Revisiting Cancer Diagnosis in Scotland: Further Insights from the Second Scottish National Cancer Diagnosis Audit

Objective . To characterise cancer diagnosis in Scottish primary care in 2018/19 and draw comparisons with diagnostic activity in 2014. Methods . A national audit of cancer diagnosis undertaken in Scottish general practices. Participating GPs collected diagnostic pathway data on patients diagnosed with cancer in 2018/19 from medical records. Tese data were supplemented by linkage to the Scottish Cancer Registry and previous audit data from 2014. Analyses explored and compared patient demographics, presentation, diagnostic routes, and intervals. Results . Seventy-three practices submitted data on 2,014 cases in 2014 and 90 practices submitted data on 2,318 cases in 2018/2019. Individual demographics and types of cancer were similar. Tere was a higher proportion of USC (urgent suspected cancer) referrals in 2019 than 2014 (42.9% vs 38.1%, p � 0 . 008) but a similar proportion of emergency presentations (19.2% vs 20.4%). Primary care (median 4 (IQR 0–22) vs 5 (0–23)) and diagnostic intervals (27 (10–59) vs 30 (13–68)) were similar in both periods. Signifcantly fewer (24.5% vs 28.3, p � 0 . 015) had a diagnostic interval > 60 days in 2019 than 2014. Harder to diagnose cancers were more likely to present as emergencies and be subject to prolonged delays in both cohorts. Conclusions . Te 2014 and 2018/19 cohorts were broadly similar. Tere is limited evidence that USC use had increased between 2014 and 2018/19. Harder to diagnose cancers are still most likely to present as emergencies and be subject to delays. Overall, it seems there were small improvements in cancer diagnosis prepandemic and a further audit could examine evidence for a postpandemic recovery.


Introduction
Over 34,000 people were diagnosed with cancer in Scotland in 2019 and incidence continues to increase [1].Te most common cancers in Scotland are breast, lung, prostate, and colorectal accounting for over half of all cancers [1].Scotland has typically had poorer survival rates than the European average [2][3][4][5][6][7], although there have recently been some modest survival improvements [2,3].
To better understand infuences on cancer outcomes in Scotland, there is value in analysing individual patient pathways to cancer diagnosis and the component intervals [8].Tese include primary care interval, the time between a patient frst presenting in primary care and secondary care referral, and the diagnostic interval, the time between a frst presentation in primary care until cancer diagnosis [8].Prolonged patient pathways vary by cancer type and prolonged intervals have been linked to increased anxiety for patients, later stage at diagnosis, increased costs, and mortality [9][10][11][12][13][14][15][16][17].
Expediting cancer diagnosis and treatment is a key priority for the Scottish Government [18].Te Scottish Cancer Referral Guidelines for Suspected Cancer, frst published in 2002 and in 2019, provide guidance to primary care practitioners on identifying those most likely to have cancer and in need of urgent referral [19,20].Te Scottish Government published its cancer strategy "Beating Cancer: Ambition and Action" in 2016 [18].Tis aims to support GPs to refer people with suspected cancer early including the introduction of Early Cancer Diagnostic Centres (ECDCs) [18].
Evaluating such strategies is vital to ensure their potential to improve cancer outcomes in Scotland [21].Te National Cancer Diagnosis Audit (NCDA), led by Cancer Research UK, aims to provide greater insight into factors impacting patients' cancer journeys within Scottish primary care.In 2020, Murchie et al. described Scottish primary care cancer diagnostic performance, with data on 2,014 cancer patients within Scotland in the year 2014 [22].Tey found most patients within Scotland present frst to a GP (71.5%) and approximately a third are referred via the "Urgent Suspicion of Cancer" pathway.Most patients are referred and diagnosed quickly, albeit with variability between cancer sites.Comparison with a similar audit conducted in England in 2014 showed little diference in diagnostic pathways despite diverging health systems [22,23].
Te current paper aims to enhance these insights by adding cases from the second NCDA in Scotland.Te current study analyses data from combined Scottish NCDA 2014 and 2018/19 audits.
Tere were three objectives for the study: (1) To describe the most recent NCDA dataset to illustrate how cancer was being diagnosed in Scotland between 2018 and 2019 (2) To determine whether important changes to cancer diagnostic pathways may have occurred in Scotland between 2014 and 2018/19 (3) To consider potential of Early Cancer Diagnostic Centres to improve cancer diagnosis in Scotland.

Methods
Te In addition, e-mail invitations were sent to practice managers at each Scottish General Practice publicising the audit.Practices responding to advert were asked to identify a lead GP with whom all further audit communication was coordinated.Diferent practices participated in the 2014 audit and 2019 audit.
Subsequently analysts from the Information and Services Division (ISD) of NHS Scotland assigned all incident cancers within participating practices (excluding non-melanoma skin cancer) that had been diagnosed in the practice population between October 2018 and September 2019 using the Scottish Cancer Registry dataset.
Participating practices completed a Caldicott data release form allowing patient information from practice-held electronic primary care records to be shared with ISD and then linked to data from the Scottish Cancer Registry using patients' CHI numbers.Te CHI number is a unique 10-character numeric identifer, allocated to each patient.A pre-prepared Excel form was securely sent to the lead GP at each participating practice who coordinated the completion of a proforma for each identifed patient.Tese were then deidentifed and returned securely to ISD.
Information collected using the forms included patient characteristics, presenting symptoms and signs, number and type of consultations prior to referral, primary care-led investigations, and the presence of documented safety-netting (either in the form of "when to represent" or further appointments organised).Information was also collected on whether the individual completing the form considered there had been an avoidable delay "GP perceived avoidable delay" (in three areas of the patient journey-the presentation, between presentation and referral or after referral), using a yes/no tick box.Referral type was also collected including "routine" (nonurgent and nonsuspected cancer referrals), "urgent" (not meeting the cancer referral guidelines but deemed to be needing to be seen urgently by referrer), "urgent suspected cancer" (suspected cancer referral and should receive treatment within 62 days of receipt of referral), emergency (same day admission for investigation), private referral, screening detected, or "others" which included multidisciplinary centre referral.Key dates, including when the patient frst presented, when referred from primary care, and when frst seen by a specialist in secondary care, were also collected.Te form included either tick boxes or drop-down menus with predefned answers, except for dates which had to be entered manually.Cancer type, date of diagnosis, stage, and grade at diagnosis were added by ISD upon receipt using linkage to the Scottish Cancer Registry dataset.Scottish Index of Multiple Deprivation (SIMD) quintile and urban-rural classifcation were also assigned based on patient residential postcodes at time of diagnosis.

Data and Statistical Analysis.
All analyses were conducted using SPSS version 27.
Te distribution of patient characteristics including sex, age group, living arrangements including housebound status, language and communication difculties, ethnicity, cancer site, SIMD (1 (most deprived)-5 (least deprived)), urban-rural 2-fold classifcation (a population of less than 3000 people is classifed as rural) [20], and presence and type of comorbidities of this cohort was compared with the 2014 cohort as described by Murchie et al. in 2020 [22].

