Aims

To investigate the preferences of members of the public regarding care, treatment and outcome, at different stages of illness towards the end of life, as decision-making capacity and swallowing ability deteriorate due to progressive neurological disease. This paper focuses on preferences for measures to sustain or to end life.

Fieldwork and recruitment

GB and USA fieldwork and recruitment was undertaken by Ipsos MORI, using previously-recruited panels paid to complete regular surveys and representative of national populations. In GB, face-to-face interviews with people aged 15 or over used computer-assisted technology during January and February 2015. In the USA, an online survey was completed by people aged 18 or over during July 2015. Respondents were informed that the stages would become progressively worse and that they were free to refuse to answer any question. To the best of our knowledge, no survey participants declined to take part in our component of their regular surveys: normal calculations of response rate are not appropriate to public panel surveys of this nature.

Informed consent process

Consent to participate in this survey is a three stage process. Stage (1) when participants are initially recruited to the Ipsos MORI panel, the interviewer explains; the surveys, the process, recruitment, anonymity and withdrawal. They then answer any questions the potential respondent has about the survey and the process. As this is happening, the interviewer assesses whether they are happy that the respondent has the decision-making capacity to proceed with the survey. If they are in any doubt, they do not proceed. If they are happy, they give the potential recruit time to consider and then take verbal consent. Once the respondent has verbally consented, they are asked for their full name (which will not be known by the interviewer until this point). This is then recorded on the handheld electronic CAPI (Computer Assisted Personal Interviewing) device to indicate that the respondent has given their informed consent. For recruitment to the online survey, information about the surveys, the process, recruitment, anonymity and withdrawal is read by the potential recruit on the computer screen. They are then given time to consider, and asked to check a box to indicate they have read and understood the information and agree to take part in an online survey. Stage (2) directly before participating in this questionnaire the respondents are read a statement (or they read it themselves on the screen) which describes: the study topic, the objective, funding, who was conducting the study, and how long it would take. The interviewer then checks verbally that the respondent is happy to proceed, or in the online survey the respondent clicks a box to indicate they are happy to proceed. Stage (3) The process of informed consent is ongoing in this survey. Each question has the option to be declined to be answered, if the participant so wishes. The participant also has the option to withdraw entirely from the survey at any point. For the in-person survey, the interviewer also continues to monitor the participant to check they do not look distressed, if they do they interviewer will give the option to stop the survey.

For young adult respondents, aged 15–17 years, parental or guardian consent was taken as well as the respondent’s own consent. For the US online survey, consent was recorded online after the respondent had read the anonymity and withdrawal information and the study statement. No persons under age 18 years took part in the US online survey. Ethical approval for this procedure was obtained from the University of Cambridge Psychology Research Ethics Committee.

Statistical analysis

Statistical data analysis was undertaken using SPSS version 22 (IBM Corp 2013). Data were weighted for gender, age, education, ethnicity, work status, and regional area to ensure national population representativeness (Table 1). Analyses and reported frequencies are based upon the weighted data set.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Unweighted survey population percentages (N = 2016) and estimated percentage national population data.

https://doi.org/10.1371/journal.pone.0172104.t001

The main focus of the analysis presented here is on response categories 1 (sustaining life at all costs, including using force or deprivation of liberty) and 4 (measures to help die peacefully). For these responses we present: (a) descriptive analysis over the 6 patient stages; (b) differences in the proportions of participants selecting the categories across the 6 stages (Cochran's Q test); and (c) differences in proportions of participants selecting the response categories between two consecutive stages (exact McNemar tests).

Further analysis investigated outlying response groups: those who selected measures to preserve life at all costs (response 1) in stage 6 (end-stage dementia), which could be considered contrary to a person’s best interests and; those who selected measures to die peacefully (response 4) in stages 1, 6 or at any stage. A logistic regression model was fitted to test the relationship between the likelihood of choosing these response categories and the independent variables of gender, country, ethnicity/race, university education, previous personal experience, living with children and age in years. Odds ratios are reported.

Survey population

Responses were obtained from 2016 people, 1025 (50.8%) USA and 991 (49.2%) UK. Unweighted survey data and the estimated representative proportions of UK and USA are shown in Table 1. Representative sampling methods were used in both countries, but USA university education was particularly unrepresentative; 41.3% USA sample had a Bachelor’s degree or higher compared to 29.0% in the US population. Nearly one quarter (23.6%, 476) of respondents lived in a household with children; 41.5% (836) reported previous experience in a professional context or with family members or friends.

