Unwelcome memento mori or best clinical practice? Community end-of-life anticipatory medication prescribing practice: a mixed methods observational study

Study design

We carried out a retrospective mixed methods observational study of deceased patients GP and community nursing held clinical records. (28,29) Reflecting the social constructionist paradigm, (30) clinical records are selective and stylised clinician accounts rather than presenting only objective facts. (31,32)

The Cambridge Positive Ageing and Cambridge Palliative and End of Life Care Patient and Public Involvement (PPI) Groups supported the study throughout, advising on research priorities, the acceptability of accessing deceased patient records without consent and interpretation of key findings. The South Cambridgeshire Research Ethics Committee granted ethics approval [Reference: 19/EE/0012]. The Health Research Authority’s Confidentiality Advisory Group [19/CAG/0014] approved the processing of confidential patient information without consent in the public interest: data were anonymised at the earliest opportunity.

Study population and setting

Participants were registered with eleven GP practices and two associated National Health Service Community Trusts providing community nursing services in Hertfordshire and Cambridgeshire. GP practices were purposively sampled from 21 practices expressing interest in participation, to obtain a maximum diversity sample in terms of patient list sizes (range 5,500 to 43,000), geographical setting (two outer London practices, three urban, six rural town / villages) and socioeconomic status (range third most deprived decile to the least deprived decile).

Each practice identified the 30 most recent expected deaths of patients aged >17 years, who had been living at home or in residential care for at least one day in the last month of life and had died from any cause except trauma, sudden death or suicide. Patients who had previously indicated a wish not to be involved in research were excluded. Patients died between 4 March 2017 and 25 September 2019: one was excluded upon confirming their cause of death after data extraction, leaving a study population of 329 patients.

Data sources and definitions

The electronic GP records and electronic and paper community nursing records of the deceased patients were retrieved and examined between May 2019 and March 2020 by BB, an experienced community and palliative care nurse. Patient demographic and clinical characteristics, documented end-of-life planning discussions and decisions, summary of events in the seven days preceding AM prescribing, recorded prescribing contexts and decision-making, and medication details were entered into a custom-built secure database (Supplemental Document 1). Relevant free text record entries were summarised. Cause and date of death were confirmed from GP practice held death certificate books or England’s General Register Office.

AMs were defined as one or more injectable medications prescribed ahead of need to be administered for symptom control in the last days of life. (7,14)

Data analysis

Data analysis combined quantitative and qualitative analyses in a mixed methods approach. (28,29) Categorical data are reported as frequencies and percentages and continuous data as median (interquartile range: IQR). The sample size of 330 patients was calculated a priori with a statistician to enable statistical analysis including Chi-square and Fisher’s exact test and multivariable logistic regression models. (29). Data analysis was performed using SPSS version 26: P<0.05 is considered statistically significant.

BB undertook qualitative analysis using inductive constant comparison incident-to-incident coding (28) of extracted data from the clinical records of patients prescribed AMs, focusing on end-of-life discussions, AM prescribing contexts and associated patient and family interactions, using NVivo version 12. Patterns and variances in records, typologies of care, and decisions in attributing significance to findings were discussed and refined with KP, SB and both PPI groups. These iterative steps informed the interpretative analysis. (28,33)

Recorded prescribing decision-making

AMs were frequently recorded as being be prescribed as part of a single end of-life planning consultation: for 111/167 (66.5%) of patients prescribed AMs, the prescription (or discussion concerning prescription) occurred during the same consultation when preferred place of death and / or cardiopulmonary resuscitation discussions were first recorded. For 78/167 (46.7%) of patients prescribed AMs, this was when the end of life appeared imminent, with rapid deterioration of physical strength over a few days, escalating symptoms, and reduced ability to eat or drink. In contrast, for 72/167 patients (43.1%), AMs were prescribed as part of longer-term forward planning, when patients had relatively stable physical function, but the focus of care had shifted to end-of-life support. For a further 17/167 (10.2%), AMs were prescribed in the context of clinical uncertainty, when oral antibiotics were prescribed for potentially reversible infections alongside AMs if their condition did not improve:

6 days before death, GP visits the patient and records: “Deterioration, taken to bed, refusing drinks for the last 1-2 days. Comfortable, responding to carers… Unclear if has a urine tract infection (UTI) or this is a pre-terminal event… Tried phoning family but no answer. Plan: Completed DNACPR form with agreement of carers. Issued anticipatory medications and oral antibiotics. Treat for UTI and encourage fluids.” Computer codes: preferred place of care and death is home, patient is ‘aware of prognosis’. [Patient 04014C]

Recorded patient and family involvement in decisions to prescribe AMs were variable: no prescribing conversations were recorded for 69/167 of patients (41.3%). AMs were typically prescribed following consultations involving conversations with patients about their preferences for having end-of-life care at home or their agreement with DNACPR decisions. For a few patients (6/167, 3.6%), it was recorded that they did not want to discuss their prognosis or consider that they were dying at the time of prescribing AMs. These patients lived alone, and prescribing decisions were framed as being in their best interests:

27 days before death, GP visits patient and records: Three recent hospital admissions in the last month for congestive cardiac failure symptom management. “Can barely get out of bed and needing carer visits four times a day… Does not want to discuss their prognosis… States wants active treatment and not wanting to engage in end-of-life care planning… Lives alone … May need to go into palliative care mode swiftly. Plan: best set [prescribe] anticipatory medications now.” Computer codes: preferred place of care and death is home. [Patient 01024D]

Patient and family involvement in decision-making was recorded for 71/167 (42.5%) of patients, the records focussing on whether they agreed with clinician decisions to prescribe AMs: 10/71 patients (6%) were prescribed AMs prior to a visit or phone call to discuss prescribing. Most records concerning AM prescribing conversations were very brief, largely limited to reporting that families had been asked to collect the medications or patient / family agreement with prescribing decisions.

