Intended for healthcare professionals


Early diagnosis of cancer: systems approach to support clinicians in primary care

BMJ 2023; 380 doi: (Published 09 February 2023) Cite this as: BMJ 2023;380:e071225

Linked Editorial

Direct access to imaging for cancer from primary care

  1. Georgia B Black, THIS Institute postdoctoral fellow1,
  2. Georgios Lyratzopoulos, professor of cancer epidemiology2,
  3. Charles A Vincent, professor of psychology3,
  4. Naomi J Fulop, professor of healthcare organisation and management1,
  5. Brian D Nicholson, academic clinical lecturer in general practice4
  1. 1Department of Applied Health Research, University College London, London, UK
  2. 2ECHO (Epidemiology of Cancer Healthcare and Outcomes), Department of Behavioural Science and Health, University College London, UK
  3. 3Department of Experimental Psychology, University of Oxford, Oxford, UK
  4. 4Nuffield Department of Primary Care Health Sciences, University of Oxford, UK
  5. Correspondence to: G B Black{at}

Georgia Black and colleagues argue that rather than focusing on the initial consultation efforts to reduce diagnosis times should look at the wider system

The substantial deficit in primary care contacts and diagnostic testing for suspected cancer during the covid-19 pandemic fed into a backlog of patients awaiting investigations.1 Modelling using NHS datasets suggests that these delays will result in around 3500 additional deaths from breast, colorectal, oesophageal, and lung cancer over the next few years in England alone.2

Delays in cancer diagnosis vary according to type of cancer. For example, the median time between presentation in primary care and referral for specialist assessment in England is 24 days for multiple myeloma compared with 0 days for breast cancer.34 Urgent referral pathways are designed to expedite cancer diagnosis for patients with concerning symptoms, and testicular and breast cancers are far more likely to be diagnosed through this route (73% and 71%, respectively). By contrast, only 6% of brain cancers and 20% of leukaemia cases are diagnosed through urgent referral pathways.5

The differential diagnosis is wider for patients with non-specific symptoms, so multiple primary care consultations are often required to establish the diagnosis.6 Referral processes also depend on the severity of presenting symptoms, and many patients with pancreatic, liver, and stomach cancers are diagnosed during emergency hospital admissions.4

Current approaches to improving diagnostic accuracy and timeliness place most weight on improving the initial primary care consultation. However, the diagnostic process is more often dynamic, involving many different encounters and healthcare professionals. A wider approach that shifts the approach away from individuals and onto the whole system is likely to have a bigger impact.

Reasons for delays

People with symptoms that may indicate underlying cancer fall into two groups: patients with alarm or “red flag” symptoms with relatively high predictive value or a specific cancer (eg, a breast lump in breast cancer or rectal bleeding in colorectal cancer) and patients who present with non-specific symptoms with a low predictive value for cancer. Clinical guidelines clearly recommend urgent referral of patients with red flag symptoms for specialist assessment, but referral is often more prolonged for those with non-specific symptoms. Such patients may also wait some time before seeking help, exacerbating delays.7

Many patients with suspected cancer experience a series of investigations, each adding additional opportunities for problems, delays, and errors.8 These can include problems accessing diagnostic tests or specialist advice but also relate to problems of poor communication and coordination of care. For example, clinical information is often incomplete, dispersed across medical notes, referral letters, test results, and clinical letters to patients. Even when clinical management follows guidelines, some delays and errors are inevitable because of limitations with existing guidelines and the inherent uncertainty in the diagnostic process.9

Delays and uncertainty in care may cause patients unnecessary anxiety about cancer risk and result in further diagnostic delays because of reluctance to seek or re-seek healthcare advice, failure to follow advice, missed appointments, or communication breakdown.101112 Pathways have now been developed to investigate non-specific symptoms such as abdominal pain that may be associated with several cancers as well as benign or other serious diseases, and these may support primary care investigation of patients who would not otherwise be eligible for urgent referral.13

Attempts to reduce delays have had limited success

Historically, most interventions aimed at reducing delays in cancer diagnosis assume that it is possible to improve diagnostic accuracy within a single initial consultation. For example, educational interventions for clinicians and the public to raise awareness of cancer signs and symptoms aim to act at first contact (table 1). Other interventions such as guidelines, decision aids, and communication tools are also predicated on supporting decision making within the presenting consultations.

