Facilitating early diagnosis of lung cancer amongst primary care patients: The views of GPs

Early diagnosis of lung cancer (LC) is a policy priority. However, symptoms are vague, associated with other morbidities, and frequently unrecognised by both patients and general practitioners (GPs). This qualitative study, part of a larger mixed methods study, explored GP views regarding the potential for early diagnosis of LC within primary care. Five focus group discussions (FGDs) were conducted with GPs (n = 16) at primary care practices (n = 5) across four counties in south England. FGDs were audio‐recorded, transcribed verbatim and analysed using a framework approach. Four broad themes emerged: patients’ reporting of symptoms; GP response to symptoms; investigating LC, and; potential initiatives for early diagnosis. GPs reported they often required high levels of suspicion to refer patients on to specialist respiratory consultations, and concerns of ‘system overload’ were prevalent. Greater access to more sensitive diagnostic investigations such as computed tomography, was argued for by some, particularly for symptomatic patients with negative chest X‐rays. GPs challenged current approaches to promoting earlier diagnosis through national symptom awareness campaigns, arguing instead that interventions targeted at high‐risk individuals might be more effective without burdening services already under pressure. Further work is needed to identify primary care patients who might most benefit from such targeted interventions.

Recent initiatives have sought to improve early diagnosis of lung cancer. For example, the national "Be Clear on Cancer: 3-week cough" campaign was conducted in England between April and June 2012, with a reported significant increase in diagnoses on the same period in the previous year (Ironmonger et al., 2015). This increased diagnosis rate was accompanied, however, by corresponding increases in additional workload, over which GPs had little control (PULSE 2014).
To promote earlier diagnosis of lung cancer without overloading services already under pressure, we need to explore ways to shorten the intervals between the patient first noticing and appraising potential symptoms as requiring clinical attention and then seeking help from their GP, while also promoting the symptoms most favourable for GPs to investigate appropriately. This study was part of a mixed methods investigation of symptomatology and help-seeking behaviour among primary care patients at high-risk (≥50 years old, recent smoking history) (Wagland et al., 2016). We found a high prevalence among participants of both symptoms associated with lung cancer and comorbidities that manifested similar symptoms. Almost half of the participants reported symptoms associated with lung cancer in a questionnaire for whom we found from a clinical notes review did not consult their GP. We also identified a small, clinically relevant group of patients (n = 61/908, 6.7%) who reported experiencing symptoms associated with lung cancer, but whom we found had not consulted their GP for ≥12 months (Wagland et al., 2016). The aim of the qualitative element of the study reported here was to explore the views of GPs regarding how best general practice might facilitate timely diagnosis of lung cancer.

| Setting and sample
Focus group discussions (FGDs) were used for data collection as they explore collective rather than individual experiences and reveal the nature and variety of participants' views (Krueger & Casey, 2000). Participants were drawn from five primary care practices across four counties within southern England. Recruitment of practices was facilitated by close collaboration with the Primary Care Research Network, to ensure a representative range of practice size and social deprivation. FGDs took place between April and August 2014 within the practice premises of collaborating GPs, and lasted 1-2 hr (mean: 1 hr 25 min). An experienced qualitative researcher (RW) conducted the focus groups as moderator, with a second researcher (AI-E) acting as observer and note-taker.

| Materials
An interview topic guide was developed to elicit participants' views regarding specific symptoms, symptom combinations, severity and chronicity of symptoms reported by patients that would raise suspicions of lung cancer. GPs gave their views regarding the importance of nine symptoms experienced by patients subsequently diagnosed with lung cancer, which comprise the IPCARD (Identifying Symptom Predictors of Chest and Respiratory Disease) questionnaire (Brindle et al., 2015). IPCARD, developed by members of the research team (Brindle et al., 2015), asks individuals about the following symptoms in lay terms to facilitate elicitation: tiredness; breathing changes; chest and upper body aches; cough; coughing up blood; non-menopausal sweats; ongoing voice changes; unintentional weight loss; and noticeably more chest infections over a 12-month period. GPs were informed of the preliminary findings from the survey and clinical notes review from the wider study (Wagland et al., 2016). Their views regarding perceived barriers and facilitators to early diagnosis were then explored and practical considerations of administering interventions to encourage primary care patients to consult for symptoms potentially indicative of lung cancer.

