One person in two will develop cancer in their lifetime, accounting for 35% of avoidable deaths in England and Wales. Reducing delays in getting a diagnosis and catching cancer earlier are government priorities. These improvements need to be made whilst being aware that an increasing number of people live with one or more long term conditions. Initial research indicates that the conditions a patient already has may impact the process of getting a diagnosis; for example, a pre-existing condition may be thought to explain a symptom that is actually due to cancer, delaying the cancer diagnosis.
The aim of this study is to find out which patients with pre-existing conditions tend to get worse outcomes in the process of getting a cancer diagnosis such as having a more advanced stage cancer at diagnosis, or waiting longer for a diagnosis and where in the process problems are occurring. To do this, we will explore a large database of about 300,000 patients GP medical records, with extra information from national cancer records and hospital records. No patients will be identifiable.
The findings from this work will help inform cancer policies and guidance for GPs to help improve their decision making with patients with pre-existing conditions. It may also lead to targeted cancer advice for patients with long term conditions. Future work in this programme will further explore these findings and develop and evaluate a practical means of facilitating quicker cancer diagnosis in patients with pre-existing conditions and symptoms of possible cancer.
Improving the timeliness of cancer diagnoses is a governmental health priority. These improvements need to be made in the context of the increasing prevalence of long term conditions, which recent evidence suggests may complicate the cancer diagnostic process. The primary aim of this study is to investigate which patients with pre-existing conditions are disadvantaged in the cancer diagnostic process, and the secondary aim is to highlight where in this process the disadvantage occurs. We will utilise anonymised electronic primary care records of approximately 300,000 patients with an incident cancer diagnosed between 2012-2018, linked to National Cancer Registration and Analysis Service (NCRAS) datasets and Hospital Episode Statistics (HES) data. Our primary outcome will be stage at diagnosis (early or advanced), with secondary outcomes including the route to diagnosis, 30-day and 1-year mortality, how long patients waited before presenting with a symptom of cancer, the interval between first presentation and referral to secondary care, and whether these referral appointments were attended. Morbidity will be operationalised in different ways: single conditions (comorbidities), a count of conditions (multimorbidity), and metrics for the overall burden of conditions (morbidity burden).
Mixed effects regression models, clustering patients by practice, will be used to explore the effect of morbidity constructs on each outcome in turn; logistic regressions for binary data, Poisson regressions for rate data, and time-to-event models appropriate to the data (e.g. flexible parametric survival models). All cancers will be included in each model, with interaction terms between cancer site and morbidity constructs used to highlight disadvantaged subgroups. Non-attendance of referral appointments will be explored as a mediator.
Uncovering patient subgroups who are disadvantaged in the cancer diagnostic process, and where this disadvantage occurs, will facilitate; improved targeting of policy, better informed clinical guidance for affected patients, and design and implementation of interventions to reduce the disadvantage.
We will examine a number of outcomes across the diagnostic pathway for patients with cancer:
Primary outcome
1. Stage at diagnosis (obtained from NCRAS cancer registry linkage): dichotomised to early (stages 0-2) vs advanced (stages 3-4), which will give an overall assessment of disadvantages in diagnostic timeliness.
Secondary outcomes
2. 30-day mortality post diagnosis (NCRAS and ONS death registration)
3. The route to diagnosis (NCRAS): emergency, Two Week Wait referral, GP referral, other)
4. The patient interval (CPES wave 5: days from symptom onset to presentation);
5. The primary care interval (days from first presentation as identified in CPRD to referral as identified in HES)
6. The diagnostic interval (days from first presentation as identified in CPRD to diagnosis as identified in NCRAS)
7. The number of pre-referral consultations (CPES)
8. The number of primary care consultations in the 12 months preceding diagnosis (CPRD). Analysis of this outcome will require a control sample without cancer.
Following reviewer feedback of manuscripts in preparation for disseminating our findings, we would like to add two further outcomes:
9. 1-year mortality post diagnosis (NCRAS and ONS death registration).
10. Whether or not referral appointments (as identified in outcome 5 above) were attended.
Gary Abel - Chief Investigator - University of Exeter
Luke Mounce - Corresponding Applicant - University of Exeter
Bianca Wiering - Collaborator - University of Exeter
Georgios Lyratzopoulos - Collaborator - University College London ( UCL )
Jose M Valderas - Collaborator - University of Exeter
Sarah Price - Collaborator - University of Exeter
William Hamilton - Collaborator - University of Exeter
HES Accident and Emergency;HES Admitted Patient Care;HES Outpatient;NCRAS Cancer Patient Experience Survey (CPES) data;NCRAS Cancer Registration Data;ONS Death Registration Data;Patient Level Index of Multiple Deprivation