 Participants 6 (a) Give the eligibility criteria, and the sources and methods of selection of 25 participants [Pages 11-12] 26 27 Variables 7 Clearly define all outcomes, exposures, predictors, potential confounders, and effect 28 modifiers. Give diagnostic criteria, if applicable [Pages 10-13] 29 Data sources/ 8* For each variable of interest, give sources of data and details of methods of 30 measurement assessment (measurement). Describe comparability of assessment methods if there is 31 32 more than one group [Page 10-11] 33 Bias 9 Describe any efforts to address potential sources of bias [Page 3] 34 Study size 10 Explain how the study size was arrived at [Page 11] 35 Quantitative variables 11 Explain how quantitative variables were handled in the analyses. If applicable, 36

37 describe which groupings were chosen and why [Pages 12-13] http://bmjopen.bmj.com/ 38 Statistical methods 12 (a) Describe all statistical methods, including those used to control for confounding 39 [Pages 12-13] 40 41 (b) Describe any methods used to examine subgroups and interactions [N/A] 42 (c) Explain how missing data were addressed [N/A] 43 (d) If applicable, describe analytical methods taking account of sampling strategy 44 [N/A] 45 on September 28, 2021 by guest. Protected copyright. 46 (e) Describe any sensitivity analyses [Page 13] 47 Results 48 Participants 13* (a) Report numbers of individuals at each stage of study—eg numbers potentially 49 50 eligible, examined for eligibility, confirmed eligible, included in the study, 51 completing follow-up, and analysed [Page 13] 52 (b) Give reasons for non-participation at each stage [N/A] 53 54 (c) Consider use of a flow diagram [N/A] 55 Descriptive data 14* (a) Give characteristics of study participants (eg demographic, clinical, social) and 56 information on exposures and potential confounders [Tables 2 and 3] 57 (b) Indicate number of participants with missing data for each variable of interest 58 59 [N/A] 60 Outcome data 15* Report numbers of outcome events or summary measures [N/A] Main results 16 (a) Give unadjusted estimates and, if applicable, confounder-adjusted estimates and

1 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 49 of 47 BMJ Open

1 2 their precision (eg, 95% confidence interval). Make clear which confounders were

3 adjusted for and why they were included [Pages 14-16] BMJ Open: first published as 10.1136/bmjopen-2019-035678 on 31 October 2020. Downloaded from 4 5 (b) Report category boundaries when continuous variables were categorized [N/A] 6 (c) If relevant, consider translating estimates of relative risk into absolute risk for a 7 meaningful time period [N/A] 8 Other analyses 17 Report other analyses done—eg analyses of subgroups and interactions, and 9 10 sensitivity analyses [Pages 14-16] 11 Discussion 12 Key results 18 Summarise key results with reference to study objectives [Page 16] 13 14 Limitations 19 Discuss limitations of the study, taking into account sources of potential bias or 15 imprecision. Discuss both direction and magnitude of any potential bias [Pages 17- 16 18] 17 18 Interpretation For20 Give peer a cautious overall review interpretation ofonly results considering objectives, limitations, 19 multiplicity of analyses, results from similar studies, and other relevant evidence 20 [Page 20] 21 Generalisability 21 Discuss the generalisability (external validity) of the study results [Page 20] 22 23 Other information 24 Funding 22 Give the source of funding and the role of the funders for the present study and, if 25 applicable, for the original study on which the present article is based [Page 22] 26 27 28 *Give information separately for exposed and unexposed groups. 29 30 Note: An Explanation and Elaboration article discusses each checklist item and gives methodological background and 31 32 published examples of transparent reporting. The STROBE checklist is best used in conjunction with this article (freely 33 available on the Web sites of PLoS Medicine at http://www.plosmedicine.org/, Annals of Internal Medicine at 34 http://www.annals.org/, and Epidemiology at http://www.epidem.com/). Information on the STROBE Initiative is 35 available at www.strobe-statement.org. 36

37 http://bmjopen.bmj.com/ 38 39 40 41 42 43 44

45 on September 28, 2021 by guest. Protected copyright. 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

2 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open BMJ Open: first published as 10.1136/bmjopen-2019-035678 on 31 October 2020. Downloaded from

Primary care practitioner diagnostic action when the patient may have cancer: an exploratory vignette study in 20 European countries ForJournal: peerBMJ Open review only Manuscript ID bmjopen-2019-035678.R1

Article Type: Original research

Date Submitted by the 18-Mar-2020 Author:

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 1 of 61 BMJ Open BMJ Open: first published as 10.1136/bmjopen-2019-035678 on 31 October 2020. Downloaded from

1 2 3 Weltermann, Birgitta; University of Bonn, Institut für Hausarztmedizin 4 Taylor, Gordon; University of Exeter, College of Medicine and Health 5 6 Primary Subject General practice / Family practice 7 Heading: 8 Secondary Subject Heading: Oncology, Public health 9 10 International health services < HEALTH SERVICES ADMINISTRATION & Keywords: 11 MANAGEMENT, Adult oncology < ONCOLOGY, PRIMARY CARE 12 13 14 15 16 For peer review only 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 http://bmjopen.bmj.com/ 34 35 36 37 38 39 40 41 on September 28, 2021 by guest. Protected copyright. 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 2 of 61

1 2

37 http://bmjopen.bmj.com/ 38 39 40 41 42 43 44

45 on September 28, 2021 by guest. Protected copyright. 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 3 of 61 BMJ Open

1 2 3 BMJ Open: first published as 10.1136/bmjopen-2019-035678 on 31 October 2020. Downloaded from 4 1 Primary care practitioner diagnostic action when the patient may have 5 6 7 2 cancer: an exploratory vignette study in 20 European countries 8 9 10 11 12 3 Authors 13 14 15 4 Michael Harris (corresponding author), Department for Health, University of Bath, Claverton 16 17 5 Down, Bath BA2 7AY, UK; telephone: +44 1761 241366; fax: none. Also: Institute of Primary 18 For peer review only 19 20 6 Health Care Bern (BIHAM), University of Bern, Bern, Switzerland 21 22 7 [email protected] 23 24 25 8 Mette Brekke, Department of General Practice and General Practice Research Unit, University 26 27 9 of Oslo, Oslo, Norway. [email protected] 28 29 30 10 Geert-Jan Dinant, Department of General Practice, Maastricht University, Maastricht, 31 32 11 Netherlands. [email protected] 33 34 35 12 Magdalena Esteva, Research Unit, Majorca Primary Health Care Department, Balearic Islands 36

37 http://bmjopen.bmj.com/ 38 13 Health Research Institute (IdISBa), Preventive Activities and Health Promotion Network, 39 40 14 Carlos III Institute of Health (RedIAPP-RETICS), Palma, Spain. [email protected] 41 42 43 15 Robert D. Hoffman, Department of Family Medicine, Tel Aviv University, Tel Aviv, Israel. 44

45 16 [email protected] on September 28, 2021 by guest. Protected copyright. 46 47 48 17 Mercè Marzo-Castillejo, Unitat de Suport a la Recerca, IDIAP Jordi Gol, Institut Català de la 49 50 18 Salut, Barcelona, Spain. [email protected] 51 52 53 19 Peter Murchie, Division of Applied Health Sciences - Academic Primary Care, University of 54 55 20 Aberdeen, Aberdeen, UK. [email protected] 56 57 58 59 60

1 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 4 of 61

1 2 21 Ana Luísa Neves, Centre for Health Policy, Imperial College, London, UK. Also CINTESIS 3 BMJ Open: first published as 10.1136/bmjopen-2019-035678 on 31 October 2020. Downloaded from 4 22 (Centre for Health Technology and Services Research) and MEDCIDS (Department of 5 6 23 Community Medicine, Information and Health Decision Sciences), Faculty of Medicine, 7 8 9 24 University of Porto, Porto, Portugal [email protected] 10 11 25 Emmanouil Smyrnakis, Laboratory of Primary Health Care, General Practice and Health 12 13 14 26 Services Research, Aristotle University of Thessaloniki, Thessaloniki, Greece. 15 16 27 [email protected] 17 18 For peer review only 19 28 Peter Vedsted, The Research Unit for General Practice, Aarhus University, Aarhus, Denmark. 20 21 29 [email protected] 22 23 24 30 Isabelle Aubin-Auger, Department of General Practice, Université Paris Diderot, Paris, France. 25 26 27 31 [email protected] 28 29 32 Joseph Azuri, Sackler Faculty of Medicine, Tel Aviv University, Tel Aviv, Israel. 30 31 32 33 [email protected] 33 34 35 34 Krzysztof Buczkowski, Department of Family Medicine, Nicolaus Copernicus University, 36

37 35 Toruń, Poland. [email protected] http://bmjopen.bmj.com/ 38 39 40 36 Nicola Buono, Department of General Practice, National Society of Medical Education in 41 42 37 General Practice (SNaMID), Caserta, Italy. [email protected] 43 44

45 38 Gergana Foreva, Medical Center BROD, Plovdiv, Bulgaria. [email protected] on September 28, 2021 by guest. Protected copyright. 46 47 48 39 Svjetlana Gašparović Babić, Croatian Health Insurance Fund, Rijeka, Croatia. 49 50 40 [email protected] 51 52 53 41 Eva Jakob, Primary Health Centre, Centro de Saúde Sarria, Sarria, Lugo, Spain. 54 55 42 [email protected] 56 57 58 59 60

2 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 5 of 61 BMJ Open

1 2 43 Tuomas Koskela, Faculty of Medicine and Health Technology, Tampere University, Tampere, 3 BMJ Open: first published as 10.1136/bmjopen-2019-035678 on 31 October 2020. Downloaded from 4 44 Finland. [email protected] 5 6 7 45 Davorina Petek, Department of Family Medicine, University of Ljubljana, Ljubljana, Slovenia. 8 9 46 [email protected] 10 11 12 47 Marija Petek Ster, Department of Family Medicine, University of Ljubljana, Ljubljana, Slovenia. 13 14 48 [email protected] 15 16 17 49 Aida Puia, Teaching Assistant, Family Medicine Department, Iuliu Hatieganu University of 18 For peer review only 19 20 50 Medicine and Pharmacy, Cluj-Napoca, Romania. [email protected] 21 22 51 Jolanta Sawicka-Powierza, Department of Family Medicine, Medical University of Bialystok, 23 24 25 52 Bialystok, Poland. [email protected] 26 27 28 53 Sven Streit, Institute of Primary Health Care Bern (BIHAM), University of Bern, Bern, 29 30 54 Switzerland. [email protected] 31 32 33 55 Hans Thulesius Department of Clinical Sciences, Lund University, Malmö, Sweden and 34 35 56 Department of Research and Development, Region Kronoberg, Sweden 36 37 57 [email protected] http://bmjopen.bmj.com/ 38 39 40 58 Birgitta Weltermann, Institut für Hausarztmedizin, University of Bonn, Bonn, Germany. 41 42 43 59 [email protected] 44 45 60 Gordon Taylor, College of Medicine and Health, University of Exeter, Exeter, UK. on September 28, 2021 by guest. Protected copyright. 46 47 48 61 [email protected] 49 50 51 62 52 53 54 55 56 57 58 59 60

3 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 6 of 61

1 2 63 Abstract 3 BMJ Open: first published as 10.1136/bmjopen-2019-035678 on 31 October 2020. Downloaded from 4 5 64 Objectives 6 7 8 65 Cancer survival rates vary widely between European countries, with differences in timeliness 9 10 11 66 of diagnosis thought to be one key reason. There is little evidence on the way in which 12 13 67 different healthcare systems influence Primary Care Practitioners’ (PCPs’) referral decisions 14 15 68 in patients that could have cancer. 16 17 18 69 This study aimed to Forexplore PCPs’peer diagnostic review actions (whether only or not they perform a key 19 20 70 diagnostic test and/or refer to a specialist) in patients with symptoms that could be due to 21 22 23 71 cancer and how they vary across European countries. 24 25 26 72 Design 27 28 29 73 A primary care survey. PCPs were given vignettes describing patients with symptoms that 30 31 74 could indicate cancer and asked how they would manage these patients. The likelihood of 32 33 34 75 taking immediate diagnostic action (a diagnostic test and/or referral) in the different 35 36 76 participating countries was analysed. Comparisons between the likelihood of taking

37 http://bmjopen.bmj.com/ 38 77 immediate diagnostic action and physician characteristics were calculated. 39 40 41 78 42 Setting 43 44 79 Centres in twenty European countries with widely varying cancer survival rates. 45 on September 28, 2021 by guest. Protected copyright. 46 47 48 80 Participants 49 50 51 81 A total of 2,086 PCPs answered the survey question, with a median of 72 PCPs per country. 52 53 54 82 Results 55 56 57 83 PCPs’ likelihood of immediate diagnostic action at the first consultation varied from 50 to 58 59 84 82% between countries. PCPs who were more experienced, were working in more remote 60

4 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 7 of 61 BMJ Open

1 2 85 areas, or worked alone or in smaller practices, were more likely to take immediate diagnostic 3 BMJ Open: first published as 10.1136/bmjopen-2019-035678 on 31 October 2020. Downloaded from 4 86 action than their peers. 5 6 7 87 Conclusions 8 9 10 88 Europe shows large between-country variations in PCPs’ diagnostic action rates for patients 11 12 89 who could have cancer, which are linked with differences in healthcare practice. 13 14 15 16 90 Keywords 17 18 For peer review only 19 91 Delivery of Health Care; Primary Health Care; Cancer; Decision Making; Survival; Europe 20 21 22 92 Word count 23 24 25 93 3770 26 27 94 28 29 30 31 32 33 34 35 36

37 http://bmjopen.bmj.com/ 38 39 40 41 42 43 44

45 on September 28, 2021 by guest. Protected copyright. 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

5 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 8 of 61

1 2 95 Article Summary 3 BMJ Open: first published as 10.1136/bmjopen-2019-035678 on 31 October 2020. Downloaded from 4 5 96 Strengths and limitations of this study 6 7 8 97 • Recruitment of PCPs from 20 European countries, four countries from each of the 9 10 11 98 Northern, Southern, Eastern, Western and Central European geographical areas, provided 12 13 99 variation in geography, health systems and levels of healthcare spending. 14 15 100 • The questionnaire was carefully developed and piloted by GPs and other PCPs, and 16 17 18 101 therefore groundedFor in their peer clinical experience.review only 19 20 102 • While the response rate was low, it was comparable to that of other equivalent surveys of 21 22 103 primary care doctors. 23 24 25 104 • There was only one survey round, and follow-up rounds may have given more 26 27 105 information about how PCPs’ decision-making change as clinical cases evolve. 28 29 106 • Most samples were taken from each local lead’s own localities, and these may not have 30 31 32 107 been representative of their nations as a whole. 33 34 35 108 36

37 109 http://bmjopen.bmj.com/ 38 39 40 41 42 43 44

45 on September 28, 2021 by guest. Protected copyright. 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

6 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 9 of 61 BMJ Open

1 2 110 Background 3 BMJ Open: first published as 10.1136/bmjopen-2019-035678 on 31 October 2020. Downloaded from 4 111 Cancer survival vary widely across Europe [1]. The fifth cycle of the European Cancer Registry 5 6 7 112 (EUROCARE)-based Study on Survival and Care of Cancer Patients shows that national one- 8 9 113 year relative survival rates for all cancer sites vary from 58.2% to 81.1% [2], but there is no 10 11 114 evidence on how much of this variation is attributable to infrastructure as opposed to 12 13 14 115 individual physician’s clinical practice. Comparison of European one-year and five-year 15 16 116 relative cancer survival rates [2–5] shows that some countries have higher survival from most 17 18 For peer review only 19 117 cancers (including Belgium, France, Sweden and Switzerland), while others have consistently 20 21 118 lower survival (including Bulgaria, Croatia, Poland and Scotland). This suggests that an 22 23 119 improvement in cancer awareness and early detection in relatively poorly performing 24 25 26 120 countries could reduce the survival gap [4]. While recent cancer survival rates show 27 28 121 improvement in most countries [5], the between-country differences remain [6]. However, 29 30 122 this is not inevitable: Denmark’s considerable efforts to improve early detection rates [7] have 31 32 33 123 resulted in a narrowing of the gap between its own relatively poor cancer survival rates and 34 35 124 those of its better performing Nordic neighbours [8]. There has been a call for studies which 36 37 125 compare practice between well and poorly performing countries, to help gain an http://bmjopen.bmj.com/ 38 39 40 126 understanding of how these disparities may be remedied [5]. 41 42 127 Although one-year and five-year relative survival can be affected by overdiagnosis and lead- 43 44 128 time biases [9,10], poorer one-year survival in some countries is thought to be rooted in 45 on September 28, 2021 by guest. Protected copyright. 46 47 129 diagnostic delay [11,12] and more advanced disease at diagnosis [13,14]. The more advanced 48 49 130 a cancer is, the more difficult it is to treat it successfully [15] and, for many but not all cancers, 50 51 131 disease stage at diagnosis is associated with survival [16,17]. There is considerable evidence 52 53 54 132 that longer time to diagnosis and treatment increases cancer mortality [18,19]. Timely 55 56 133 diagnosis of cancer is, therefore, a cornerstone of health policy throughout Europe [20]. 57 58 134 However, there is a substantial challenge in deciding where and how to achieve this [21], as it 59 60

7 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 10 of 61

1 2 135 is uncertain whether late diagnosis is due to cancer patients presenting later, not being 3 BMJ Open: first published as 10.1136/bmjopen-2019-035678 on 31 October 2020. Downloaded from 4 136 referred quickly enough from primary care, or whether they are inefficiently investigated, 5 6 137 diagnosed and treated in secondary care [15]. This may be a particular issue where cancer 7 8 9 138 patients present without ‘red-flag’ symptoms, as how the Primary Care Practitioner (PCP) acts 10 11 139 will depend to a large extent on local health service organisation [22]. Balanced with a desire 12 13 140 for more timely diagnosis of cancer is the need to avoid increased healthcare costs, as well as 14 15 16 141 to minimise overdiagnosis and overtreatment, though these latter concerns are particularly 17 18 142 related to cancer screeningFor [23–25].peer review only 19 20 143 There is little evidence on how different healthcare systems influence PCPs’ referral decisions 21 22 23 144 [21]. However, a large variety of non-clinical factors affect these referral decisions [22]. These 24 25 145 include the extent of gatekeeping, funding systems, access to special investigations, concerns 26 27 146 over litigation, and barriers to accessing specialist advice, as well as the availability of fast- 28 29 30 147 track programmes for suspected cancer. Whether the responsibility for early detection of 31 32 148 cancer is principally in primary or secondary care varies between countries. The International 33 34 149 Cancer Benchmarking Partnership (ICBP) [26] examined differences in cancer awareness and 35 36

37 150 beliefs between six countries with comparable wealth in an attempt to explain differences in http://bmjopen.bmj.com/ 38 39 151 cancer survival [27]. It found a positive association between national cancer survival rates 40 41 152 and the readiness of PCPs in those countries to investigate potential cancer symptoms [28]. 42 43 44 153 However, there has not yet been an investigation of how PCPs’ diagnostic actions (a key

45 on September 28, 2021 by guest. Protected copyright. 46 154 diagnostic test and/or referral to a specialist) with respect to potential cancer symptoms vary 47 48 49 155 across Europe, amongst countries with a wide range of socio-economic development, 50 51 156 healthcare systems and healthcare spending. Also, while there is evidence that PCP gender 52 53 157 can effect specialist referral rates [29,30] as may the extent of their professional experience 54 55 56 158 [30] and whether or not they work in rural areas [31,32], this has not been assessed with 57 58 159 regard to referrals for patients that could have cancer. 59 60

8 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 11 of 61 BMJ Open

1 2 160 We therefore aimed to explore the diagnostic action rates of PCPs for patients with symptoms 3 BMJ Open: first published as 10.1136/bmjopen-2019-035678 on 31 October 2020. Downloaded from 4 161 that could be due to four types of cancer (lung, ovarian, breast and colorectal), how they 5 6 162 compared across European countries, and to explore the effect of PCPs’ demographics on their 7 8 9 163 diagnostic action rates for these cancers, 10 11 12 13 164 Methods 14 15 16 165 Design 17 18 For peer review only 19 166 We provided clinical vignettes to PCPs from twenty European countries with markedly 20 21 167 different socio-economic and healthcare systems. The vignettes described patients presenting 22 23 168 with symptoms that could indicate cancer. Recruitment started in November 2015 and was 24 25 26 169 completed at the end of 2016. 27 28 29 170 Outcome measures 30 31 171 The primary outcome was a comparison of PCPs’ immediate diagnostic action rates (a key 32 33 34 172 diagnostic test and/or referral to a specialist) for patients with symptoms that could be due to 35 36 173 cancer across the participating European countries.

37 http://bmjopen.bmj.com/ 38 174 The secondary outcome was an assessment of the effect of PCPs’ demographics on their 39 40 41 175 diagnostic action rates. 42 43 44 176 Study population

45 on September 28, 2021 by guest. Protected copyright. 46 177 The Örenäs Research Group (ÖRG) is a European collaborative of primary care researchers, 47 48 49 178 formed in 2013 to study the factors influencing national variations in the early diagnosis of 50 51 179 cancer in primary care. The research was conducted in 25 ÖRG centres in 20 countries across 52 53 180 Europe: Bulgaria, Croatia, Denmark, England, Finland, France, Germany, Greece, Israel, Italy, 54 55 56 181 Netherlands, Norway, Poland, Portugal, Romania, Scotland, Slovenia, Spain, Sweden and 57 58 182 Switzerland. Medical doctors were eligible for the survey if they were working mainly in 59 60

9 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 12 of 61

1 2 183 primary care. These doctors, here referred to collectively as ‘Primary Care Practitioners’, 3 BMJ Open: first published as 10.1136/bmjopen-2019-035678 on 31 October 2020. Downloaded from 4 184 included General Practitioners (GPs) and other doctors who had other specialist training but 5 6 185 worked in the community and could be accessed directly by patients without referral. 7 8 9 186 10 Development of the questionnaire 11 12 187 We used clinical vignettes as they have been validated as a measure of clinical practice 13 14 188 [33,34]. Following a literature review, ÖRG investigators developed a questionnaire designed 15 16 17 189 to elicit PCPs’ demographic information and their actions for patients that could have cancer. 18 For peer review only 19 190 The questionnaire with five clinical vignettes (three new ones, and two designed and 20 21 191 validated by the ICBP [35] and used with permission) was piloted by the ÖRG local leads in 22 23 24 192 January 2015 to check validity. More information about this process is given elsewhere [36]. 25 26 193 No changes to the demographic questions were made. One of the vignettes was found to be 27 28 194 invalid and was removed. The next version of the questionnaire, in English, was then piloted 29 30 31 195 by 49 PCPs in 16 ÖRG member countries in July 2015. No changes to the questions were made 32 33 196 following this second pilot. 34 35 197 ÖRG leads arranged for translations of the questionnaire into their local languages where 36

37 http://bmjopen.bmj.com/ 38 198 these were not English, a total of 19 translations from the original English. Translation and 39 40 199 validation by backtranslation were done in a standardised way [37] and are described 41 42 200 elsewhere [38]. 43 44

45 on September 28, 2021 by guest. Protected copyright. 46 201 Description of the questionnaire 47 48 202 The questionnaire consisted of 47 items and was divided into four sections: (a) demographic 49 50 203 questions (questions about years since graduation, gender, type and rural/urban location of 51 52 53 204 practice and number of doctors working in the practice); (b) referral availability questions 54 55 205 (questions about tests and specialist opinions that were either directly or indirectly available 56 57 206 to the respondent); (c) four fictitious clinical vignettes and (d) 20 health system factor 58 59 60 207 questions. The patients’ names in the vignettes were pseudonyms. Each of the vignettes

10 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 13 of 61 BMJ Open

1 2 208 provided information on the patient’s presenting symptoms, previous medical history, 3 BMJ Open: first published as 10.1136/bmjopen-2019-035678 on 31 October 2020. Downloaded from 4 209 medication, clinical findings and other relevant information). The version of the questionnaire 5 6 210 for PCPs in England, on which those for the other countries were based, is given in Appendix 7 8 9 211 1. 10 11 212 The demographic questions were: 12 13 213 • How many years is it since you graduated as a doctor? Under 10 years/10-19 14 15 16 214 years/20-29 years/30-39 years/40 years or over/I prefer not to say 17 18 215 • Are you - Female?For Male? peer I prefer notreview to say only 19 20 216 • What type of practice do you work in? Urban/Rural/Island or remote/Mixed 21 22 23 217 • How many doctors work in your practice/health centre in total? 1/2/3/4-5/6-7/8- 24 25 218 9/10 or more 26 27 219 The vignettes were chosen to have a low but significant possibility of cancer: 28 29 30 220 1. A 62-year-old male smoker with chronic obstructive pulmonary disease and now a two- 31 32 221 week history of a productive cough; positive predictive value (PPV) for lung cancer: 3.6% 33 34 222 [39]; 35 36

37 223 2. A 53-year-old woman with lower abdominal pain and abdominal distension; PPV for http://bmjopen.bmj.com/ 38 39 224 ovarian cancer: 3.1% [40]; 40 41 225 3. A 35-year-old breastfeeding woman with an abnormal nipple discharge and eczematous 42 43 44 226 changes around the nipple; PPV for breast cancer: 1.2% [41];

45 on September 28, 2021 by guest. Protected copyright. 46 227 4. A 22-year-old man with coeliac disease who now has abdominal pain, rectal bleeding and 47 48 49 228 diarrhoea; PPV for colorectal cancer: 3.4% [39]. 50 51 229 For each patient a range of five possible management decisions was given, with an invitation 52 53 230 to choose as many as needed: 54 55 56 231 • ‘I would write an appropriate prescription for the patient.’ 57 58 232 • ‘I would arrange to see the patient again for follow-up and reassessment.’ 59 60

11 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 14 of 61

1 2 233 • ‘I would not arrange formal follow-up, but would tell the patient under what 3 BMJ Open: first published as 10.1136/bmjopen-2019-035678 on 31 October 2020. Downloaded from 4 234 circumstances he should see me again.’ 5 6 235 • ‘I would organise an investigation at this consultation.’ 7 8 9 236 • ‘I would refer the patient to a specialist at this consultation.’ 10 11 237 Those that chose to investigate the patient were able to select from a range of possible 12 13 238 diagnostic tests. The response of primary interest was a PCPs’ management choice that would 14 15 16 239 be likely to identify a cancer as a cause of the patients’ symptoms, by either opting to request 17 18 240 a significant diagnosticFor test orpeer by referring review to a specialist. only The tests used in the analysis were: 19 20 241 a plain chest X-ray or lung computerised tomography (CT) for the lung vignette; a tumour 21 22 23 242 marker, diagnostic ultrasound or CT for the ovarian vignette; an ultrasound of the breast or 24 25 243 mammography for the breast vignette; and diagnostic ultrasound, sigmoidoscopy, 26 27 244 colonoscopy or CT colonography for the colorectal vignette. A factor analysis of the results of 28 29 30 245 the survey section on the effect of health system factors is reported separately [42]. 31 32 246 We chose to make the survey anonymous, as this has been shown to improve the honesty of 33 34 247 responses to questions that may be considered sensitive [43]. 35 36

37 http://bmjopen.bmj.com/ 38 248 Sample size 39 40 249 We aimed for a total sample size of at least 1000 PCPs, with at least 50 responses from each of 41 42 250 the participating countries. There was a pragmatic decision to use this sample size, as some 43 44

45 251 ÖRG local leads were unsure as to whether they would be able to recruit more than this on September 28, 2021 by guest. Protected copyright. 46 47 252 number because, in their participating countries, PCPs had little experience of research using 48 49 253 on-line surveys. 50 51 52 53 254 Recruitment of participants 54 55 255 Each ÖRG local lead was asked to email an invitation to take part in the survey to the PCPs in 56 57 256 their local health district or jurisdiction, and to recruit at least 50 participants. In six countries 58 59 60 257 (Denmark, Norway, Portugal, Romania, Slovenia, Sweden), the invitation was distributed to a

12 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 15 of 61 BMJ Open

1 2 258 national sample. The recruitment email stated that the research aimed to identify which 3 BMJ Open: first published as 10.1136/bmjopen-2019-035678 on 31 October 2020. Downloaded from 4 259 health system factors affect PCPs’ decisions to refer patients for further investigation. The 5 6 260 possibility of cancer as being a cause of the vignette symptoms was not mentioned in either 7 8 9 261 the recruitment email or the survey. 10 11 12 262 Distribution of the questionnaire 13 14 263 The questionnaire was designed using SurveyMonkey (SurveyMonkey, California, USA). 15 16 17 264 Because of the study’s wide geographical coverage, on-line delivery of the questionnaire was 18 For peer review only 19 265 used; this methodology has previously been successfully used in research involving cancer 20 21 266 care professionals [44,45]. Local leads were asked to send two follow-up reminders to 22 23 24 267 encourage completion of the survey. 25 26 27 268 Statistical analysis 28 29 269 Demographic questions and those relating to decisions to arrange a diagnostic test and to 30 31 32 270 refer to a specialist were analysed using descriptive statistics. As it was considered that some 33 34 271 PCPs would not organise a diagnostic test because they were referring to a specialist, and 35 36 272 conversely some PCPs would not refer to a specialist because they were organising a 37 http://bmjopen.bmj.com/ 38 39 273 diagnostic test, we used a composite measure of a decision to arrange a diagnostic test and/or 40 41 274 refer to a specialist, i.e. the likelihood of taking immediate diagnostic action for cancer. For 42 43 275 each individual PCP, mean immediate diagnostic action rates were calculated from the four 44

45 on September 28, 2021 by guest. Protected copyright. 46 276 individual vignette responses. From those, mean immediate diagnostic action rates were 47 48 277 calculated for each country. For comparisons between countries, medians and ranges were 49 50 278 calculated. We performed a multiple regression analysis to investigate the relationship 51 52 53 279 between PCP demographics and likelihood of immediate diagnostic action. The Pearson 54 55 280 correlation coefficient was calculated to examine correlations between PCPs’ likelihood of 56 57 281 organising a diagnostic test and their likelihood of referring to a specialist, and between 58 59 60

13 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 16 of 61

1 2 282 national response rates and PCPs’ likelihood of taking immediate diagnostic action in those 3 BMJ Open: first published as 10.1136/bmjopen-2019-035678 on 31 October 2020. Downloaded from 4 283 countries. Calculations were performed using IBM SPSS v25. 5 6 7 284 Patient and public involvement 8 9 10 285 There was no patient or public involvement in this study. 11 12 13 14 286 Results 15 16 287 A total of 2,086 PCPs completed the questionnaire. There was a median of 72 respondents per 17 18 For peer review only 19 288 country, range 59-446 (Table 1). 20 21 289 (Place Table 1 here) 22 23 290 The response rate for two countries was unknown. For the other 18 countries, the median 24 25 26 291 response rate was 24.8% (range 7.1% to 65.6%). Participants’ demographic distributions are 27 28 292 shown in Table 2. 29 30 293 (Place Table 2 here) 31 32 33 34 294 Organising a diagnostic test 35 36 295 The range of PCPs who stated that they would organise a diagnostic test at this first

37 http://bmjopen.bmj.com/ 38 296 consultation varied from 35.6%-80.1%, median 53.0% (Figure 1). 39 40 41 297 (Place Figure 1 here) 42 43 44 298 Referring the patient to a specialist

45 on September 28, 2021 by guest. Protected copyright. 46 299 Across the participating countries, a median of 34.2%, range 12.3%-64.7%, of PCPs decided to 47 48 49 300 refer the patient to a specialist at the first consultation (Figure 2). 50 51 301 (Place Figure 2 here) 52 53 54 302 Arranging a diagnostic test and/or referring the patient to a specialist 55 56 57 303 There was a strong correlation between PCPs’ likelihood of organising a diagnostic test and 58 59 304 their likelihood of referring to a specialist: r=0.77, P=<0.001. Across the surveyed countries, 60

14 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 17 of 61 BMJ Open

1 2 305 the proportion of PCPs who would take diagnostic action at this first consultation varied from 3 BMJ Open: first published as 10.1136/bmjopen-2019-035678 on 31 October 2020. Downloaded from 4 306 50.0% to 82.1%, median 59.9% (Figure 3). 5 6 307 (Place Figure 3 here) 7 8 9 308 Overall, the likelihood of immediate diagnostic action varied across the four cases: lung 10 11 309 vignette 54.8%, ovarian vignette 56.7%, breast vignette 58.1%, colorectal vignette 78.1%. 12 13 310 Although there was a wide variation in response rates across the countries, there was no 14 15 16 311 significant linear relationship between national response rates and PCPs’ likelihood of taking 17 18 312 immediate diagnosticFor action peerin those countries review (r=0.4, P =0.055,only Figure 4). 19 20 313 (Place Figure 4 here) 21 22 23 24 314 Other PCP actions 25 26 315 A median of 39.2% of PCPs indicated that they would issue a prescription to the patients in 27 28 316 their vignettes, though the range across the 20 countries was wide (21.7%-73.3%). This is 29 30 31 317 shown in Figure 5. 32 33 318 (Place Figure 5 here) 34 35 319 Between 43.7% and 77.6% of PCPs would arrange to see the patients again, median 61.6% 36

37 http://bmjopen.bmj.com/ 38 320 (Figure 6). 39 40 321 (Place Figure 6 here) 41 42 322 A smaller proportion, median 12.9%, range 1.5-28.9%, indicated that they would not arrange 43 44

45 323 formal follow-up but would tell the patient under what circumstances they should return on September 28, 2021 by guest. Protected copyright. 46 47 324 (Figure 7). 48 49 325 (Place Figure 7 here) 50 51 52 53 326 Effect of PCP demographics on their likelihood of immediate diagnostic action 54 55 327 The results of the regression analysis indicated that gender, years since graduation and size of 56 57 328 practice explained 3.0% of the variance in likelihood of immediate diagnostic action and that 58 59 60 329 this model was a significant predictor, F(3,2044) = 20.1, P=<0.001. Each of the three factors

15 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 18 of 61

1 2 330 contributed significantly to the model: gender (female PCPs were more likely to take 3 BMJ Open: first published as 10.1136/bmjopen-2019-035678 on 31 October 2020. Downloaded from 4 331 diagnostic action: B=0.038, P<0.011); years since graduation (higher diagnostic action rates 5 6 332 for more experienced doctors, B=0.021, P=001); and size of practice (higher diagnostic action 7 8 9 333 rates for PCPs in smaller practices B=0.024, P=<0.001). 10 11 334 In the twelve countries with respondents who self-identified as working in remote or island 12 13 335 practices, those PCPs were significantly more likely to take immediate diagnostic action than 14 15 16 336 their colleagues (71.4% vs. 60.7%, P=0.021). 17 18 For peer review only 19 20 337 Discussion 21 22 23 338 Principal findings 24 25 26 339 When faced with vignettes of patients with symptoms that could be due to cancer, PCPs’ 27 28 340 stated actions varied markedly across the different European countries. In all the 29 30 341 participating countries, at least half of PCPs would have taken immediate diagnostic action 31 32 33 342 (either organised a key diagnostic test or referred the patient to a specialist, or both). 34 35 343 PCPs who were more likely to arrange a diagnostic test were also more likely to refer their 36 37 344 patients to a specialist at the same time. PCPs who had graduated more recently were less http://bmjopen.bmj.com/ 38 39 40 345 likely to take diagnostic action in these vignettes than their more experienced peers, but PCPs 41 42 346 working in more remote locations were more likely to take diagnostic action than their 43 44 347 colleagues in other localities. PCPs working alone or in smaller practices were more likely to 45 on September 28, 2021 by guest. Protected copyright. 46 47 348 take diagnostic action than those in larger practices, as were female PCPs. 48 49 50 349 Strengths and limitations of this study 51 52 350 One of the strengths of our study is the wide spectrum of participating centres, with four 53 54 55 351 countries from each of the Central, Eastern, Northern, Southern and Western European 56 57 352 geographical areas, providing variation in geography, socioeconomic and health systems, and 58 59 353 levels of healthcare spending. It included the views of PCPs who are not usually involved in 60

