This review of primary care interventions to reduce delays in secondary care cancer referrals found no evidence that any intervention directly reduced delays. Limited evidence suggested that complex interventions had potential to reduce delays. This was a good review, but the conclusions need to be considered in the light of the poor study quality and risk of language bias.

To identify interventions in primary care that reduce the delay in referring cancer patients to secondary care.

MEDLINE, EMBASE, CINAHL, PsycINFO, HMIC, Web of Science, ERIC and The Cochrane Library were searched from inception to March 2010 for studies published in English; search terms were reported. Reference lists of included papers and reviews and citations were checked. Experts in the field were contacted.

Eligible studies were primary studies of any design that evaluated interventions conducted in a primary care setting with the aim of reducing delays in cancer referral. Studies had to report reductions in delays in cancer diagnosis, include fully trained primary healthcare professionals as participants and could involve any type of cancer. Studies were excluded if they aimed to improve screening or self-examination rates, were of professionals other than fully trained general practitioners or practice nurses, were of the two-week wait rule or did not provide separate results for the effects on primary healthcare professionals.

Most of the included studies focused on skin cancer; others assessed breast, colorectal and prostate cancers. Around half of the studies were carried out in Europe (five were in the UK). Other studies were from Australia, USA, Canada, Egypt and Tunisia and Pakistan. No study directly reported the primary outcome of reducing delay in the cancer care pathway. Reported outcomes included improved knowledge or observational skills, improved diagnostic accuracy and improved referral rates. The interventions included education, audit and feedback, decision-support software and guideline use, diagnostic tools and other skills training.

One reviewer screened the titles, abstracts and full papers. A second reviewer checked 25% of the abstracts and all of the full papers.

Study quality was assessed using the Newcastle-Ottawa Scale adapted for use with randomised controlled trials (RCTs) and other studies in three sections that covered participant selection, comparability and outcome. RCTs could achieve a maximum score of 11 and other study designs could achieve 10. Scores of 7 of more for RCTs and 6 or more for other designs were considered good quality.

Studies were scored by two reviewers independently and then compared to reach a final score.

Information on the interventions and outcomes were extracted by two reviewers independently and then compared to reach consensus.

Results were presented in a narrative synthesis grouped by the type of intervention.

Twenty-two studies were included (number of participants not reported). Six studies (all RCTs) were classed as good quality. Issues with study quality included selection bias, less than one year’s follow-up, small sample size, high drop-out rate and a lack of detail about the administration of the intervention.

Education (nine studies): Five before-and-after studies used traditional lectures or training sessions to try to improve knowledge of risk factors and diagnostic ability among health professionals. None of these studies were rated as good quality. All involved similar assessments (before and immediately post-test) and showed some improvement in knowledge.

RCTs used various educational interventions that included lectures, videos and an internet triage package, compared to no intervention control groups. Trials were of mixed quality and provided inconsistent evidence. The two RCTs that reported a positive outcome of improved diagnosis and management did not see this improvement maintained to the end of the trial.

Audit and feedback (three studies): One high quality RCT where participants either did or did not receive feedback reported improvement in the recording of clinical information but not in diagnostic accuracy. Low quality before-and-after studies showed significant improvement in outcomes.

Decision support software and guidelines (five studies): Three high quality RCTs evaluated computer-based interventions designed to improve the quality of referrals to secondary care. None of the studies found an improvement in the appropriateness of referrals. One study of a distance learning intervention reported a reduction in urology referrals.

One of two low quality before-and-after studies found that participants made fewer incorrect clinical decisions after using software for the triage of skin lesions and the other found increased use of breast cancer guidelines but no effect on knowledge or attitudes.

Diagnostic and assessment tools (four studies): None of the studies that aimed to improve particular skills or use of equipment to increase diagnostic accuracy were considered to be good quality. One before-and-after study of skin cancer found that dermoscopic teleconsultations led to more relevant referrals than clinical examination. A feasibility study of a clinical prediction rule for colorectal cancer found that uptake of the tool was low.

One RCT gave primary healthcare professionals a one-day training course in a rule for assessing skin cancer and randomised them to assess skin lesions by standard clinical examination or dermoscopy. Dermoscopy improved triage, referral sensitivity and accuracy and led to participants being less likely to miss skin cancers. The other RCT assessed an educational intervention for improving melanoma diagnosis using surface microscopy compared to no training. Participants’ diagnostic ability on clinical photographs improved significantly after microscopy training.

Other skills training: One low quality RCT assessed whether arts-based skill training could improve observational skills. After training by an artist, observational skills improved significantly more than for a control group who received a lecture on skin diseases.

There was no evidence that any intervention directly reduced primary care delay in the diagnosis of cancer. There was limited evidence to suggest that complex interventions, including audit and feedback and specific skills training, had potential to do so.

This review stated the inclusion criteria for study design, intervention, participants and outcome. The literature search covered a several relevant sources and included contact with experts in the field. The search was restricted to studies published in English so there was a risk of language bias. The authors acknowledged this as a limitation as they did not have resources for translation services. Most of the review methods (apart from title and abstract screening) were performed by two independent reviewers to reduce errors or bias. The studies were quality assessed and the overall score and details of the problems with each study were reported but the complete results were not reported for each study. The narrative presentation of results was appropriate given variation between the studies. Information such as the sample size of each study would have been helpful. The authors provided a thoughtful discussion of the limitations of the evidence and their review.

On the whole this was a good review, but the conclusions need to be considered in the light of the poor study quality and risk of language bias.

Practice: The authors stated that audit and feedback and specific skills training appeared to be most effective in improving longer-term outcomes. Interventions were most useful when tailored to the needs of the population.

Research: The authors stated a need for future research to attempt to measure delay directly to determine which interventions were most effective in reducing referral times from primary to secondary care. Such studies needed large participant populations and associated funding, longer follow-up periods, improved participant retention and improved reporting.

NHS North Staffordshire; NHS Stoke; NHS Executive West Midlands R&D Office; North Staffordshire and Cheshire R&D Consortium.