| Analysis
Focus group discussions were transcribed verbatim and analysed using the computer programme Nvivo10 for assistance in structuring a framework analysis (Ritchie & Lewis, 2003). Framework analysis aims at facilitating applied research and its requirements to meet specific information needs and actionable outcomes and is conducted through a series of stages. The focus group moderator led the analysis, ensuring deep familiarisation with the data.
To facilitate data analysis, three researchers (RW, EJ, AIE) independently analysed one FGD and discussed findings with the full research team to agree upon an index of emerging themes. Thereafter, regular 2-weekly discussions took place between the three researchers to review the development of a thematic framework and ensure analytical rigour. A thematic framework was developed, initially drawn from the topic guide, to identify the key concepts central to the symptomology associated with identifying patients with lung cancer, judging symptom severity and the potential for early diagnosis. During the analysis, other categories were derived from the data, including critiques of national symptom awareness campaigns; issues related to investigating lung cancer and fatalistic attitudes amongst some GPs.
The whole data set was then indexed according to these categories (indexing) and comparisons made both within and between them according to their thematic content, and data within categories summarised (charting). Relationships and associations between the categories were then identified (mapping and interpretation) that explained GPs' views regarding the facilitation of timely diagnosis of lung cancer during the diagnostic interval.

| Ethical approval
Ethical approval for the study was secured by the National Research  Four broad themes emerged from the data: patients' reporting of symptoms; GPs response to specific symptoms; issues related to investigating suspected lung cancer, and; issues related to the potential effectiveness of early diagnosis interventions (Table 2).

| Patients' reporting of symptoms
The subjective nature of symptom experience made appraising the severity and seriousness of symptoms reported by patients difficult for GPs, especially for symptoms such as fatigue and chest pain for which limited objective measurements exist. GPs described how many older patients with smoking histories had several comorbidities, and as the extract below illustrates, GPs recognised that consequently those patients often believed it was normal to experience symptoms such as cough, fatigue and breathlessness, and consequently did not report them even when they worsened.

GP1: Until you ask the question, [patients] change their
boundaries of what they do, to accommodate how they feel. They're not actually aware sometimes that their shortness of breath has got worse, they just don't walk so far, or they don't do this or that, and until you actually ask the question, they wouldn't come and see you because they're not aware of it. GP2: They didn't think it was a problem, they just changed the way that they live their lives, because I find that with COPD patients, … They don't think they're any worse than they were a year before, but when you actually ask that question, they're quite considerably worse but they're not aware of it. GP1: It's actually asking the question rather than waiting for them to present with a problem that they don't perceive as a problem. symptoms, such as voice changes, patients were unlikely to report them.
In addition, GPs reported that smokers often experienced guilt for symptoms deemed "self-inflicted", and were consequently reluctant to report them to GPs. The reported duration and severity of symptomology by patients was often vague and reported differently between consultations (the story changes), further complicating diagnosis.