16 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 19 of 61 BMJ Open

1 2 354 research. The questionnaire was carefully developed and piloted by GPs and other PCPs, and 3 BMJ Open: first published as 10.1136/bmjopen-2019-035678 on 31 October 2020. Downloaded from 4 355 therefore grounded in their clinical experience. We used clinical vignettes, which have been 5 6 356 shown to be effective for cross-national studies [45,46]. There is evidence that responses to 7 8 9 357 vignettes in surveys correspond well to clinical practice [28], and such surveys have 10 11 358 previously been used to study primary care investigation preferences in patients who could 12 13 359 have cancer [34]. 14 15 16 360 While vignette studies on mixed physician populations can achieve high response rates 17 18 361 [45,46], low survey Forresponse peer rates are common review in primary only care [35,36] and are known to 19 20 362 vary between countries. However, those in our study compared favourably with those of a 21 22 23 363 recent ICBP survey, in which response rates varied from 5.5% to 45.6% [47]. 24 25 364 Most samples were taken from the local lead’s own localities, and these may not have been 26 27 365 representative of their nations as a whole [48]. The recruitment method used in this study 28 29 30 366 resulted in variable response rates, leading to a risk of non-response bias [49]. However, the 31 32 367 goal of 50 survey participants per country and more than 1000 respondents in total was 33 34 368 achieved. There was a trend to a relationship between national response rates and likelihood 35 36

37 369 of taking immediate diagnostic action, and while this did not reach significance, it may have http://bmjopen.bmj.com/ 38 39 370 influenced the findings. 40 41 371 We have no data on non-responders as the survey was anonymous. However, the respondent 42 43 44 372 anonymity might have reduced the risk of social desirability bias. While respondents gave us

45 on September 28, 2021 by guest. Protected copyright. 46 373 some demographic data, in many participating countries there were no equivalent national 47 48 49 374 data on PCP demographics. This, and the low response rates in some countries, means that we 50 51 375 were unable to be sure how representative their results were of their wider PCP populations. 52 53 376 The varying response rates may be due to differences in how the local study leads selected 54 55 56 377 PCPs for the study, and variations in PCPs’ willingness to take part in online survey research. 57 58 378 It is possible that the PCPs with the most interest in diagnostic decision-making were the 59 60 379 most likely to respond. Our respondents were only able to select their management decisions

17 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 20 of 61

1 2 380 from the options that we gave them; this means they might have selected an option that they 3 BMJ Open: first published as 10.1136/bmjopen-2019-035678 on 31 October 2020. Downloaded from 4 381 would not have thought of without prompting, which may have affected the diagnostic testing 5 6 382 and referral rates. There was only one survey round, and follow-up rounds may have given 7 8 9 383 more information about how PCPs’ decision-making change as clinical cases evolve. 10 11 12 384 Interpretation of the results 13 14 385 Diagnostic testing and referral rates for these vignettes differed widely between participating 15 16 17 386 countries. Whereas we might expect that referring a patient would result in PCPs being less 18 For peer review only 19 387 likely to organise a diagnostic test on their patient, and vice versa, this was not confirmed in 20 21 388 the survey: the more likely PCPs were to refer a patient, the more likely they were to organise 22 23 24 389 a diagnostic test at the same time. The reasons for this are unclear, but it may be that PCPs 25 26 390 who are more worried about patients with unexplained symptoms are more likely to take all 27 28 391 the measures available to help them make a diagnosis. 29 30 31 392 While we found that PCPs working in remote locations were more likely to take immediate 32 33 393 diagnostic action, this may be due to confounding, as doctors with young families have been 34 35 394 found to be less likely to work rurally [50]. 36

37 http://bmjopen.bmj.com/ 38 39 395 Comparison with existing literature 40 41 396 While our study which shows a link between PCP diagnostic actions and practice location, size 42 43 397 of practice and PCP experience, an ICBP study found no health system characteristics that 44

45 on September 28, 2021 by guest. Protected copyright. 46 398 explained their findings [51]. However, the ICBP study only studied six, relatively wealthy, 47 48 399 countries (Australia, Canada, Denmark, Norway, Sweden and the United Kingdom). Our 49 50 400 finding of a lower likelihood of diagnostic action for PCPs working in larger practices 51 52 53 401 corresponds with other study findings [52]. Our evidence that PCPs working in more remote 54 55 402 locations were more likely to take diagnostic action for these vignettes links across to 56 57 403 evidence that such remote living is associated with more rapid cancer diagnosis and 58 59 60 404 treatment following GP referral [53]. While our data show higher diagnostic action rates in

18 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 21 of 61 BMJ Open

1 2 405 PCPs with more years since graduation, the opposite was found in a Finnish study [30], and no 3 BMJ Open: first published as 10.1136/bmjopen-2019-035678 on 31 October 2020. Downloaded from 4 406 difference was found in a United Kingdom study [54]. However, those two studies did not 5 6 407 specifically study referrals for suspected cancer: it may be that experienced PCPs are more 7 8 9 408 likely to recognise symptoms that suggest a possibility of cancer, even in the absence of ‘red- 10 11 409 flag’ symptoms, because they are more likely to have previously seen patients with those 12 13 410 symptoms who were subsequently found to have a serious diagnosis. 14 15 16 411 The extent to which respondents were gatekeepers, and needed to authorise their patients’ 17 18 412 access to specialist careFor and peerdiagnostic testsreview [55], may have only been a factor in their diagnostic 19 20 413 actions. There has been a suggestion that stronger gatekeeper systems are linked with lower 21 22 23 414 one-year relative cancer survival than non-gatekeeper systems [56], possibly because 24 25 415 gatekeeping can result in cost and resource decisions which reduce the likelihood of early 26 27 416 referral [57]. However, there are important variations in the level of gatekeeping between 28 29 30 417 countries, with no simple binary model as to whether or not a country has a ‘GP-as-gate- 31 32 418 keeper’ system, and a European study found no association between cancer survival and a 33 34 419 probability of presentation to a GP [36]. 35 36

37 http://bmjopen.bmj.com/ 38 420 Implications for research and practice 39 40 421 This study has shown considerable national variation in PCPs’ actions when faced with 41 42 422 patients who have a low but significant risk of cancer, and the reasons for this need to be 43 44

45 423 investigated. While it might be expected that PCPs who are more likely to arrange a diagnostic on September 28, 2021 by guest. Protected copyright. 46 47 424 test would be less likely to refer their patient in the same consultation, we found the opposite 48 49 425 was the case, and research is needed to explain this. 50 51 52 53 426 54 Conclusions 55 56 427 When given vignettes of patients with a low but significant possibility of cancer, more than 57 58 428 half of PCPs across Europe would take diagnostic action (a key diagnostic test and/or referral 59 60

19 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 22 of 61

1 2 429 to a specialist), most often by ordering diagnostic tests. However, there are substantial 3 BMJ Open: first published as 10.1136/bmjopen-2019-035678 on 31 October 2020. Downloaded from 4 430 between-country differences, which are linked with the variations in health care practice. 5 6 431 7 8 9 10 11 12 13 14 15 16 17 18 For peer review only 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36

37 http://bmjopen.bmj.com/ 38 39 40 41 42 43 44

45 on September 28, 2021 by guest. Protected copyright. 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

20 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 23 of 61 BMJ Open

1 2 432 List of abbreviations 3 BMJ Open: first published as 10.1136/bmjopen-2019-035678 on 31 October 2020. Downloaded from 4 433 CI Confidence Interval 5 6 7 434 CT Computerised Tomography 8 9 435 GP General practitioner 10 11 436 ICSS International Cancer Survival Standards 12 13 14 437 ICBP International Cancer Benchmarking Partnership 15 16 438 ÖRG Örenäs Research Group 17 18 For peer review only 19 439 PCP Primary Care Physician 20 21 440 PPV Positive Predictive Value 22 23 24 25 441 Declarations 26 27 28 442 Ethics and consent to participate 29 30 31 443 Ethical approval for the study was given by the University of Bath Research Ethics Approval 32 33 444 Committee for Health (approval date: 24th November 2014; REACH reference number: EP 34 35 36 445 14/15 66). Other countries’ study leads either achieved local ethical approval or gave

37 http://bmjopen.bmj.com/ 38 446 statements that formal ethical approval was not needed in their jurisdictions (see 39 40 447 supplementary file). Consent was implied by agreeing to take part in the survey. 41 42 43 448 Availability of data 44

45 on September 28, 2021 by guest. Protected copyright. 46 449 To avoid the risk of identification of individual participants, the datasets generated and 47 48 450 analysed during the current study are not publicly available. However, they are available 49 50 51 451 (with participants’ identifying information redacted) from the corresponding author on 52 53 452 reasonable request. 54 55 56 453 Competing interests 57 58 454 None declared. 59 60

21 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 24 of 61

1 2 455 Patient consent for publication 3 BMJ Open: first published as 10.1136/bmjopen-2019-035678 on 31 October 2020. Downloaded from 4 456 Not required. 5 6 7 457 Funding 8 9 10 458 This study received no specific grant from any funding agency in the public, commercial or 11 12 459 not-for-profit sectors. ALN’s time is supported by the National Institute for Health Research 13 14 460 (NIHR) Imperial Patient Safety Translation Research Centre, with her infrastructure support 15 16 17 461 provided by the NIHR Imperial Biomedical Research Centre (BRC). 18 For peer review only 19 20 462 Author contributions 21 22 463 IA-A, JA, KB, MB, NB, G-JD, ME, GF, SGB, MH, RH, EJ, TK, MM-C, PM, ALN, AP, DP, MPS, JS-P, ES, 23 24 25 464 SS, GT, HT, PV and BW participated in the study design. All authors except GT were involved in 26 27 465 the data collection. All authors contributed to the manuscript and approved the final version. 28 29 466 MH had overall responsibility for the study design, recruitment of local leads, analysis of data 30 31 32 467 and interpretation of results. GT advised on the study design and the statistical analysis. 33 34 35 468 Acknowledgements 36 37 469 We would like to thank all the PCPs who piloted the questionnaire and those who completed http://bmjopen.bmj.com/ 38 39 40 470 the survey. We would also like to thank the European GP Research Network for its support. 41 42 471 We are grateful to Prof. Barbara Silverman and Prof. Lital Keinan for the data on cancer 43 44 472 survival rates in Israel, and to Dr Yochai Schonmann for his work on those data. Two of the 45 on September 28, 2021 by guest. Protected copyright. 46 47 473 vignettes were used by kind permission of the ICBP; we also thank Dr Peter Murchie and Dr 48 49 474 Rhona Auckland, who generously provided the other two vignettes. Prof. Antonius Schneider 50 51 475 kindly organised the Technical University of Munich’s data collection. 52 53 54 476 55 56 477 57 58 59 60

22 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 25 of 61 BMJ Open

1 2 478 References 3 BMJ Open: first published as 10.1136/bmjopen-2019-035678 on 31 October 2020. Downloaded from 4 5 479 1 Møller H, Linklater KM, Robinson D. A visual summary of the EUROCARE-4 results: a UK 6 480 perspective. Br J Cancer 2009;101:S110-4. 7 8 481 2 EUROCARE. EUROCARE-5. Istituto Nazionale Tumori (Milan) and Istituto Superiore di 9 482 Sanità (Rome) 2014. 10 11 12 483 3 EUROCARE. EUROCARE-4. Istituto Nazionale Tumori (Milan) and Istituto Superiore di 13 484 Sanità (Rome) 2011. 14 15 485 4 Thomson CS, Forman D. Cancer survival in England and the influence of early diagnosis: 16 486 what can we learn from recent EUROCARE results? Br J Cancer 2009;101 Suppl 2:S102– 17 487 S109. doi:10.1038/sj.bjc.6605399 18 For peer review only 19 20 488 5 De Angelis R, Sant M, Coleman MP, et al. Cancer survival in Europe 1999-2007 by country 21 489 and age: results of EUROCARE-5 - a population-based study. Lancet Oncol 2014;15:23–34. 22 23 490 6 Coleman MP, Forman D, Bryant H, et al. Cancer survival in Australia, Canada, Denmark, 24 491 Norway, Sweden, and the UK, 1995-2007 (the International Cancer Benchmarking 25 492 Partnership): an analysis of population-based cancer registry data. Lancet 2011;377:127– 26 493 38. 27 28 29 494 7 Storm HH, Gislum M, Engholm G. [Cancer survival before and after initiating the Danish 30 495 Cancer Control plan]. Ugeskr Laeger 2008;170:3065–9. 31 32 496 8 Allemani C, Matsuda T, Di Carlo V, et al. Global surveillance of trends in cancer survival 33 497 2000–14 (CONCORD-3): analysis of individual records for 37 513 025 patients diagnosed 34 498 with one of 18 cancers from 322 population-based registries in 71 countries. Lancet 35 499 Published Online First: 31 January 2018. doi:https://doi.org/10.1016/S0140- 36

37 500 6736(17)33326-3 http://bmjopen.bmj.com/ 38 39 501 9 Zahl PH, Jorgensen KJ, Gotzsche PC. Overestimated lead times in cancer screening has led to 40 502 substantial underestimation of overdiagnosis. Br J Cancer 2013;109:2014-9. doi: 41 503 10.1038/bjc.2013.427. Epub 2013 Aug 20. 42 43 504 10 Carter JL, Coletti RJ, Harris RP. Quantifying and monitoring overdiagnosis in cancer 44 505 screening: a systematic review of methods. BMJ 2015;350:g7773.:10.1136/bmj.g7773. 45 on September 28, 2021 by guest. Protected copyright. 46 47 506 11 Maringe C, Walters S, Rachet B, et al. Stage at diagnosis and colorectal cancer survival 48 507 in six high-income countries: a population-based study of patients diagnosed during 2000- 49 508 2007. Acta Oncol 2013;52:919-32. doi: 10.3109/0284186X.2013.764008. Epub 2013 Apr 50 509 15. 51 52 510 12 Walters S, Maringe C, Coleman MP, et al. Lung cancer survival and stage at diagnosis in 53 511 Australia, Canada, Denmark, Norway, Sweden and the UK: a population-based study, 2004- 54 55 512 2007. Thorax 2013;68:551–64. doi:10.1136/thoraxjnl-2012-202297 56 57 513 13 Richards MA. The size of the prize for earlier diagnosis of cancer in England. Br J Cancer 58 514 2009;101 Suppl 2:S125–S129. doi:10.1038/sj.bjc.6605402 59 60

23 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 26 of 61

1 2 515 14 Woods LM, Coleman MP, Lawrence G, et al. Evidence against the proposition that “UK 3 516 cancer survival statistics are misleading”: simulation study with National Cancer Registry BMJ Open: first published as 10.1136/bmjopen-2019-035678 on 31 October 2020. Downloaded from 4 517 data. BMJ 2011;342:d3399.:10.1136/bmj.d3399. 5 6 518 15 Butler J, Foot C, Bomb M, et al. The International Cancer Benchmarking Partnership: An 7 519 international collaboration to inform cancer policy in Australia, Canada, Denmark, Norway, 8 520 Sweden and the United Kingdom. Health Policy 2013;112:148–55. 9 10 521 doi:https://doi.org/10.1016/j.healthpol.2013.03.021 11 12 522 16 Green T, Atkin K, Macleod U. Cancer detection in primary care: insights from general 13 523 practitioners. Br J Cancer 2015;112:S41. doi:10.1038/bjc.2015.41 14 524 https://www.nature.com/articles/bjc201541#supplementary-information 15 16 525 17 Jacobsen MM, Silverstein SC, Quinn M, et al. Timeliness of access to lung cancer 17 526 diagnosis and treatment: A scoping literature review. Lung Cancer 2017;112:156–64. 18 For peer review only 19 527 doi:10.1016/j.lungcan.2017.08.011 20 21 528 18 Tørring ML, Frydenberg M, Hansen RP, et al. Evidence of increasing mortality with 22 529 longer diagnostic intervals for five common cancers: a cohort study in primary care. Eur J 23 530 Cancer 2013;49:2187–98. doi:10.1016/j.ejca.2013.01.025 24 25 531 19 Neal RD, Tharmanathan P, France B, et al. Is increased time to diagnosis and treatment 26 27 532 in symptomatic cancer associated with poorer outcomes? Systematic review. Br J Cancer 28 533 2015;112 Suppl 1:S92–107. doi:10.1038/bjc.2015.48 29 30 534 20 Malmström M, Rasmussen BH, Bernhardson B-M, et al. “It is important that the process 31 535 goes quickly, isn’t it?” A qualitative multi-country study of colorectal or lung cancer 32 536 patients’ narratives of the timeliness of diagnosis and quality of care. Eur J Oncol Nurs 33 537 2018;34:82–8. doi:10.1016/j.ejon.2018.04.002 34 35 36 538 21 Foot C, Harrison T. How to improve cancer survival: explaining England’s relatively

37 539 poor rates. 2011. http://bmjopen.bmj.com/ 38 39 540 22 Harris M, Frey P, Esteva M, et al. How health system factors influence referral decisions 40 541 in patients that may have cancer: European symposium report. J Cancer Res Ther 41 542 2016;4(1):7–10. doi:10.14312/2052-4994.2016-2 42 43 543 23 Davies L, Petitti DB, Martin L, et al. Defining, Estimating, and Communicating 44

45 544 Overdiagnosis in Cancer Screening. Ann Intern Med 2018;169:36–43. doi:10.7326/M18- on September 28, 2021 by guest. Protected copyright. 46 545 0694 47 48 546 24 Jung M. Breast, prostate, and thyroid cancer screening tests and overdiagnosis. Curr 49 547 Probl Cancer 2017;41:71–9. doi:10.1016/j.currproblcancer.2016.11.006 50 51 548 25 Dubben H-H. [Early detection of prostate cancer: harm verified, benefit not verifiable]. 52 549 Bundesgesundheitsblatt Gesundheitsforschung Gesundheitsschutz 2014;57:318–26. 53 54 550 doi:10.1007/s00103-013-1904-1 55 56 551 26 International Cancer Benchmarking Partnership. ICBP Newsletter. 2011. 57 58 552 27 Forbes LJL, Simon AE, Warburton F, et al. Differences in cancer awareness and beliefs 59 553 between Australia, Canada, Denmark, Norway, Sweden and the UK (the International 60

24 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 27 of 61 BMJ Open

1 2 554 Cancer Benchmarking Partnership): do they contribute to differences in cancer survival? Br 3 555 J Cancer 2013;108:292–300. BMJ Open: first published as 10.1136/bmjopen-2019-035678 on 31 October 2020. Downloaded from 4 5 556 28 Rose PW, Rubin G, Perera-Salazar R, et al. Explaining variation in cancer survival 6 557 between 11 jurisdictions in the International Cancer Benchmarking Partnership: a primary 7 558 care vignette survey. BMJ Open 2015;5:e007212. doi: 10.1136/bmjopen-2014-007212. 8 9 559 29 Ringberg U, Fleten N, Deraas TS, et al. High referral rates to secondary care by general 10 11 560 practitioners in Norway are associated with GPs’ gender and specialist qualifications in 12 561 family medicine, a study of 4350 consultations. BMC Health Serv Res 2013;13:147. 13 562 doi:10.1186/1472-6963-13-147 14 15 563 30 Vehviläinen AT, Kumpusalo EA, Voutilainen SO, et al. Does the doctors’ professional 16 564 experience reduce referral rates? Evidence from the Finnish referral study. Scand J Prim 17 565 Health Care 1996;14:13–20. doi:10.3109/02813439608997063 18 For peer review only 19 20 566 31 Jiwa M, Gordon M, Arnet H, et al. Referring patients to specialists: A structured vignette 21 567 survey of Australian and British GPs. BMC Fam Pract 2008;9:2. doi:10.1186/1471-2296-9-2 22 23 568 32 Sladden MJ, Thomson AN. How do general practitioners manage rectal bleeding? Aust 24 569 Fam Physician 1998;27:78–82. 25 26 570 33 Peabody JW, Luck J, Glassman P, et al. Comparison of vignettes, standardized patients, 27 28 571 and chart abstraction: a prospective validation study of 3 methods for measuring quality. 29 572 JAMA 2000;283:1715–22. doi:10.1001/jama.283.13.1715 30 31 573 34 Peabody JW, Luck J, Glassman P, et al. Measuring the quality of physician practice by 32 574 using clinical vignettes: a prospective validation study. Ann Intern Med 2004;141:771–80. 33 575 doi:10.7326/0003-4819-141-10-200411160-00008 34 35 576 35 Rose PW, Hamilton W, Aldersey K, et al. Development of a survey instrument to 36

37 577 investigate the primary care factors related to differences in cancer diagnosis between http://bmjopen.bmj.com/ 38 578 international jurisdictions. BMC Fam Pr 2014;15:122.:10.1186/1471-2296-15-122. 39 40 579 36 Harris M, Frey P, Esteva M, et al. How the probability of presentation to a primary care 41 580 clinician correlates with cancer survival rates: a European survey using vignettes. Scand J 42 581 Prim Health Care 2017;:27–34. doi:10.1080/02813432.2017.1288692 43 44 582 37 Brett J, Staniszewska S, Mockford C, et al. Mapping the impact of patient and public 45 on September 28, 2021 by guest. Protected copyright. 46 583 involvement on health and social care research: a systematic review. Health Expect 47 584 2014;17:637–50. doi:10.1111/j.1369-7625.2012.00795.x 48 49 585 38 Harris M, Taylor G, Örenäs Research Group. How health system factors affect primary 50 586 care practitioners’ decisions to refer patients for further investigation: protocol for a pan- 51 587 European ecological study. BMC Health Serv Res 2018;18:338. doi:10.1186/s12913-018- 52 588 3170-2 53 54 55 589 39 Hamilton W. The CAPER studies: five case-control studies aimed at identifying and 56 590 quantifying the risk of cancer in symptomatic primary care patients. Br J Cancer 2009;101 57 591 Suppl 2:S80-6. doi:10.1038/sj.bjc.6605396 58 59 60

25 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 28 of 61

1 2 592 40 Hamilton W, Peters TJ, Bankhead C, et al. Risk of ovarian cancer in women with 3 593 symptoms in primary care: population based case-control study. BMJ 2009;339:b2998. BMJ Open: first published as 10.1136/bmjopen-2019-035678 on 31 October 2020. Downloaded from 4 594 doi:10.1136/bmj.b2998 5 6 595 41 Walker S, Hyde C, Hamilton W. Risk of breast cancer in symptomatic women in primary 7 596 care: a case-control study using electronic records. Br J Gen Pract 2014;64:e788-93. 8 597 doi:10.3399/bjgp14X682873 9 10 11 598 42 Harris M, Vedsted P, Esteva M, et al. Identifying important health system factors that 12 599 influence primary care practitioners’ referrals for cancer suspicion: a European cross- 13 600 sectional survey. BMJ Open 2018;8. doi:10.1136/bmjopen-2018-022904 14 15 601 43 Gnambs T, Kaspar K. Disclosure of sensitive behaviors across self-administered survey 16 602 modes: a meta-analysis. Behav Res Methods 2015;47:1237–59. doi:10.3758/s13428-014- 17 603 0533-4 18 For peer review only 19 20 604 44 Dobrow MJ, Orchard MC, Golden B, et al. Response audit of an Internet survey of health 21 605 care providers and administrators: implications for determination of response rates. J Med 22 606 Internet Res 2008;10:e30. doi:10.2196/jmir.1090 23 24 607 45 Peabody JW, DeMaria L, Smith O, et al. Large-Scale Evaluation of Quality of Care in 6 25 608 Countries of Eastern Europe and Central Asia Using Clinical Performance and Value 26 27 609 Vignettes. Glob Health Sci Pract 2017;5:412–29. doi:10.9745/GHSP-D-17-00044 28 29 610 46 Peabody JW, Liu A. A cross-national comparison of the quality of clinical care using 30 611 vignettes. Health Policy Plan 2007;22:294–302. doi:10.1093/heapol/czm020 31 32 612 47 Banks J, Hollinghurst S, Bigwood L, et al. Preferences for cancer investigation: a 33 613 vignette-based study of primary-care attendees. Lancet Oncol 2014;15:232–40. 34 614 doi:10.1016/S1470-2045(13)70588-6 35 36

37 615 48 Greenacre Z. The Importance of Selection Bias in Internet Surveys. Open J Stat http://bmjopen.bmj.com/ 38 616 2016;6:397–404. doi:http://dx.doi.org/10.4236/ojs.2016.63035 39 40 617 49 Pit SW, Vo T, Pyakurel S. The effectiveness of recruitment strategies on general 41 618 practitioner’s survey response rates - a systematic review. BMC Med Res Methodol 42 619 2014;14:76. doi:10.1186/1471-2288-14-76 43 44 620 50 McGrail MR, Russell DJ, O’Sullivan BG. Family effects on the rurality of GP’s work 45 on September 28, 2021 by guest. Protected copyright. 46 621 location: a longitudinal panel study. Hum Resour Health 2017;15. doi:10.1186/s12960-017- 47 622 0250-z 48 49 623 51 Rose P, Rubin G, Perrera-Salazar R, et al. Explaining variation in cancer survival 50 624 between eleven jurisdictions in the International Cancer Benchmarking Partnership: a 51 625 primary care survey. BMJ Open 2015. 52 53 626 52 Verstappen WH, ter Riet G, Dubois WI, et al. Variation in test ordering behaviour of 54 55 627 GPs: professional or context-related factors? Fam Pract 2004;21:387–95. 56 628 doi:10.1093/fampra/cmh408 57 58 629 53 Turner M, Fielding S, Ong Y, et al. A cancer geography paradox? Poorer cancer 59 630 outcomes with longer travelling times to healthcare facilities despite prompter diagnosis 60

26 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 29 of 61 BMJ Open

1 2 631 and treatment: a data-linkage study. Br J Cancer 2017;117:439–49. 3 632 doi:10.1038/bjc.2017.180 BMJ Open: first published as 10.1136/bmjopen-2019-035678 on 31 October 2020. Downloaded from 4 5 633 54 Wilkin D, Smith AG. Variation in general practitioners’ referral rates to consultants. J R 6 634 Coll Gen Pr 1987;37:350–3. 7 8 635 55 Franks P, Clancy CM, Nutting PA. Gatekeeping revisited--protecting patients from 9 636 overtreatment. N Engl J Med 1992;327:424–9. doi:10.1056/nejm199208063270613 10 11 12 637 56 Vedsted P, Olesen F. Are the serious problems in cancer survival partly rooted in 13 638 gatekeeper principles? Br J Gen Pr 2011;61:512–3. 14 15 639 57 Neal RD. Commentary. Cancer diagnosis - the role of urgent referral guidelines. Br J Gen 16 640 Pract 2010;60:127. doi:10.3399/bjgp10X483427 17 18 641 For peer review only 19 20 21 642 22 23 24 643 25 26 27 28 29 30 31 32 33 34 35 36

37 http://bmjopen.bmj.com/ 38 39 40 41 42 43 44

45 on September 28, 2021 by guest. Protected copyright. 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

27 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open: first published as 10.1136/bmjopen-2019-035678 on 31 October 2020. Downloaded from

BMJ Open Page 30 of 61

1 2 Table 1. Number of respondents per country, response rates, mean national cancer survival rates for the four cancers of interest, and 3 4 healthcare expenditures (given as PPP Current International Dollars per capita). 5 6 7 8 9 10 Number of Number of Response 1-year 5-year 11 12 respondentsFor (% ofpeer PCPs reviewrate (%) relative onlyrelative 13 14 15 all respondents) invited cancer cancer 16 http://bmjopen.bmj.com/ 17 survival* (%) survival* 18 19 (%) 20 21 22 Respondents Bulgaria 59 (2.8) 90 65.5 59.6 38.4 23 24 per country on September 28, 2021 by guest. Protected copyright. 25 Croatia 67 (3.2) 292 22.9 63.7 44.7 26 27 (in 28 Denmark 107 (5.1) 400 26.8 69.0 45.4 29 alphabetical 30 31 England 65 (3.1) 300 21.7 65.2 42.7 32 order) 33 34 Finland 65 (3.1) 178 36.5 73.2 50.3 35 36 37 France 59 (2.8) 550 10.7 74.9 49.8 38 39 40 41 42 28 43 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 BMJ Open: first published as 10.1136/bmjopen-2019-035678 on 31 October 2020. Downloaded from

Page 31 of 61 BMJ Open

1 2 Germany 103 (4.9) 242 42.6 73.5 50.3 3 4 5 Greece 68 (3.3) 318 21.4 Data not available 6 7 ** * 8 Israel 75 (3.6) 339 22.1 79.2 58.3 9 10 11 Italy 63 (3.0) 200 31.5 72.9 49.4 12 For peer review only 13 14 Netherlands 113 (5.4) 1601 7.1 72.0 49.1 15 16 Norway 90 (4.3) 500 18.0 72.8 49.9 http://bmjopen.bmj.com/ 17 18 19 Poland 152 (7.3) 422 36.0 65.8 41.5 20 21 22 Portugal 65 (3.1) 227 28.6 71.0 48.2 23 24 on September 28, 2021 by guest. Protected copyright. 25 Romania 177 (8.5) Not known Data not available 26 27 28 Scotland 65 (3.1) 350 18.6 66.5 43.7 29 30 31 Slovenia 104 (5.0) 352 29.5 69.5 44.8 32 33 34 Spain 446 (21.4) Not known 70.3 46.8 35 36 37 Sweden 79 (3.8) 400 19.8 75.9 51.5 38 39 40 41 42 29 43 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 BMJ Open: first published as 10.1136/bmjopen-2019-035678 on 31 October 2020. Downloaded from

BMJ Open Page 32 of 61

1 2 Switzerland 64 (3.1) 100 64.0 75.7 50.2 3 4 5 Total 2086 (100) 6 7 8 9 10 * Calculated using International Cancer Survival Standards (ICSS). 11 12 For peer review only 13 ** Calculated from data provided by B. Silverman, Israel Ministry of Health (personal communication, 7 September 2017) and Y. Schonmann, 14 15 London School of Hygiene & Tropical Medicine (personal communication, 7 September 2018). 16 http://bmjopen.bmj.com/ 17 18 19 20 21 22 23 24 on September 28, 2021 by guest. Protected copyright. 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 30 43 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 Page 33 of 61 BMJ Open

Table 2. Demographic distributions of respondents. 1 2 3 BMJ Open: first published as 10.1136/bmjopen-2019-035678 on 31 October 2020. Downloaded from 4 5 6 Number (%) 7 8 9 Gender Female 1274 (61.1) 10 11 12 Male 790 (37.9) 13 14 15 Not stated 22 (1.1) 16 17 18 For peer review only 19 20 Years since graduation <10 years 331 (15.5) 21 22 23 10-19 years 553 (26.9) 24 25 26 20-29 years 609 (29.2) 27 28 29 30-39 years 499 (23.9) 30 31 32 40 years or over 76 (3.6) 33 34 35 Not stated 18 (0.9) 36

37 http://bmjopen.bmj.com/ 38 39 40 41 Site of practice Urban 1238 (59.3) 42 43 Rural 485 (23.3) 44

45 on September 28, 2021 by guest. Protected copyright. 46 Remote or Island 56 (2.7) 47 48 49 Mixed 295 (14.1) 50 51 52 Not stated 12 (0.6) 53 54 55 56 57 58 Number of doctors in 1 286 (13.7) 59 60

31 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 34 of 61

practice 2 233 (11.2) 1 2 3 3 226 (10.8) BMJ Open: first published as 10.1136/bmjopen-2019-035678 on 31 October 2020. Downloaded from 4 5 6 4-5 347 (16.6) 7 8 9 6-7 259 (12.4) 10 11 12 8-9 172 (8.2) 13 14 15 10 or more 542 (26.0) 16 17 Not stated 21 (1.0) 18 For peer review only 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36

37 http://bmjopen.bmj.com/ 38 39 40 41 42 43 44

45 on September 28, 2021 by guest. Protected copyright. 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

32 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 35 of 61 BMJ Open Figure Legends 1 2 3 Figure 1. Percentage of PCPs in each country who would organise a diagnostic test. BMJ Open: first published as 10.1136/bmjopen-2019-035678 on 31 October 2020. Downloaded from 4 5 6 Figure 2. Percentage of PCPs in each country who would refer the patients to a specialist. 7 8 9 Figure 3. Percentage of PCPs in each country who would organise an investigation and/or 10 11 refer the patients to a specialist. 12 13 14 Figure 4. Association between national response rates and PCPs’ likelihood of taking 15 16 immediate diagnostic action. 17 18 For peer review only 19 Figure 5. Percentage of PCPs in each country who would issue a prescription. 20 21 22 Figure 6. Percentage of PCPs in each country who would arrange to see the patients again. 23 24 25 Figure 7. Percentage of PCPs in each country who would not arrange formal follow-up. 26 27 28 29 30 31 32 33 34 35 36

37 http://bmjopen.bmj.com/ 38 39 40 41 42 43 44

45 on September 28, 2021 by guest. Protected copyright. 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

33 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 36 of 61

1 2

45 on September 28, 2021 by guest. Protected copyright. 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 37 of 61 BMJ Open

1 2

37 http://bmjopen.bmj.com/ 38 39 40 41 42 43 44

45 on September 28, 2021 by guest. Protected copyright. 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 38 of 61

1 2

37 http://bmjopen.bmj.com/ 38 39 40 41 42 43 44

45 on September 28, 2021 by guest. Protected copyright. 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open: first published as 10.1136/bmjopen-2019-035678 on 31 October 2020. Downloaded from BMJ Open http://bmjopen.bmj.com/ For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml For peer review only on September 28, 2021 by guest. Protected copyright. Page 39 of 61 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 BMJ Open Page 40 of 61

1 2

3 100 BMJ Open: first published as 10.1136/bmjopen-2019-035678 on 31 October 2020. Downloaded from 4 5 6 7 8 80 9 10 11 12 60 13 14 15 16 17 40 18 For peer review only 19 20

21 Would write a prescription (%) 22 20 23 24 25 26

0 Netherlands Norway Greece Poland Denmark Sweden Israel Slovenia Switzerland Bulgaria Germany Romania Croatia England Finland Spain Scotland Italy Portugal France 27 28 29 30 31 32 Country 33 34 35 36

37 http://bmjopen.bmj.com/ 38 39 40 41 42 43 44

45 on September 28, 2021 by guest. Protected copyright. 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 Page 1

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 41 of 61 BMJ Open

1 2

3 100 BMJ Open: first published as 10.1136/bmjopen-2019-035678 on 31 October 2020. Downloaded from 4 5 6 7 8 80 9 10 11 12 60 13 14 15 16 17 40 18 For peer review only 19 20 21 22 20 23 24 Would see patient again for follow-up (%) 25 26

0 Italy Sweden Finland France Denmark Israel Spain Greece Portugal Bulgaria Norway Romania Scotland England Germany Slovenia Netherlands Switzerland Poland Croatia 27 28 29 30 31 32 Country 33 34 35 36

37 http://bmjopen.bmj.com/ 38 39 40 41 42 43 44

45 on September 28, 2021 by guest. Protected copyright. 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 Page 1

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 42 of 61

1 2

3 100 BMJ Open: first published as 10.1136/bmjopen-2019-035678 on 31 October 2020. Downloaded from 4 5 6 7 8 80 9 10 11 12 60 13 14 15 16 17 40 18 For peer review only 19 20 No formal follow-up (%) 21 22 20 23 24 25 26

0 Croatia Romania Bulgaria Greece Switzerland Italy Spain Poland Germany Slovenia France Denmark Israel Portugal England Norway Netherlands Sweden Scotland Finland 27 28 29 30 31 32 Country 33 34 35 36

37 http://bmjopen.bmj.com/ 38 39 40 41 42 43 44

45 on September 28, 2021 by guest. Protected copyright. 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 Page 1

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 43 of 61 BMJ Open

1 2

37 http://bmjopen.bmj.com/ 38 39 40 41 42 43 44

45 on September 28, 2021 by guest. Protected copyright. 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 44 of 61

1 2

37 http://bmjopen.bmj.com/ 38 39 40 41 42 43 44 on September 28, 2021 by guest. Protected copyright. 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 45 of 61 BMJ Open

1 2

37 http://bmjopen.bmj.com/ 38 39 40 41 42 43 44

45 on September 28, 2021 by guest. Protected copyright. 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 46 of 61

1 2

37 http://bmjopen.bmj.com/ 38 39 40 41 42 43 44

45 on September 28, 2021 by guest. Protected copyright. 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 47 of 61 BMJ Open

1 2

37 http://bmjopen.bmj.com/ 38 39 40 41 42 43 44

45 on September 28, 2021 by guest. Protected copyright. 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 48 of 61

1 2

45 on September 28, 2021 by guest. Protected copyright. 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 49 of 61 BMJ Open

1 2

37 http://bmjopen.bmj.com/ 38 39 40 41 42 43 44

45 on September 28, 2021 by guest. Protected copyright. 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 50 of 61

1 2

37 http://bmjopen.bmj.com/ 38 39 40 41 42 43 44

45 on September 28, 2021 by guest. Protected copyright. 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 51 of 61 BMJ Open

1 2

37 http://bmjopen.bmj.com/ 38 39 40 41 42 43 44

45 on September 28, 2021 by guest. Protected copyright. 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 52 of 61

1 2

37 http://bmjopen.bmj.com/ 38 39 40 41 42 43 44

45 on September 28, 2021 by guest. Protected copyright. 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 53 of 61 BMJ Open

1 2

37 http://bmjopen.bmj.com/ 38 39 40 41 42 43 44

45 on September 28, 2021 by guest. Protected copyright. 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 54 of 61

1 2

37 http://bmjopen.bmj.com/ 38 39 40 41 42 43 44

45 on September 28, 2021 by guest. Protected copyright. 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 55 of 61 BMJ Open

1 2

37 http://bmjopen.bmj.com/ 38 39 40 41 42 43 44

45 on September 28, 2021 by guest. Protected copyright. 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 56 of 61

1 2

37 http://bmjopen.bmj.com/ 38 39 40 41 42 43 44 45 on September 28, 2021 by guest. Protected copyright. 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 57 of 61 BMJ Open

1 2

37 http://bmjopen.bmj.com/ 38 39 40 41 42 43 44 45 on September 28, 2021 by guest. Protected copyright. 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 58 of 61

1 2

37 http://bmjopen.bmj.com/ 38 39 40 41 42 43 44

45 on September 28, 2021 by guest. Protected copyright. 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 59 of 61 BMJ Open

1 2

37 http://bmjopen.bmj.com/ 38 39 40 41 42 43 44

45 on September 28, 2021 by guest. Protected copyright. 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 60 of 61

1 2 Supplementary file. Ethical and other approvals obtained in each Örenäs

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 61 of 61 BMJ Open

Portugal N/A In Portugal, research ethics approval 1 was not required as no biomedical 2 intervention was performed.

