Two week wait referrals for possible cancer: how much of the variation between GP practices is due to CCGs and hospital trusts?

Talk Code: 
Chris Burton
Luke O'Neill, Phillip Oliver
Author institutions: 
University of Sheffield


Ensuring that patients with symptoms of possible cancer have access to prompt diagnostic tests and treatment is a key concern for healthcare systems. Variation between providers in referral or access to such services is an important performance issue. In the UK and Europe, particular attention has been drawn to variation in referral from primary to secondary care for patients with clinical features possibly indicative of cancer. We aimed to examine how much of this variation at practice level was attributable to local healthcare organisations and provision (CCGs and Acute Hospital Trusts)


We used multilevel linear regression on publicly available data over the 5 years from 2013- 2017. We included all practices in England with 50 or more two-week wait referrals over the period. For each practice we obtained standardised referral rates and calculated sensitivity and specificity of referrals for a diagnosis of cancer. We mapped each practice to a CCG and each CCG to an Acute Hospital Trust (AHT). We used three types of model: (a) random intercept, practice within CCG; (b) random slope and intercept, practice within CCG and (c) random intercept practice within CCG within AHT. The first two model types were run on all practices and the third only on practices where two or more CCGs mapped to the same AHT. All models were run before and after adjusting for practice characteristics. For models with cancer detection rate (sensitivity) as the outcome we also adjusted for specificity. We estimated the proportion of variation as either intraclass correlation coefficient, R-squared or both.


CCGs accounted for 21% of the variation between general practices in the standardised fast-track referral rate and 35% of the unadjusted variation in cancer detection rate (sensitivity). After including practice characteristics (specificity, deprivation, and proportion of patients aged over 65), CCGs accounted for 30% of the variation in cancer detection rate (compared to 13% accounted for by practice characteristics). In areas where a hospital trust was the main provider for multiple CCGs, trusts accounted for approximately twice as much variation as CCGs (between 62% and 70% of the explained variation).


This is the first large-scale finding that a substantial proportion of the variation between GP practices in referrals is attributable to their local healthcare systems. Efforts to reduce variation need to focus not just on individual practices but on local diagnostic service provision and culture at the interface of primary and secondary care.

Submitted by: 
Chris Burton
Funding acknowledgement: