Ethical considerations

The collection of data from clinical records after death required additional research ethics permissions. Permission to access patient records after death, without their prior consent, lies outside the scope of NHS Local Research Ethics Committees. While the further ethical permission from the National Information Governance Board for Health and Social Care (now Health Research Authority Confidentiality Advisory Group) took several months, it enabled the creation of a novel data set of community care provision in the final 6 months of life.

Patient-identifiable data (name, NHS number, dates of birth or death) remained on practice premises, with only unique study identifier codes available to researchers off-site. To further ensure confidentiality and reduce potential bias from prior knowledge, data extraction from clinical records and bereaved carer interviews were conducted by different research team members. The study advisory group of clinicians, researchers and lay members advised that bereaved carers might be uncomfortable that their loved one’s medical records had been accessed, despite our having ethical permission to do so. Patient and Public Involvement (PPI) members of the advisory group made a significant contribution to the submission of ethical applications. The role of PPI in ethical applications ensures that the research is accessible and applicable to a patient group, and can lend credence to applications for research approvals in sensitive areas such as palliative care.

GP practices were asked to identify individuals who had been close to the deceased patients and to minimise gatekeeping by including all those suitable for interview.15 16 Potential participants were approached by practices on behalf of the research team 6 months after bereavement, avoiding significant dates such as birthdays or Christmas. Information concerning bereavement support services was given at interview, highlighting the role of GP practices, with a telephone call made a few days after interviews to check on participants’ well-being.

Conducting research with bereaved people is highly sensitive, potentially placing a burden on participants at a vulnerable time. Several bereaved carer participants reported benefit from taking part and sharing their experiences: none reported interviews to have been unhelpful, adding to evidence that many bereaved people welcome the opportunity to be involved in research and may receive benefit from doing so.17 Timing of interviews 6 months after death, avoiding sensitive dates and providing bereavement support information helped create a safe environment for participants to discuss the care of their loved one.

Recruiting GP practices and DN teams

Few difficulties were encountered in recruiting the 20 GP practices, in large part due to strong support from Clinical Research Network (CRN) colleagues, who were familiar with local practices and arranged payments to GP practices. Purposive sampling recruited a diverse sample of practices across the East of England by: urban, rural and market town populations; practice age structure, ethnicity and social deprivation.

DN team recruitment proved much more challenging, despite DN managers being involved in the funding application and study design throughout. While warmly supportive in principle, fieldwork took place during a time of considerable change and uncertainty in DN teams with changing providers, high workloads and staffing shortages. In contrast to GP practices, CRN colleagues had little success in their recruitment approaches to DN teams. DN teams were more willing to participate when approached through the recruited GP surgeries. Personal explanations of the limited commitment involved reassured several who had initially been put off by the study paperwork. It eventually proved possible to recruit each DN team linked to the 20 participating GP practices, although considerable flexibility in recruitment approaches was needed.

The context within which recruitment is undertaken cannot be ignored. As stated above, DN teams were operating within an environment of considerable uncertainty and pressure which impaired their ability to commit to a research project. The research team ensured that expectations of time commitment were clear and realistic and were flexible to accommodate the realities of the local DN landscape. It should be noted that these observations relate to UK NHS practice.

Despite successful recruitment of all GP surgeries and DN teams, participation of individual team members varied. GP practice agreement to participate was often made by a managing partner or one GP. This did not always guarantee participation of all GPs in the practice. In all, 123 GPs from 20 practices were involved in the study. Although we sought participation by the GP most involved with a patient, an alternative GP was sought when the patient’s main GP declined to participate. In some DN teams only one nurse knew the deceased patients due to colleagues moving posts or long-term illness. This often resulted in one nurse taking on all aspects of the study.

Documenting primary care provision

In phase 1 data were extracted concerning GP and DN care provision for patients in the final 6 months of life from their GP and DN records. All participating GP practices readily provided access to their GP records. Administrative staff printed copies of the electronic GP records and give access to missing data such as medications.