More detailed AM decision-making conversations were recorded when patients or families were concerned about possible symptoms (29/167, 17.4%). In a few cases (5/167, 3%), patient or families were recorded as not agreeing with a decision to prescribe AMs: three patients and one family were resistant to the idea of prescribing AMs, and one patient was “aggrieved” on discovering they had been prescribed AMs without being asked. In these cases, clinicians recorded that it was in the patient’s best interests to have medications available and documented persuading them to accept prescriptions:

3 days before death, GP registrar visits patient at home: “Discussed DNACPR and patient does not want resuscitation. States would like to pass away peacefully… Discussed anticipatory medications. [Patient states] does not want or need any medications. Explained these medications were only for if in distress… [Patient] remained adamant that does not need them… Discussed each group of medication and intended benefit in detail… I advised that they might not need them, and we will of course adhere to their wishes, but we do not want them to suffer… Agreed to [having] them.” [Patient 04025K]

Timing of prescribing

AMs were prescribed between 0 and 1212 days before death. Patients who died from cancer were prescribed AMs a median of 21.5 days before death (IQR 7 to 42 days, range 0 to 375 days); for those who died from non-cancer illnesses, AMs were prescribed a median of 12 days before death (IQR 4 to 47 days, range 1 to 1212 days). Seven patients were prescribed AMs a year or more before death, of which six had a non-cancer diagnosis. The median prescribing timing was 17 days before death across the eleven GP practices, with range of median of 6 to 33 days between the practices. See Table 3.

Clinicians prescribing medications

For 71/167 (42.5%) of patients issued AMs, requests for prescriptions came from clinicians different to the prescriber: specialist palliative care team members (42/167, 25.1%), community nurses (20/167, 12%); GP practice-based paramedics (3/167, 1.8%); care home staff (2/167, 1.2%). GPs in all eleven GP practices (37/167, 22.2%) prescribed AMs following requests from palliative care specialist or community nurse colleagues without recorded contact with the patient or family.

The majority of AMs (127/167, 76%) were prescribed by GPs: other prescribers included hospital doctors (25/167, 15%), nurse prescribers (7/167, 4.2%), out of hours doctors (6/167, 3.6%) and specialist palliative care doctors (2/167, 1.2%).

Symptom control prescribing

The majority of patients (154/167, 92.2%) were prescribed AMs for all five common end-of-life symptoms: pain, breathlessness, nausea and vomiting, agitation and respiratory tract secretions. Similar drugs and dose ranges were prescribed for all five symptoms following end-of-life electronic record template recommendations for 105/167 (62.9%) of patients. See Table 4.

Anticipatory syringe drivers

For 49/167 (29.3%) of patients, a prescription for a continuous subcutaneous infusion of end-of-life care drugs was also issued ahead of need. These “anticipatory syringe drivers” were usually for the same medications as AM injections, often with larger dose ranges. The frequency and timing of anticipatory syringe driver prescriptions varied widely between GP practices, ranging from 1/16 patients (6.3%) to 10/14 patients (71.4%), with prescribing timing a median of 5.5 days before death across the eleven GP practices (range 2 to 27 days). See Table 5.

Strengths and limitations

Caution is needed in interpreting what records can tell us about patient and family participation in prescribing decisions. Records only contain a small part of any clinical encounter, the emphasis frequently being on clinical decisions and prescribing matters. (32) The lack of recorded information on patient understanding of AMs and their preferences is problematic as records are considered authoritative and influence subsequent care decisions. (31,32)

The generalisability of the results is enhanced by the identification of sequential deaths and purposive sampling of GP practices and community nursing services to obtain a maximum diversity sample of team cultures and practices. (28,29,53) Rich descriptions of practice aid understanding and transferability of our results. (53) Our methods enabled detailed qualitative and quantitative analysis of recorded events and their context, which would not have been possible through analysis of the available large primary care datasets. (26,49,54) AMs prescribing data, context and decision-making are not routinely recorded in a way which is systematically retrievable by using electronic algorithms. Consequently, details of care in the body of free text records are often overlooked in large database studies and valuable insights into practice are missed. (26,54)

We collected complete data from patient electronic clinical records: five patient community nursing paper prescription charts were missing. Prescribing events and contexts were confirmable from electronic records, and we present a full data set for all the variables analysed. As clinical records are not designed to collect research data, some patient characteristics were not routinely recorded, such as cohabitation status or perceived risks of opioids being misused or diverted, factors that may influence AM prescribing (20). Data concerning the administration of AMs for these patients will be presented in a forthcoming paper.