2
European Journal of Cancer Care Key variables detailing referral type leading to diagnosis, pathway intervals, and presence or absence of perceived avoidable delay were grouped by sex, age group (0-24, 25-49, 50-64, 65-74, 75-84 and ≥85 years), cancer site, SIMD, urban-rural 2-fold classifcation, and comorbidity count.Te comorbidity count was categorised as 0, 1, 2, and ≥3 based on the number of separate coexisting medical conditions of individuals included from the primary care record.N < 5 was used to minimise disclosure risk when illustrating these data in the tables and represents either 0, 1, 2, 3, or 4 patients only in this category.
Two key pathway intervals were analysed in the whole cohort: primary care interval and the diagnostic interval [8].Te primary care interval was calculated from date of frst presentation to primary care with symptoms considered relevant to the ultimate cancer diagnosis as judged by the GP completing the proforma to the date of frst referral to secondary care.Te diagnostic interval was calculated from date of frst presentation as above to the date of diagnosis recorded in the Scottish Cancer Registry.Intervals of <0 and >730 days were excluded, as per analysis of previous audits [15,16].Te median and 25 th and 75 th centiles are described for the updated 2019 audit, alongside the percentage of patients with a primary care interval or diagnostic interval of >60 or >90 days.Te proportion of patients with primary care interval or diagnostic interval of >60 or >90 days were compared between the original 2014 dataset and 2019 dataset with unadjusted and adjusted logistic regression (adjusted for sex, age, cancer site, number of comorbidities, SIMD, and urban/rural classifcation).To further explore potential sex diferences in the pathway primary care and diagnostic intervals were also calculated for males and females separately for the commonest cancers, colon, lung and bronchus, and others, with intervals also calculated separately for the sex specifc cancers: female breast, ovarian and gynaecological, and prostate.
In a supplementary analysis, to establish potential beneft from Early Cancer Diagnostic Centres (ECDCs) which were piloted in three Health Boards in Scotland between June 2021 and June 2022, cancer pathway intervals were calculated for patients who, based on their symptoms and signs at diagnosis, would have been eligible for an ECDC referral according to those being applied at three current ECDC sites in Scotland during the pilot [24].Patients with one or more of the following symptoms were selected: pain, nausea/ vomiting, loss of appetite or early satiety, fatigue, jaundice, weight loss, or others as signifying "GP Gut feeling."Patients with one or more of the following clinical fndings were also selected: nonspecifc anaemia, hypercalcaemia, raised ESR, thrombocytosis, abnormal LFTs, and symptomatic FIT.From this subset, patients with an additional alarm feature or who had been referred as "Urgent-Suspected Cancer" were excluded.Te ECDC and non-ECDC samples were described and median (interquartile range) pathway delays (as reported in previous evaluations of ECDC-type initiatives conducted elsewhere) were calculated.

Patient and Practice Characteristics.
In 2014, 73 (7.7% of a total of 948) Scottish General Practices submitted data and 90 (9.9% of a total of 911) in 2018/2019 provided 2,014 and 2,318 cancer diagnoses, respectively.Nineteen practices participated in both audits.
Tere were no notable diferences likely to be clinically important between the two audit cohorts in the distribution of individual patient characteristics such as age, sex, cancer type, and comorbidities (Table 1); however, there was a signifcant diference between cohorts in area-based deprivation (SIMD) and urban-rural classifcation (SIMD and rurality) likely driven by diferences in the GP practices which participated in the 2014 and 2019 audits.Tere was also a signifcant diference between cohorts in the distribution of ethnicity and language ability between cohorts although apparently due to a small increase in non-white subjects (16 vs 31) and non-native but fuent English speakers.Patient characteristics of the 2014 audit have previously been described [22].In the 2019 audit, the median age at diagnosis of included patients was 69 years and 51.1% were women.Stage at diagnosis was available for 71.8% of individuals.Patients were predominately white and native English speakers.Most individuals were not housebound (86.2%) and just over half described cohabiting (55.5%).A quarter of individuals had no comorbidities and just over 20% had three or more comorbidities.Te most common comorbidity described was hypertension (34.7%) followed by cardiovascular disease (21%) and then arthritis/musculoskeletal disease (15.9%).With respect to the Scottish Index of Multiple Deprivation (SIMD), 19.9% of individuals were from the most deprived and 22.9% from the least deprived quintiles.Most patients, 78.9%, were urban-dwellers and 21.1% were rural dwellers.