37 http://bmjopen.bmj.com/ 38 39 40 41 42 43 44

45 on September 28, 2021 by guest. Protected copyright. 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 62 of 61

1 2 Primary care practitioner diagnostic action when the patient may have cancer: a vignette survey in 20

3 European countries BMJ Open: first published as 10.1136/bmjopen-2019-035678 on 31 October 2020. Downloaded from 4 5 STROBE Statement – checklist of items that should be included in reports of cross-sectional studies 6 Item 7 No Recommendation 8 Title and abstract 1 (a) Indicate the study’s design with a commonly used term in the title or the abstract 9 10 [Within the title page 1 and Design section of the abstract page 4] 11 (b) Provide in the abstract an informative and balanced summary of what was done 12 and what was found [pages 4-5] 13 14 Introduction 15 Background/rationale 2 Explain the scientific background and rationale for the investigation being reported 16 [Pages 7-8] 17 Objectives 3 State specific objectives, including any prespecified hypotheses [Page 9] 18 For peer review only 19 Methods 20 Study design 4 Present key elements of study design early in the paper [Page 9] 21 Setting 5 Describe the setting, locations, and relevant dates, including periods of recruitment, 22 23 exposure, follow-up, and data collection [Pages 9-10] 24 Participants 6 (a) Give the eligibility criteria, and the sources and methods of selection of 25 participants [Pages 12-13] 26 27 Variables 7 Clearly define all outcomes, exposures, predictors, potential confounders, and effect 28 modifiers. Give diagnostic criteria, if applicable [Pages 9] 29 Data sources/ 8* For each variable of interest, give sources of data and details of methods of 30 measurement assessment (measurement). Describe comparability of assessment methods if there is 31 32 more than one group [Page 10-12] 33 Bias 9 Describe any efforts to address potential sources of bias [N/A] 34 Study size 10 Explain how the study size was arrived at [Page 12] 35 Quantitative variables 11 Explain how quantitative variables were handled in the analyses. If applicable, 36

37 describe which groupings were chosen and why [Pages 13-14] http://bmjopen.bmj.com/ 38 Statistical methods 12 (a) Describe all statistical methods, including those used to control for confounding 39 [Pages 13-14] 40 41 (b) Describe any methods used to examine subgroups and interactions [N/A] 42 (c) Explain how missing data were addressed [N/A] 43 (d) If applicable, describe analytical methods taking account of sampling strategy 44 [N/A] 45 on September 28, 2021 by guest. Protected copyright. 46 (e) Describe any sensitivity analyses [N/A] 47 Results 48 Participants 13* (a) Report numbers of individuals at each stage of study—eg numbers potentially 49 50 eligible, examined for eligibility, confirmed eligible, included in the study, 51 completing follow-up, and analysed [Page 14] 52 (b) Give reasons for non-participation at each stage [N/A] 53 54 (c) Consider use of a flow diagram [N/A] 55 Descriptive data 14* (a) Give characteristics of study participants (eg demographic, clinical, social) and 56 information on exposures and potential confounders [Table 2] 57 (b) Indicate number of participants with missing data for each variable of interest 58 59 [N/A] 60 Outcome data 15* Report numbers of outcome events or summary measures [N/A] Main results 16 (a) Give unadjusted estimates and, if applicable, confounder-adjusted estimates and

1 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 63 of 61 BMJ Open

1 2 their precision (eg, 95% confidence interval). Make clear which confounders were

3 adjusted for and why they were included [Pages 14-16] BMJ Open: first published as 10.1136/bmjopen-2019-035678 on 31 October 2020. Downloaded from 4 5 (b) Report category boundaries when continuous variables were categorized [N/A] 6 (c) If relevant, consider translating estimates of relative risk into absolute risk for a 7 meaningful time period [N/A] 8 Other analyses 17 Report other analyses done—eg analyses of subgroups and interactions, and 9 10 sensitivity analyses [N/A] 11 Discussion 12 Key results 18 Summarise key results with reference to study objectives [Page 16] 13 14 Limitations 19 Discuss limitations of the study, taking into account sources of potential bias or 15 imprecision. Discuss both direction and magnitude of any potential bias [Pages 16- 16 17] 17 18 Interpretation For20 Give peer a cautious overall review interpretation ofonly results considering objectives, limitations, 19 multiplicity of analyses, results from similar studies, and other relevant evidence 20 [Page 19] 21 Generalisability 21 Discuss the generalisability (external validity) of the study results [Page 19-20] 22 23 Other information 24 Funding 22 Give the source of funding and the role of the funders for the present study and, if 25 applicable, for the original study on which the present article is based [Page 22] 26 27 28 *Give information separately for exposed and unexposed groups. 29 30 Note: An Explanation and Elaboration article discusses each checklist item and gives methodological background and 31 32 published examples of transparent reporting. The STROBE checklist is best used in conjunction with this article (freely 33 available on the Web sites of PLoS Medicine at http://www.plosmedicine.org/, Annals of Internal Medicine at 34 http://www.annals.org/, and Epidemiology at http://www.epidem.com/). Information on the STROBE Initiative is 35 available at www.strobe-statement.org. 36

37 http://bmjopen.bmj.com/ 38 39 40 41 42 43 44

45 on September 28, 2021 by guest. Protected copyright. 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

2 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open BMJ Open: first published as 10.1136/bmjopen-2019-035678 on 31 October 2020. Downloaded from

Article Type: Original research

Date Submitted by the 25-Jun-2020 Author:

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 1 of 65 BMJ Open BMJ Open: first published as 10.1136/bmjopen-2019-035678 on 31 October 2020. Downloaded from

1 2

37 http://bmjopen.bmj.com/ 38 39 40 41 42 43 44

45 on September 28, 2021 by guest. Protected copyright. 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 3 of 65 BMJ Open

1 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 4 of 65

2 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 5 of 65 BMJ Open

3 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 6 of 65

1 2 63 Abstract 3 BMJ Open: first published as 10.1136/bmjopen-2019-035678 on 31 October 2020. Downloaded from 4 5 64 Objectives 6 7 8 65 Cancer survival rates vary widely between European countries, with differences in timeliness 9 10 11 66 of diagnosis thought to be one key reason. There is little evidence on the way in which 12 13 67 different healthcare systems influence Primary Care Practitioners’ (PCPs’) referral decisions 14 15 68 in patients that could have cancer. 16 17 18 69 This study aimed to Forexplore PCPs’peer diagnostic review actions (whether only or not they perform a key 19 20 70 diagnostic test and/or refer to a specialist) in patients with symptoms that could be due to 21 22 23 71 cancer and how they vary across European countries. 24 25 26 72 Design 27 28 29 73 A primary care survey. PCPs were given vignettes describing patients with symptoms that 30 31 74 could indicate cancer and asked how they would manage these patients. The likelihood of 32 33 34 75 taking immediate diagnostic action (a diagnostic test and/or referral) in the different 35 36 76 participating countries was analysed. Comparisons between the likelihood of taking

4 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 7 of 65 BMJ Open

1 2 82 Results 3 BMJ Open: first published as 10.1136/bmjopen-2019-035678 on 31 October 2020. Downloaded from 4 5 83 PCPs’ likelihood of immediate diagnostic action at the first consultation varied from 50% to 6 7 84 82% between countries. PCPs who were more experienced were more likely to take 8 9 85 immediate diagnostic action than their peers. 10 11 12 86 Conclusions 13 14 15 87 When given vignettes of patients with a low but significant possibility of cancer, more than 16 17 18 88 half of PCPs across EuropeFor would peer take diagnostic review action, onlymost often by ordering diagnostic 19 20 89 tests. However, there are substantial between-country variations. 21 22 23 90 Keywords 24 25 26 91 Delivery of Health Care; Primary Health Care; Cancer; Decision Making; Survival; Europe 27 28 29 92 Word count 30 31 32 93 3749 33 34 35 94 36

37 http://bmjopen.bmj.com/ 38 39 40 41 42 43 44

45 on September 28, 2021 by guest. Protected copyright. 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

5 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 8 of 65

1 2 95 Article Summary 3 BMJ Open: first published as 10.1136/bmjopen-2019-035678 on 31 October 2020. Downloaded from 4 5 96 Strengths and limitations of this study 6 7 8 97 • Recruitment of PCPs from 20 European countries, four countries from each of the 9 10 11 98 Northern, Southern, Eastern, Western and Central European geographical areas, provided 12 13 99 variation in geography, health systems and levels of healthcare spending. 14 15 100 • The questionnaire was carefully developed and piloted by GPs and other PCPs, and 16 17 18 101 therefore groundedFor in their peer clinical experience.review only 19 20 102 • While the response rate varied between countries and was low, it was comparable to that 21 22 103 of other equivalent surveys of primary care doctors. 23 24 25 104 • There was only one survey round, and follow-up rounds may have given more 26 27 105 information about how PCPs’ decision-making change as clinical cases evolve. 28 29 106 • Most samples were taken from each local lead’s own localities, and these may not have 30 31 32 107 been representative of their nations as a whole. 33 34 35 108 36

37 109 http://bmjopen.bmj.com/ 38 39 40 41 42 43 44

45 on September 28, 2021 by guest. Protected copyright. 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

6 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 9 of 65 BMJ Open

1 2 110 Background 3 BMJ Open: first published as 10.1136/bmjopen-2019-035678 on 31 October 2020. Downloaded from 4 111 Cancer survival rates vary widely across Europe [1]. The fifth cycle of the European Cancer 5 6 7 112 Registry (EUROCARE)-based Study on Survival and Care of Cancer Patients shows that 8 9 113 national one-year relative survival rates for all cancer sites vary from 58.2% to 81.1% [2], but 10 11 114 there is no evidence on how much of this variation is attributable to infrastructure as opposed 12 13 14 115 to individual physician’s clinical practice. Comparison of European one-year and five-year 15 16 116 relative cancer survival rates [2–5] shows that some countries have higher survival from most 17 18 For peer review only 19 117 cancers (including Belgium, France, Sweden and Switzerland), while others have consistently 20 21 118 lower survival (including Bulgaria, Croatia, Poland and Scotland). This suggests that an 22 23 119 improvement in cancer awareness and early detection in relatively poorly performing 24 25 26 120 countries could reduce the survival gap [4]. While recent cancer survival rates show 27 28 121 improvement in most countries [5], the between-country differences remain [6]. However, 29 30 122 this is not inevitable: Denmark’s considerable efforts to improve early detection rates [7] have 31 32 33 123 resulted in a narrowing of the gap between its own relatively poor cancer survival rates and 34 35 124 those of its better performing Nordic neighbours [8]. There has been a call for studies which 36 37 125 compare practice between well and poorly performing countries, to help gain an http://bmjopen.bmj.com/ 38 39 40 126 understanding of how these disparities may be remedied [5]. 41 42 127 Although one-year and five-year relative survival can be affected by overdiagnosis and lead- 43 44 128 time biases [9,10], poorer one-year survival in some countries is thought to be rooted in 45 on September 28, 2021 by guest. Protected copyright. 46 47 129 diagnostic delay [11,12] and more advanced disease at diagnosis [13,14]. The more advanced 48 49 130 a cancer is, the more difficult it is to treat it successfully [15] and, for many but not all cancers, 50 51 131 disease stage at diagnosis is associated with survival [16,17]. There is considerable evidence 52 53 54 132 that longer time to diagnosis and treatment increases cancer mortality [18,19]. Timely 55 56 133 diagnosis of cancer is, therefore, a cornerstone of health policy throughout Europe [20]. 57 58 134 However, there is a substantial challenge in deciding where and how to achieve this [21], as it 59 60

7 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 10 of 65

8 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 11 of 65 BMJ Open

1 2 160 We therefore aimed to explore how the diagnostic action rates of PCPs for patients with 3 BMJ Open: first published as 10.1136/bmjopen-2019-035678 on 31 October 2020. Downloaded from 4 161 symptoms that could be due to four types of cancer (lung, ovarian, breast and colorectal) 5 6 162 compared across European countries, and to explore the effect of PCPs’ demographics on their 7 8 9 163 diagnostic action rates for these cancers. 10 11 12 13 164 Methods 14 15 16 165 Design 17 18 For peer review only 19 166 We provided clinical vignettes to PCPs from twenty European countries with markedly 20 21 167 different socio-economic and healthcare systems. The vignettes described patients presenting 22 23 168 with symptoms that could indicate cancer. Recruitment started in November 2015 and was 24 25 26 169 completed at the end of 2016. 27 28 29 170 Outcome measures 30 31 171 The primary outcome was a comparison of PCPs’ immediate diagnostic action rates (a key 32 33 34 172 diagnostic test and/or referral to a specialist) for patients with symptoms that could be due to 35 36 173 cancer across the participating European countries.

37 http://bmjopen.bmj.com/ 38 174 The secondary outcome was an exploratory analysis to investigate the relationship between 39 40 41 175 PCPs’ demographics and their diagnostic action rates. 42 43 44 176 Study population

9 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 12 of 65

45 on September 28, 2021 by guest. Protected copyright. 46 201 Description of the questionnaire 47 48 202 The questionnaire consisted of 47 items and was divided into four sections: (a) demographic 49 50 203 questions (questions about years since graduation, gender, type and rural/urban location of 51 52 53 204 practice and number of doctors working in the practice); (b) referral availability questions 54 55 205 (questions about tests and specialist opinions that were either directly or indirectly available 56 57 206 to the respondent); (c) four fictitious clinical vignettes; and (d) 20 health system factor 58 59 60 207 questions. The patients’ names in the vignettes were pseudonyms. Each of the vignettes

10 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 13 of 65 BMJ Open

1 2 208 provided information on the patient’s presenting symptoms, previous medical history, 3 BMJ Open: first published as 10.1136/bmjopen-2019-035678 on 31 October 2020. Downloaded from 4 209 medication, clinical findings and other relevant information. A factor analysis of the results of 5 6 210 the survey section on the effect of health system factors is reported separately [39]. 7 8 9 211 The version of the questionnaire for PCPs in England, on which those for the other countries 10 11 212 were based, is given in Appendix 1. We chose to make the survey anonymous, as this has been 12 13 213 shown to improve the honesty of responses to questions that may be considered sensitive 14 15 16 214 [40]. 17 18 215 The demographic questionsFor were:peer review only 19 20 216 • How many years is it since you graduated as a doctor? Under 10 years/10-19 21 22 23 217 years/20-29 years/30-39 years/40 years or over/I prefer not to say 24 25 218 • Are you - Female?/Male?/I prefer not to say 26 27 219 • What type of practice do you work in? Urban/Rural/Island or remote/Mixed 28 29 30 220 • How many doctors work in your practice/health centre in total? 1/2/3/4-5/6-7/8- 31 32 221 9/10 or more 33 34 222 The vignettes were chosen to have a low but significant possibility of cancer: 35 36

37 223 1. A 62-year-old male smoker with chronic obstructive pulmonary disease and now a two- http://bmjopen.bmj.com/ 38 39 224 week history of a productive cough; positive predictive value (PPV) for lung cancer: 3.6% 40 41 225 [41]; 42 43 44 226 2. A 53-year-old woman with lower abdominal pain and abdominal distension; PPV for

45 on September 28, 2021 by guest. Protected copyright. 46 227 ovarian cancer: 3.1% [42]; 47 48 49 228 3. A 35-year-old breastfeeding woman with an abnormal nipple discharge and eczematous 50 51 229 changes around the nipple; PPV for breast cancer: 1.2% [43]; 52 53 230 4. A 22-year-old man with coeliac disease who now has abdominal pain, rectal bleeding and 54 55 56 231 diarrhoea; PPV for colorectal cancer: 3.4% [41]. 57 58 232 For each patient a range of five possible management decisions was given, with an invitation 59 60 233 to choose as many as needed:

11 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 14 of 65

1 2 234 • ‘I would write an appropriate prescription for the patient.’ 3 BMJ Open: first published as 10.1136/bmjopen-2019-035678 on 31 October 2020. Downloaded from 4 235 • ‘I would arrange to see the patient again for follow-up and reassessment.’ 5 6 236 • ‘I would not arrange formal follow-up, but would tell the patient under what 7 8 9 237 circumstances she/he should see me again.’ 10 11 238 • ‘I would organise an investigation at this consultation.’ 12 13 239 • ‘I would refer the patient to a specialist at this consultation.’ 14 15 16 240 Those that chose to investigate the patient were able to select from a range of possible 17 18 241 diagnostic tests. TheFor response peer of primary review interest was a PCPs’only management choice that would 19 20 242 be likely to identify a cancer as a cause of the patients’ symptoms, by either opting to request 21 22 23 243 a significant diagnostic test or by referring to a specialist. The tests used in the analysis were: 24 25 244 a plain chest X-ray or lung computerised tomography (CT) for the lung vignette; a tumour 26 27 245 marker, diagnostic ultrasound or CT for the ovarian vignette; an ultrasound of the breast or 28 29 30 246 mammography for the breast vignette; and diagnostic ultrasound, sigmoidoscopy, 31 32 247 colonoscopy or CT colonography for the colorectal vignette. 33 34 35 248 Sample size 36

37 http://bmjopen.bmj.com/ 38 249 We aimed for a total sample size of at least 1000 PCPs, with at least 50 responses from each of 39 40 250 the participating countries. There was a pragmatic decision to use this sample size, as some 41 42 251 ÖRG local leads were unsure as to whether they would be able to recruit more than this 43 44

45 252 number because, in their participating countries, PCPs had little experience of research using on September 28, 2021 by guest. Protected copyright. 46 47 253 on-line surveys. 48 49 50 254 Recruitment of participants 51 52 53 255 Each ÖRG local lead was asked to email an invitation to take part in the survey to the PCPs in 54 55 256 their local health district or jurisdiction, and to recruit at least 50 participants. In six countries 56 57 257 (Denmark, Norway, Portugal, Romania, Slovenia, Sweden), the invitation was distributed to a 58 59 60 258 national sample. The recruitment email stated that the research aimed to identify which

12 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 15 of 65 BMJ Open

1 2 259 health system factors affect PCPs’ decisions to refer patients for further investigation. The 3 BMJ Open: first published as 10.1136/bmjopen-2019-035678 on 31 October 2020. Downloaded from 4 260 possibility of cancer as being a cause of the vignette symptoms was not mentioned in either 5 6 261 the recruitment email or the survey. 7 8 9 262 10 Distribution of the questionnaire 11 12 263 The questionnaire was designed using SurveyMonkey (SurveyMonkey, California, USA). 13 14 264 Because of the study’s wide geographical coverage, on-line delivery of the questionnaire was 15 16 17 265 used; this methodology has previously been successfully used in research involving cancer 18 For peer review only 19 266 care professionals [44,45]. Local leads were asked to send two follow-up reminders to 20 21 267 encourage completion of the survey. 22 23 24 268 25 Statistical analysis 26 27 269 Demographic questions and those relating to vignette decision-making were analysed using 28 29 270 descriptive statistics. For each individual PCP, mean vignette decision-making rates were 30 31 32 271 calculated from the four individual vignette responses. Merging the vignette data has face 33 34 272 validity as we aimed to explore PCPs’ diagnostic actions in patients with symptoms that could 35 36 273 be due to cancer, and the four cancers between them account for 37% of new cancer cases in 37 http://bmjopen.bmj.com/ 38 39 274 Europe [46]. Also, other authors have merged vignette data where the aim is to compare the 40 41 275 action of different groups of healthcare professionals, as in this study, as opposed to 42 43 276 comparing the effect of different vignettes on their action [47,48]. For comparisons between 44

45 on September 28, 2021 by guest. Protected copyright. 46 277 countries, means, standard deviations and ranges were calculated. As it was considered that 47 48 278 some PCPs would not organise a diagnostic test because they were referring to a specialist, 49 50 279 and conversely some PCPs would not refer to a specialist because they were organising a 51 52 53 280 diagnostic test, we used a composite measure of a decision to arrange a diagnostic test and/or 54 55 281 refer to a specialist, i.e. the likelihood of taking immediate diagnostic action for cancer. To 56 57 282 compare these rates between countries, we fitted an ANOVA model to investigate whether the 58 59 60 283 differences were statistically significant. We fitted a mixed effects model, adjusted for country,

13 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 16 of 65

1 2 284 to investigate the relationship between PCP demographics (as independent variables) and 3 BMJ Open: first published as 10.1136/bmjopen-2019-035678 on 31 October 2020. Downloaded from 4 285 likelihood of immediate diagnostic action (the dependent, continuous variable). The Pearson 5 6 286 correlation coefficient was calculated to examine correlations between PCPs’ likelihood of 7 8 9 287 organising a diagnostic test and their likelihood of referring to a specialist, and between 10 11 288 national response rates and PCPs’ likelihood of taking immediate diagnostic action in those 12 13 289 countries. Calculations were performed using IBM SPSS v25 and, for the mixed effects model, 14 15 16 290 Stata SE v15.1. 17 18 For peer review only 19 291 Patient and public involvement 20 21 292 There was no patient or public involvement in this study. 22 23 24 25 293 Results 26 27 28 294 A total of 2,086 PCPs completed the questionnaire. There was a median of 72 respondents per 29 30 295 country, range 59-446 (Table 1). 31 32 33 296 (Place Table 1 here) 34 35 297 The response rate for two countries was unknown. For the other 18 countries, the median 36 37 298 response rate was 24.8% (range 7.1% to 65.6%). Participants’ demographic distributions are http://bmjopen.bmj.com/ 38 39 40 299 shown in Table 2. 41 42 300 (Place Table 2 here) 43 44 45 301 Organising a diagnostic test on September 28, 2021 by guest. Protected copyright. 46 47 48 302 The range of PCPs who stated that they would organise a diagnostic test at this first 49 50 303 consultation varied from 35.6% to 80.1%, mean 54.1%, standard deviation (SD) 11.2 (Figure 51 52 304 1). 53 54 55 305 (Place Figure 1 here) 56 57 58 59 60

14 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 17 of 65 BMJ Open

1 2 306 Referring the patient to a specialist 3 BMJ Open: first published as 10.1136/bmjopen-2019-035678 on 31 October 2020. Downloaded from 4 307 Across the participating countries, a mean of 33.6%, SD 13.7, range 12.3% to 64.7%, of PCPs 5 6 308 decided to refer the patient to a specialist at the first consultation (Figure 2). 7 8 9 309 (Place Figure 2 here) 10 11 12 310 Arranging a diagnostic test and/or referring the patient to a specialist 13 14 311 There was a strong correlation between PCPs’ likelihood of organising a diagnostic test and 15 16 17 312 their likelihood of referring to a specialist: r=0.77, P=<0.001. Across the surveyed countries, 18 For peer review only 19 313 the proportion of PCPs who would take diagnostic action at this first consultation varied from 20 21 314 50.0% to 82.1%, mean 62.6%, SD 10.3 (Figure 3). This variation was statistically significant 22 23 24 315 (P=<0.001). 25 26 316 (Place Figure 3 here) 27 28 317 Overall, the likelihood of immediate diagnostic action varied across the four cases: lung 29 30 31 318 vignette 54.8%, ovarian vignette 56.7%, breast vignette 58.1%, colorectal vignette 78.1%. 32 33 319 Although there was a wide variation in response rates across the countries, there was no 34 35 320 significant linear relationship between national response rates and PCPs’ likelihood of taking 36

37 http://bmjopen.bmj.com/ 38 321 immediate diagnostic action in those countries (r=0.4, P=0.055, Figure 4). 39 40 322 (Place Figure 4 here) 41 42 43 323 Other PCP actions 44

45 on September 28, 2021 by guest. Protected copyright. 46 324 A mean of 41.2%, SD 13.8, of PCPs indicated that they would issue a prescription to the 47 48 325 patients in their vignettes, though the range across the 20 countries was wide (21.7% to 49 50 326 73.3%). This is shown in Figure 5. 51 52 53 327 (Place Figure 5 here) 54 55 328 Between 43.7% and 77.6% of PCPs would arrange to see the patients again, mean 61.8%, SD 56 57 329 9.1 (Figure 6). 58 59 60 330 (Place Figure 6 here)

15 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 18 of 65

1 2 331 A smaller proportion, mean 12.9%, SD 7.9, range 1.6 to 28.8%, indicated that they would not 3 BMJ Open: first published as 10.1136/bmjopen-2019-035678 on 31 October 2020. Downloaded from 4 332 arrange formal follow-up but would tell the patient under what circumstances they should 5 6 333 return (Figure 7). 7 8 9 334 (Place Figure 7 here) 10 11 12 335 Effect of PCP demographics on their likelihood of immediate diagnostic action 13 14 336 The results of a mixed effects model analysis indicated that number of years since graduation 15 16 17 337 was a significant predictor of likelihood of immediate diagnostic action (Table 3): PCPs who 18 For peer review only 19 338 had graduated within the previous 10 years were significantly less likely to investigate or 20 21 339 refer than those who had graduated longer ago. Neither PCP gender nor size of practice had a 22 23 24 340 significant effect. 25 26 341 In the twelve countries with respondents who self-identified as working in remote or island 27 28 342 practices, those PCPs were significantly more likely to take immediate diagnostic action than 29 30 31 343 their colleagues (71.4% vs. 60.7%, P=0.021). 32 33 34 35 344 Discussion 36

37 http://bmjopen.bmj.com/ 38 345 Principal findings 39 40 41 346 When faced with vignettes of patients with symptoms that could be due to cancer, PCPs’ 42 43 347 stated actions varied markedly across twenty European countries. In all the participating 44 45 348 countries, at least half of PCPs would have taken immediate diagnostic action (either on September 28, 2021 by guest. Protected copyright. 46 47 48 349 organised a key diagnostic test or referred the patient to a specialist, or both). 49 50 350 PCPs who were more likely to arrange a diagnostic test were also more likely to refer their 51 52 351 patients to a specialist at the same time. PCPs who had graduated more recently were less 53 54 55 352 likely to take diagnostic action in these vignettes than their more experienced peers, and PCPs 56 57 353 working in more remote locations were more likely to take diagnostic action than their 58 59 354 colleagues in other localities. 60

16 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 19 of 65 BMJ Open

1 2 355 Strengths and limitations of this study 3 BMJ Open: first published as 10.1136/bmjopen-2019-035678 on 31 October 2020. Downloaded from 4 356 One of the strengths of our study is the wide spectrum of participating centres, with four 5 6 357 countries from each of the Central, Eastern, Northern, Southern and Western European 7 8 9 358 geographical areas, providing variation in geography, socioeconomic and health systems, and 10 11 359 levels of healthcare spending. It included the views of PCPs who are not usually involved in 12 13 360 research. The questionnaire was carefully developed and piloted by GPs and other PCPs, and 14 15 16 361 therefore grounded in their clinical experience. We used clinical vignettes, which have been 17 18 362 shown to be effectiveFor for cross-national peer studiesreview [45,49]. Thereonly is evidence that responses to 19 20 363 vignettes in surveys correspond well to clinical practice [28], and such surveys have 21 22 23 364 previously been used to study primary care investigation preferences in patients who could 24 25 365 have cancer [34]. 26 27 366 While vignette studies on mixed physician populations can achieve high response rates 28 29 30 367 [45,49], low survey response rates are common in primary care [35,36] and are known to 31 32 368 vary between countries. However, those in our study compared favourably with those of a 33 34 35 369 recent ICBP survey, in which response rates varied from 5.5% to 45.6% [50]. There may be 36

37 370 non-response bias, as those that did not complete the questionnaire may be systematically http://bmjopen.bmj.com/ 38 39 371 different from those who participated. 40 41 42 372 Most samples were taken from the local lead’s own localities, and these may not have been 43 44 373 representative of their nations as a whole [51]. The recruitment method used in this study

45 on September 28, 2021 by guest. Protected copyright. 46 374 resulted in variable response rates, leading to a risk of non-response bias [52]. However, the 47 48 49 375 goal of 50 survey participants per country and more than 1000 respondents in total was 50 51 376 achieved. There was a trend to a relationship between national response rates and likelihood 52 53 377 of taking immediate diagnostic action, and while this did not reach significance, it may have 54 55 56 378 influenced the findings. 57 58 59 60

17 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 20 of 65

1 2 379 We have no data on non-responders as the survey was anonymous. However, the respondent 3 BMJ Open: first published as 10.1136/bmjopen-2019-035678 on 31 October 2020. Downloaded from 4 380 anonymity might have reduced the risk of social desirability bias. While respondents gave us 5 6 381 some demographic data, in many participating countries there were no equivalent national 7 8 9 382 data on PCP demographics. This, and the low response rates in some countries, means that we 10 11 383 were unable to be sure how representative their results were of their wider PCP populations. 12 13 384 The varying response rates may be due to differences in how the local study leads selected 14 15 16 385 PCPs for the study, and national variations in PCPs’ willingness to take part in online survey 17 18 386 research. For peer review only 19 20 387 It is possible that the PCPs with the most interest in diagnostic decision-making were the 21 22 23 388 most likely to respond. Our respondents were only able to select their management decisions 24 25 389 from the options that we gave them; this means they might have selected an option that they 26 27 390 would not have thought of without prompting, which may have affected the diagnostic testing 28 29 30 391 and referral rates. There was only one survey round, and follow-up rounds may have given 31 32 392 more information about how PCPs’ decision-making change as clinical cases evolve. 33 34 35 393 Interpretation of the results 36

37 http://bmjopen.bmj.com/ 38 394 Diagnostic testing and referral rates for these vignettes differed widely between participating 39 40 395 countries. Whereas we might expect that referring a patient would result in PCPs being less 41 42 396 likely to organise a diagnostic test on their patient, and vice versa, this was not confirmed in 43 44

45 397 the survey: the more likely PCPs were to refer a patient, the more likely they were to organise on September 28, 2021 by guest. Protected copyright. 46 47 398 a diagnostic test at the same time. The reasons for this are unclear, but it may be that PCPs 48 49 399 who are more worried about patients with unexplained symptoms are more likely to take all 50 51 52 400 the measures available to help them make a diagnosis. 53 54 401 While we found that PCPs working in remote locations were more likely to take immediate 55 56 402 diagnostic action, this may be due to confounding, as doctors with young families have been 57 58 59 403 found to be less likely to work rurally [53]. 60

18 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 21 of 65 BMJ Open

1 2 404 Comparison with existing literature 3 BMJ Open: first published as 10.1136/bmjopen-2019-035678 on 31 October 2020. Downloaded from 4 405 While our study shows a link between PCP diagnostic actions and both practice location and 5 6 406 duration of PCP experience, an ICBP study found no health system characteristics that 7 8 9 407 explained their findings [54]. However, the ICBP study only studied six, relatively wealthy, 10 11 408 countries (Australia, Canada, Denmark, Norway, Sweden and the United Kingdom). Another 12 13 409 study found a lower likelihood of diagnostic action for PCPs working in larger practices [55]. 14 15 16 410 Our evidence that PCPs working in more remote locations were more likely to take diagnostic 17 18 411 action for these vignettesFor links peer across to review evidence that such only remote living is associated with 19 20 412 more rapid cancer diagnosis and treatment following GP referral [56]. While our data show 21 22 23 413 higher diagnostic action rates in PCPs with more years since graduation, the opposite was 24 25 414 found in a Finnish study [30], and no difference was found in a United Kingdom study [57]. 26 27 415 However, those two studies did not specifically study referrals for suspected cancer: it may be 28 29 30 416 that experienced PCPs are more likely to recognise symptoms that suggest a possibility of 31 32 417 cancer, even in the absence of ‘red-flag’ symptoms, because they are more likely to have 33 34 35 418 previously seen patients with those symptoms who were subsequently found to have a 36

37 419 serious diagnosis. http://bmjopen.bmj.com/ 38 39 420 The extent to which respondents were gatekeepers, and needed to authorise their patients’ 40 41 42 421 access to specialist care and diagnostic tests [58], may have been a factor in their diagnostic 43 44 422 actions. There has been a suggestion that stronger gatekeeper systems are linked with lower

45 on September 28, 2021 by guest. Protected copyright. 46 423 one-year relative cancer survival than non-gatekeeper systems [59], possibly because 47 48 49 424 gatekeeping can result in cost and resource decisions which reduce the likelihood of early 50 51 425 referral [60]. However, there are important variations in the level of gatekeeping between 52 53 426 countries, with no simple binary model as to whether or not a country has a ‘GP-as-gate- 54 55 56 427 keeper’ system, and a European study found no association between cancer survival and a 57 58 428 probability of presentation to a GP [36]. 59 60

19 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 22 of 65

1 2 429 Implications for research and practice 3 BMJ Open: first published as 10.1136/bmjopen-2019-035678 on 31 October 2020. Downloaded from 4 430 This study has shown considerable national variation in PCPs’ actions when faced with 5 6 431 patients who have a low but significant risk of cancer, and the reasons for this need to be 7 8 9 432 investigated. While it might be expected that PCPs who are more likely to arrange a diagnostic 10 11 433 test would be less likely to refer their patient in the same consultation, we found the opposite 12 13 434 was the case, and research is needed to explain this. 14 15 16 17 435 Conclusions 18 For peer review only 19 20 436 When given vignettes of patients with a low but significant possibility of cancer, more than 21 22 437 half of PCPs across Europe would take diagnostic action (a key diagnostic test and/or referral 23 24 25 438 to a specialist), most often by ordering diagnostic tests. However, there are substantial 26 27 439 between-country differences. 28 29 30 440 31 32 33 34 35 36

37 http://bmjopen.bmj.com/ 38 39 40 41 42 43 44

45 on September 28, 2021 by guest. Protected copyright. 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

20 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 23 of 65 BMJ Open

1 2 441 List of abbreviations 3 BMJ Open: first published as 10.1136/bmjopen-2019-035678 on 31 October 2020. Downloaded from 4 442 CI Confidence Interval 5 6 7 443 CT Computerised Tomography 8 9 444 GP General practitioner 10 11 445 ICSS International Cancer Survival Standards 12 13 14 446 ICBP International Cancer Benchmarking Partnership 15 16 447 ÖRG Örenäs Research Group 17 18 For peer review only 19 448 PCP Primary Care Physician 20 21 449 PPV Positive Predictive Value 22 23 24 25 450 Declarations 26 27 28 451 Ethics and consent to participate 29 30 31 452 Ethical approval for the study was given by the University of Bath Research Ethics Approval 32 33 453 Committee for Health (approval date: 24th November 2014; REACH reference number: EP 34 35 36 454 14/15 66). Other countries’ study leads either achieved local ethical approval or gave

