Abstract
Background
Emergency diagnosis of lung cancer is common and associated with worse prognosis.
Aim
To compare prediagnostic healthcare use between emergency-diagnosed patients and patients referred routinely or urgently.
Design and setting
Population-based linked English primary care, hospital admission, imaging, and cancer registration data were analysed for patients with lung cancer (2007–2018).
Method
Monthly prediagnosis rates of consultations (for any clinical reason and selected symptoms) and chest imaging by diagnostic route (emergency, routine referral, and urgent referral) were measured. Multivariable Poisson regression estimated route-specific event rates and inflection points.
Results
The study examined 4473 patients with lung cancer with features representative of the nationwide incident cohort, of whom 33% (n = 1491) were emergency diagnosed. Most (emergency diagnosis, n = 1473/1491; routine referral, n = 1023/1026; urgent referral, n = 1252/1259; and other, n = 684/697) had consulted in primary care in the year prediagnosis, independent of diagnostic route. Consultation and imaging rates increased from 5- and 4-months prediagnosis, respectively, with shorter diagnostic windows in emergency-diagnosed than referred route patients. Compared with emergency-diagnosed patients, referred route patients had higher prediagnostic consultations rates for cough (adjusted incidence rate ratio [aIRR] compared with emergency-diagnosed 1.90, 95% confidence interval [CI] = 1.58 to 2.30 for routinely and 1.94, 95% CI = 1.61 to 2.33 for urgently referred) and chest X-ray use (aIRR 1.91, 95% CI = 1.53 to 2.38 for routinely and 1.77, 95% CI = 1.42 to 2.21 for urgently referred).
Conclusion
Similar or shorter diagnostic windows suggest similar potential for earlier diagnosis among emergency-diagnosed and referred route patients alike. Earlier detection may be supported through improved management of non-specific symptoms, timely follow-up of imaging, and greater access to chest computed tomography. Future research should measure missed diagnostic opportunities to identify clinical actions to further reduce emergency lung cancer diagnoses.
How this fits in
It has been postulated that emergency diagnoses of cancer (which occurs frequently and confers a poorer prognosis) may relate to suboptimal diagnostic management in primary care, but evidence to support or refute this hypothesis is sparse. This study found that emergency-diagnosed patients with lung cancer were less likely to present with relevant respiratory symptoms and had fewer chest imaging investigations before diagnosis compared with patients diagnosed via referred routes, indicating an important role of disease factors in emergency diagnosis. The study also found that the rate of prediagnostic healthcare activity increased at the same time or closer to diagnosis for emergency-diagnosed patients compared with those diagnosed through referred routes, suggesting a similar potential for earlier diagnosis for all diagnostic routes. Improved imaging follow-up, greater access and use of computed tomography, and greater use of safety netting in primary care could help diagnose lung cancer earlier across all diagnostic routes.
Introduction
Patients with lung cancer are often diagnosed through an emergency presentation route, for example, after an emergency hospital admission or an emergency department attendance. 1 Emergency diagnosis is strongly associated with worse survival, independent of stage at diagnosis. 2 The extent to which emergency diagnoses are avoidable remains unclear, as tumour, patient, and health system factors may all play a role. 3 Reductions may be possible through improved help-seeking or by enhancing the diagnostic process post-presentation. However, rapidly progressing symptoms in the context of aggressive tumours may also contribute. 3
In the UK, the NHS provides universal, publicly funded care, with GPs central to cancer recognition and referral. The National Institute for Health and Care Excellence (NICE) guidelines outline symptoms, risk factors, and thresholds for urgent referral (the ‘2-week-wait’ pathway) to support timely diagnosis. 4 Despite these systems, many cancers are still diagnosed late, often via emergency presentations.
The concept of a ‘diagnostic window’ has been used in cancer and other diseases, 5–9 representing prediagnostic periods where healthcare use for a patient group increases from baseline, during which there is potential for earlier diagnosis. The time point where rates start to rise is known as the inflection point. Diagnostic windows can be defined by rising rates of events such as GP consultations, hospital visits, prescriptions, or investigations. The presence of diagnostic windows has been described in patients with lung cancer in Denmark, 10,11 England, 8,12 Australia, 13 and New Zealand, 14 without examining differences by diagnostic route. Such analyses can elucidate likely mechanisms leading to emergency diagnosis, as shown for haematological, 15 colorectal, 8,9 and lung cancers. 10,14
Prior studies have shown patients with emergency-diagnosed colon cancer are less likely to present with alarm symptoms like rectal bleeding than those diagnosed via referrals. 9 Emergency-diagnosed patients are more likely to have advanced or unknown stage and unspecified tumour type. However, this research relies on recorded and coded primary care symptoms, limiting the analysis to those captured in the data, which may not completely represent patients’ true symptom experiences. These patterns suggest emergency-diagnosed patients may have more aggressive tumours, although patient- and system-level factors may also contribute.
Additionally, patients may have shorter diagnostic intervals because they experience more severe symptoms and aggressive tumour presentation, and so may receive specialist care and treatment sooner. This concept is referred to as the ‘waiting time paradox’, whereby patients with shorter diagnostic intervals have worse outcomes than those with longer diagnostic intervals. 16
In lung cancer, lower prediagnostic healthcare use rates and shorter diagnostic windows among emergency-diagnosed patients would point to tumour factors influencing the diagnostic route. In contrast, higher prediagnostic healthcare use rates and longer diagnostic windows among emergency-diagnosed patients would suggest opportunities for improving the diagnostic process. This study examined whether prediagnostic healthcare use differed between patients diagnosed as emergencies or through referred routes.
Method
Study population and data sources
Patients with lung cancer were identified from a random sample of 1 million patients registered with Clinical Practice Research Datalink (CPRD) GOLD (UK primary care data source, November 2021 build) aged 30–99 years with ≥1 year of registration during January 2007 to October 2018. Patients diagnosed with lung cancer were diagnosed during January 2007 to October 2018, defined using International Classification of Diseases (ICD)-10 codes C33 or C34 in the cancer registry records. Data were linked to the National Cancer Registration and Analysis Service (NCRAS), Hospital Episode Statistics Admitted Patient Care (HES APC), and imaging data (HES DID, available from April 2012 onwards) (see Supplementary Figure S1). Index of Multiple Deprivation (IMD) quintiles (a small-area measure of socioeconomic status) were assigned based on the patient’s residence postcode.