Referral Type. Table
European Journal of Cancer Care Individuals aged 0-24 and 85 years and over had a lower percentage of USC referrals (26.9% and 29.3%) and higher percentage of emergency referrals (42.3% and 36.1%,respectively).Individuals with no comorbidities had a higher percentage of urgent suspicion of cancer referrals compared with individuals with three or more comorbidities (47.7% and 35.3%, respectively) with the inverse seen for emergency referral (15.1% and 24.9%, respectively).Table 3 [22].Te median primary care interval was 4 days (IQR 0-22 days), with 10.9% of individuals having a primary care interval >60 days and 7.5% of individuals with a primary care interval >90 days.Te primary care interval was shortest in individuals with breast cancer (0, IQR 0-1) and those diagnosed with melanoma (0, IQR 0-2) and longest in individuals with an unknown primary (55, IQR 14-76) or multiple myeloma (27,.Tere were no signifcant diferences between male and female patients for non-sex specifc cancers which are further described in supplementary Table 2.For colon cancer, the median primary care interval for men was 8 days (IQR 1-22 days) and for women, it was 6 days (IQR 1-21 days), p � 0.885.For lung and bronchus cancer, the median primary care interval for men was 7 days (IQR 0-30.5 days) and for women, it was 11 days (IQR 0-26 days), p � 0.575.For all other non-sex specifc cancers combined the median primary care interval for men was 4 days (IQR 0-27 days) and for women, it was 3 days (IQR 0-28.5 days), p � 0.753.
Tere did not appear to be any clinically important diferences in primary care interval based on SIMD status (3 days in least deprived vs 4 days in most deprived), although individuals living in an urban setting had a shorter median primary care interval of 3 days (IQR 0-20) compared with 7 days (IQR 0-29) in individuals living within a rural setting.
In 39.4% of cases, GPs perceived there had been an avoidable delay.Tis ranged from 57.4% of individuals with rectal cancer to 21.1% of individuals with leukaemia (supplementary Table 2).GP perceived avoidable delay appeared lower in the most deprived areas (31.8%) and higher in rural settings (48.3%). 1, 3, and 6 compare data between the 2014 and the 2019 audit.Table 1 shows that the individual demographics and types of cancer were similar in both audits.Tere was a change in the pattern of referrals from 2014 to 2019 (p � 0.008) with an increase in "urgent suspected cancer" and corresponding decrease in "routine," "emergency," and "others" referrals (Table 3).Te median primary care interval was 5 days (IQR 0-23 days) in 2014 and 4 days (IQR 0-22 days) in 2019.Tere was no signifcant diference in the proportion of individuals

Discussion
4.1.Summary of Key Findings.In this study, we describe the updated National Cancer Diagnosis Audit and compare the fndings with the previous audit carried out in 2014.Te proportion of patients being referred via a USC referral was signifcantly higher in 2019 audit with a similar proportion of patients being diagnosed following an emergency presentation.Tere do not appear to have been any notable and clinically signifcant changes in primary care and diagnostic intervals between the two audit periods.However, we documented a signifcant reduction in the proportion of patients with diagnostic interval >60 days in the latest audit cohort.Tere remains clear variation between cancer site in the use of "urgent suspected cancer referral" pathway, the length of the primary care interval/diagnostic interval, and the presence of avoidable delays.Tere was no diference in the primary care and diagnostic interval in nongender specifc cancers.In 2019, overall, the median primary care interval was 4 days, and the diagnostic interval was 27 days.A subset of patients who may have been eligible for ECDC referral were identifed amongst whom lung, colon, and liver and bile tract cancers were the commonest diagnoses.Overall, their median primary care interval was 7 days and diagnostic interval 24 days.

Context with Other
Literature.Almost half (42.9%) of the patients were referred through the "Urgent-Suspected Cancer" (USC) pathway, a 4.8% increase from 38.1% in 2014.
Tere is some evidence that mortality is lower amongst patients from general practices which make more USC referrals, particularly for the four main cancer (breast, lung, colorectal, and prostate) types [25][26][27][28].Tis efect is thought to result from earlier stage at diagnosis and supports increasing use of USC referrals and lowering of referral thresholds [29].In this context our results are encouraging since they suggest that GPs began to use the USC route more frequently between the two audit periods which could be contributing to a greater proportion of cancers being diagnosed at an earlier stage in Scotland.Tis point will, however, require to be confrmed using the much larger dataset provided by the Scottish Cancer Registry.Te Scottish Cancer Referral guidelines were updated in early European Journal of Cancer Care 9 2019 with some thresholds being reduced [19].Due to the overlap between the guideline update and the audit, it is difcult to be certain that the update directly impacted USC referrals, but it may represent an encouraging trend.Conversely, random case-mix of easier and harder to diagnose cancer occurring between practices and time periods could account for some of the diferences seen [30].Of note, the percentage of individuals referred via USC in Scotland continues to be below that described in the English NCDA, a point worthy of further exploration [23].
Our fndings are consistent with previous research that cancers with "nonspecifc" presenting symptoms are less likely to be referred via the USC route including leukaemia, pancreatic cancer, and multiple myeloma [31].Whereas, cancers which present with more identifable alarm symptoms, such as an abnormal mole or breast lump, are more likely to be referred urgently.
An "inverse n" pattern of USC referrals with age was observed, whereby the oldest and youngest individuals were less likely to have an USC referral which may refect variation in cancer types but might also suggest a lower index of suspicion for patients at the extremes of age [31].In addition, the Scottish Cancer Referral Guidelines have age thresholds within their referral criteria for some cancer types which may impact on the number of USC referrals in diferent age groups.Individuals with higher number of comorbidities appeared to be less likely to have an USC referral.Tis may refect symptoms being attributed to comorbid conditions, such as the overlapping nature of COPD symptoms and lung cancer.Nearly one ffth of new cancer diagnoses were via emergency presentation, consistent with earlier research, and were most common for individuals with brain cancer, pancreatic cancer, multiple myeloma, and leukaemia [23,32].Tere was no change in the percentage of emergency presentations between the time periods investigated.Emergency presentation has been linked with increased mortality and worse patient experience [33,34].However, avoiding such presentations is challenging and they may in fact refect appropriate patient care in many cases [32].
Tere was an encouraging signifcant reduction in the proportion of individuals with a diagnostic interval >60 days in the 2019 cohort and a nonsignifcant reduction in individuals with a diagnostic interval >90 days.Tis is a welcome fnding as it is recognised that prolonged diagnostic intervals can worsen cancer outcomes [17,35].Tis may refect a small increase in USC referrals, but equally could relate to enabling GPs to refer directly for imaging such as CT chest, abdomen, and pelvis.However, the relationship between diagnostic interval and cancer outcomes is complex [35].Shorter diagnostic intervals may result from expedited diagnosis caused by disease severity/emergency presentation and may have worse outcomes.As seen in this dataset, leukaemia, pancreatic, and brain cancer all had short diagnostic intervals but high rates of emergency referrals.Terefore, it is important to interpret diagnostic intervals in the context of a complex and dynamic process from patient presentation to ultimate cancer diagnosis and outcome.In addition, the importance of the diagnostic interval can vary by cancer type with a delay in a rapidly progressive cancer leading to worse outcomes than a more indolent cancer type.
Despite some encouraging trends in referral rates and diagnostic intervals demonstrated in the updated 2019 cohort in over a third of cases GPs described the patient experiencing an avoidable delay within the patient journey.
In considering the pathway intervals for those patients included in the cohort who had nonspecifc signs and symptoms at presentation and would likely qualify for ECDC referral, some limited comparisons could be made with results from similar initiatives in other countries [36][37][38][39][40]. Te most meaningful comparison perhaps is with the secondary care interval reported in the English ACE MDC evaluation which was longer than that observed in the current cohort [8,40].Tis raises the possibility that ECDC may not signifcantly expedite the diagnosis of cancers which present in Scotland with nonspecifc symptoms.It will be important, therefore, that the evaluation of the ECDC pilot in Scotland closely considers patient perspectives and the benefts of promptly ruling out cancer.

