Affiliation:
1. King’s College London, The School of Life Sciences and Medicine, Cicely Saunders Institute, Department of Palliative Care, Policy and Rehabilitation, London, UK
2. Knowledge & Intelligence (South West) and National End of Life Care Intelligence Network, Public Health England, Grosvenor House, Bristol, UK
3. Life Events and Population Sources Division, Office for National Statistics, Newport, Wales, UK
Abstract
BackgroundPlace of death (PoD) has been used as an outcome measure for end-of-life care. Analysis of variations in PoD can improve understanding about service users’ needs and thus better target health-care services.Objectives(1) To describe PoD in England by demographic, socioeconomic and temporal variables; (2) to determine how much of the variation in PoD can be explained by potential explanatory variables at the area level, and building on this to develop individual-level multivariable regression models; and (3) to evaluate factors associated with PoD and to construct risk assessment models to inform practice.MethodsA population-based study of all registered deaths between 1984 and 2010 in England (n = 13,154,705). The outcome was the PoD. Explanatory variables included age, gender, cause of death (CoD), marital status, year of death, whether or not the death was in a holiday period (Christmas, Easter, New Year), season of death, the location of usual residence and area-level deprivation. The proportion of explained variation in PoD was estimated using the weighted aggregate-level linear regression. Factors associated with PoD were investigated using generalised linear models. The risk assessment models were constructed using the 2006–9 data; the performance was evaluated using the 2010 data.ResultsHospital was the most common PoD in 2001–10 [overall 57.3%; range – cancer 46.1% to chronic obstructive pulmonary disease (COPD) 68.3%], followed by home [overall 19.0%; range – cerebrovascular disease (CBD) 6.7% to cardiovascular disease 27.4%] or care home (overall 17.2%; range – cancer 10.1% to neurological conditions 35.2%), depending on CoD. Over the period, the proportion of hospital deaths for people who died from non-cancer increased (57.1–61.2%) and care home deaths reduced (21.2% down to 20.0%); a reverse pattern was seen for those who died of cancer (hospital: reduced, 48.6–47.3%; care home: increased, 9.3–10.1%). Hospice deaths varied considerably by CoD (range – CBD 0.2% to cancer 17.1%), and increased slightly overall from 4.1% in 1993–2000 to 5.1% in 2001–10. Multivariable analysis found that hospital deaths for all causes combined were more likely for people aged 75+ years [proportion ratios (PRs) 0.863–0.962 vs. aged 25–54 years], those who lived in London (PRs 0.872–0.988 vs. North West), those who were divorced, single and widowed (PRs 0.992–1.001 vs. married), those who lived in more deprived areas (PRs 0.929–1.000 more deprived vs. less deprived) and those who died in autumn, winter or at New Year. We were able to develop risk assessment models but the areas under the receiver operating characteristic curve indicating poor predictive performance, ranging from 0.552 (COPD) to 0.637 (CBD).ConclusionsHospital remains the most common PoD, followed by home and care home. Hospices play an important role for people who died from cancer but little for other diseases. Place of death is strongly associated with the underlying CoD. The variation in PoD by region, age, marital status and area deprivation suggests that inequities exist, which services and clinical commissioning groups could seek to address.FundingThe National Institute for Health Services and Delivery Research programme.
Funder
National Institute for Health Research
Publisher
National Institute for Health Research
Subject
General Economics, Econometrics and Finance