Explanatory and outcome variables
The main outcome was the diagnostic route by which patients with lung cancer were diagnosed, assigned using NCRAS’s routes-to-diagnosis algorithm. 17 Diagnosis via an emergency presentation was compared with the two main referred routes: urgent referral for suspected cancer (hereafter denoted as ‘urgent’ referral, and historically known as 2-week-wait referral) or GP routine referral. 18
Primary explanatory variables were types of healthcare use in the 24 months before diagnosis, excluding the 30 days immediately prediagnosis. First, monthly mean event rates per patient were calculated. Second, diagnostic window length was measured, defined as the month when the inflection point occurred (that is, when healthcare use started to increase from baseline).
Monthly rates (a month is defined as 30 days) of clinical primary care consultations (limited to one per day, excluding administrative encounters); 19,20 and consultations for six selected relevant symptoms, comprising three respiratory symptoms (cough, dyspnoea, and haemoptysis, assessed individually and grouped) included in referral guidelines for suspected lung cancer, 21 and three non-localising symptoms (appetite loss, weight loss, and fatigue, assessed as a group) associated with lung cancer were examined. 22 Symptoms were defined in CPRD using previously published Read (version 2 [v2]) codelists. 23 Same-day repeat events of the same type were excluded. Chest imaging recorded in HES DID, including chest X-ray and chest computed tomography (CT), using previously published National Interim Clinical Imaging Procedure and Systematised Nomenclature of Medicine codelists were also examined. 24
Covariates
Several patient-level factors were included as covariates, including age at diagnosis, gender, and patient’s area-level IMD (2015). Ethnicity was identified from NCRAS, categorised as ‘White’, ‘any other ethnicity’ (‘Asian’, ‘Black’, ‘mixed’, or ‘other’), and ‘unknown’. 25 Smoking status at time of diagnosis was identified using Read v2 codes in CPRD, categorised as ‘smoker’ (‘current’ or ‘ex-smoker’) and ‘non-smoker’. 26,27 Chronic obstructive pulmonary disease (COPD) status was identified using Read v2 and ICD-10 codes in CPRD and HES APC, categorised as ‘no COPD’, ‘new-onset COPD’ (diagnosed within 24 months before lung cancer), and ‘pre-existing COPD’ (diagnosed >24 months before lung cancer diagnosis) (see Supplementary Figure S2). 12,28 Elixhauser comorbidity score was identified using ICD-10 codes in HES APC, excluding codes for COPD and lung cancer, categorised into scores of 0, 1 ,2, or ≥3. 29–31 Cancer-related factors included cancer stage at diagnosis (categorised as ‘advanced’ for stages 3–4 and ‘non-advanced’ for stages 1–2) 32 and morphology (non-small-cell and small-cell lung cancer). 33 A restriction window of 12 months prediagnosis was applied to the Elixhauser scores (see Supplementary Figure S3). 34 Death within a year post-diagnosis defined using the death date recorded in CPRD was also described. Codelists used in this study are available online (see Data section).
Statistical analysis
The study examined healthcare use across segmented prediagnosis periods (months 24–1, 24–12, 12–6, and 6–1 before diagnosis) and, in line with previous studies, 6,9 results for the 12 months prediagnosis where most changes occurred are reported (see Supplementary Figure S4).
Crude rates were calculated as the mean number of events per patient per month. For each month, total events was summed and divided by the number of patients with follow-up. Crude rates were stratified by diagnostic route. For urgently and routinely referred groups, age-standardised rates are also reported using the emergency group’s age distribution as the standard to directly allow for between-route comparisons. The cumulative monthly proportion of patients with ≥1 relevant event in the prediagnostic year was calculated.
Mixed-effects Poisson models (including a random effect for patient) 35 were fitted to compare prediagnostic rates (mean monthly events per patient) by diagnostic route, using a crude model and adjusted for patient factors (age, gender, ethnicity, IMD, COPD status, smoking status, and Elixhauser score). The emergency-diagnosed group was used as the reference category. Including morphology and stage did not materially affect results and therefore was excluded. The month immediately prediagnosis was excluded from the crude rate modelling.
The maximum likelihood estimation method 20,36 was used to estimate the inflection points for each healthcare use type by diagnostic route for the entire 12-month period before diagnosis. This method fits a series of Poisson regression models (as above), each fitting a unique inflection-point variable corresponding to a different prediagnosis month. The model with the month with highest log-likelihood denotes the month corresponding to the inflection point. The 95% confidence intervals (CIs) were estimated using bootstrapping. Periods before the inflection points are referred to as background periods.
Results
Cohort description
After exclusions, 4473 patients with lung cancer were included in analysis (Figure 1): 1491 patients (33%) were diagnosed as emergencies, 1259 (28%) through urgent referral, 1026 (23%) through routine referral, and 697 (16%) through other routes (Table 1). Similar proportions of patients presented to primary care before diagnosis across all routes (Table 2). Emergency-diagnosed patients were on average older and more likely to have advanced or unknown stage at diagnosis, tumour of unspecified morphology, pre-existing or new-onset COPD, higher Elixhauser comorbidity score, and to die within a year post-diagnosis compared with referred route patients (Table 1). The regional distribution of patients is presented in Supplementary Table S1.
Figure 1. Flowchart describing the cohort selection process. a2059 patients were diagnosed after April 1 2012, and were therefore included in analyses relating to chest imaging. IMD = Index of Multiple Deprivation. NCRAS = National Cancer Registration and Analysis Service.
Table 1. Comparison of characteristics of the studied patients with lung cancer by diagnostic route (N = 4473).