37 http://bmjopen.bmj.com/ 38 455 statements that formal ethical approval was not needed in their jurisdictions (see 39 40 456 supplementary file). Consent was implied by agreeing to take part in the survey. 41 42 43 457 Availability of data 44

45 on September 28, 2021 by guest. Protected copyright. 46 458 To avoid the risk of identification of individual participants, the datasets generated and 47 48 459 analysed during the current study are not publicly available. However, they are available 49 50 51 460 (with any possible identifying information redacted) from the corresponding author on 52 53 461 reasonable request. 54 55 56 462 Competing interests 57 58 463 None declared. 59 60

21 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 24 of 65

1 2 464 Patient consent for publication 3 BMJ Open: first published as 10.1136/bmjopen-2019-035678 on 31 October 2020. Downloaded from 4 465 Not required. 5 6 7 466 Funding 8 9 10 467 This study received no specific grant from any funding agency in the public, commercial or 11 12 468 not-for-profit sectors. ALN’s time is supported by the National Institute for Health Research 13 14 469 (NIHR) Imperial Patient Safety Translation Research Centre, with her infrastructure support 15 16 17 470 provided by the NIHR Imperial Biomedical Research Centre (BRC). 18 For peer review only 19 20 471 Author contributions 21 22 472 IA-A, JA, KB, MB, NB, G-JD, ME, GF, SGB, MH, RH, EJ, TK, MM-C, PM, ALN, AP, DP, MPS, JS-P, ES, 23 24 25 473 SS, GT, HT, PV and BW participated in the study design. All authors except GT were involved in 26 27 474 the data collection. All authors contributed to the manuscript and approved the final version. 28 29 475 MH had overall responsibility for the study design, recruitment of local leads, analysis of data 30 31 32 476 and interpretation of results. GT advised on the study design and the statistical analysis. 33 34 35 477 Acknowledgements 36 37 478 We would like to thank all the PCPs who piloted the questionnaire and those who completed http://bmjopen.bmj.com/ 38 39 40 479 the survey. We would also like to thank the European GP Research Network for its support. 41 42 480 We are grateful to Prof. Barbara Silverman and Prof. Lital Keinan for the data on cancer 43 44 481 survival rates in Israel, and to Dr Yochai Schonmann for his work on those data. Two of the 45 on September 28, 2021 by guest. Protected copyright. 46 47 482 vignettes were used by kind permission of the ICBP; we also thank Dr Peter Murchie and Dr 48 49 483 Rhona Auckland, who generously provided the other two vignettes. Prof. Antonius Schneider 50 51 484 kindly organised the Technical University of Munich’s data collection. 52 53 54 485 55 56 486 57 58 59 60

22 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 25 of 65 BMJ Open

1 2 487 References 3 BMJ Open: first published as 10.1136/bmjopen-2019-035678 on 31 October 2020. Downloaded from 4 5 488 1 Møller H, Linklater KM, Robinson D. A visual summary of the EUROCARE-4 results: a UK 6 7 489 perspective. Br J Cancer 2009;101:S110-4. 8 9 10 11 490 2 EUROCARE. EUROCARE-5. Istituto Nazionale Tumori (Milan) and Istituto Superiore di 12 13 491 Sanità (Rome) 2014. 14 15 16 492 3 EUROCARE. EUROCARE-4. Istituto Nazionale Tumori (Milan) and Istituto Superiore di 17 18 For peer review only 19 493 Sanità (Rome) 2011. 20 21 22 494 4 Thomson CS, Forman D. Cancer survival in England and the influence of early diagnosis: 23 24 495 what can we learn from recent EUROCARE results? Br J Cancer 2009;101 Suppl 2:S102– 25 26 27 496 S109. doi:10.1038/sj.bjc.6605399 28 29 30 497 5 De Angelis R, Sant M, Coleman MP, et al. Cancer survival in Europe 1999-2007 by country 31 32 498 and age: results of EUROCARE-5 - a population-based study. Lancet Oncol 2014;15:23–34. 33 34 35 36 499 6 Coleman MP, Forman D, Bryant H, et al. Cancer survival in Australia, Canada, Denmark,

37 http://bmjopen.bmj.com/ 38 500 Norway, Sweden, and the UK, 1995-2007 (the International Cancer Benchmarking 39 40 501 Partnership): an analysis of population-based cancer registry data. Lancet 2011;377:127– 41 42 43 502 38. 44

45 on September 28, 2021 by guest. Protected copyright. 46 503 7 Storm HH, Gislum M, Engholm G. [Cancer survival before and after initiating the Danish 47 48 504 Cancer Control plan]. Ugeskr Laeger 2008;170:3065–9. 49 50 51 52 505 8 Allemani C, Matsuda T, Di Carlo V, et al. Global surveillance of trends in cancer survival 53 54 506 2000–14 (CONCORD-3): analysis of individual records for 37 513 025 patients diagnosed 55 56 57 507 with one of 18 cancers from 322 population-based registries in 71 countries. Lancet 58 59 60

23 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 26 of 65

1 2 508 Published Online First: 31 January 2018. doi:https://doi.org/10.1016/S0140- 3 BMJ Open: first published as 10.1136/bmjopen-2019-035678 on 31 October 2020. Downloaded from 4 509 6736(17)33326-3 5 6 7 510 9 Zahl PH, Jorgensen KJ, Gotzsche PC. Overestimated lead times in cancer screening has led to 8 9 10 511 substantial underestimation of overdiagnosis. Br J Cancer 2013;109:2014-9. doi: 11 12 512 10.1038/bjc.2013.427. Epub 2013 Aug 20. 13 14 15 513 10 Carter JL, Coletti RJ, Harris RP. Quantifying and monitoring overdiagnosis in cancer 16 17 18 514 screening: a systematicFor review peer of methods. review BMJ 2015; 350:g7773.only :10.1136/bmj.g7773. 19 20 21 515 11 Maringe C, Walters S, Rachet B, et al. Stage at diagnosis and colorectal cancer survival 22 23 516 in six high-income countries: a population-based study of patients diagnosed during 2000- 24 25 26 517 2007. Acta Oncol 2013;52:919-32. doi: 10.3109/0284186X.2013.764008. Epub 2013 Apr 27 28 518 15. 29 30 31 519 12 Walters S, Maringe C, Coleman MP, et al. Lung cancer survival and stage at diagnosis in 32 33 34 520 Australia, Canada, Denmark, Norway, Sweden and the UK: a population-based study, 2004- 35 36 521 2007. Thorax 2013;68:551–64. doi:10.1136/thoraxjnl-2012-202297

37 http://bmjopen.bmj.com/ 38 39 522 13 Richards MA. The size of the prize for earlier diagnosis of cancer in England. Br J Cancer 40 41 42 523 2009;101 Suppl 2:S125–S129. doi:10.1038/sj.bjc.6605402 43 44

45 524 14 Woods LM, Coleman MP, Lawrence G, et al. Evidence against the proposition that “UK on September 28, 2021 by guest. Protected copyright. 46 47 525 cancer survival statistics are misleading”: simulation study with National Cancer Registry 48 49 50 526 data. BMJ 2011;342:d3399.:10.1136/bmj.d3399. 51 52 53 527 15 Butler J, Foot C, Bomb M, et al. The International Cancer Benchmarking Partnership: An 54 55 56 528 international collaboration to inform cancer policy in Australia, Canada, Denmark, Norway, 57 58 529 Sweden and the United Kingdom. Health Policy 2013;112:148–55. 59 60 530 doi:https://doi.org/10.1016/j.healthpol.2013.03.021

24 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 27 of 65 BMJ Open

1 2 531 16 Green T, Atkin K, Macleod U. Cancer detection in primary care: insights from general 3 BMJ Open: first published as 10.1136/bmjopen-2019-035678 on 31 October 2020. Downloaded from 4 532 practitioners. Br J Cancer 2015;112:S41. doi:10.1038/bjc.2015.41 5 6 533 https://www.nature.com/articles/bjc201541#supplementary-information 7 8 9 10 534 17 Jacobsen MM, Silverstein SC, Quinn M, et al. Timeliness of access to lung cancer 11 12 535 diagnosis and treatment: A scoping literature review. Lung Cancer 2017;112:156–64. 13 14 536 doi:10.1016/j.lungcan.2017.08.011 15 16 17 18 537 18 Tørring ML, FrydenbergFor peer M, Hansen review RP, et al. Evidence only of increasing mortality with 19 20 538 longer diagnostic intervals for five common cancers: a cohort study in primary care. Eur J 21 22 539 Cancer 2013;49:2187–98. doi:10.1016/j.ejca.2013.01.025 23 24 25 26 540 19 Neal RD, Tharmanathan P, France B, et al. Is increased time to diagnosis and treatment 27 28 541 in symptomatic cancer associated with poorer outcomes? Systematic review. Br J Cancer 29 30 542 2015;112 Suppl 1:S92–107. doi:10.1038/bjc.2015.48 31 32 33 34 543 20 Malmström M, Rasmussen BH, Bernhardson B-M, et al. “It is important that the process 35 36 544 goes quickly, isn’t it?” A qualitative multi-country study of colorectal or lung cancer

37 http://bmjopen.bmj.com/ 38 545 patients’ narratives of the timeliness of diagnosis and quality of care. Eur J Oncol Nurs 39 40 41 546 2018;34:82–8. doi:10.1016/j.ejon.2018.04.002 42 43 44 547 21 Foot C, Harrison T. How to improve cancer survival: explaining England’s relatively

45 on September 28, 2021 by guest. Protected copyright. 46 548 poor rates. 2011. 47 48 49 50 549 22 Harris M, Frey P, Esteva M, et al. How health system factors influence referral decisions 51 52 550 in patients that may have cancer: European symposium report. J Cancer Res Ther 53 54 55 551 2016;4(1):7–10. doi:10.14312/2052-4994.2016-2 56 57 58 59 60

25 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 28 of 65

1 2 552 23 Davies L, Petitti DB, Martin L, et al. Defining, Estimating, and Communicating 3 BMJ Open: first published as 10.1136/bmjopen-2019-035678 on 31 October 2020. Downloaded from 4 553 Overdiagnosis in Cancer Screening. Ann Intern Med 2018;169:36–43. doi:10.7326/M18- 5 6 554 0694 7 8 9 10 555 24 Jung M. Breast, prostate, and thyroid cancer screening tests and overdiagnosis. Curr 11 12 556 Probl Cancer 2017;41:71–9. doi:10.1016/j.currproblcancer.2016.11.006 13 14 15 557 25 Dubben H-H. [Early detection of prostate cancer: harm verified, benefit not verifiable]. 16 17 18 558 BundesgesundheitsblattFor Gesundheitsforschung peer review Gesundheitsschutz only 2014;57:318–26. 19 20 559 doi:10.1007/s00103-013-1904-1 21 22 23 560 26 International Cancer Benchmarking Partnership. ICBP Newsletter. 2011. 24 25 26 27 561 27 Forbes LJL, Simon AE, Warburton F, et al. Differences in cancer awareness and beliefs 28 29 562 between Australia, Canada, Denmark, Norway, Sweden and the UK (the International 30 31 563 Cancer Benchmarking Partnership): do they contribute to differences in cancer survival? Br 32 33 34 564 J Cancer 2013;108:292–300. 35 36

37 565 28 Rose PW, Rubin G, Perera-Salazar R, et al. Explaining variation in cancer survival http://bmjopen.bmj.com/ 38 39 566 between 11 jurisdictions in the International Cancer Benchmarking Partnership: a primary 40 41 42 567 care vignette survey. BMJ Open 2015;5:e007212. doi: 10.1136/bmjopen-2014-007212. 43 44

45 568 29 Ringberg U, Fleten N, Deraas TS, et al. High referral rates to secondary care by general on September 28, 2021 by guest. Protected copyright. 46 47 569 practitioners in Norway are associated with GPs’ gender and specialist qualifications in 48 49 50 570 family medicine, a study of 4350 consultations. BMC Health Serv Res 2013;13:147. 51 52 571 doi:10.1186/1472-6963-13-147 53 54 55 56 572 30 Vehviläinen AT, Kumpusalo EA, Voutilainen SO, et al. Does the doctors’ professional 57 58 573 experience reduce referral rates? Evidence from the Finnish referral study. Scand J Prim 59 60 574 Health Care 1996;14:13–20. doi:10.3109/02813439608997063

26 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 29 of 65 BMJ Open

1 2 575 31 Jiwa M, Gordon M, Arnet H, et al. Referring patients to specialists: A structured vignette 3 BMJ Open: first published as 10.1136/bmjopen-2019-035678 on 31 October 2020. Downloaded from 4 576 survey of Australian and British GPs. BMC Fam Pract 2008;9:2. doi:10.1186/1471-2296-9-2 5 6 7 577 32 Sladden MJ, Thomson AN. How do general practitioners manage rectal bleeding? Aust 8 9 10 578 Fam Physician 1998;27:78–82. 11 12 13 579 33 Peabody JW, Luck J, Glassman P, et al. Comparison of vignettes, standardized patients, 14 15 580 and chart abstraction: a prospective validation study of 3 methods for measuring quality. 16 17 18 581 JAMA 2000;283:1715–22.For peer doi:10.1001/jama.283.13.1715 review only 19 20 21 582 34 Peabody JW, Luck J, Glassman P, et al. Measuring the quality of physician practice by 22 23 583 using clinical vignettes: a prospective validation study. Ann Intern Med 2004;141:771–80. 24 25 26 584 doi:10.7326/0003-4819-141-10-200411160-00008 27 28 29 585 35 Rose PW, Hamilton W, Aldersey K, et al. Development of a survey instrument to 30 31 586 investigate the primary care factors related to differences in cancer diagnosis between 32 33 34 587 international jurisdictions. BMC Fam Pr 2014;15:122.:10.1186/1471-2296-15-122. 35 36

37 588 36 Harris M, Frey P, Esteva M, et al. How the probability of presentation to a primary care http://bmjopen.bmj.com/ 38 39 589 clinician correlates with cancer survival rates: a European survey using vignettes. Scand J 40 41 42 590 Prim Health Care 2017;:27–34. doi:10.1080/02813432.2017.1288692 43 44

45 591 37 Brett J, Staniszewska S, Mockford C, et al. Mapping the impact of patient and public on September 28, 2021 by guest. Protected copyright. 46 47 592 involvement on health and social care research: a systematic review. Health Expect 48 49 50 593 2014;17:637–50. doi:10.1111/j.1369-7625.2012.00795.x 51 52 53 594 38 Harris M, Taylor G, Örenäs Research Group. How health system factors affect primary 54 55 56 595 care practitioners’ decisions to refer patients for further investigation: protocol for a pan- 57 58 596 European ecological study. BMC Health Serv Res 2018;18:338. doi:10.1186/s12913-018- 59 60 597 3170-2

27 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 30 of 65

1 2 598 39 Harris M, Vedsted P, Esteva M, et al. Identifying important health system factors that 3 BMJ Open: first published as 10.1136/bmjopen-2019-035678 on 31 October 2020. Downloaded from 4 599 influence primary care practitioners’ referrals for cancer suspicion: a European cross- 5 6 600 sectional survey. BMJ Open 2018;8. doi:10.1136/bmjopen-2018-022904 7 8 9 10 601 40 Gnambs T, Kaspar K. Disclosure of sensitive behaviors across self-administered survey 11 12 602 modes: a meta-analysis. Behav Res Methods 2015;47:1237–59. doi:10.3758/s13428-014- 13 14 603 0533-4 15 16 17 18 604 41 Hamilton W. ForThe CAPER peer studies: reviewfive case-control only studies aimed at identifying and 19 20 605 quantifying the risk of cancer in symptomatic primary care patients. Br J Cancer 2009;101 21 22 606 Suppl 2:S80-6. doi:10.1038/sj.bjc.6605396 23 24 25 26 607 42 Hamilton W, Peters TJ, Bankhead C, et al. Risk of ovarian cancer in women with 27 28 608 symptoms in primary care: population based case-control study. BMJ 2009;339:b2998. 29 30 609 doi:10.1136/bmj.b2998 31 32 33 34 610 43 Walker S, Hyde C, Hamilton W. Risk of breast cancer in symptomatic women in primary 35 36 611 care: a case-control study using electronic records. Br J Gen Pract 2014;64:e788-93.

37 http://bmjopen.bmj.com/ 38 612 doi:10.3399/bjgp14X682873 39 40 41 42 613 44 Dobrow MJ, Orchard MC, Golden B, et al. Response audit of an Internet survey of health 43 44 614 care providers and administrators: implications for determination of response rates. J Med

45 on September 28, 2021 by guest. Protected copyright. 46 615 Internet Res 2008;10:e30. doi:10.2196/jmir.1090 47 48 49 50 616 45 Peabody JW, DeMaria L, Smith O, et al. Large-Scale Evaluation of Quality of Care in 6 51 52 617 Countries of Eastern Europe and Central Asia Using Clinical Performance and Value 53 54 55 618 Vignettes. Glob Health Sci Pract 2017;5:412–29. doi:10.9745/GHSP-D-17-00044 56 57 58 619 46 International Agency for Research on Cancer. Estimated number of new cases in 2018, 59 60 620 Europe, both sexes, all ages. https://gco.iarc.fr/today/home (accessed 19 Jun 2020).

