In our study, approximately half of the GPs declared that they always refer elderly cancer patients to a cancer team (this was slightly more frequent for early stages than for advanced disease). More than three quarters of referring and non-referring GPs reported being influenced by the five following patients-linked factors: 1) wish or reluctance on the part of the patient; 2) wish of the family if present; 3) presence or absence of serious comorbidity; 4) unsuitability of conducting invasive investigations; and 5) the degree of mental and physical autonomy. Three quarters or more of the GPs were influenced by two disease-linked factors, the seriousness of the cancer symptoms and expected side effects and tolerance of treatment. Finally, being used to collaborating with specialist cancer teams was the only organisational element reported to be influential by more than 75% of the GPs. With regard to the two specific cancer cases presented in clinical case vignettes, the GPs did not have the same approach for the two patients, nor when compared to the general referral situation. Confronted with a case of prostate cancer, the GPs tended to refer the patient to a specialist (75%), whereas when faced with a sigmoid colon cancer, 45% of GPs reported that they would refer to a specialist and 40% to an oncologist. For both these types of cancer, the seriousness of the symptoms appears to influence GPs the most. Regardless of the stage of the cancer, organisation difficulties were an independent factor influencing the GP's decision whether or not to refer elderly patients.
In studies reported in the literature and conducted in other regions in France and in Canada [10
], higher referral rates of elderly patients presenting with cancer have been reported according to patient's age [12
] and to stages of the disease [10
]. Our referral rates are slightly lower in the general situation, but we have similar high referral rates for both clinical vignettes. Overall, slightly more GPs declared that they "always refer" for early stages than for advanced disease which is similar to results previously reported [10
]. In Canada, as in France today, patients encountering a health problem consult in the first instance the family practitioner (GP) who decides if referral to a specialist is required. It is possible that in the present study the figure for referral is underestimated on account of the absence in France for a clear definition of a "team of cancer specialists", so that the GPs may not have included specialists practising oncology under the term.
In terms of the factors influencing GPs' referral decisions, we found the same factors to be cited in majority (patient's wishes and tumour-related factors) as observed previously [10
]. When independent factors associated with the decision to "always" refer an elderly patient to a cancer team for early stages of the disease are considered, GPs reporting being influenced by the anatomical localisation of the cancer reported referred their patients less often. In the limited literature available, the anatomical localisation has never been documented as a factor influencing GPs' decisions. However, this factor seems likely to play a part in decisions on the way a cancer patient is to be cared for, and this aspect was confirmed in this study. In particular, our results show that the GPs did not refer patients in the same way in the general situation (responses to questionnaire Section B) and in the two clinical situations proposed. Another decisional factor found, irrespective of disease extension, was the difficulty involved in organising care. This factor is also found in another two studies conducted on a sample of GPs where organisational difficulties were found to influence decisions [10
]. This has also been observed in studies conducted among elderly breast and colon cancer patients where the patients reported preferring to receive care close to their homes [13
]. Indeed, cancer care that is often complex and requires frequent consultations may be difficult to organise for elderly patients living at a distance from the specialised cancer centres. In this context, the GP may choose to refer a patient to a closer facility such as, a geriatric or medical ward [6
]. Despite this, the practice setting (urban or rural) was not found as a significant factor in any of the models in the present study. For advanced cancer, GPs reporting that they "did not always refer" their patients were those not having attended courses in oncology, and those influenced by the chronological age of their patients, the stage of advancement of the disease, or organisational difficulties for care. For this group, age itself was a determining factor of the referral decision. This association has not been documented in the literature up until now. This difference may not have been observed in the Canadian study as it was performed some time ago and in a different health system [10
]. In the French study, the stage was not taken into account when studying referral decisions [12
]. Oncology training was found to increase referral rates in our study, but we did not observe an association with geriatric training that has previously been reported in the literature as decreasing referral rates [10
The GPs did not refer their patients to the same specialists in the two clinical case vignettes. For prostate cancer patients, GPs referred their patients to a urological specialist. For patients with colon cancer, GPs referred their patients to a gastro-enterologist or to an oncologist. Specialists' attitudes towards oncogeriatrics and established collaboration relationships can have an important impact on the initial management of patients [15
Finally, there are two main limitations to this study to keep in mind when interpreting results. The first relates to the type of survey, with this being a postal survey conducted among GPs in the South-West region of France, Aquitaine. After two reminders, an acceptable response rate for this type of study was obtained (30%). At first regard, this may seem low and indicate a selection bias, but since only 50% of graduated doctors are thought to be in general practice in France [16
], probably only 50% of the GPs listed in the regional database we used to generate our participant list were actually concerned by the study. Kurtz et al [12
] showed a higher GP response rate but this difference can be explained by the different types of questionnaire (they did not present clinical case vignettes) and they employed a more direct regional communication method to obtain GP responses. In our group of respondents, there were slightly more males and GPs with a rural practice than in the regional GP database. An explanation could be that these GPs see more elderly patients so they were more interested in participating in the study.
The second limitation concerns the fact that the clinical cases vignettes covered two specific disease localisations and the GPs were probably influenced by the prognosis for these specific cancers.