| Variable | Emergency diagnosis ( n = 1491, 33%), n (%)a | Routine referral ( n = 1026, 23%), n (%)a | Urgent referral ( n = 1259, 28%), n (%)a | Other (n = 697, 16%),n (%)a,b | P-value c |
| Age at diagnosis, years | <0.001 | ||||
| Mean (SD) | 75 (11) | 72 (10) | 71 (10) | 72 (11) | |
| Median (IQR), range | 76 (68–84), 32–99 | 73 (65–80), 39–96 | 71 (63–78), 36–97 | 72 (65–79), 38–98 | |
| Age at diagnosis, years | <0.001 | ||||
| 30–64 | 258 (17) | 235 (23) | 343 (27) | 172 (25) | |
| 65–74 | 391 (26) | 360 (35) | 441 (35) | 231 (33) | |
| 75–84 | 508 (34) | 334 (33) | 363 (29) | 220 (32) | |
| 85–99 | 334 (22) | 97 (9) | 112 (9) | 74 (11) | |
| Stage of cancer at diagnosis | <0.001 | ||||
| Advanced | 785 (53) | 473 (46) | 695 (55) | 289 (41) | |
| Not advanced | 106 (7) | 210 (20) | 214 (17) | 142 (20) | |
| Unknown | 600 (40) | 343 (33) | 350 (28) | 266 (38) | |
| Gender | 0.130 | ||||
| Female | 722 (48) | 470 (46) | 558 (44) | 337 (48) | |
| Male | 769 (52) | 556 (54) | 701 (56) | 360 (52) | |
| Ethnicity | <0.001 | ||||
| White | 1388 (93) | 962 (94) | 1198 (95) | 618 (89) | |
| Any other ethnicity | 81 (5) | 45 (4) | 48 (4) | 45 (6) | |
| Unknown | 22 (1) | 19 (2) | 13 (1) | 34 (5) | |
| IMD | 0.500 | ||||
| 1 (least deprived) | 239 (16) | 191 (19) | 200 (16) | 134 (19) | |
| 2 | 271 (18) | 205 (20) | 239 (19) | 137 (20) | |
| 3 | 319 (21) | 209 (20) | 267 (21) | 148 (21) | |
| 4 | 325 (22) | 215 (21) | 284 (23) | 144 (21) | |
| 5 (most deprived) | 337 (23) | 206 (20) | 269 (21) | 134 (19) | |
| Smoker status | <0.001 | ||||
| Smoker | 1315 (88) | 920 (90) | 1164 (92) | 605 (87) | |
| Non-smoker | 176 (12) | 106 (10) | 95 (8) | 92 (13) | |
| COPD status | 0.001 | ||||
| New-onset COPD (during months –24 to 0) | 238 (16) | 148 (14) | 150 (12) | 84 (12) | |
| Pre-existing COPD (at –24 months or earlier) | 356 (24) | 222 (22) | 252 (20) | 152 (22) | |
| No COPD | 897 (60) | 656 (64) | 857 (68) | 461 (66) | |
| Elixhauser comorbidity score | <0.001 | ||||
| 0 | 144 (10) | 259 (25) | 439 (35) | 174 (25) | |
| 1 | 161 (11) | 184 (18) | 301 (24) | 121 (17) | |
| 2 | 257 (17) | 195 (19) | 202 (16) | 116 (17) | |
| ≥3 | 929 (62) | 388 (38) | 317 (25) | 286 (41) | |
| Survival at 1 year from diagnosis | <0.001 | ||||
| Died | 1176 (79) | 528 (51) | 634 (50) | 351 (50) | |
| Alive | 315 (21) | 498 (49) | 625 (50) | 346 (50) | |
| Morphology | <0.001 | ||||
| SCLC | 133 (9) | 85 (8) | 170 (14) | 59 (8) | |
| NSCLC | 684 (46) | 718 (70) | 915 (73) | 462 (66) | |
| Unspecified | 674 (45) | 223 (22) | 174 (14) | 176 (25) |
aUnless otherwise stated. bOther routes include: inpatient elective, other managed pathway, DCO (diagnosis by death certificate only), and unknown. 17 cOne-way ANOVA for continuous age at diagnosis; otherwise Pearson’s χ2 test. COPD = chronic obstructive pulmonary disease. IQR = interquartile range. NSCLC = non-small-cell lung cancer. SCLC = small-cell lung cancer. SD = standard deviation.
Table 2. Summary of the proportion of patients and mean events per patient per month of the included healthcare use measures 12 to 1 month prediagnosis by diagnostic route.
| Healthcare event type | Emergency diagnosis ( n = 1491, 33%) | Routine referral ( n = 1026, 23%) | Urgent referral ( n = 1259, 28%) | Othera ( n = 697, 16%) | P-valueb |
| Proportion of patients with ≥1 event, n (%) | |||||
| Consultations | |||||
| Any | 1473 (98.8) | 1023 (99.7) | 1252 (99.4) | 684 (98.1) | <0.001 |
| With any selected six relevant symptomsc | 859 (57.6) | 669 (65.2) | 891 (70.8) | 400 (57.4) | <0.001 |
| With any selected three respiratory symptomsd | 791 (53.1) | 610 (59.5) | 833 (66.2) | 355 (50.9) | <0.001 |
| With cough | 472 (31.7) | 433 (42.2) | 606 (48.1) | 223 (32.0) | <0.001 |
| With dyspnoea | 533 (35.7) | 327 (31.9) | 367 (29.2) | 210 (30.1) | <0.001 |
| With haemoptysis | 35 (2.3) | 71 (6.9) | 104 (8.3) | 25 (3.6) | <0.001 |
| Imaging eventse | |||||
| Any chest imaging | 604 (88.8) | 449 (91.1) | 552 (92.8) | 244 (83.8) | <0.001 |
| Chest X-ray | 577 (84.9) | 399 (80.9) | 513 (86.2) | 224 (77.0) | <0.001 |
| Chest CT | 341 (50.1) | 353 (71.6) | 465 (78.2) | 179 (61.5) | <0.001 |
| Events per patient per month, mean (95% CI)f | |||||
| Consultations | |||||
| Any | 1.10 (1.02 to 1.17) | 1.21 (1.11 to 1.30) | 0.93 (0.86 to 1.00) | 0.99 (0.89 to 1.09) | — |
| With any selected six relevant symptomsc | 0.06 (0.05 to 0.07) | 0.09 (0.07 to 0.11) | 0.08 (0.06 to 0.10) | 0.07 (0.05 to 0.09) | — |
| With any selected three respiratory symptomsd | 0.05 (0.05 to 0.07) | 0.05 (0.04 to 0.07) | 0.05 (0.04 to 0.07) | 0.06 (0.04 to 0.08) | — |
| With cough | 0.03 (0.02 to 0.03) | 0.05 (0.03 to 0.06) | 0.05 (0.04 to 0.06) | 0.04 (0.02 to 0.05) | — |
| With dyspnoea | 0.03 (0.02 to 0.04) | 0.03 (0.02 to 0.04) | 0.02 (0.01 to 0.03) | 0.03 (0.01 to 0.04) | — |
| With haemoptysis | 0.00 (0.00 to 0.00) | 0.01 (0.00 to 0.01) | 0.00 (0.00 to 0.01) | 0.00 (0.00 to 0.00) | — |
| Imaging eventse | |||||
| Any chest imaging | 0.04 (0.02 to 0.05) | 0.06 (0.04 to 0.07) | 0.04 (0.03 to 0.05) | 0.05 (0.03 to 0.06) | — |
| Chest X-ray | 0.03 (0.02 to 0.04) | 0.04 (0.03 to 0.05) | 0.03 (0.02 to 0.04) | 0.03 (0.02 to 0.05) | — |
| Chest CT | 0.01 (0.00 to 0.01) | 0.02 (0.01 to 0.03) | 0.01 (0.01 to 0.01) | 0.01 (0.01 to 0.02) | — |
aOther routes include: inpatient elective, other managed pathway, DCO (diagnosis by death certificate only), and unknown. 17 bOne-way ANOVA. cSymptoms include appetite loss, weight loss, fatigue, cough, dyspnoea, and haemoptysis. dSymptoms include cough, dyspnoea, and haemoptysis. eProportions calculated out of all patients diagnosed after 1 April 2012 (emergency presentation, n = 680; routine referral, n = 493; urgent referral, n = 595; and other, n = 291). fExcluding the month immediately before diagnosis. CT = computed tomography.