Table 1

Interventions designed to improve early diagnosis of cancer

View this table:

The proportion of all cancers diagnosed at an early stage in the UK has increased by 2% since 2013, yet the results for individual cancers (and particularly those presenting with non-specific symptoms) are less encouraging.16 For example, the proportion of patients diagnosed with Hodgkin’s lymphoma while in stage I remained at just over 13% during 2013-19.16 Although trends in stage at diagnosis are hard to interpret as they may represent both increasing earlier diagnosis of symptomatic disease and increasing overdiagnosis, the effect of interventions to reduce late stage diagnosis remains uncertain. A campaign to increase awareness of colorectal cancer in the UK was shown to have little effect on stage at diagnosis and had no measurable effect on cancer survival.17 Knowledge is not the only barrier to seeking care; patients may normalise symptoms and self-manage until their symptoms get worse.1819

Few interventions acknowledge the reality that primary healthcare professionals will probably not suspect cancer at the first consultation, or, if cancer is suspected, the risk will be considered relatively low. Managing the most likely diagnosis may therefore be appropriate. The efficiency of diagnostic decision aids, risk prediction tools, and safety netting relies on clinician suspicion of cancer; clinicians are less likely to use these tools or to engage fully with them if they do not have a high suspicion of cancer.20 Most cancer patients with non-specific symptoms have multiple consultations before getting a diagnosis, and they experience periods of unmonitored time in the process.21 Relatively few interventions are targeted at reducing delays during these periods for reasons such as over-reassurance from a benign explanation for their symptoms, misunderstanding instructions to monitor symptoms at home, or fear of bothering the doctor.10

Potential benefits of systems approach

A systems approach to improving the quality of healthcare involves thinking about all the interconnected components (eg, healthcare professionals and organisations, technology, equipment, and workplace culture) that act together to achieve a common goal.2223 Interventions designed from this viewpoint embrace the complexity and unpredictability of clinical work rather than trying to reduce predictable errors in one area.24 Typical system interventions support effective team (rather than individual) delivery of care, monitoring for early warning signals, or anticipating risks or opportunities.25 For example, use of a wireless patient physiological surveillance system improved patient flow and detection of patient deterioration as well as reducing staff administration time in a general care setting.26

In the case of cancer diagnosis, a systems approach could improve monitoring for minor errors and delays in primary care investigations, whether or not a clinician suspects cancer or another serious disease.27 A systems approach places less responsibility on individuals by using information technology and proactive monitoring by healthcare teams to enhance the diagnostic process. It could also help circumvent typical problems such as escalating workloads. Figure 1 includes an example of a system capable of mitigating delays and errors in a potential cancer diagnosis.

Fig 1
Fig 1

Approach to avoid delays in management of a patient who is not immediately referred for cancer investigations

Improvements made at a system level are likely to improve diagnosis in general, and not just for cancer. Such an approach takes a pragmatic view of what is achievable in a single consultation, accepting that a diagnosis could reasonably be missed. For patients who are not immediately referred, for example, a systems approach could incorporate automated tracking to reduce delays or errors incurred through multiple follow-up visits (eg, to see if symptoms respond to initial intervention) and unmonitored time between visits.6

A systems approach shares responsibility for monitoring patient progress rather than placing it on the primary care doctor or the patient. Missed actions or communications might initiate an automated system response (eg, sending a reminder or transferring responsibility to someone else), for instance. This is particularly relevant as patients may consult with multiple clinicians, including locum or trainee practitioners.

Some system level interventions have already been implemented successfully and have reduced errors and delays in the care pathway. In the United States, for example, software was developed to alert clinicians when a positive faecal occult blood test result had not been followed up after three weeks.28 In a 2015 cluster randomised control trial, a similar algorithm developed to examine electronic records and identify patients with abnormal imaging results reduced time to diagnostic evaluation among patients being investigated for colorectal (from median 200 to 104 days) and prostate cancer (192 to 144 days).29 A 2018 qualitative study also showed the acceptability of using text messaging triggered automatically by the electronic patient record to help primary care patients with low risk symptoms to monitor their symptoms at home.30

Systems approaches do not always require technology: patient navigators (people who facilitate patients to attend healthcare appointments) can mitigate non-attendance by providing support, reminders, and patient education, and have been shown to improve screening uptake across a range of studies.31 Patients who cannot use smartphones or electronic devices could be supported through existing validated paper based tools such as information summary sheets, delivered at the point of consultation to avoid deepening health inequalities.32

Adopting a systems approach to cancer diagnosis will require further research and development and may require multiple infrastructure changes, such as improving patient access to phlebotomy and pharmacy.33 This would need to be coupled with technological support to issue automated reminders (by text message or telephone call), linking results to the correct patient record, automated identification of errors from the electronic health record (eg, missed prescriptions, missed or delayed blood, imaging, or endoscopy tests) and a clear protocol for who is responsible for following up.34 It would also include protocols for relaying blood or imaging results (both normal and abnormal) that do not depend on patient initiative.

