Summary
Regression modelling indicated no statistically significant association between ethnicity and primary palliative care identification and coding, after adjustment for age, sex, frailty, deprivation, and cancer diagnosis. Overall, nearly one-third (32%) of patients had a primary palliative care record before death (n = 64 887). After identification, around one-third of patients survived for up to 90 days, around one-third survived between 91 days and 1 year, and around one-third survived over 1 year. Patients from minority ethnic groups were significantly more likely to have a first record of specialist palliative care activity, compared with patients from White groups. After adjustment, a mixed-effects Cox proportional hazards model (n = 58 681) indicated that, compared with patients from White groups, patients from Asian, Black, African, or Caribbean groups survived significantly longer from first record to death, and patients from ‘Other’ groups had significantly shorter survival times.
This study found that patients from minority ethnic groups do not appear disadvantaged in accessing primary palliative care support before death compared with patients from White groups. However, differences were observed in survival for some minority ethnic patients, and patients from minority ethnic groups were more likely to have a first record of specialist palliative care. Around one-third of patients had a primary palliative care record, which is substantially lower than population-based estimates of palliative care need, potentially reflecting inadequacies in identification and/or coding.
Strengths and limitations
This is the first national study to examine the relationship between ethnicity and primary palliative care identification and coding in England using routine data, which is underutilised in palliative care.49 There are a number of limitations. Those without a recorded ethnicity were excluded, which may have affected the results. Using only two- and five-group ethnicity categorisations may have obscured intragroup variability.61 This study relies on clinical coding. Clinical coding may not always be accurate and complete,62 and there may be additional challenges for clinical coding within primary care.63 However, QOF codes, such as ‘palliative care’ during the period of this study, are more complete than other primary care clinical codes.64 Ethnic group data codes may also be inaccurate or incomplete.61 Clinical codes do not show the full picture of a patient’s care. Information about a patient’s palliative care needs may not be coded, but instead recorded qualitatively in the ‘free text’ of a patient’s medical record. It is unknown who input data. Primary care data may have been input by specialist palliative care teams, which could have affected the results, as coding practices between specialist and primary care may differ. The multilevel logistic regression model accounted for the hierarchical structure of data, with patients nested within GP practices. However, it was not confirmed whether any patients switched GP practices. The eFI57 is a composite measure, combining multiple deficits from electronic health records. Adjusting for eFI could lead to model overfit, which could have affected results. Additionally, the dataset is pre-COVID-19 and primary care practice has since changed in the UK.65,66
Comparison with other literature
After model adjustment, there was no statistically significant association between ethnicity and primary palliative care identification and coding. This finding is in contrast with much of the UK and international literature on ethnicity and specialist palliative care, which has shown ethnic disparities in access to services,43,67–71 prescribing,48,72 advance planning,71,73,74 and use of acute care at the end of life.48,71 The reasons why the findings from this study differ from those observed in specialist palliative care are unknown. It may reflect a real difference in equity of access to care. In the UK, primary care has a universal point of access, and therefore fewer barriers and less gatekeeping. It could also be the result of better recording practices, due to incentivised coding practices within primary care. The recording of ‘palliative care’ patients was previously an incentivised quality indicator, part of the NHS GP contract until April 2025.24,25 However, as discussed, clinical coding in primary care can be challenging and may not always be complete. Additionally, correct identification and coding do not necessarily translate into patient experience. Bajwah et al’s study into specialist palliative care services during the COVID-19 pandemic found services were equal on paper, but inequitable in practice; patients from minority ethnic groups may have experienced adverse impacts such as communication challenges and unmet faith needs.75 The provision of primary palliative care is a complex multi-component intervention, of which, identification and coding are only partial components. Further research is required, including realist methods to understand and evaluate the complex mechanisms of providing equitable high-quality care.
Though there was no association with ethnicity, other sociodemographic and disease factors may be important. While not the primary exposures of interest, the regression modelling indicated some significant associations between primary palliative care and cancer diagnosis, deprivation, frailty, sex, and age. Evidence suggests there may be relationships between these variables and healthcare access;43,76–81 future research could investigate these factors and their relationships to primary palliative care access.
Patients from minority ethnic groups were more likely to have a first primary palliative care record of specialist palliative care. One explanation might be hesitancy to access primary care, in line with UK research showing that patients from minority ethnic groups report poorer experiences of primary care.82 Another possible explanation could be first presentation in hospital, or late disease presentation,83,84 but this is at odds with the current study’s finding that patients from Asian, Black, African, or Caribbean groups survived longer than those from White groups. Rather, these findings reflect those from another UK study, which reported patients from Black, African, Caribbean, and Black British groups had the longest survival between hospice referral and death, compared with other groups.59
A UK hospital data study found ‘Other’ ethnicity was poorly and inconsistently recorded during short periods of urgent care, and suggested patients may be coded as ‘Other’ when their ethnicity is unknown.85 This may be reflected in the findings from this study, in which patients from ‘Other’ ethnic groups had the shortest survival times. This is particularly pertinent as levels of urgent care use at the end of life are high; 61% of people in England have ≥1 emergency admissions in the last year of life.86
This study showed 32% of patients had a primary palliative care record before death; this reflects a smaller-scale study of primary palliative care registration in Scotland, which found 29% of patients had a record.87 These figures are substantially lower than most population-based estimates of palliative care need.31–35 This potential gap could result from factors including lack of resources, inadequate identification tools, or low GP skills and confidence.30,88,89 It may also be the result of inconsistent or poor recording, as a UK survey reported 86% of patients had primary care contact in the months before death.90 It could also result from challenges in identifying need in some communities, where end-of-life discussions are sensitive, such as frail older people, those with non-malignant disease, and people from minority ethnic communities.8,36–38 More research understanding the complex processes of identifying palliative care need within primary care is required. Additionally, patients may receive specialist palliative care elsewhere, such as in a hospital or hospice, which should be communicated to primary care. Research shows hospice and hospital discharge letters frequently lack required information.91,92