Diagnostic window length
Diagnostic windows were present for all patient groups defined by diagnostic route and were consistently shorter or of equal length for emergency-diagnosed compared with referred route patients (Figure 2 and Table 3). For example, consultations for any reason began to increase from 5 months prediagnosis for both emergency-diagnosed and urgently referred patients, and from 7 months for routinely referred patients. Similarly, corresponding inflection points for consultations with any of three respiratory symptoms (cough, dyspnoea, or haemoptysis) were 5, 10, and 6 months, for emergency-diagnosed, routinely referred, and urgently referred patients, respectively.
Figure 2. Rates of healthcare use per patient per month 12 months to 1 day before diagnosis by three main diagnostic routes, age-standardised in reference to the emergency-diagnosed group. Y-axis scale differs substantially between the panels/type of healthcare event visualised. The rates were modelled using Poisson regression with data from 12 to 1 month before diagnosis adjusted for patient factors (age at diagnosis, gender, ethnicity, Index of Multiple Deprivation, smoking status, chronic obstructive pulmonary disease status, and Elixhauser comorbidity score at time of diagnosis). Any selected six symptoms include: appetite loss, weight loss, fatigue, cough, dyspnoea, and haemoptysis. Any selected three respiratory symptoms include: cough, dyspnoea, and haemoptysis. CT = computed tomography.
Table 3. Summary of the measures of healthcare use by diagnostic route including the prediagnostic rates and inflection points within 12 to 1 month before lung cancer diagnosis by diagnostic route compared with emergency-diagnosed patients.
| Healthcare event type | Adjusted IRR (95% CI) a | Month of inflection-point estimate (95% CI) | ||||
| Emergency diagnosed | Routinely referred | Urgently referred | Emergency diagnosed | Routinely referred | Urgently referred | |
| Primary care consultations | 1 (Ref) | 1.34 (1.26 to 1.43)b | 1.11 (1.04 to 1.18)b | 5 (4.99 to 5.01) | 7 (6.96 to 7.04)c | 5 (4.99 to 5.01)d |
| Consultations with any selected six relevant symptomse | 1 (Ref) | 1.71 (1.51 to 1.94)b | 1.59 (1.40 to 1.79)b | 5 (4.95 to 5.05) | 9 (8.93 to 9.07)c | 6 (5.97 to 6.03)d |
| Consultations with any selected three respiratory symptomsf | 1 (Ref) | 1.72 (1.51 to 1.96)b | 1.62 (1.42 to 1.84)b | 5 (4.95 to 5.05) | 10 (9.93 to 10.07)c | 6 (5.97 to 6.03)c |
| Consultations with cough | 1 (Ref) | 1.90 (1.58 to 2.30)b | 1.94 (1.61 to 2.33)b | 7 (6.94 to 7.06) | 10 (9.93 to 10.07)c | 7 (6.96 to 7.04)d |
| Consultations with dyspnoea | 1 (Ref) | 1.33 (1.11 to 1.61)g | 1.07 (0.88 to 1.30)g | 3 (2.96 to 3.04) | 9 (8.88 to 9.12)c | 5 (4.96 to 5.04)c |
| Consultations with haemoptysis | 1 (Ref) | 3.25 (1.33 to 7.96)g | 2.92 (1.19 to 7.18)g | 5 (4.88 to 5.12) | 10 (9.93 to 10.07)c | 7 (6.94 to 7.06)c |
| Any chest imaging events | 1 (Ref) | 2.45 (1.97 to 3.05)b | 1.95 (1.56 to 2.25)b | 4 (3.98 to 4.02) | 7 (6.97 to 7.03)c | 6 (5.98 to 6.02)d |
| Chest X-rays | 1 (Ref) | 1.91 (1.53 to 2.38)b | 1.77 (1.42 to 2.21)b | 4 (3.97 to 4.03) | 7 (6.96 to 7.04)c | 6 (5.96 to 6.04)d |
| Chest CT | 1 (Ref) | 3.39 (2.47 to 4.66)b | 2.07 (1.48 to 2.90)b | 4 (3.99 to 4.01) | 6 (5.96 to 6.04)d | 6 (5.97 to 6.03)d |
aThe adjusted IRR values show the Poisson modelling results using data from 12 to 1 month before diagnosis adjusted for patient factors (age at diagnosis, gender, ethnicity, Index of Multiple Deprivation, smoking status, chronic obstructive pulmonary disease status, and Elixhauser comorbidity score at time of diagnosis). bSignificant positive association (P<0.05). cInflection points occurring ≥2 months before those for emergency presentations. dA difference in the inflection points <2 months, considered equal compared with emergency presentations. eSymptoms include appetite loss, weight loss, fatigue, cough, dyspnoea, and haemoptysis. fSymptoms include cough, dyspnoea, and haemoptysis. gResults with weak evidence (P>0.05). IRR = incidence rate ratio. Ref = reference.
Overall trends in the prediagnostic rates of healthcare use by route
In the 12 months leading up to (excluding the month immediately before) diagnosis, emergency-diagnosed patients had consultation or imaging rates not discernible from those of patients diagnosed through referred routes (Table 2). Emergency-diagnosed patients had somewhat lower age-standardised rates of prediagnostic consultations or imaging in the year prediagnosis compared with urgently or routinely referred patients, for whom adjusted incidence rate ratios (aIRRs) for both consultations and imaging were consistently higher, often significantly, than those of emergency-diagnosed patients (Supplementary Tables S2–S10, Table 3, and Figure 3). Additional analyses for segmented prediagnostic periods (months 24–1, 24–12, 12–6, and 6–1) across all models and adjustment types are presented in Supplementary Tables S11–S16, showing broadly consistent healthcare use patterns by diagnostic route as reported in the main analysis.
Figure 3. Unadjusted and adjusted Poisson model results showing the incidence rate ratios (IRRs, log-scale) for healthcare use by type 12 months to 1 month before diagnosis (with emergency diagnosed as the reference category). Log transformed data are shown for the economy of visualisation; exact values are included in Table 3. Any selected six symptoms include: appetite loss, weight loss, fatigue, cough, dyspnoea, and haemoptysis. Any selected three respiratory symptoms include: cough, dyspnoea, and haemoptysis. CT = computed tomography.