Access to DN records was more challenging. Most teams used patient-held paper records. Some were untraceable, being either unlocatable in external archiving facilities or thought to be still in the deceased patient’s home. Occasionally portions of DN paper notes were missing. Fuller records were obtained from the few DN teams that used electronic records in addition to paper notes in the home. Based on GP records, the DN records that we had access to and GP reports, we estimate that 202 (51%) of the 400 deceased patients received DN care on at least one occasion during the final 6 months of life, of which we were able to obtain data for 134 (66%; 33% of total). In time, increasing use of shared electronic health records should mitigate these difficulties.18

GP practices used several computer systems, only some of which contained information from other services such as specialist palliative care, speech and language therapy and physiotherapy. In practices using other information technology systems, evidence of such services’ involvement was only available in letters or GP comments. Analysis therefore focused solely on GP and DN care. A full picture of care would require accessing the records of all potential services, including hospital and hospice records, which was outside the remit of the study. Data were missing for cause of death for 34 patients: this was obtained from the General Register Office.

Data were extracted into a study-specific database for all recorded patient interactions with healthcare services in the last 6 months of life. They were categorised by location (GP surgery consultation, telephone consultation, outpatient appointment, inpatient admission, and so on), healthcare professional involved (GP, DN, consultant, healthcare assistant, and so on) and days before death. For 22/400 (5.5%) patients, GP records were not available for the entire final 6 months of life, having been previously registered with another practice whose records were not accessible. This was particularly the case for those transferred to a nursing home at the end of life. This meant that while 100% of clinical records were provided by the GP practice with which the patient was registered at death, a portion of these records for the final 6 months of life was not available to the research team.

GPs and DNs completed a brief questionnaire for each patient concerning if/when/why the patient had been identified as approaching the end of life, key indicators of decline and additional community services such as specialist palliative care nursing and social care. Replies were obtained for 388/400 (97%) GP and 118/202 (58%) DN questionnaires.

Obtaining healthcare professionals’ perspectives

After data extraction from records, focus groups were held in 10 practices, discussing in each the care of five practice patients, purposively selected as a maximum variety sample (cancer/non-cancer, gender, age group and degree of GP and DN involvement). Between 2 and 10 participants, mainly GPs and DNs, with some other clinical staff took part: timetabling for scheduled educational or multidisciplinary team meetings ensured optimal attendance.

In the other 10 practices semistructured interviews were held with bereaved carers, GPs and DNs for two patients per practice (one cancer, one non-cancer) with whom there had been significant primary care team input. Carer interviews explored perspectives on care provision for the deceased, both as recipients of care and as proxies for the deceased. GP interviews were problematic to schedule: practice staff at times booked appointments without the GP knowing the purpose and occasionally arranged a 10 min surgery appointment. It was common for one DN to engage with the study on behalf of their team. At times he or she was the sole DN completing all questionnaires and participating in the focus group or the two interviews. Some DN teams covered more than one participating GP practice, creating a DN participant burden that delayed scheduling interviews.

Obtaining bereaved carer perspectives

Practice teams found it difficult to identify bereaved carers. After searching deceased patients’ records and checking informal practice knowledge, no lay carer was identified for some: for others carers’ names were known but not their addresses as they were registered with different GPs or non-resident with the patient. Carer recruitment was initially very poor: in the first five practices, of the 21 bereaved carers invited, 4 (19%) agreed, 6 (29%) declined and 11 (52%) did not respond. The study user group was invaluable here, suggesting that potential participants may have felt intimidated by the bulky and official-looking study information pack: they suggested that a less formal initial brief letter be sent in handwritten envelopes, with the full participant information sheet subsequently mailed to those interested in participating. They also advised that non-response to the initial approach letter did not indicate refusal and that a follow-up letter after 1 month was appropriate. These changes, approved by the Local Research Ethics Committee, had a profound effect on participation: of the subsequent 15 carers invited, 87% agreed (two after the follow-up letter), 13% did not respond and none declined. We achieved an improved carer response rate by avoiding overburdening people with official-looking information at the first approach and sending reminders giving additional opportunities to respond. This resulted in markedly improved response rates overall, yielding 17 carer participants, 47% of those approached.

Thus with flexibility and persistence it proved possible to recruit bereaved carers, GPs and DNs, obtaining their differing perspectives on individual patients’ care.