Patients are likely to support this approach as they often cite discontinuity, fragmentation, and inflexibility of care as concerns.35 Patients could be reassured by knowing that someone is monitoring their situation and providing clarity about what is expected of them.3637 For example, patients want reassurance about whether they are allowed to contact specialists when their referral is delayed.12

Moving forward

The introduction of system measures to expedite diagnosis can have unintended consequences for professionals. These include risk compensation (acting in a more risky way when you know interventions are in place) and alert fatigue (when there are too many pop-ups, triggers, and other interventions), although improvements in visual design may mitigate some of these effects.3839

Limitations in current technology and health system resources may present a barrier to system change, although a systems approach may reduce long term demand.22 There is also widespread resistance to technology among healthcare professionals and tolerance of current flawed systems because of familiarity.40 There is a danger that automated systems could create inequalities for patients with atypical presentations not included in guidelines and those with incomplete patient records, generating new barriers to access. It is unclear whether this would exacerbate inequalities that already exist in the UK.41 We envisage that a systems approach would free up clinical capacity in primary care and potentially overcome other barriers to care such as clinical bias and administrative error. However, increasing demands for the diagnosis and management of patients whose symptoms are not caused by serious illness could absorb any efficiency gains.

Research and investment are needed in new and existing technologies that support system safety, underpinned by sociotechnical models that acknowledge the complex relationships between humans, technology, and organisational cultures. Not all errors and failures in diagnostic pathways are unavoidable, but they can be minimised by designing a system that increases the resilience of the diagnostic process by proactively monitoring, detecting, and resolving problems as they occur.

Key messages

  • Non-specific presenting symptoms make the timely diagnosis of cancer in primary care challenging

  • Initial misattribution of symptoms to other diseases is common, leading to prolonged diagnostic intervals and worse outcomes in some patients

  • Systems approaches that automate information sharing, symptom monitoring, and detection of missed or incomplete actions could reduce unnecessary delays in diagnosis


GBB is funded by a Fellowship from The Institute of Healthcare Improvement Studies (University of Cambridge) (RG88620/PD-2019-02-004). GL acknowledges support by Cancer Research UK Advanced Clinician Scientist Fellowship Award (C18081/A18180). CAV is supported by the Health Foundation. NJF is an NIHR Senior Investigator. BDN is an NIHR GP Academic Clinical Lecturer and Cancer Research UK postdoctoral fellow (RCCPDF\100005).##


  • Contributors and sources: GBB is a social psychologist whose research focuses on patient safety in diagnostic pathways and the effect of socioeconomic inequalities and exclusion from healthcare. GL is an epidemiologist, specialising in cancer healthcare and early diagnosis research. CV is a clinical psychologist who has conducted extensive research on the causes of harm to patients, the consequences for patients and staff, and methods of improving the safety of healthcare. NF is a social scientist and health services researcher with expertise in qualitative and mixed methods studies of change and improvement at the interfaces between health policy and service delivery, management, and organisation of healthcare. BDN’s research portfolio centres on improving diagnosis for patients with non-specific or vague symptoms of cancer. GBB conceptualised the article and drafted the narrative. All authors developed components of the article and identified examples from clinical practice. GBB is the guarantor.

  • Patient and public involvement: Members of the public were consulted about their views and priorities for research about diagnosis in primary care in a stakeholder event in December 2018. They highlighted the importance of continuity and ‘system thinking’ based on their experiences as patients. For example, members of the public highlighted that it can be difficult to return to primary care for a second or third consultation, and to be unsure about who to contact for test results.

  • Competing interests: We have read and understood BMJ policy on declaration of interests and have no relevant interests to declare.

  • Provenance and peer review: Not commissioned; externally peer reviewed.