Rates of consultations for any reason and rates of chest imaging
After adjustment, prediagnostic consultation rates for any clinical reason were lower in emergency-diagnosed patients (aIRR for routinely referred versus emergency diagnosed 1.34, 95% CI = 1.26 to 1.43; aIRR for urgently referred versus emergency diagnosed 1.11, 95% CI = 1.04 to 1.18). Similar but more pronounced differences by diagnostic route were observed for chest imaging activity (aIRR for routinely referred versus emergency diagnosed 2.45, 95% CI = 1.97 to 3.05; aIRR for urgently referred versus emergency diagnosed 1.95, 95% CI = 1.56 to 2.25) (Table 3 and Figure 3).
Rates of consultations for selected respiratory symptoms
Emergency-diagnosed patients had lower prediagnostic consultation rates for any of three respiratory symptoms (cough, dyspnoea, and haemoptysis) compared with routinely or urgently referred patients. This association remained significant after adjustment, for both routinely (aIRR versus emergency diagnosed 1.72, 95% CI = 1.51 to 1.96) and urgently referred patients (aIRR versus emergency diagnosed 1.62, 95% CI = 1.42 to 1.84). The results were similar for prediagnostic consultation rates with any of the selected six relevant symptoms. Considering consultations for each symptom individually, adjusted prediagnostic consultation rates for cough were lower among emergency-diagnosed patients (aIRR for routinely referred versus emergency diagnosed 1.90, 95% CI = 1.58 to 2.30; aIRR for urgently referred versus emergency-diagnosed 1.94, 95% CI = 1.61 to 2.33). There was inconsistent or weak statistical evidence for variation by diagnostic route in adjusted prediagnostic consultation rates for haemoptysis or dyspnoea (Table 3 and Figure 3).
Discussion
Summary
Diagnostic windows of appreciable length were observed in patients with lung cancer, defined by both consultations and imaging activity. In the year prediagnosis, emergency-diagnosed patients had lower adjusted rates of consultations and chest imaging than those diagnosed via urgent or routine referrals. Improving referral pathways and post-imaging follow-up may support earlier diagnosis across routes, although the overall similar rates of prediagnostic primary care contacts across routes limit opportunities for targetted interventions. Surprisingly, as denoted by diagnostic windows of similar length, this potential did not vary substantially between diagnostic routes.
Strengths and limitations
A representative primary care cohort with linked healthcare data was analysed but the analysis was limited to variables recorded in electronic patient records. This leaves scope for residual confounding from unmeasured variables. Specifically, the findings are susceptible to ‘confounding by indication’, whereby unaccounted-for case-mix differences may influence both diagnostic route and diagnostic windows. The analysis relied on coded rather than free-text data, 37 and imaging results were unavailable. The findings are limited to those diagnosed with lung cancer, as the study aimed to compare diagnostic windows by route within this group, it did not examine healthcare use patterns in a reference population. Although more recent datasets, such as CPRD Aurum, could be used in future research, cancer registry linkage is currently available only up to 2021. Therefore, major changes in the findings are unlikely, apart from potential effects related to the COVID-19 pandemic. Incorporating additional data such as referral source, prescriptions, and test results could improve understanding of prediagnostic healthcare use and missed diagnostic opportunities.
The maximum likelihood estimation method used to estimate the inflection points assumes a linear trend in healthcare use post-inflection point. The model assumed linear relationships between the covariates and healthcare use, and did not include possible interaction terms between variables. 38 A case-only design without population controls was used, as the comparisons related to patients with lung cancer diagnosed through different routes. 38
Emergency-diagnosed patients have higher 1-year mortality even after adjusting for stage and morphology. 39,40 Although comparing treatment by diagnostic route may clarify factors leading to poorer outcomes, such analyses were beyond the scope of this study.
Comparisons with existing literature
Consistent with prior studies (measuring primary care consultations, prescriptions, blood tests, and chest X-rays), increased healthcare use was observed in the current study beginning 4–6 months prediagnosis. 11,13,40 However, the current study additionally profiled diagnostic windows and related healthcare activity by diagnostic route for the first time, to the best of the authors’ knowledge, in patients with lung cancer. A similar approach has been recently reported for diagnostic windows for patients with haematological cancers diagnosed through different routes. 15
The notion that emergency-diagnosed patients consult infrequently in primary care is not supported by the current study’s findings, showing similar background consultation patterns across routes. 41 Furthermore, the finding of lower healthcare use (any consultations, consultations with relevant symptoms, and chest imaging activity) during the diagnostic window among emergency-diagnosed patients aligns with studies on other cancer sites. 8,9,14,15,31
Diagnostic window lengths for emergency-diagnosed patients were similar or shorter than for referred route patients. Patients with lung cancer often present with non-specific symptoms such as cough, dyspnoea, respiratory infections, and chest pain. 42–44 Consistent with previous studies, the current study found that haemoptysis, the respiratory symptom with the highest positive predictive value for lung cancer, was relatively rare. 45,46 These findings concord with those observed for colorectal cancer, where similar diagnostic window length and fewer alarm symptoms (such as rectal bleeding) were reported in emergency-diagnosed patients. 8,9
Fewer chest imaging events were observed among emergency-diagnosed patients. Patients with higher primary care engagement and guideline-concordant care are more likely to receive timely imaging. 11,24,47
Tumour stage and morphology are associated with emergency diagnosis, 14,39 but adjusting for them did not materially alter the findings regarding event rates.
Implications for research and practice
Emergency-diagnosed patients consulted less with the selected symptoms and underwent fewer prediagnostic imaging investigations than those diagnosed through urgent or routine referrals. This may reflect genuine differences in tumour presentation across diagnostic routes, or patient-related factors such as health literacy (for example, poorer recognition of significant symptoms) and delayed help-seeking despite symptom experience. 48–51 Health literacy and intervals from symptom onset to help-seeking are not captured in routine health record data, yet are likely to contribute to the observed differences and should be explored in future research.
This study provides population-level evidence on diagnostic routes for lung cancer and highlights potential for earlier diagnosis. Substantial diagnostic windows across all routes suggest the need for generic improvements, such as better risk assessment of non-specific symptoms and greater imaging use. For emergency-diagnosed patients, fewer symptomatic consultations mean fewer primary care opportunities to act on signs that might have prompted non-emergency referral. Further work should examine where specific actions, particularly for chest imaging results, could have expedited diagnosis.