28 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 31 of 65 BMJ Open

1 2 621 47 Aydin S, Crone MR, Siebelink BM, et al. Recognition of anxiety disorders in children: a 3 BMJ Open: first published as 10.1136/bmjopen-2019-035678 on 31 October 2020. Downloaded from 4 622 cross-sectional vignette-based survey among general practitioners. BMJ Open 5 6 623 2020;10:e035799. doi:10.1136/bmjopen-2019-035799 7 8 9 10 624 48 Hudelson P, Perron NJ, Perneger T. Using clinical vignettes to assess doctors’ and 11 12 625 medical students’ ability to identify sociocultural factors affecting health and health care. 13 14 626 Med Teach 2011;33:e564–71. doi:10.3109/0142159X.2011.602994 15 16 17 18 627 49 Peabody JW, ForLiu A. A cross-nationalpeer review comparison ofonly the quality of clinical care using 19 20 628 vignettes. Health Policy Plan 2007;22:294–302. doi:10.1093/heapol/czm020 21 22 23 629 50 Banks J, Hollinghurst S, Bigwood L, et al. Preferences for cancer investigation: a 24 25 26 630 vignette-based study of primary-care attendees. Lancet Oncol 2014;15:232–40. 27 28 631 doi:10.1016/S1470-2045(13)70588-6 29 30 31 632 51 Greenacre Z. The Importance of Selection Bias in Internet Surveys. Open J Stat 32 33 34 633 2016;6:397–404. doi:http://dx.doi.org/10.4236/ojs.2016.63035 35 36

37 634 52 Pit SW, Vo T, Pyakurel S. The effectiveness of recruitment strategies on general http://bmjopen.bmj.com/ 38 39 635 practitioner’s survey response rates - a systematic review. BMC Med Res Methodol 40 41 42 636 2014;14:76. doi:10.1186/1471-2288-14-76 43 44

45 637 53 McGrail MR, Russell DJ, O’Sullivan BG. Family effects on the rurality of GP’s work on September 28, 2021 by guest. Protected copyright. 46 47 638 location: a longitudinal panel study. Hum Resour Health 2017;15. doi:10.1186/s12960-017- 48 49 50 639 0250-z 51 52 53 640 54 Rose P, Rubin G, Perrera-Salazar R, et al. Explaining variation in cancer survival 54 55 56 641 between eleven jurisdictions in the International Cancer Benchmarking Partnership: a 57 58 642 primary care survey. BMJ Open 2015. 59 60

29 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 32 of 65

1 2 643 55 Verstappen WH, ter Riet G, Dubois WI, et al. Variation in test ordering behaviour of 3 BMJ Open: first published as 10.1136/bmjopen-2019-035678 on 31 October 2020. Downloaded from 4 644 GPs: professional or context-related factors? Fam Pract 2004;21:387–95. 5 6 645 doi:10.1093/fampra/cmh408 7 8 9 10 646 56 Turner M, Fielding S, Ong Y, et al. A cancer geography paradox? Poorer cancer 11 12 647 outcomes with longer travelling times to healthcare facilities despite prompter diagnosis 13 14 648 and treatment: a data-linkage study. Br J Cancer 2017;117:439–49. 15 16 17 649 doi:10.1038/bjc.2017.180 18 For peer review only 19 20 650 57 Wilkin D, Smith AG. Variation in general practitioners’ referral rates to consultants. J R 21 22 651 Coll Gen Pr 1987;37:350–3. 23 24 25 26 652 58 Franks P, Clancy CM, Nutting PA. Gatekeeping revisited--protecting patients from 27 28 653 overtreatment. N Engl J Med 1992;327:424–9. doi:10.1056/nejm199208063270613 29 30 31 654 59 Vedsted P, Olesen F. Are the serious problems in cancer survival partly rooted in 32 33 34 655 gatekeeper principles? Br J Gen Pr 2011;61:512–3. 35 36

37 656 60 Neal RD. Commentary. Cancer diagnosis - the role of urgent referral guidelines. Br J Gen http://bmjopen.bmj.com/ 38 39 657 Pract 2010;60:127. doi:10.3399/bjgp10X483427 40 41 42 43 658 44

45 on September 28, 2021 by guest. Protected copyright. 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

30 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open: first published as 10.1136/bmjopen-2019-035678 on 31 October 2020. Downloaded from

Page 33 of 65 BMJ Open

1 2 Table 1. Number of respondents per country, response rates, mean national cancer survival rates for the four cancers of interest. 3 4 5 6 7 8 Number of Number of Response 1-year 5-year 9 10 respondents (% of PCPs rate (%) relative relative 11 12 all respondents)For peerinvited reviewcancer onlycancer 13 14 * * 15 survival (%) survival 16 http://bmjopen.bmj.com/ 17 (%) 18 19 20 Respondents Bulgaria 59 (2.8) 90 65.5 59.6 38.4 21 22 per country 23 Croatia 67 (3.2) 292 22.9 63.7 44.7 24 on September 28, 2021 by guest. Protected copyright. 25 (in 26 Denmark 107 (5.1) 400 26.8 69.0 45.4 27 alphabetical 28 England 65 (3.1) 300 21.7 65.2 42.7 29 order) 30 31 Finland 65 (3.1) 178 36.5 73.2 50.3 32 33 34 France 59 (2.8) 550 10.7 74.9 49.8 35 36 37 Germany 103 (4.9) 242 42.6 73.5 50.3 38 39 40 41 42 31 43 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 BMJ Open: first published as 10.1136/bmjopen-2019-035678 on 31 October 2020. Downloaded from

BMJ Open Page 34 of 65

1 2 Greece 68 (3.3) 318 21.4 Data not available 3 4 5 Israel 75 (3.6) 339 22.1 79.2** 58.3* 6 7 8 Italy 63 (3.0) 200 31.5 72.9 49.4 9 10 11 Netherlands 113 (5.4) 1601 7.1 72.0 49.1 12 For peer review only 13 14 Norway 90 (4.3) 500 18.0 72.8 49.9 15 16 Poland 152 (7.3) 422 36.0 65.8 41.5 http://bmjopen.bmj.com/ 17 18 19 Portugal 65 (3.1) 227 28.6 71.0 48.2 20 21 22 Romania 177 (8.5) Not known Data not available 23 24 on September 28, 2021 by guest. Protected copyright. 25 Scotland 65 (3.1) 350 18.6 66.5 43.7 26 27 28 Slovenia 104 (5.0) 352 29.5 69.5 44.8 29 30 31 Spain 446 (21.4) Not known 70.3 46.8 32 33 34 Sweden 79 (3.8) 400 19.8 75.9 51.5 35 36 37 Switzerland 64 (3.1) 100 64.0 75.7 50.2 38 39 40 41 42 32 43 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 BMJ Open: first published as 10.1136/bmjopen-2019-035678 on 31 October 2020. Downloaded from

Page 35 of 65 BMJ Open

1 2 Total 2086 (100) 3 4 5 6 7 * Calculated using International Cancer Survival Standards (ICSS). 8 9 10 ** Calculated from data provided by B. Silverman, Israel Ministry of Health (personal communication, 7 September 2017) and Y. Schonmann, 11 12 London School of Hygiene & Tropical ForMedicine (personal peer communication, review 7 September 2018). only 13 14 15 16 http://bmjopen.bmj.com/ 17 18 19 20 21 22 23 24 on September 28, 2021 by guest. Protected copyright. 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 33 43 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 BMJ Open Page 36 of 65

37 http://bmjopen.bmj.com/ 38 39 40 41 Site of practice Urban 1238 (59.3) 42 43 Rural 485 (23.3) 44

45 on September 28, 2021 by guest. Protected copyright. 46 Remote or Island 56 (2.7) 47 48 49 Mixed 295 (14.1) 50 51 52 Not stated 12 (0.6) 53 54 55 56 57 58 Number of doctors in 1 286 (13.7) 59 60

34 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 37 of 65 BMJ Open

37 http://bmjopen.bmj.com/ 38 39 40 41 42 43 44

45 on September 28, 2021 by guest. Protected copyright. 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

35 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 38 of 65 Table 3. Mixed effects model to investigate the relationship between PCP demographics and 1 2 likelihood of immediate diagnostic action, adjusted for country. 3 BMJ Open: first published as 10.1136/bmjopen-2019-035678 on 31 October 2020. Downloaded from 4 5 Demographic factor Margin (95% Standard error 6 7 confidence interval) 8 9 10 PCP gender 11 12 13 Female 63.6 (58.9-68.3) 2.4 14 15 16 Male 62.7 (57.9-67.6) 2.5 17 18 For peer review only 19 Years since graduation 20 21 22 less than 10 56.1 (50.5-61.6) 2.8 23 24 25 10-19 64.3 (59.2-69.3) 2.6 26 27 20-29 63.6 (58.5-68.6) 2.6 28 29 30 30-39 66.1 (61.0-71.3) 2.6 31 32 33 40 or over 66.3 (58.1-74.6) 4.2 34 35 36 Number of doctors in

37 http://bmjopen.bmj.com/ 38 PCP’s practice 39 40 41 1 64.8 (59.0-70.6) 3.0 42 43 44 2 64.6 (58.6-70.5) 3.0 45 on September 28, 2021 by guest. Protected copyright. 46 47 3 62.4 (56.4-68.3) 3.0 48 49 50 4-5 63.3 (57.8-68.7) 2.8 51 52 53 6-7 60.8 (55.1-66.6) 2.9 54 55 56 8-9 61.3 (54.9-67.7) 3.3 57 58 59 10 or more 63.9 (58.4-69.3) 2.8 60

36 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 39 of 65 BMJ Open Figure Legends 1 2 3 Figure 1. Percentage of PCPs in each country who would organise a diagnostic test. BMJ Open: first published as 10.1136/bmjopen-2019-035678 on 31 October 2020. Downloaded from 4 5 6 Figure 2. Percentage of PCPs in each country who would refer the patient to a specialist. 7 8 9 Figure 3. Percentage of PCPs in each country who would organise an investigation and/or 10 11 refer the patients to a specialist. 12 13 14 Figure 4. Association between national response rates and PCPs’ likelihood of taking 15 16 immediate diagnostic action. 17 18 For peer review only 19 Figure 5. Percentage of PCPs in each country who would issue a prescription. 20 21 22 Figure 6. Percentage of PCPs in each country who would arrange to see the patients again. 23 24 25 Figure 7. Percentage of PCPs in each country who would not arrange formal follow-up. 26 27 28 29 30 31 32 33 34 35 36

37 http://bmjopen.bmj.com/ 38 39 40 41 42 43 44

45 on September 28, 2021 by guest. Protected copyright. 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

37 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 40 of 65

1 2

45 on September 28, 2021 by guest. Protected copyright. 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 41 of 65 BMJ Open

1 2

37 http://bmjopen.bmj.com/ 38 39 40 41 42 43 44

45 on September 28, 2021 by guest. Protected copyright. 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 42 of 65

1 2

37 http://bmjopen.bmj.com/ 38 39 40 41 42 43 44

45 on September 28, 2021 by guest. Protected copyright. 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open: first published as 10.1136/bmjopen-2019-035678 on 31 October 2020. Downloaded from BMJ Open http://bmjopen.bmj.com/ For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml For peer review only on September 28, 2021 by guest. Protected copyright. Page 43 of 65 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 BMJ Open Page 44 of 65

1 2

3 100 BMJ Open: first published as 10.1136/bmjopen-2019-035678 on 31 October 2020. Downloaded from 4 5 6 7 8 80 9 10 11 12 60 13 14 15 16 17 40 18 For peer review only 19 20

21 Would write a prescription (%) 22 20 23 24 25 26

37 http://bmjopen.bmj.com/ 38 39 40 41 42 43 44

45 on September 28, 2021 by guest. Protected copyright. 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 Page 1

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 45 of 65 BMJ Open

1 2

37 http://bmjopen.bmj.com/ 38 39 40 41 42 43 44

45 on September 28, 2021 by guest. Protected copyright. 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 Page 1

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 46 of 65

1 2

37 http://bmjopen.bmj.com/ 38 39 40 41 42 43 44

45 on September 28, 2021 by guest. Protected copyright. 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 Page 1

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 47 of 65 BMJ Open

1 2

37 http://bmjopen.bmj.com/ 38 39 40 41 42 43 44

45 on September 28, 2021 by guest. Protected copyright. 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 48 of 65

1 2

37 http://bmjopen.bmj.com/ 38 39 40 41 42 43 44 on September 28, 2021 by guest. Protected copyright. 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 49 of 65 BMJ Open

1 2

37 http://bmjopen.bmj.com/ 38 39 40 41 42 43 44

45 on September 28, 2021 by guest. Protected copyright. 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 50 of 65

1 2

37 http://bmjopen.bmj.com/ 38 39 40 41 42 43 44

45 on September 28, 2021 by guest. Protected copyright. 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 51 of 65 BMJ Open

1 2

37 http://bmjopen.bmj.com/ 38 39 40 41 42 43 44

45 on September 28, 2021 by guest. Protected copyright. 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 52 of 65

1 2

45 on September 28, 2021 by guest. Protected copyright. 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 53 of 65 BMJ Open

1 2

37 http://bmjopen.bmj.com/ 38 39 40 41 42 43 44

45 on September 28, 2021 by guest. Protected copyright. 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 54 of 65

1 2

37 http://bmjopen.bmj.com/ 38 39 40 41 42 43 44

45 on September 28, 2021 by guest. Protected copyright. 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 55 of 65 BMJ Open

1 2

37 http://bmjopen.bmj.com/ 38 39 40 41 42 43 44

45 on September 28, 2021 by guest. Protected copyright. 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 56 of 65

1 2

37 http://bmjopen.bmj.com/ 38 39 40 41 42 43 44

45 on September 28, 2021 by guest. Protected copyright. 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 57 of 65 BMJ Open

1 2

37 http://bmjopen.bmj.com/ 38 39 40 41 42 43 44

45 on September 28, 2021 by guest. Protected copyright. 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 58 of 65

1 2

37 http://bmjopen.bmj.com/ 38 39 40 41 42 43 44

45 on September 28, 2021 by guest. Protected copyright. 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 59 of 65 BMJ Open

1 2

37 http://bmjopen.bmj.com/ 38 39 40 41 42 43 44

45 on September 28, 2021 by guest. Protected copyright. 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 60 of 65

1 2

37 http://bmjopen.bmj.com/ 38 39 40 41 42 43 44 45 on September 28, 2021 by guest. Protected copyright. 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 61 of 65 BMJ Open

1 2

37 http://bmjopen.bmj.com/ 38 39 40 41 42 43 44 45 on September 28, 2021 by guest. Protected copyright. 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 62 of 65

1 2

37 http://bmjopen.bmj.com/ 38 39 40 41 42 43 44

45 on September 28, 2021 by guest. Protected copyright. 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 63 of 65 BMJ Open

1 2

37 http://bmjopen.bmj.com/ 38 39 40 41 42 43 44

45 on September 28, 2021 by guest. Protected copyright. 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 64 of 65

1 2 Supplementary file. Ethical and other approvals obtained in each Örenäs

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 65 of 65 BMJ Open

Portugal N/A In Portugal, research ethics approval 1 was not required as no biomedical 2 intervention was performed.

37 http://bmjopen.bmj.com/ 38 39 40 41 42 43 44

45 on September 28, 2021 by guest. Protected copyright. 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 66 of 65

1 2 Primary care practitioner diagnostic action when the patient may have cancer: a vignette survey in 20

3 European countries BMJ Open: first published as 10.1136/bmjopen-2019-035678 on 31 October 2020. Downloaded from 4 5 STROBE Statement – checklist of items that should be included in reports of cross-sectional studies 6 Item 7 No Recommendation 8 Title and abstract 1 (a) Indicate the study’s design with a commonly used term in the title or the abstract 9 10 [Within the title page 1 and Design section of the abstract page 4] 11 (b) Provide in the abstract an informative and balanced summary of what was done 12 and what was found [pages 4-5] 13 14 Introduction 15 Background/rationale 2 Explain the scientific background and rationale for the investigation being reported 16 [Pages 7-8] 17 Objectives 3 State specific objectives, including any prespecified hypotheses [Page 9] 18 For peer review only 19 Methods 20 Study design 4 Present key elements of study design early in the paper [Page 9] 21 Setting 5 Describe the setting, locations, and relevant dates, including periods of recruitment, 22 23 exposure, follow-up, and data collection [Pages 9-10] 24 Participants 6 (a) Give the eligibility criteria, and the sources and methods of selection of 25 participants [Pages 12-13] 26 27 Variables 7 Clearly define all outcomes, exposures, predictors, potential confounders, and effect 28 modifiers. Give diagnostic criteria, if applicable [Page 9] 29 Data sources/ 8* For each variable of interest, give sources of data and details of methods of 30 measurement assessment (measurement). Describe comparability of assessment methods if there is 31 32 more than one group [Page 10-12] 33 Bias 9 Describe any efforts to address potential sources of bias [N/A] 34 Study size 10 Explain how the study size was arrived at [Page 12] 35 Quantitative variables 11 Explain how quantitative variables were handled in the analyses. If applicable, 36

1 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 67 of 65 BMJ Open

1 2 their precision (eg, 95% confidence interval). Make clear which confounders were

3 adjusted for and why they were included [Pages 14-16] BMJ Open: first published as 10.1136/bmjopen-2019-035678 on 31 October 2020. Downloaded from 4 5 (b) Report category boundaries when continuous variables were categorized [N/A] 6 (c) If relevant, consider translating estimates of relative risk into absolute risk for a 7 meaningful time period [N/A] 8 Other analyses 17 Report other analyses done—eg analyses of subgroups and interactions, and 9 10 sensitivity analyses [N/A] 11 Discussion 12 Key results 18 Summarise key results with reference to study objectives [Page 16] 13 14 Limitations 19 Discuss limitations of the study, taking into account sources of potential bias or 15 imprecision. Discuss both direction and magnitude of any potential bias [Pages 17- 16 18] 17 18 Interpretation For20 Give peer a cautious overall review interpretation ofonly results considering objectives, limitations, 19 multiplicity of analyses, results from similar studies, and other relevant evidence 20 [Page 18] 21 Generalisability 21 Discuss the generalisability (external validity) of the study results [Page 19-20] 22 23 Other information 24 Funding 22 Give the source of funding and the role of the funders for the present study and, if 25 applicable, for the original study on which the present article is based [Page 22] 26 27 28 *Give information separately for exposed and unexposed groups. 29 30 Note: An Explanation and Elaboration article discusses each checklist item and gives methodological background and 31 32 published examples of transparent reporting. The STROBE checklist is best used in conjunction with this article (freely 33 available on the Web sites of PLoS Medicine at http://www.plosmedicine.org/, Annals of Internal Medicine at 34 http://www.annals.org/, and Epidemiology at http://www.epidem.com/). Information on the STROBE Initiative is 35 available at www.strobe-statement.org. 36

37 http://bmjopen.bmj.com/ 38 39 40 41 42 43 44

45 on September 28, 2021 by guest. Protected copyright. 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

2 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open BMJ Open: first published as 10.1136/bmjopen-2019-035678 on 31 October 2020. Downloaded from

Article Type: Original research

Date Submitted by the 04-Aug-2020 Author: