WELCOME-GP: A randomised controlled trial of the effectiveness of a targeted postal cancer awareness intervention for increasing attendance at general practice

ABSTRACT Objective To assess the effectiveness of a targeted postal promotion for improving cancer symptom awareness and increasing help-seeking in general practice, on subsequent general practitioner (GP) consultation rates in a population which has made infrequent use of consultations at their local practice. Setting 23 general practices in England. Design Randomised controlled trial comparing a mailed leaflet providing information on six key cancer symptoms plus a covering letter signed by their general practitioner designed to reduce barriers to primary care help-seeking (intervention arm), with usual care (control arm). Participants 1,513 adults aged 50-84 years (783 individually randomised to the intervention arm and 730 individually randomised to the control arm) who had not consulted their GP in the last 12 months and had at least two other risk factors for late presentation with cancer, identified by practice staff between November 2016 and May 2017. 749 individuals in the intervention arm and 705 in the control arm were included in the intention to treat analyses. Outcome measure The primary outcome was number of GP consultations in the six months subsequent to mailing of the intervention. Results There was a significantly higher rate of consultation in the intervention arm: 436 consultations compared to 335 in the control arm (RR = 1.40, 95% CI 1.11-1.77, p=0.004). However, there was no difference in the numbers of persons consulting their GP, with 165 in each group. Conclusions Targeted interventions of this nature can change behaviour. This intervention stimulated a greater number of consultations but not a greater number of patients consulting. There is a need to develop interventions which can be more effective on the broader less engaged population. Trial registration The trial was registered prospectively on the International Standard Randomised Controlled Trial Number (ISRCTN) registry (ISRCTN95610478). Funding This research is funded by the National Institute for Health Research (NIHR) Policy Research Programme, conducted through the Policy Research Unit in Cancer Awareness, Screening and Early Diagnosis, 106/0001. The views expressed are those of the author(s) and not necessarily those of the NIHR or the Department of Health and Social Care. Additional support, including NHS service support costs, were provided by the National Institute for Health Research Clinical Research Network (NIHR CRN) (UKCRN ID 31163).

other risk factors for late presentation with cancer, identified by practice staff between November 2016 and May 2017. 749 individuals in the intervention arm and 705 in the control arm were included in the intention to treat analyses.

Outcome measure
The primary outcome was number of GP consultations in the six months subsequent to mailing of the intervention.

INTRODUCTION
It has long been observed that for some common cancers, survival in the UK is lower than in comparable high-income countries. [1] It is thought that this is due largely to later stage at diagnosis, which occurs for a number of reasons. Later stage at diagnosis is likely to be partly related to system delays following presentation, but may also arise from low symptom awareness, negative beliefs about cancer and reluctance to 'waste the doctor's time', [2,3] all of which could extend the patient interval. [4] Within the UK, there is an ecological association between levels of awareness of cancer, including cancer symptom awareness, and cancer survival, with higher survival being observed in areas of heightened awareness. [5] In addition, there remain geographic and socioeconomic inequalities in cancer survival in the UK, although the gap has narrowed somewhat in recent years. [6,7] There have been several campaigns, such as 'Be Clear on Cancer,' to raise the level of symptom awareness and encourage help-seeking for symptoms. [8] Some have shown encouraging results with respect to diagnosis of cancer at an earlier stage. [9] Others, however, have shown an increase in consultations and referrals, but no effect on diagnoses of cancer or stage of disease. [10] This gives rise to the concern that population-level, mass media campaigns may increase consultations among the 'worried well', rather than reaching the population most in need of earlier presentation.
The above considerations led to speculation that targeted rather than whole-population cancer symptom awareness interventions might be more effective in reducing the patient interval, the time from initial presentation of symptoms to first contact with primary care, and thereby improve stage at symptomatic presentation. Targeting in this context is not at those at high risk of specific cancers, but at those whose circumstances or lifestyle imply that, if they do develop cancer, they are at increased risk of diagnosis at a later stage. This population includes those of lower socioeconomic status, smokers, those with chronic comorbidities, certain ethnic groups, and those who tend to use emergency services rather than primary care. [11][12][13][14][15] Patients in these groups may be more likely to attend primary care in a manner which is incongruous with both their symptoms and clinical need. [12]

METHODS
Here we report on a randomised controlled trial, WELCOME-GP (Writing to Encourage Late Consultation Outpatients to Make Engagement with their GP) of a postal intervention designed to promote awareness of six cancer 'red flag' symptoms, informed by the Model of Pathways To Treatment, [16] and to allay fears of wasting the doctor's time. The intervention was targeted at a population which may be less engaged with primary care (i.e. those who have not made recent use of consultations at their general practice with factors that suggest this may be incongruous with their clinical need) and with attributes known to be risk factors for late stage presentation of cancer. [11][12][13][14][15] The main aim of the trial was to assess the extent to which the intervention promoted re-engagement with general practice.
The average rate of general practitioner (GP) consultation in the age group 50-84 is around 8 per person per year. It was assumed that, in the target population without intervention, the rate would be only 0.8 per year, or 40 consultations per 100 people over a six-month period. If the intervention were to increase the rate of consultation from 40 per 100 to 55 per 100 (i.e. 15 percentage points), and if the number of consultations were independent Poisson distributed, 547 persons per group would confer 90% power for the comparison (5% significance level, two-sided testing). Because consultations are not necessarily independent, it was proposed to adopt, as a fail-safe, 700 subjects per group, a total of 1,400, with the intention to recruit 70 subjects from each of 20 general practices, allocated 1:1 between the intervention and usual care within each practice.
. CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted June 11, 2020. . https://doi.org/10.1101/2020.06.09.20126540 doi: medRxiv preprint General practices were approached in three areas of England (the North East, Greater Manchester and London) with the assistance of four CRNs (clinical research networks) operating in those areas, commencing in May 2016. A total of 24 general practices agreed to take part, with one subsequently withdrawing.
Patients were eligible for inclusion if they were aged 50-84, registered with one of the participating practices, had not had a GP consultation in the last 12 months and satisfied at least two of the following conditions:  Low socioeconomic status (two lowest 2015 Index of Multiple Deprivation (IMD) quintiles based on residential post code [17])  Missed last breast, bowel, cervical cancer or abdominal aortic aneurism (AAA) screen  History of use of emergency or out of hours services instead of primary care  Missed last appointment for chronic disease monitoring/management  Living alone (indexed by being the only person registered with the general practice at their address) as a marker of social isolation  Smoker (ever) However, we were unable to identify with confidence patients who had used emergency or out of hours services from the practice databases. Therefore, no patients were recruited on this basis.
Persons were also ineligible if they already had a diagnosis of cancer, if the GP considered recruitment inappropriate in view of the patient's state of health or cognitive capacity, or if the GP felt that the patient would not wish to be included in the research.
Staff within each practice were asked to identify eligible patients, with the target of identifying 70 patients. Each practice was provided with a computer-generated randomisation schedule provided by the research team. Practice staff randomly allocated patients to the intervention or to usual care; there was no concealment of allocation. Some practices failed to reach the target, whilst others slightly exceeded it. Additionally, a small number of patients were found to be ineligible after allocation, and were removed from the study. A total of 1,513 patients were randomised. Simple randomisation was used, so the allocation was uneven, with 783 in the intervention group and 730 allocated to usual care.
At the end of the six month follow-up period, practice staff were asked to report the number of consultations by each patient who continued to be registered throughout the period. For each such consultation, they were also asked to report the month of the consultation, the reason for the consultation, and whether or not it resulted in any investigation or referral to secondary care. Intention to treat analyses were carried out for all eligible patients for whom data was available at follow-up.
Due to the nature of the intervention, it was necessary that patients in both groups were unaware that they were taking part in a research study. Practice staff were aware of which patients were allocated to each group and the data analysis team were also not blinded to the allocation. As patients were not aware that they were taking part in the study the findings of the study will not be disseminated to individual patients, but will be shared with the participating practices.

Intervention
The Model of Pathways to Treatment [16] provides a framework for understanding the pathways between an individual first detecting a bodily change and undergoing treatment. This begins with 'appraisal' and 'help-seeking' intervals during which an individual appraises a bodily change as abnormal and decides to consult a health care professional. Crucially, the individual must believe there to be a reason to present with their symptom, often motivated by heuristics such as concern about a worsening symptom or its interference with their daily activities. [18] The present intervention can be contextualised within these two . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted June 11, 2020. . https://doi.org/10.1101/2020.06.09.20126540 doi: medRxiv preprint intervals, in that it was intended to prompt appraisal through symptom awareness, provide individuals with a 'cue to action' for visiting their GP, and use messaging to counteract known attitudinal barriers to helpseeking. Furthermore, the personalised, GP-led approach drew on the success of primary care endorsement in the cancer screening context. [19,20] The intervention consisted of a letter, signed by one of the GPs in the patient's practice, noting that the recipient had not been seen at the practice for some time, drawing attention to an enclosed leaflet on six symptoms potentially suspicious for cancer, and reassuring the recipient that if he or she should consult the GP with any of these symptoms, or indeed for any other reason, this would not be considered wasting the doctor's time, a known reason for patients with vague symptoms delaying to present. [3] The letter and leaflet are given in Supplementary Appendix 1 and Supplementary Appendix 2 respectively. The symptoms covered in the leaflet were: blood in urine; blood in stool; persistent cough; haemoptysis; difficulty swallowing; and unexplained weight loss. These were chosen as they featured in in the National Institute for Health and Care Excellence (NICE) recommendations for referral for suspected cancer [21], and were relatively easy to recognise by the patient compared to other non-specific symptoms.
The first practice mailed the intervention materials to all patients allocated to the intervention arm on 11 November 2016, and the last on 31 May 2017.

Outcomes
Since the aim of the trial was to assess the effect on re-engagement with general practice, the primary endpoint was the rate of GP consultations in the six months following randomisation (i.e. the date that letters were mailed). Secondary endpoints were total general practice activity (consultations, referrals, investigations and diagnoses) and the use of emergency and out-of-hours services in the study and control groups.
Since, as noted above, it could not be established with confidence whether or not patients had made use of emergency services, this particular analysis was not possible. It also became clear during the course of the study that it would not be possible to collect data on subsequent diagnoses, since these may have occurred some time after the end of the six month follow-up.
However, a number of additional analyses were carried out. The primary and secondary analyses were repeated, considering only those consultations which had taken place for one or more of the six symptoms identified in the leaflet. A further comparison was carried out of the number of consultations which took place during the same calendar month as randomisation.
Data were analysed by zero-inflated Poisson regression, [22] a technique which takes account of the large number of persons with no consultations at all during the six-month observation period. Relative risks (RR) of consultation with 95% confidence intervals (CI) were calculated. Robust variance estimators were used, allowing for clustering by person. Stata version 15.1 was used for data analysis. Table 1 shows the age, sex, eligibility criteria satisfied and region of residence of the patients randomised. Figure 1 shows the trial profile. Of the 1,513 randomised, 1,454 (96%), (749 in the intervention arm and 705 in the control arm) completed six months within the practice in which they were randomised, and therefore had a primary endpoint for analysis.

Primary analyses
. CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted June 11, 2020. . https://doi.org/10.1101/2020.06.09.20126540 doi: medRxiv preprint  Table 2 shows general practice consultations and onward referrals in the two trial arms, and the reasons given for consultations. There was a significantly higher rate of consultation in the intervention arm with 436 consultations compared to 335 in the control arm (RR = 1.40, 95% CI 1.11-1.77, p=0.004). There was, however, no difference between arms in the numbers of individual persons consulting, with 165 in each group. is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

Additional analyses
The number of consultations for the symptoms described in the leaflet was higher in the intervention than in the control arm (38 vs 22), although this was not significant (RR = 1.74, 95% CI 0.81-3.74, p=0.16). Onward referral in relation to symptoms in the leaflet did not differ by arm (8 vs 8; RR=0.88, 95% CI 0.25-3.07, p=0.84). There were more subsequent investigations in the intervention arm for symptoms in the leaflet (34 vs 19) but this difference was also not significant (RR=1.94, 95% CI 0.90-4.21, p=0.09). There were no significant differences between the intervention and control arms with regard to the reasons given for consultation. Table 3 shows the number of consultations in the calendar month of randomisation and in subsequent months. Table 4 shows the month in which the first consultation took place. In the intervention arm, there were significantly more consultations during the calendar month of randomisation than in the control arm, at 16 vs 8 (RR=2.50, 95% CI 1.07-5.86, p=0.035). However, there was no significant difference, between the arms, in the month of first consultation (0.3 months earlier in intervention arm, 95% CI -0.7-0.1, p=0.10). CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted June 11, 2020. . https://doi.org/10.1101/2020.06.09.20126540 doi: medRxiv preprint Whilst the date of randomisation in each practice was known, exact dates of consultations were not. Sensitivity analyses were carried out to determine whether the date of the month on which randomisation occurred had any impact on these results. The association between number of consultations in the month of randomisation and arm continued to be significant, whilst that between month of first consultation and arm continued not to be.

DISCUSSION
This trial took place in a population which was interacting less with primary care and which had factors known to be associated with a lesser tendency to help-seek and late presentation of cancer. [11][12][13][14][15]23,24] The results show that a primary care-based intervention can change consulting behaviour in this population, but not necessarily in the way expected. In the intervention group there were significantly more general practice consultations than in the control group, but there was no increase in the number of persons consulting. Thus the same number of persons visited their GP in both groups, but those in the intervention group visited more frequently.
The intervention included a leaflet highlighting six 'red flag' cancer symptoms. These are important symptoms which less engaged patients may be less likely to recognise as important, and for which they may therefore be less likely to seek help at an early stage. [25] There were twice as many consultations for these symptoms in the intervention group than the control group, but this was not statistically significant. However, given the relatively small proportion of the population targeted in this trial, it is at least possible that were more primary care providers incentivised to deliver such a targeted intervention, and the results of this trial were repeatable nationally, that this might result in statistically and clinically significant numbers of increased primary care consultations for cancer 'red flag' symptoms and referrals.
Those patients that did present to their GP in the intervention group presented more frequently than those in the control group. Thus, the intervention appears to have stimulated greater numbers of consultations in those patients who may have consulted their GP in any case. Irrespective of the symptom(s) these patients presented with, if the increase in consultation rates observed amongst this target patient group were to be sustained over a longer period of time then this may result in earlier diagnoses of cancer. This is because GPs generally enquire about 'red flag' symptoms when patients present with other lower-risk symptoms of possible cancer. Increased consultation rates also provide GPs with an opportunity to promote primary and secondary prevention of cancer and other disease by supporting patients to attend non-symptomatic screening, to manage other chronic conditions more effectively, and to adopt healthier behaviours through initiatives such as Making Every Contact Count. [26] This could include promoting attendance at cancer screening services, which are less well attended by this patient group, [27] as well as opportunistic blood pressure and HbA1c monitoring, promoting regular attendance at primary care chronic disease management clinics and signposting to sources of support for healthier lifestyle modifications. These are activities GPs already routinely offer to patients who present frequently, and which may partially explain inequalities in outcomes in groups that access services less. By reducing the inequity in access to GP support, referral and signposting to other services, interventions such as the one we have tested may play a role in addressing lower cancer survival in lower socioeconomic groups and other groups who are less engaged with primary care, as well as reducing inequality in outcomes for other chronic diseases.
There were several strengths to this trial. Firstly we were able to recruit a relatively large sample of patients from typically hard to reach groups registered at both inner-city and rural practices across England, which gives us confidence that the results can be applied to the national population. Secondly, this intervention . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted June 11, 2020. . https://doi.org/10.1101/2020.06.09.20126540 doi: medRxiv preprint was relatively cheap and was acceptable to most practices approached. Thirdly, the unusual design of the study, with no study specific prior informed consent process, meant the trial was not limited by selection bias for patients who are most interested in and prepared to alter behaviour. This is a significant limitation of many RCTs of health behavioural interventions. While we were able to measure presenting behaviour, a limitation of this study was that we were unable to assess the duration of effect and the impact on clinical end-points. Another limitation was the ambitious concept of changing behaviour of a particularly difficult to reach group with a single postal communication. As such, we do not know whether a second letter would increase effect or if there is a 'dose response' to this intervention. There are several stages in the process of cancer awareness, willingness to seek help, referral and diagnosis, and we considered we were not at the stage of sufficient knowledge to assess the impact of a psychosocial intervention on diagnoses. However, the intervention did change consulting behaviour in a difficult to reach population, although not in the way expected.
Thus, the intervention is only a partial solution to the problem of incongruous consultation behaviour in primary care, particularly bearing in mind the pressures on primary care capacity. It suggests that targeting promotion of cancer awareness and help-seeking for symptoms may have a contribution to make in increasing and modifying help-seeking behaviours. This may contribute towards reducing inequity in access to care and the inequality in cancer survival. Similar targeted interventions may have applications in mitigating other health inequalities by facilitating more equitable access to other areas of primary care within the NHS (such as pharmacies, psychological wellbeing services and dentists). [28,29] Other applications of postal interventions may include facilitating more equitable access to support for health and wellbeing promotion and for the primary, secondary and tertiary prevention of disease. [30] Conclusions This trial demonstrates the potential and feasibility of an inexpensive targeted postal symptom awareness intervention for altering consultation behaviour and reducing barriers to help-seeking in general practice. There is a need to evolve similar interventions with the potential to influence a wider range of potential patients. This type of targeted intervention for those with less access to primary care support is particularly relevant in the UK given the commitment of the NHS to providing equitable care. [31] . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

WHAT IS ALREADY KNOWN ON THIS TOPIC
Later stage of cancer diagnosis, which is associated with poorer survival rates, may arise from low symptom awareness and from delays in engaging with primary care.
Population-wide campaigns to increase awareness and encourage help-seeking have shown mixed results in terms of stage at diagnosis and numbers of primary care consultations.
There have been concerns that the increase in consultations may have taken place primarily among the "worried well", rather than those who would benefit most from early presentation.

WHAT THIS STUDY ADDS
This randomised controlled trial, targeted at a population whose circumstances would suggest that, if they develop cancer, they are at increased risk of a later stage of diagnosis, did not find any increase in the number of patients consulting their GP following a postal intervention that promoted cancer awareness and help-seeking.
It did however find that, among those who did see their GP, those who had received the intervention had a significantly higher rate of consultation.

Trial Registration
This trial was registered prospectively on the International Standard Randomised Controlled Trial Number (ISRCTN) registry (ISRCTN95610478).

Ethical Approval
Ethical approval was granted by the West of Scotland Research Ethics Service (WoSREC) (reference 16/WS/0110). The study was approved by the Health Research Authority (HRA) (reference 201992).

Funding Statement
This research is funded by the National Institute for Health Research (NIHR) Policy Research Programme, conducted through the Policy Research Unit in Cancer Awareness, Screening and Early Diagnosis, 106/0001. The views expressed are those of the author(s) and not necessarily those of the NIHR or the Department of Health and Social Care. SLQ is supported by a Cancer Research UK postdoctoral fellowship (C50664/A24460). The study funders were not involved in the design or analysis of the study. Additional support, including NHS service support costs, were provided by the National Institute for Health Research Clinical Research Network (NIHR CRN) (UKCRN ID 31163).

Competing Interests
Funding for the study is detailed above. All authors have completed the ICMJE uniform disclosure form at www.icmje.org/coi_disclosure.pdf and declare: grants from the National Institute of Health Research for the submitted work (SWD, PS, JPL, DV, and DP), a fellowship from Cancer Research UK (SQ); no financial relationships with any organisations that might have an interest in the submitted work in the previous three years; no other relationships or activities that could appear to have influenced the submitted work.
. CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

Figures
. CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted June 11, 2020. . https://doi.org/10.1101/2020.06.09.20126540 doi: medRxiv preprint . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted June 11, 2020. . https://doi.org/10.1101/2020.06.09.20126540 doi: medRxiv preprint

Supplementary Appendices
Supplementary Appendix 1. Intervention letter . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted June 11, 2020. I enclose a leaflet about symptoms which are usually harmless but can sometimes be a sign of something more serious. We recommend patients with any of the symptoms listed in the leaflet come to the surgery to discuss them with their doctor.
I know many people worry about wasting the doctor's time. However, I would like to assure you that if you do come to see us with any of these symptoms you would NOT be wasting our time.
Please call us on XXXXXXXXXXX if you think you might have any of these symptoms and would like to book an appointment to see your doctor. Please also consult us anytime about any aspect of your health which is worrying you. Our patients are important to us.
Thank you for taking the time to read this letter and the enclosed leaflet.
Yours sincerely, DR NAME SURNAME XXXXXXXXXXX NHS Health Centre <DATE> . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted June 11, 2020. . https://doi.org/10.1101/2020.06.09.20126540 doi: medRxiv preprint