As around a fifth of lung cancers are not detectable via chest X-ray at the time of symptomatic presentation, 47 the observed diagnostic windows defined by chest imaging are likely to reflect false-negative investigations, underscoring the need for proactive follow-up (‘safety netting’) and wider chest CT access (for example, as a follow-up chest imaging test, after initial negative chest X-ray when symptoms and diagnostic suspicion prevail). 50 Expanding GP access to chest CTs, which have greater sensitivity than chest X-rays, could improve early lung cancer detection and reduce emergency diagnoses. 51 Similar findings in urological cancers show that use of less-sensitive imaging modalities can hinder timely diagnosis. 52
Although rare, delays in acting on abnormal chest X-ray findings may represent missed diagnostic opportunities; future research should aim to quantify whether and how often such safety incidents might be occurring. 40 Evaluating adherence to referral/imaging guidelines and detailed event analysis may identify missed diagnostic opportunities and improve care across diagnostic routes. 24,53 Such studies can inform interventions targeting potentially avoidable emergency diagnoses.
Lung cancer screening could reduce emergency diagnoses, as shown in colorectal cancer. 53–56 In the current cohort, 2171/4473, 49% of patients (including n = 573/1491, 38% of emergency-diagnosed patients), were aged 55–74 years, within the eligible screening range depending on smoking history. However, participation in screening is lower among socioeconomically deprived groups, who are more likely to be diagnosed via an emergency route. 57 Thus, although screening may reduce emergency presentations, equitable uptake is essential to avoid widening inequalities in lung cancer diagnosis.
In conclusion, comparing prediagnostic healthcare use between emergency-diagnosed and referred patients with lung cancer revealed no clear aetiological targets to reduce emergency diagnosis. Instead, the study’s findings suggest potential to expedite diagnosis across all groups. In particular, through improved management of non-specific symptoms, timely follow-up of imaging, and greater access to chest CTs for suspected cancer.
Funding
This study was funded by the National Institute for Health and Care Research (NIHR) Policy Research Programme (reference: PR-PRU-NIHR206132); the Cancer Research UK (CRUK; project award: EDDPJT-Nov24/100036); and the International Alliance for Cancer Early Detection (ACED), a partnership between CRUK (reference: C18081/A31373, RREDD-EHR project), Canary Center at Stanford University, the University of Cambridge, Oregon Health and Science University Knight Cancer Institute, University College London, and the University of Manchester. Matthew E Barclay is supported by an ACED Pathway Award Fellowship (reference: EDDAPA-2022/100002). The views expressed are those of the author(s) and not necessarily those of the NIHR or the Department of Health and Social Care.
Ethical approval
The study was approved by the Medicines and Healthcare products Regulatory Agency (UK) Independent Scientific Advisory Committee (reference: 18_299R), under Section 251 of the NHS Social Care Act 2006.
Provenance
Freely submitted; externally peer reviewed.
Data
Codelists used in this study are available online at https://github.com/martaberglund/EPlungcancer-phenotypes.git. Data cannot be shared by the researchers, but access to Clinical Practice Research Datalink (CPRD) data is available subject to protocol approval via CPRD’s Research Data Governance process; see https://cprd.com/data-access for further details.
Acknowledgements
This study is based on data from the Clinical Practice Research Datalink (CPRD) obtained under licence from the UK Medicines and Healthcare products Regulatory Agency. The data is provided by patients and collected by the NHS as part of their care and support. The interpretation and conclusions contained in this study are those of the author/s alone. Symptoms were defined using libraries of Read codes originally developed by Professor Willie Hamilton and Dr Sarah Price at Exeter University, with additional codes added by colleagues Georgios Lyratzopoulos, Cristina Renzi, Becky White, Matthew Barclay, and Dr Meena Rafiq at UCL.
Competing interests
All authors declare no support from any organisation for the submitted work; Matthew E Barclay receives personal fees from GRAIL Inc for Independent Data Monitoring Committee (IDMC) membership unrelated to this study and Becky White received personal funding from the British Liver Trust in 2023 for an unrelated data consultancy project, otherwise there are no financial relationships with any organisations that might have an interest in the submitted work in the previous 3 years and no other relationships or activities that could appear to have influenced the submitted work.
References
- 1.McPhail S, Swann R, Johnson SA, et al. Risk factors and prognostic implications of diagnosis of cancer within 30 days after an emergency hospital admission (emergency presentation): an international cancer benchmarking partnership (ICBP) population-based study. Lancet Oncol. 2022;23(5):587–600. doi: 10.1016/S1470-2045(22)00127-9. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.Lyratzopoulos G, Saunders CL, Abel GA. Are emergency diagnoses of cancer avoidable? A proposed taxonomy to motivate study design and support service improvement. Future Oncol. 2014;10(8):1329–1333. doi: 10.2217/fon.14.80. [DOI] [PubMed] [Google Scholar]
- 3.Zhou Y, Abel GA, Hamilton W, et al. Diagnosis of cancer as an emergency: a critical review of current evidence. Nat Rev Clin Oncol. 2017;14(1):45–56. doi: 10.1038/nrclinonc.2016.155. [DOI] [PubMed] [Google Scholar]
- 4.National Institute for Health and Care Excellence (NICE) London: NICE; 2025. [5 Mar 2026]. Suspected cancer: recognition and referral. NG12. accessed. [PubMed] [Google Scholar]
- 5.Rafiq M, White B, Barclay M, et al. A UK population-based case-control study of blood tests before cancer diagnosis in patients with non-specific abdominal symptoms. Br J Cancer. 2025;132(5):450–461. doi: 10.1038/s41416-024-02936-9. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.White B, Renzi C, Rafiq M, et al. Does changing healthcare use signal opportunities for earlier detection of cancer? A review of studies using information from electronic patient records. Cancer Epidemiol. 2022;76:102072. doi: 10.1016/j.canep.2021.102072. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Whitfield E, White B, Denaxas S, Lyratzopoulos G. Diagnostic windows in non-neoplastic diseases: a systematic review. Br J Gen Pract. 2023 doi: 10.3399/BJGP.2023.0044. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Sheringham JR, Georghiou T, Chitnis XA, Bardsley M. Comparing primary and secondary health-care use between diagnostic routes before a colorectal cancer diagnosis: cohort study using linked data. Br J Cancer. 2014;111(8):1490–1499. doi: 10.1038/bjc.2014.424. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Renzi C, Lyratzopoulos G, Card T, et al. Do colorectal cancer patients diagnosed as an emergency differ from non-emergency patients in their consultation patterns and symptoms? A longitudinal data-linkage study in England. Br J Cancer. 2016;115(7):866–875. doi: 10.1038/bjc.2016.250. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Lyhne Christensen N, Gouliaev A, McPhail S, et al. Lung cancer among the elderly in Denmark – a comprehensive population-based cohort study. Lung Cancer (Auckl) 2024;191:107555. doi: 10.1016/j.lungcan.2024.107555. [DOI] [PubMed] [Google Scholar]