Missing data

For each stage, respondents could answer “don’t know” or not answer at all. The missing data for each stage ranged from 1.9% to 2.8%. Combined missing and ‘don’t know’ data are reported.

Response behaviour and patient vignette

Table 2 and Fig 1 summarise responses using weighted data.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Percentage of respondents’ end of life preferences across six stages of the vignette scenario (N = 2016).

https://doi.org/10.1371/journal.pone.0172104.g001

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Cross-tabulation illustrating total respondents’ answers across the six-stage vignette using weighted data (n = 2016).

https://doi.org/10.1371/journal.pone.0172104.t002

Maintain life at any cost, including the use of force and restraints even when resisted

The proportion of respondents choosing this option peaked at 30.2% at Stage 2 (starting to lose short-term memory), dropped to the lowest point of 13.8% at stage 3 (choking on some food and drink), then remained largely stable over the last three stages (Fig 2 below).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Percentage of respondents choosing maintain life at any cost across six stages (N = 2016).

https://doi.org/10.1371/journal.pone.0172104.g002

There was a significant difference in respondents choosing response 1 in all 6 stages (Cochran's Q test, p <0.001), with a significant difference in those choosing response 1 across the first three stages, but not in the last 3 stages (exact McNemar’s tests (p<0.001). The largest increase in those choosing response 1, from 17.7% to 30.2% occurs between stage 1 (living in a care home, to stage 2 (beginning to lose short term memory). The largest decrease, from 30.2% to 13.8%, occurs between stage 2 (beginning to lose short term memory) and stage 3 (choking on food and drink). These changes in proportions choosing response 1 do not differ by country. (S1 Table).

“Measures to help me die peacefully”

The proportions of respondents selecting this option increased significantly between each stage as the person’s condition deteriorated, from 3.9% to 37.0% (Cochran's Q test, p <0.001) and at each stage (exact McNemar's tests, all p <0.001). The largest increase, 10.3% to 23.0%, occurred between stages 3 (some memory loss and choking on food), and 4 (very confused and lost ability to communicate). The second largest increase, 26.6% to 37.0%, occurred between stages 5 (no decision-making capacity and recent aspiration pneumonia) and 6 (end-stage disease, bed bound and asleep). This option was selected on at least one occasion by 44.4% (860) of respondents, and never selected by 55.6% (1079) at any stage (excluding missing cases). (Fig 3 below, S2 Table).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. Percentage of respondents choosing “measures to help me die peacefully” across six stages (N = 2016).

https://doi.org/10.1371/journal.pone.0172104.g003

The logistic regression model gave no evidence that gender, country or university education affected whether respondents chose “measures to help me die peacefully” (p>0.1 for all in S3 Table and Table 3 below). Ethnicity was the strongest predictor, with respondents of “black” race/ethnicity much less likely than those of “white” race/ethnicity to choose this response (OR = 0.45). Living in a household with children also reduced the probability of choosing this response (OR = 0.72), while older participants (OR = 1.09 for each 10-year increase in age) and those with previous personal experience (OR = 1.34) were more likely to choose this response. Re-running the analyses including professional experience showed that in contrast to personal experience, professional experience had no statistically association to the response behaviour, in the context of the other predictors. (S7, S8, S9 and S10 Tables).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 3. Logistic regression for “measures to help me die peacefully” chosen at least once (n = 1834).

https://doi.org/10.1371/journal.pone.0172104.t003

Outlying groups analysis

Among the heterogeneous responses obtained, three groups were of particular interest. Firstly, 16.8% (337) selected interventions to sustain life at all costs including force and restraints, in end stage disease. Logistic regression identified age, ethnicity and living with children as significant predictors: older participants (OR = 0.78, p<0.001), those living with children (OR = 1.49, p<0.01) and respondents who identified as “black” (OR = 2.57, p<0.001). (S4 Table).

Secondly, a small proportion of 3.9% (78) chose “measures to help me die peacefully” in the first stage (living in care home, missing meal times for unknown reasons). Logistic regression identified only country as a significant predicator for this response behaviour: USA respondents were significantly less likely than GB respondents to select this option (OR = 0.35, p < .001). (S5 Table).

Thirdly, 37.0% (745) chose “measures to help me die peacefully” in end-stage disease. Logistic regression identified ethnicity, experience, age and living with children as significant predictors: more likely among older participants (OR = 1.11, p<0.001) and those with past experience (OR = 1.4, p<0.01), and less likely among respondents who did not self-identify as “white” (OR = 0.26, p<0.001) and those living with children (OR = 0.72, p<0.01). (S6 Table).

Heterogeneity of views and the challenge for legislators

This large-scale survey of public opinion found very similar patterns in the views of British and American respondents. One in six (16.8%) expressed a preference for being tube fed when the condition had progressed to near the end of life, a stage where the quality of life available would, to many people, appear minimal. This suggests that a significant minority have a stable moral intuition that life should be supported, even by invasive medical treatment, regardless of the family or medical team’s perception of the individual’s quality of life. This is consistent with a strong, vitalist adherence to the principle of inviolability of human life. On the other hand, over one third (37%) chose “provide measures to help me die peacefully” as their preference for themselves in response to at least one of the stages. This suggests that another, larger, minority hold the view that death would be preferable to being sustained by ANH in these circumstances. This option was kept deliberately broad, to include those opting for a non-interventionist palliative care approach as well as others who would hold that PAD, PAS or VE are morally permissible and desirable under these circumstances.

The challenge for legislators in liberal democracies such as the UK and the US is to enact legal frameworks that represent the views of the majority without unjustly inhibiting the freedom of action of minorities, or putting the lives of people who wish to live at risk.

Valuing life over liberty

A preference for measures to preserve life at all costs, potentially involving deprivation of liberty peaked in response to stage 2. Around 30% of respondents would prefer to be forced to attend mealtimes, if experiencing short-term memory problems. However, half of these respondents would not wish this coercion to continue if the condition progressed such that being fed by mouth would require greater coercion such as the use of physical restraint. These respondents may be expressing a nuanced moral intuition: generally favouring preservation of life over respecting individual choice, particularly for potentially vulnerable people, provided that this can be achieved without resorting to physical force.

Valuing liberty over life

The pattern of respondents’ rejection of deprivation of liberty indicates a widespread moral intuition that it is wrong to use force to interfere with the apparently autonomous choices of people with full decision-making capacity. 82% would not want to be made to attend mealtimes if they lived in a care home. However, only around 70% would not want to be forced to attend meals if they developed short-term memory difficulties, suggesting that the moral intuition to respect individual choice is tempered when there is some question about the patient's ability to appreciate the consequences of their choice, even if their capacity to make decisions remained sufficiently intact for the law to require that those decisions be respected.

86% of respondents would not want to be forcibly fed if they could no longer safely swallow ordinary food, higher than at any other stage. This could indicate that, even when there is a question about a person’s ability to consistently appreciate the consequences of not eating, there is a widespread intuition that nutrition should not be imposed if this requires the use of physical force or seems likely to cause distress, even if the potential benefits are considerable. This is consistent with the principle of 'first do no harm' in the Hippocratic oath, or ‘non-maleficence’ in modern medical ethics.[5] The proportion rejecting coercion (force-feeding or tube-feeding) in response to stages 4–6 remains stable at 82% to 83%. This suggests another widespread moral intuition: that when people patients have lost the ability to make decisions for themselves, preservation of life at all costs is not the aim and life-sustaining treatment should be withheld if the risks and burdens are considered to outweigh the benefits.

Choosing to die

That participants living in households with their children were less likely to choose “measures to help me die peacefully” is consistent with evidence that dependent children reduces suicide risk.[26] Participants self- identifying as belonging to black or minority ethnic groups were also less likely to choose this option, suggesting that moral intuitions regarding care at the end of life, including the permissibility of VE and PAS, are culturally influenced. This introduces additional complexity for legislators in multi-cultural societies, particularly where the ethnic and cultural composition of the legislature or medical profession is different to that of the general population.

Two factors increased the likelihood of respondents expressing a preference for “measures to help me die peacefully”: older age and personal or professional experience of similar illness. This may indicate generational differences in attitudes to end of life care, including VE and PAS, and/or that peoples’ views shift as they witness family and friends ageing and dying. It is of note that in the UK Parliament, the House of Lords (mean age 69 years),[27] voted in favour of legalising AD in January 2015, but the House of Commons (mean age 50 years)[28] voted against in September 2015.

Support for the importance of age as a predictor of attitudes towards PAD comes from a recent cross-sectional survey conducted in the USA in 2015.[29] Periyakoil et al found that older respondents were significantly more supportive of PAD, and that those who reported spiritual or religious beliefs were significantly supportive of PAD. Periyakoil et al also, found that ethnicity and gender were not significant predictors.[29] In contrast, and as discussed above, the findings from this research revealed ethnicity as a strong predictor of choosing ‘measures to help me die peacefully’. Respondents who self-identified as ‘black’ or non-white were less likely to select this option. This difference may be because of the narrower geographical location of respondents in Periyakoil et al (only Hawaii and California) compared with a representative sample of all US states in this research, or it could be because of the different descriptions in each survey–something known to affect outcome in studies on PAD.[30] Periyakoil et al focused solely on PAD, were as the survey instrument in this research examined a broader concept including PAS, PAD and VE.

It appears that most people who would consider death to be preferable to artificially sustained life, including some wishing a form of PAS or VE for themselves, would only prefer this at a late stage of illness, when the quantity and quality of remaining life is limited. The largest increases in preference for “measures to help me die peacefully” occurred at stage 4 (person loses capacity to make decisions about treatment) and stage 6 (condition progresses to terminal phase).

The ‘Oregon model’ of PAD would not be possible for people in either of those situations. As well as the capacity criterion not being met at both stages, in stage 4 the disability criterion (terminal illness) would not be met, while in stage 6, death would be expected to occur naturally within six months, but the person would be unable to self-administer lethal drugs and their only option for hastening death would be euthanasia.

In jurisdictions that permit PAS or VE, their availability depends upon the wording of the disability and capacity criteria. People with incurable cancer commonly retain decision-making capacity until the late stages: they would have the option of PAD under the Oregon model, and PAS or VE under the Dutch model. People with progressive neurological diseases such as multiple sclerosis or motor neurone disease, however, may experience “unbearable suffering” well in advance of the terminal phase. PAS and VE would be available to them in the Netherlands and Belgium, but not the US or Luxembourg. By the terminal phase their physical disabilities would, in all likelihood, prevent them from ending their own life by PAD: VE would be the only option for hastening death, and would be permitted in Luxembourg but not in the US. The risk of permitting PAS but not VE for unbearable suffering is that people would have to choose between using PAS to end their life earlier than they may have wished to or losing altogether the opportunity to hasten their own death in the terminal phase of illness.

People with dementia experience a long and unpredictable illness course: the relatively early loss of decision-making capacity would require either the Swiss model of allowing PAS in early illness, or the Dutch model of Advance Directives. The Swiss approach is not compatible with the wishes expressed by the 37% of study respondents preferring measures to help them die in end-stage dementia, but the Dutch approach is.

From a consequentialist perspective, there is no significant difference in the end-stages of dementia between respecting an Advance Directive to refuse ANH and upholding an Advance Directive to consent to VE. However, both of these approaches are open to the criticism that, by allowing a decision to hasten death made in the past, by someone who does not know what it is like to have dementia, does not extend the reach of self-determination. By privileging that past decision, there is a risk of hastening the death of someone who may still enjoy a reasonable quality of life and wants to continue living. Some argue further, that dementia so profoundly changes personality and sense of self that the person who makes an Advance Directive before the condition progresses is not the same person as the one whose death is hastened when they have lost capacity.[31]

An alternative position, that there is a meaningful distinction to be made between (permissible) Advance Directives to refuse ANH and (impermissible) advance consent to VE, is based upon a moral distinction between acts and omissions. In requiring doctors to respect advance refusals, the law does not compel doctors to act in ways contrary to their professional obligation to help and avoid harming patients. Should the law require doctors to respect advance consent for VE, this would compel a doctor to undertake an act that appears to harm their patient. Arguably, VE may not be harmful, if this is what the patient wants and if it ends, rather than causes, suffering. However this is less clear with VE authorised by Advance Directive, as the person may have changed their mind after developing dementia.

The Oregon model of PAD would permit the provision of help to die for a small number of people, whilst the disability criterion based on terminal illness and the consent criterion based on contemporaneous, capacitous consent limit the potential for harm via the ‘practical slippery-slope’. However, these same safeguards would make help to die unavailable in the situations where, according to our findings, some people may want it. Potentially, this discriminates against people with conditions that limit either their physical ability to self-administer lethal drugs, or their capacity to consent to PAS or euthanasia when they reach the terminal phase of illness.

Advance directives provide a potential solution. Advance decisions to refuse ANH are legally binding, even if ANH is believed to be in the patient’s best interests. The Dutch model could be adapted to protect the choices of other people who want an approach that their physician may disagree with, to be delivered after they have lost decision-making capacity, be that the continuation of ANH or euthanasia.

The limitations of this study are that the survey instrument did not ask people directly about their religious/spiritual views, which other recent research has shown to be an important predictive factor.[29] Another limitation is the lack of exploration of the impact of household income and socioeconomic status of respondents in light of the different healthcare systems in the USA and GB. Further research could investigate these issues in greater depth.