| GP response to specific symptoms
NICE guidelines (NICE, 2005;NICE, 2015) recommend an urgent chest X-ray (CXR) for patients with the following symptoms: unexplained haemoptysis; cough; fatigue; shortness of breath; chest pain, and; weight loss. The guidelines also recommend CXRs be offered to patients who present with recurrent chest infection. During FGDs, GPs were asked about two additional symptoms incorporated within IPCARD and potentially indicative of lung cancer: ongoing voice changes and non-menopausal sweating (Brindle, Pope, Corner, Leydon, & Banerjee, 2012). GPs were asked to consider their experience of reviewing patients with these symptoms and the relative weight they gave to each symptom (see Table 3).
General practitioners considered many of these as "red flag" symptoms, but breathing changes, repeated chest infections, unintentional considered of most concern if the only ones present were haemoptysis, unintentional weight loss and persistent cough lasting longer than 6 weeks, although weight loss might be indicative of any tumour type.
Other important signs identified during the FGDs included persistent hoarseness, and disturbed sleep caused by any of the other symptoms.
General practitioners described that while some patients were frequent attenders, others rarely if ever consulted GPs whatever symptoms they experienced. As the extract below illustrates, GPs in two FGDs argued that patients who rarely consulted the practice but then suddenly reported symptoms should trigger a high level of GP concern, irrespective of specific symptoms.  Nevertheless, CXRs were perceived as blunt instruments and concern existed amongst GPs regarding the optimum timing of this investigation.

| Investigating patients for lung cancer
Several participants were concerned that CXRs were often insufficiently sensitive to identify a lung tumour until quite large, and possibly inoperable. At the same time, as the GPs in the following extract argue, if CXRs were conducted when the mass was too small to be detected, then the negative result may serve as a false sense of security for both patient and GP. concern. It would make the process clearer for us and cost the system less. (FGD 4) Agreement about direct access to CT scanning was not universal, however, with some GPs indicating false negatives were similarly possible or that referring patients for CTs prior to secondary referral could slow rather than speed the process.
Despite the reportedly low threshold of suspicion to conduct a CXR, several GPs were less certain as to how they would proceed if the CXR result were inconclusive but symptoms persisted. As the following extract shows, a higher threshold of evidence appeared necessary for most GPs before referring patients to specialist secondary care teams, partly due to a concern that the system would sink if they referred all patients about whom they had low level suspicion.

| Potential for early diagnosis
Opinions differed amongst GPs with regards the potential for effec-