- 11.Guldbrandt LM, Møller H, Jakobsen E, Vedsted P. General practice consultations, diagnostic investigations, and prescriptions in the year preceding a lung cancer diagnosis. Cancer Med. 2017;6(1):79–88. doi: 10.1002/cam4.965. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Wickramasinghe B, Renzi C, Barclay M, et al. Pre-diagnostic prescribing patterns in dyspnoea patients with as-yet-undiagnosed lung cancer: a longitudinal study of linked primary care and cancer registry data. Cancer Epidemiol. 2023;86:102429. doi: 10.1016/j.canep.2023.102429. [DOI] [PubMed] [Google Scholar]
- 13.Rafiq M, Drosdowsky A, Solomon B, et al. Trends in primary care blood tests prior to lung and colorectal cancer diagnosis - a retrospective cohort study using linked Australian data. Cancer Med. 2024;13(14):e70006. doi: 10.1002/cam4.70006. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Gurney J, Davies A, Stanley J, et al. Emergency presentation prior to lung cancer diagnosis: a national-level examination of disparities and survival outcomes. Lung Cancer (Auckl) 2023;179:107174. doi: 10.1016/j.lungcan.2023.03.010. [DOI] [PubMed] [Google Scholar]
- 15.Virgilsen LF, Vedsted P, Jensen H, et al. Diagnostic window prior to a haematological cancer diagnosis and the association with patient pathways: a nationwide register-based cohort study on healthcare utilization in Denmark. Eur J Haematol. 2025;114(2):353–364. doi: 10.1111/ejh.14315. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Tørring ML, Murchie P, Hamilton W, et al. Evidence of advanced stage colorectal cancer with longer diagnostic intervals: a pooled analysis of seven primary care cohorts comprising 11 720 patients in five countries. Br J Cancer. 2017;117(6):888–897. doi: 10.1038/bjc.2017.236. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Elliss-Brookes L, McPhail S, Ives A, et al. Routes to diagnosis for cancer - determining the patient journey using multiple routine data sets. Br J Cancer. 2012;107(8):1220–1226. doi: 10.1038/bjc.2012.408. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Neal RD, Smith L. Urgent cancer referrals: how well are they working and can they be improved? Br J Gen Pract. 2021 doi: 10.3399/bjgp21X716801. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.Zhang CX, Boukari Y, Pathak N, et al. Migrants’ primary care utilisation before and during the COVID-19 pandemic in England: an interrupted time series analysis. Lancet Reg Health Eur. 2022;20:100455. doi: 10.1016/j.lanepe.2022.100455. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Price S, Wiering B, Mounce LTA, et al. Examining methodology to identify patterns of consulting in primary care for different groups of patients before a diagnosis of cancer: an exemplar applied to oesophagogastric cancer. Cancer Epidemiol. 2023;82:102310. doi: 10.1016/j.canep.2022.102310. [DOI] [PubMed] [Google Scholar]
- 21.NICE London: NICE; 2024. [5 Mar 2026]. Lung cancer: diagnosis and management. NG122. accessed. [Google Scholar]
- 22.Hamilton W, Hajioff S, Graham J, Schmidt-Hansen M. Suspected cancer (part 2--adults): reference tables from updated NICE guidance. BMJ. 2015;350:h3044. doi: 10.1136/bmj.h3044. [DOI] [PubMed] [Google Scholar]
- 23.Moore SF, Price SJ, Chowienczyk S, et al. The impact of changing risk thresholds on the number of people in England eligible for urgent investigation for possible cancer: an observational cross-sectional study. Br J Cancer. 2021;125(11):1593–1597. doi: 10.1038/s41416-021-01541-4. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24.Koo MM, Mounce LTA, Rafiq M, et al. Guideline concordance for timely chest imaging after new presentations of dyspnoea or haemoptysis in primary care: a retrospective cohort study. Thorax. 2024;79(3):236–244. doi: 10.1136/thorax-2022-219509. [DOI] [PubMed] [Google Scholar]
- 25.NHS Data Model and Dictionary Service Class: ETHNIC CATEGORY. https://v2.datadictionary.nhs.uk/data_dictionary/classes/e/ethnic_category_de.asp@shownav=0.html. [5 Mar 2026]. https://v2.datadictionary.nhs.uk/data_dictionary/classes/e/ethnic_category_de.asp@shownav=0.html accessed.
- 26.Marston L, Carpenter JR, Walters KR, et al. Smoker, ex-smoker or non-smoker? The validity of routinely recorded smoking status in UK primary care: a cross-sectional study. BMJ Open. 2014;4(4):e004958. doi: 10.1136/bmjopen-2014-004958. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27.Booth HP, Prevost AT, Gulliford MC. Validity of smoking prevalence estimates from primary care electronic health records compared with national population survey data for England, 2007 to 2011. Pharmacoepidemiol Drug Saf. 2013;22(12):1357–1361. doi: 10.1002/pds.3537. [DOI] [PubMed] [Google Scholar]
- 28.Whitfield E. Atlas-phenotypes: published version of codelists. 2024. https://zenodo.org/records/13710755. [5 Mar 2026]. https://zenodo.org/records/13710755 accessed.
- 29.Quan H, Sundararajan V, Halfon P, et al. Coding algorithms for defining comorbidities in ICD-9-CM and ICD-10 administrative data. Med Care. 2005;43(11):1130–1139. doi: 10.1097/01.mlr.0000182534.19832.83. [DOI] [PubMed] [Google Scholar]
- 30.Elixhauser A, Steiner C, Harris DR, Coffey RM. Comorbidity measures for use with administrative data. Med Care. 1998;36(1):8–27. doi: 10.1097/00005650-199801000-00004. [DOI] [PubMed] [Google Scholar]
- 31.Virgilsen LF, Falborg AZ, Vedsted P, et al. Unplanned cancer presentation in patients with psychiatric disorders: a nationwide register-based cohort study in Denmark. Cancer Epidemiol. 2022;81:102293. doi: 10.1016/j.canep.2022.102293. [DOI] [PubMed] [Google Scholar]
- 32.NHS England Cancer registration statistics, England 2020. 2022. https://digital.nhs.uk/data-and-information/publications/statistical/cancer-registration-statistics/england-2020/cancer-incidence-by-stage. [5 Mar 2026]. https://digital.nhs.uk/data-and-information/publications/statistical/cancer-registration-statistics/england-2020/cancer-incidence-by-stage accessed.
- 33.Barclay ME, Lyratzopoulos G, Walter FM, et al. Incidence of second and higher order smoking-related primary cancers following lung cancer: a population-based cohort study. Thorax. 2019;74(5):466–472. doi: 10.1136/thoraxjnl-2018-212456. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 34.Maringe C, Fowler H, Rachet B, Luque-Fernandez MA. Reproducibility, reliability and validity of population-based administrative health data for the assessment of cancer non-related comorbidities. PLoS One. 2017;12(3):e0172814. doi: 10.1371/journal.pone.0172814. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 35.Harrison XA, Donaldson L, Correa-Cano ME, et al. A brief introduction to mixed effects modelling and multi-model inference in ecology. PeerJ. 2018;6:e4794. doi: 10.7717/peerj.4794. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 36.Rafiq M, Abel G, Renzi C, Lyratzopoulos G. Steroid prescribing in primary care increases prior to hodgkin lymphoma diagnosis: a UK nationwide case-control study. Cancer Epidemiol. 2022;81:102284. doi: 10.1016/j.canep.2022.102284. [DOI] [PubMed] [Google Scholar]
- 37.Price SJ, Stapley SA, Shephard E, et al. Is omission of free text records a possible source of data loss and bias in clinical practice research datalink studies? A case-control study. BMJ Open. 2016;6(5):e011664. doi: 10.1136/bmjopen-2016-011664. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 38.Whitfield E. Leveraging electronic health records to evaluate the potential for earlier diagnosis. 2024. https://discovery.ucl.ac.uk/id/eprint/10202584. [5 Mar 2026]. https://discovery.ucl.ac.uk/id/eprint/10202584 accessed.
- 39.Christensen NL, Gouliaev A, McPhail S, et al. Routes to diagnosis in Danish lung cancer patients: emergency presentation, age and smoking history-a population-based cohort study. Clin Lung Cancer. 2024;25(7):e348–e356. doi: 10.1016/j.cllc.2024.05.009. [DOI] [PubMed] [Google Scholar]
- 40.Singh H, Hirani K, Kadiyala H, et al. Characteristics and predictors of missed opportunities in lung cancer diagnosis: an electronic health record-based study. J Clin Oncol. 2010;28(20):3307–3315. doi: 10.1200/JCO.2009.25.6636. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 41.Laake J-P, Vulkan D, Quaife SL, et al. Targeted encouragement of GP consultations for possible cancer symptoms: a randomised controlled trial. Br J Gen Pract. 2021 doi: 10.3399/bjgp20X713489. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 42.Zakkak N, Barclay ME, Swann R, et al. The presenting symptom signatures of incident cancer: evidence from the English 2018 national cancer diagnosis audit. Br J Cancer. 2024;130(2):297–307. doi: 10.1038/s41416-023-02507-4. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 43.Iyen-Omofoman B, Tata LJ, Baldwin DR, et al. Using socio-demographic and early clinical features in general practice to identify people with lung cancer earlier. Thorax. 2013;68(5):451–459. doi: 10.1136/thoraxjnl-2012-202348. [DOI] [PubMed] [Google Scholar]
- 44.Price S, Spencer A, Zhang X, et al. Trends in time to cancer diagnosis around the period of changing national guidance on referral of symptomatic patients: a serial cross-sectional study using UK electronic healthcare records from 2006-17. Cancer Epidemiol. 2020;69:101805. doi: 10.1016/j.canep.2020.101805. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 45.Chowienczyk S, Price S, Hamilton W. Changes in the presenting symptoms of lung cancer from 2000-2017: a serial cross-sectional study of observational records in UK primary care. Br J Gen Pract. 2020 doi: 10.3399/bjgp20X708137. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 46.Hamilton W, Sharp D. Diagnosis of lung cancer in primary care: a structured review. Fam Pract. 2004;21(6):605–611. doi: 10.1093/fampra/cmh605. [DOI] [PubMed] [Google Scholar]
- 47.Bradley SH, Neal R, Callister M, et al. General practice chest x-ray rate is associated with earlier lung cancer diagnosis and reduced all-cause mortality: a retrospective observational study. Br J Gen Pract. 2024 doi: 10.3399/BJGP.2024.0466. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 48.Sele Sætre LM, Balasubramaniam K, Aaby A, et al. Health literacy and healthcare seeking with lung cancer symptoms among individuals with different smoking statuses: a population‐based study. Eur J Cancer Care. 2024;2024(1):7919967. [Google Scholar]
- 49.Holden CE, Wheelwright S, Harle A, Wagland R. The role of health literacy in cancer care: a mixed studies systematic review. PLoS One. 2021;16(11):e0259815. doi: 10.1371/journal.pone.0259815. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 50.Friedemann Smith C, Lunn H, Wong G, Nicholson BD. Optimising GPs’ communication of advice to facilitate patients’ self-care and prompt follow-up when the diagnosis is uncertain: a realist review of “safety-netting” in primary care. BMJ Qual Saf. 2022;31(7):541–554. doi: 10.1136/bmjqs-2021-014529. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 51.Bradley SH, Baldwin D, Bhartia BSK, et al. Achieving earlier diagnosis of symptomatic lung cancer. Br J Gen Pract. 2025 doi: 10.3399/bjgp25X740493. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 52.Zhou Y, Abel GA, Hamilton W, et al. Imaging activity possibly signalling missed diagnostic opportunities in bladder and kidney cancer: a longitudinal data-linkage study using primary care electronic health records. Cancer Epidemiol. 2020;66:101703. doi: 10.1016/j.canep.2020.101703. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 53.Te Marvelde L, McNair P, Whitfield K, et al. Alignment with indices of a care pathway is associated with improved survival: an observational population-based study in colon cancer patients. EClinicalMedicine. 2019;15:42–50. doi: 10.1016/j.eclinm.2019.08.009. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 54.Askari A, Nachiappan S, Currie A, et al. Who requires emergency surgery for colorectal cancer and can national screening programmes reduce this need? Int J Surg. 2017;42:60–68. doi: 10.1016/j.ijsu.2017.04.050. [DOI] [PubMed] [Google Scholar]
- 55.Hewitt RJ, Bartlett EC, Ganatra R, et al. Lung cancer screening provides an opportunity for early diagnosis and treatment of interstitial lung disease. Thorax. 2022;77(11):1149–1151. doi: 10.1136/thorax-2022-219068. [DOI] [PubMed] [Google Scholar]
- 56.Decker KM, Lambert P, Nugent Z, et al. Time trends in the diagnosis of colorectal cancer with obstruction, perforation, and emergency admission after the introduction of population-based organized screening. JAMA Netw Open. 2020;3(5):e205741. doi: 10.1001/jamanetworkopen.2020.5741. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 57.Redondo-Sánchez D, Petrova D, Rodríguez-Barranco M, et al. Socio-economic inequalities in lung cancer outcomes: an overview of systematic reviews. Cancers (Basel) 2022;14(2):398. doi: 10.3390/cancers14020398. [DOI] [PMC free article] [PubMed] [Google Scholar]