Strengths and Limitations.
A strength of this audit is it provides a unique insight into data collected from primary care itself on the diagnostic processes by GPs.Te audit links staging and cancer diagnosis data with information directly collected from primary care records by clinical experts in the form of GPs and then data were subsequently handled by highly skilled data professionals and robust cleaning undertaken.However, it must be recognised that each practice completed its own patient data collection on details such as frst presentation to primary care with symptoms suggestive of cancer and the presence of avoidable delays.Tus, there was the potential for data collection bias whereby variation exists between GPs interpretation of the questions and completion of the data collection form.Te cases were allocated to GPs centrally.Te overall return rate of completed audit records by GP practices was 77%.Tere was a notable drop of completed audit records during the COVID-19 pandemic which may be due to pressures within the practices during this time.Nevertheless, it is possible that GPs completed audits on patients which were likely to refect more favourably on the practice, thus infuencing the length of primary care and diagnostic intervals and the presence of avoidable delays.
When comparing the two audit cycles, it must also be recognised that diferent practices participated in both audits and so any comparison must be interpreted with caution.Furthermore, the data were provided by volunteer practices and GPs.Tis raises the possibility that participating practices and GPs difered from those who did not participate and may have introduced biases into the data.Te identity of participating practices and GPs was not made available to researchers meaning we could not explore these diferences in detail, but it is an important point that should be borne in mind whilst considering the results.It could be, for example, that participating practices were those who had GPs with a particular interest in cancer care, were encouraged to participate due to recent difcult cases, had 10 European Journal of Cancer Care recently undergone restructuring, or were training practices.Overall, there is the possibility that participating practices difered in important respects with those that did not participate, but the need to assure practice anonymity prevents us from exploring this point in detail.Whilst there was no diference in the age, cancer type, number of comorbidities, and stage at diagnosis between the two audit cohorts, there were signifcant diferences in the SIMD and urban/rural mix and we have published a separate paper exploring the observed infuence of socioeconomic status and place of residence on diagnostic pathways in the combined audits [41].We have thus focussed on any substantive diferences between the two cohorts in their referral type, primary care, and diagnostic interval.Tere is, therefore, the possibility that case-mix could infuence these comparisons.A further point to note is that the 2 nd NCDA was completed prior to the 2020-22 global COVID-19 pandemic which will have altered the cancer diagnostic landscape in Scotland, and probably for the worse [42].It seems sensible to suggest that methods followed in these two audit cycles, if repeated in future, could provide a useful refection on the cancer diagnostic performance within the Scottish NHS which could inform policy-makers on the extent to which the service has recovered.

Conclusions
Patients comprising the 2014 and 2019 cohorts were broadly similar with respect to type of cancer and individual demographics.Te proportion of USC referrals increased from 38.1% to 42.9% between the two time periods and slightly fewer patients experienced prolonged delays.Together, this suggests improvement in NHS Scotland's prepandemic cancer performance likely resulting from increased adherence to urgent suspected guidelines by GPs.We believe that these gains can be consolidated by disseminating the reported results to GP colleagues via training and messaging from Scottish Government and other relevant professional organizations.Our data suggest that prolonged delays and emergency presentation are still more likely for cancers such as multiple myeloma that are often considered harder to diagnose due to nonspecifc symptoms at presentation and have previously been documented to be more likely to have multiple consultations prior to referral [43].Tis reinforces the importance of ongoing and extended eforts to introduce ECDCs in other areas of Scotland.It is important too that ECDCs are widely publicized to the public, so that they can be usefully discussed with GPs during relevant consultations.We would recommend that further regular audits using similar methods should form part of NHS Scotland's post-COVID recovery plans.Repeated audit cycles will also enable a large and valuable dataset to accrue.Tis may enable further research in to whether characteristics of the primary care practice (such as list size or geographical coverage) infuences cancer diagnoses in Scotland.

Data Availability
In compliance with all regulatory and legal requirements, data were stored, accessed, and analysed with the National Safe Haven maintained by NHS National Services Scotland.Outputs were subject to disclosure checks with members of the Electronic Data Research and Innovation (eDRIS) team of the Information and Statistics Division, Scotland, prior to release to the research team for inclusion in this manuscript.Any future access to this dataset would require application to the Scottish Government.

Table 1 :
compares Patient characteristics and comparison of 2014 and 2019 cohorts.

Table 2 :
Type of referral for 2019 only.

Table 3 :
Type of referral and percentages compared between time periods.

Table 4 :
Primary care interval (PCI) descriptive statistics with percentages of patients with PCI of >60 and >90 days for 2019 (excluding screening detected patients 1 ).Te 2019 survey included an initial question: "Was the patient's cancer known to be detected by NHS cancer screening services relevant to this diagnosis?"any patient where the GP had responded yes to this question were excluded from further analysis and accounts for the diference in valid N between the analysis on referrals compared with PCI/DI.
Conversely, subsequent intervals were all signifcantly shorter in the ECDC sample compared to those who may not have been eligible for ECDC referral (supplementary Table4).

Table 5 :
Diagnostic interval descriptive statistics with percentages of patients with DI of >60 and >90 days for 2019 (excluding screen detected).

Table 6 :
Primary care intervals and diagnostic intervals where the interval was greater than 60 and 90 days, respectively, for both time periods (excluding screen detected).