| DISCUSSION
Limited work has previously investigated GP's views of their own role in early detection (Green, Atkin, & Macleod, 2015). This study conducted focus groups with GPs from participating general practice sites to elicit their views with regards facilitating targeted interventions.
Participant GPs identified what they perceived as the three most relevant symptoms for diagnosing possible lung cancer as: recent, significant weight loss; persistent cough for longer than 6 weeks; and haemoptysis. There was no consensus between FGDs on those symptoms most indicative of early lung cancer. Previous research has found that lung cancer symptoms may be experienced only as vague or mild (Smith, Pope, & Botha, 2005), may be confounded by high levels of comorbidities (Stolper et al., 2011), and patients subsequently diagnosed with lung cancer reported good health prior to diagnosis (Brindle et al., 2012). Studies have shown patients with lung cancer commonly experience multiple and synchronous symptoms (Hamilton et al., 2005;Walter et al. 2015), and may be symptomatic for many months before presentation (Corner, Hopkinson, Fitzsimmons, Barclay, & Muers, 2005). Thus, patients are often unable to recognise all their symptoms (Smith et al., 2009), appraise symptoms as not warranting help-seeking (Corner & Brindle, 2011), or else normalise symptoms, attributing their cause either to ageing processes or comorbidities (Corner et al., 2005;Tod & Joanne, 2010). Members of this research team have previously argued that to better elicit lung cancer symptoms, GPs may need to ask patients closed questions using non-disease terminology (Brindle et al., 2012).
General practitioners emphasised that clinical histories and GP "hunches" were as important as specific symptoms in identifying patients who should be further investigated for lung cancer (Stolper et al., 2011). GPs also argued that symptomatic patients who rarely attended the practice, irrespective of their presenting symptoms, would trigger their concern. The CAPER studies and QCancer algorithms have provided risk prediction models for cancer types, including lung cancer (Hamilton, 2009;Hippisley-Cox & Coupland, 2011). However, of a sample of patients who subsequently developed lung cancer, between 17% and 34% of symptoms presented in the previous 24 months were not caused by the cancer (Biswas, Ades, & Hamilton, 2015). Also, while hemoptysis is the strongest symptom predictor of lung cancer, only a fifth of patients experience it (Walter et al., 2015).
In response to symptoms, GPs reported they required low thresholds of suspicion of lung cancer before referring patients for CXRs; the standard initial investigation for symptoms indicative of lung cancer (NICE, 2005;NICE, 2011;NICE, 2015). Nevertheless, GPs expressed limited confidence in the diagnostic capacity of CXRs, and previous studies found few CXRs identified signs of lung cancer in patients 6 months or more prior to their actual diagnosis (Hamilton et al., 2005;Stapley, Sharp, & Hamilton, 2006). Uncertainty also existed amongst GPs regarding both how best to proceed if a patient's symptoms persist despite a negative CXR, and the level of suspicion appropriate before specialist respiratory referral should be made. The recently revised NICE guidance for suspected cancer referrals has reduced the expected probability threshold for a cancer diagnosis to trigger a secondary referral, from the previous predictive and prognostic value (PPV) threshold of approximately 5%-3% (NICE, 2015). Effectively, this means an extension in the number of patients needed to be referred (NNtR) for one cancer diagnosis from approximately 20 to 33. Nevertheless, GPs felt pressure not to refer patients with vague symptoms on to specialist secondary care without sufficient diagnostic evidence, and cited concern for "system overload". It is therefore unclear how GPs will respond to a lower PPV threshold for cancer referrals and whether the guideline change will result in any meaningful behaviour change, hence referrals might remain restricted at practice level. Further research should explore whether and how GPs utilise the reduced PPV threshold. Although some GPs were ambivalent about greater access to additional diagnostic procedures such as CT, others argued it would give them greater confidence in the PPV of suspected lung cancer cases, better facilitating timely diagnosis. Investigating this potential is another area for future work.
While acknowledging the impact of the "3-week cough" campaign, GPs argued such national initiatives often heightened demand on limited resources, but may have little impact upon those patients most at risk who ignore them. The existence of a symptomatic, nonconsulting group of primary care patients identified by our wider study would support this view (Wagland et al., 2016). Although the "3-week cough campaign" led to a 9% increase in lung cancer diagnoses compared with the same period in the previous year (Ironmonger et al., 2015), there was a corresponding increase of >200,000 additional GP attendances and 30% increase in 2-week waits recorded (PULSE 2014). Evidence also indicates such increases are particularly apparent in affluent rather than deprived areas (Green et al., 2015).
All GPs in our sample recognised the importance for seeking methods for timely diagnoses of lung cancer, despite concerns about increased workloads. GPs within all FGDs agreed interventions targeting patients at high-risk of lung cancer and who rarely attend primary care, might be at least as effective and cost-effective as targeting specific symptoms. Potential methods, given this study's findings, would include practice-level interventions that allowed GPs control over identifying and contacting "high-risk" patients, facilitating planning for subsequent workload increases. Further work is needed to identify profiles of primary care patients who would benefit most from such targeted interventions.

| Strength and limitations
Participant GPs had a broad range of experience (mean: 19 practice years), represented practices of different sizes, with both high and low levels of social deprivation, in both rural and urban settings, and consensus existed across FGDs on most themes.

| CONCLUSION
General practitioners questioned current approaches to promoting earlier diagnosis through national campaigns. Given the problematic "symptom signature" and corresponding difficulties for both GPs and patients to recognise symptoms of lung cancer, future interventions promoting early diagnosis of lung cancer should include the targeting of "high-risk" individuals. Some GPs also argued for greater access to more sensitive diagnostic investigations, in particular CT scans, to enhance the PPV of secondary referrals.
Allowing GPs to target "at-risk" patients on their lists would allow them to plan for the more limited workload increases these would entail compared with large national symptom campaigns over which they have limited control. Further work is required to identify primary care patients who would most benefit from such targeted interventions.