The median secondary care interval decreased overall after the introduction of urgent referrals for suspected cancer. With a decrease of 10 days, breast cancer patients saw the largest decrease. Studies have shown that breast cancer patients with treatment delays of three months or more had a lower 5-year survival rate than patients with shorter delays . Our finding of a median reduction of 10 days may not result in reduced mortality in general, but the patients who waited the longest (e.g. the 75 percentile was 74 days at other hospitals than Vejle before the national introduction of the urgent referral guidelines) may have gained more from the introduction of the urgent referral guidelines. Finally, there may be other benefits valued by patients as well as professionals.
Both before and after the national introduction of the urgent referral guidelines, patients had a shorter secondary care interval in cases where initial symptoms were categorised as alarm symptoms by the GP or the GP indicated cancer suspicion in the referral documents. Obviously, patients with a clearer pathological picture are easier to diagnose. This also supports the assumption that the GP’s symptom interpretation and reaction to suspicion of cancer is instrumental in shortening the secondary care interval for cancer patients. However, at other hospitals than Vejle, the secondary care interval also decreased for patients with no indication of cancer suspicion and no alarm symptoms, indicating improved performance of the system as a whole.
The tendency towards a decrease in the secondary care interval at other hospitals seemed to appear when the need for such decrease was officially recognised, i.e. before the formal implementation of the urgent referral guidelines. Furthermore, it should be noted that Vejle Hospital was able to reduce its secondary care intervals even further when the introduction of the urgent referral system was introduced at national level.
The study design does not allow any determination of causation. The fact that diagnoses without urgent referrals did not decrease statistically significant after the introduction of urgent referrals indicates that the decreasing secondary care intervals could be related to the introduction of urgent referrals. On the contrary, the fact that Vejle Hospital actually managed to decrease their secondary care intervals, also for patients with no indication of cancer suspicion in the referrals from their GP, indicates that the decrease is also related to other factors.
Strengths and weaknesses
Selection bias in this study was induced during the initial inclusion procedure. We handled this by adding a second inclusion and hereby completing the cohort of patients. However, this may have induced information bias since the GPs of Sample 1 received the questionnaire within one month after diagnosis, while the GPs of Sample 2 received the questionnaire between one and two years after diagnosis. GP recall bias could influence the secondary care interval and the categorisation of patients according to initial symptom presentation and cancer suspicion for the first referral to secondary health care. Recall bias is not considered to have much impact on length of secondary care interval nor on number of suspected cancer referrals since the GPs have access to this information through the patients’ electronic medical records, which are considered robust data sources. Apart from guiding the GPs with examples of alarm symptoms, it was not specified precisely which symptoms should be categorised as alarm symptoms. Thus, the interpretation of the same symptom could differ among GPs. This source of information bias is not considered to be major for two reasons. First, alarm symptoms of cancer are not clear-cut and will therefore be influenced by individual GP interpretation. Second, alarm symptoms of cancer are known to the GPs so that if they do suspect cancer such symptoms (e.g. rectal bleeding) will be categorised as alarm symptoms. Another possible source of information bias is the retrospective design of the study. The GP knows that the patient has cancer when filling out the questionnaire which may influence the answers. However, it would be costly to set up a prospective study given the few cancer patients diagnosed by each GP every year. Furthermore, it has been shown that the GPs are willing to answer these questions because they see a learning potential in going through the care pathway retrospectively . In conclusion, even though the retrospective design may be a source of bias, the data are considered a reliable foundation for the analyses.
GP-induced selection bias may have occurred if patients of non-responding GPs had a different secondary care interval than patients of responding GPs. If GPs with a longer doctor interval (time from the patient’s first contact with the GP until first cancer-related investigation) tended to be more reluctant to respond, this could lead to shorter secondary care intervals. This bias may therefore lead us to make a type II error by underestimating the association between the introduction of the urgent referral guidelines and the secondary care interval.
The considerable sample size strengthens the statistical precision of this study. We did not have complete information on the secondary care interval for almost one third of the patients due to missing treatment dates. We used the date of admission in these cases, which introduced a systematic underestimation of the secondary care interval. Since more patients in other hospitals than Vejle as well as patients without alarm symptoms and without indication of cancer suspicion had incomplete secondary care intervals, we underestimated the secondary care interval for patients with the longest secondary care interval, and this may have provided absolute minimum differences. Since we were still able to detect considerable changes, we have found that this bias was not considerable enough to significantly alter our conclusions.
No information is available on who were actually referred urgently after the introduction of urgent referral. The GPs’ reported data on whether patients presented with alarm symptoms and clearly indicated cancer suspicion in connection with first referral are therefore our best sources of information for determining whether the patients were regarded as urgent or not. However, since the study included a period before and a period after the urgent referral was implemented, it is meaningful to compare the GPs’ assessment of the initial symptoms and whether or not the GPs indicated cancer suspicion in their referrals to secondary health care since we have this information for both periods.
The conclusions of this study are considered generalisable to other regions in Denmark and to other health care systems involving GPs serving as gatekeepers to secondary health care. The possible impact of the urgent referral will depend on the local context.
Findings in relation to other studies
The overall median secondary care interval of 42 days (IQI: 22; 80 days) for other hospitals than Vejle Hospital before the introduction of urgent referrals is fairly consistent with the results from another Danish study from 2005, which reported a secondary care interval of 46 days (IQI: 26;78) . This study used the same definition of the secondary care interval and also collected data by use of questionnaires sent to the GPs. A comparison of our results with the findings of a recent study by the Danish National Board of Health shows that the secondary care intervals reported in our study were shorter than those previously reported, except for colorectal cancer . The National Board of Health counted the days from the receipt of referrals at the hospitals, whereas we measured from the day on which the referrals were forwarded by the GPs. Furthermore, the National Board of Health counted the days until the patients gave consent to treatment, whereas we included the time until treatment was actually initiated. Further, we included the GPs’ suspicion of cancer and having noted so in the referral documents. The present study was hence able to demonstrate the importance of the referral process and the GPs’ knowledge, awareness and action. In addition, we were able to compare Vejle Hospital with other hospitals and identify aspects highlighting the possible impact of urgent referral on time to treatment.
In other healthcare settings, studies have reported a direct, negative impact on the length of the secondary care interval for non-urgent patients from the introduction of urgent referral for suspected cancer [19, 27–29]. However, we saw no indication of such major adverse effect in our data.