Variation in number of general practitioner consultations before hospital referral for cancer: findings from the 2010 National Cancer Patient Experience Survey in England.

Background: Information from patient surveys can help to identify patient groups and cancers with the greatest potential for improvement in the experience and timeliness of cancer diagnosis. We aimed to examine variation in the number of pre-referral consultations with a general practitioner between patients with different cancers and sociodemographic characteristics.

Methods: We analysed data from 41,299 patients with 24 different cancers who took part in the 2010 National Cancer Patient Experience Survey in England. We examined variation in the number of general practitioner consultations with cancer symptoms before hospital referral to diagnose cancer. Logistic regression was used to identify independent predictors of three or more pre-referral consultations, adjusting for cancer type, age, sex, deprivation quintile, and ethnic group.

Findings: We identified wide variation between cancer types in the proportion of patients who had visited their general practitioner three or more times before hospital referral (7·4% [625 of 8408] for breast cancer and 10·1% [113 of 1124] for melanoma; 41·3% [193 of 467] for pancreatic cancer and 50·6% [939 of 1854] for multiple myeloma). In multivariable analysis, with patients with rectal cancer as the reference group, those with subsequent diagnosis of multiple myeloma (odds ratio [OR] 3·42, 95% CI 3·01-3·90), pancreatic cancer (2·35, 1·91-2·88), stomach cancer (1·96, 1·65-2·34), and lung cancer (1·68, 1·48-1·90) were more likely to have had three or more pre-referral consultations; conversely patients with subsequent diagnosis of breast cancer (0·19; 0·17-0·22), melanoma (0·34, 0·27-0·43), testicular cancer (0·47, 0·33-0·67), and endometrial cancer (0·59, 0·49-0·71) were more likely to have been referred to hospital after only one or two consultations. The probability of three or more pre-referral consultations was greater in young patients (OR for patients aged 16-24 years vs 65-74 years 2·12, 95% CI 1·63-2·75; p<0·0001), those from ethnic minorities (OR for Asian vs white 1·73, 1·45-2·08; p<0·0001; OR for black vs white 1·83, 1·51-2·23; p<0·0001), and women (OR for women vs men 1·28, 1·21-1·36; p<0·0001). We identified strong evidence of interactions between cancer type and age group and sex (p<0·0001 for both), and between age and ethnicity (p=0·0013). The model including these interactions showed a particularly strong sex effect for bladder cancer (OR for women vs men 2·31, 95% CI 1·98-2·69) and no apparent ethnic group differences in young patients aged 16-24 years, whilst the only cancers without an apparent age gradient were testicular cancer and mesothelioma.

Interpretation: Our findings could help to prioritise and stratify early diagnosis initiatives and research, focusing on patients with cancers and sociodemographic characteristics with the largest potential for improvement.

Funding: None.