End-of-life care research is sensitive by the nature of the subject matter it seeks to investigate. Researchers working in this area must also consider the role of research networks, research governance authorities and gatekeeping clinicians, when planning studies.
The data in this study provide an insight into the efficiency of recruitment of bereaved carers or older adult patients into interview studies via primary care. The number of interviewees identified per general practice varied widely, and was unrelated to the size of the population from which the interviewees were drawn, or the socioeconomic characteristics of the area in which the practice was located. In the areas where research network staff negotiated between researcher and GP practice, fewer interviews were completed and detailed information on the research process was not readily available. Refusal to grant research governance permission was unpredictable, and in this case, removed an area of high socioeconomic deprivation from the study.
The variation in the number of potential interviewees identified by different practices was so wide, it suggests that there was significant selection going on at that level. This is not surprising in the case of carers, as we did not expect that GPs would be aware of the identity of the carer for all their deceased patients, or that they would necessarily feel comfortable contacting all of the known carers. The practices varied in size from fewer than two thousand to more than fifteen thousand registered patients, yet the majority managed to find at most, one carer to be interviewed. Similarly, the proportion of patients invited to be interviewed varied by a factor of 38 between practices. The carers were drawn disproportionately from less disadvantaged areas. This may reflect the socioeconomic circumstances of the regions involved, but it is possible that GPs selected people who they felt would be better interviewees, because of their education, or availability, or people with whom they had built closer relations during the decedent’s last illness. It is also plausible that people in less deprived areas felt more comfortable about responding to the invitation to be interviewed. We have no data to suggest that the social patterning in the group of carers is related to the way in which research network staff were involved in recruitment.
Comparison with other work
Data on recruitment are seldom presented in detail, despite this being a fundamental influence on the potential value of research outputs. The response rate of around 25% for carers was lower than some postal surveys of carers with broader inclusion criteria, and other disease-specific studies [20–23]. Ewing and colleagues have previously reported on the barriers presented by professional gatekeeping and NHS governance approvals [11]. When they used a similar opt-in approach to recruiting patients, approximately 1% of invited patients were interviewed. Our higher participation rate (15%) seems most likely to be due to patient related factors, rather than the provision of financial rewards to practices, in the form of modest research support costs. Staff understanding of the study inclusion criteria was not perceived to be a barrier to recruitment [24], though we collected no data to test this assumption.
Low participation rates may be accommodated more readily in qualitative enquiry, when the aim is to extend our understanding of patient and caregiver experiences. However, it would be a serious limitation to any study where obtaining a representative sample was crucial. Qualitative studies with older people near the end-of-life often do not reveal how many people were approached to find their interviewees, and the denominator population is also unknown for studies that recruit via local or national advertising.
Strengths and limitations
Three different researchers were involved in the work reported here, and we cannot be certain that variation in their tenacity or sensitivity to the clinical context did not contribute to differential recruitment rates. The inclusion and exclusion criteria were specific to our studies, and may have made our recruitment of participants particularly challenging. Both the aim of interviewing carers between four and nine months after the death and the requirement for the decedent to have moved between places of care, limited our population of potential interviewees. Nevertheless, our choice of patients and decedents aged over 75 years captures the majority of deaths in England and Wales, and the conditions specified account for the common causes of predictable death. It is a strength of this analysis that we have information on a large number of participants from qualitative interview studies. And by working across regions, we were able to evaluate the different modes of working with research networks.
Implications
The nature and extent of gatekeeping by health professionals is an important issue for researchers, as data protection or privacy legislation prevent direct approaches to patients. Gaining the trust of referring clinicians is thought to be important for success [25], and that may be missing when a research network employee acts as an intermediary between researcher and health professional.
The regulatory framework may leave some GPs with difficult judgements to make. Any desire they have to shield vulnerable older adults from potentially unwanted approaches by researchers, must be reconciled with avoiding paternalism, and a policy rhetoric that promotes empowering patients to make their own choices. In the UK, the combination of organisational reform, increasing workloads and greater involvement of private providers in NHS care present a particularly challenging context for researchers in primary care.
The UK Royal College of General Practitioners is currently encouraging its members to identify the 1% of patients on their lists who are likely to die within the following twelve months, in order to improve their care [26]. Even taking into account the fact that our search was restricted to aged patients with cancer, the number identified was considerably lower than could be expected, if every hundredth registered patient is in the last year of life. Identification of people with a short prognosis is still an important barrier to overcome. Financial compensation for general practices who took part in the research was modest, and did not result in a high number of patients being identified. It is possible that payment per participant would be a more effective incentive, though this is not likely to be judged acceptable for end-of-life studies.
People from lower socioeconomic groups or areas are often classified as ‘hard to reach.’ This study provides some data to support that assertion, as the recruited sample of carers was made up of disproportionately fewer people from poorer areas. Allowing sufficient time for purposive recruitment by socioeconomic status may help to overcome this problem, along with raising awareness amongst recruiting health professionals. Without particular attention to social patterning in study recruitment, research has the potential to inadvertently perpetuate inequalities in access to care. NHS authorities play a crucial role in study regulation, and the process by which they allow access to different populations should be transparent and follow nationally agreed criteria.