As much as possessing the competencies to provide care to treat illnesses and to maintain good quality health and longevity, doctors are expected to possess knowledge, skills and attitudes necessary to provide compassionate and appropriate care at the end-of-life. The need for adequate competencies in providing end-of-life care has been accentuated by the current global pandemic of COVID-19 that is claiming hundreds of thousands of lives irrespective of race, religion or creed.

An initial step in beginning a discussion on end-of-life care is the disclosure of a poor prognosis to the patient. Studies have shown that physicians frequently do not discuss prognosis with the patient or their caregiver, they present fewer facts and less detail concerning prognostic information compared with other types of information, do not adequately explore the patient’s understanding of his/her condition and are reluctant to provide a frank estimate of survival even if the patient requests this information [1,2,3,4,5]. In Sri Lanka, doctors commonly inform the family instead of the patient of a poor outcome of his or her illness despite evidence that patients prefer to be told whether they had a life-limiting illness and to discuss about prognosis and end-of-life decisions [6, 7].

Delivery of appropriate end-of-life care is dependent on whether doctors possess the requisite knowledge and attitudes apropos end-of-life issues, the concept of good death and the principles of medical ethics that guide decision-making in the last stages of the patient’s life which are often emotionally and ethically challenging. Thus, we conducted a study to assess the knowledge on the four principles of medical ethics, attitudes about communicating end-of-life decisions to patients, do-not-attempt cardiopulmonary resuscitation (DNACPR), advance directives, withholding and withdrawing life-sustaining therapy and good death, and their associated factors among doctors working in tertiary care hospitals in Sri Lanka.


A cross-sectional study was conducted among allopathic medical doctors currently working in government tertiary care hospitals (teaching hospitals, provincial hospitals, district general hospitals) in Sri Lanka. At the time of the study, there were 6958 doctors employed in 34 tertiary-care hospitals distributed within the 25 administrative districts of Sri Lanka [8] and included doctors of all categories ranging from senior (consultant, senior registrar/chief resident) to junior (registrar/resident, senior house-officer) personnel. ‘Senior house-officer’ is the period of employment in a clinical setting without enrolment into any postgraduate course. Of the doctors working in tertiary care hospitals, intern house-officers who are at the beginning of their medical career and doctors who had not worked in an in-patient clinical setting (paediatric, medical, obstetrics or surgical wards or intensive care units) during the last one year were excluded from the study.

The sample of doctors to be recruited for the study from tertiary care hospitals was identified from a comprehensive register maintained of post-intern doctors working in government hospitals in Sri Lanka. This list is routinely updated and includes their current workstation and contact details including electronic mail addresses. The required sample size was 322 in order to detect an expected proportion of doctors having knowledge on the principles of medical ethics of 70.1% based on an Indian study among tertiary care doctors [9], alpha error of 1.96 and precision of 0.05. However, a number in excess to the calculated sample size (approximately tenfold of the required sample size, i.e., n = 3400) was randomly selected from the list of tertiary care hospital doctors in the register and invited to participate in the study. This was done because a high non-response rate was anticipated when data collection was not through face-to-face encounters [10] and because we were not aware as to how many on the register would fulfil the eligibility criteria including working in an in-ward clinical setting during the past one year.

Data were collected using a self-administered questionnaire, which included open- and close-ended questions as well as hypothetical clinical scenarios to elicit demographic characteristics, work and training experience, and knowledge and attitudes on end-of-life care, the core principles of medical ethics (autonomy, beneficence, non-maleficence and justice) [11], physician aid-in-dying and good death (Table 1 and Additional File 1: S1). Whether any of the participants had been caregivers for a terminally ill relative during the last one year was also recorded.

Table 1 Questions used in the survey (for the complete questionnaire, please see Additional File)

The questionnaire was developed by an expert group comprising a Neurologist, Physician, Oncologist and a Clinical Psychologist. The questionnaire was pre-tested in a convenience sample of 10 tertiary-care hospital doctors representing the main four clinical specialities (medicine, surgery, paediatrics and obstetrics) who were not recruited for the study proper. The tool underwent judgemental validity to assesses whether the conceptual definitions have been appropriately converted into operational terms or not [12]. For this purpose, face validity was assessed by a non-expert group representing the target population; and content and consensual validity by an independent expert panel of relevant specialities. Each item of the tool was assessed for its relevance, appropriateness of the wording used and acceptability in the local context. Modified Delphi technique was used to achieve consensus among the experts following the ratings given for each item in the questionnaire.

The questionnaire was primarily administered via email using Google forms and the recipients were requested to complete it only if they qualified based on the stated eligibility criteria. This was complemented by sending a link to the Google form via social media and by posting a hard copy of the questionnaire with a self-addressed envelope to return the completed form. All participants were asked to complete the questionnaire only once, using any one of the preferred medium. Three reminders were sent by electronic and social media at two-week intervals, and not returning the form after three months was considered as either not willing to participate in the study or the doctor considering him/herself ineligible to participate in the study. Anonymity of the participants was ensured by not obtaining personal identification details in any of the forms and by using a newly created Google account that was discarded after collection of data. Returning the completed Google form was considered as implied consent of the participants. Prior to data collection, a pilot survey was conducted in a convenient sample of doctors to assess the feasibility and practical issues related to the Google form.

Ethics clearance was obtained from the Ethics Review Committee of the Faculty of Medicine, University of Colombo, Sri Lanka.

Data analysis

Google form data were compiled into an Excel sheet, and all data were analysed using Statistical Package for Social Sciences (SPSS) version 20. The main analysis included data only of the doctors who had not cared for a relative with terminal illness during the preceding year. Data from those with a terminally ill relative were used only to compare the knowledge on the concept of ‘good death’ with that obtained by doctors included in the main analysis.

The key to how responses to the questions were analyzed are given in Additional File 2: S2, which defines the terms used in Fig. 1 and Table 2.

Fig. 1
figure 1

Selected knowledge and attitudes related to end-of-life care among doctors

Table 2 Factors determining knowledge and attitudes in end-of-life care (corresponding question numbers in the questionnaire are denoted as Qx in the left column)

For assessing knowledge on the concept of ‘good death’, participants were tested on 12 previously identified principles of a good death [13]. In addition to the 12 correct principles, 2 incorrect responses were included into the list of options to identify acquiescence bias [14]. One mark was allocated for each correct response, 0 for no answer and minus one mark for each incorrect response, giving a total score range of − 14 to + 14. This scoring system was designed for this study and had not been previously validated.

Data were summarised using mean and standard deviation (SD) for quantitative data, and proportions used for qualitative data. Further, the individual factors associated with knowledge and attitudes relevant to some important aspects of end-of-life care were assessed using univariate analysis, and thereafter in logistic regression analysis using backward LR method. In the models tested on each aspect, adjusted odds ratios (adj. OR) were calculated to ascertain the independent role of sex, religion, postgraduate training, ICU training, in-service/postgraduate training in end-of-life care and work experience (independent variables) on having adequate knowledge and favourable attitudes on the selected aspects (dependent).


The questionnaire was sent as a Google form to 3400 post-intern tertiary care hospital doctors who had email addresses in the register. Of them, 450 doctors who were eligible and consenting for the study responded from all 25 administrative districts in Sri Lanka. The response rate could not be accurately calculated since we were not aware how many of the 3400 that we invited to the study met the eligibility criteria. Among the responders, there were 60 doctors who had been family care-givers for at least one terminally ill first-degree relative and were excluded from the main analysis. The socio-demographic and professional training profile of the doctors in the main analysis (n = 390) is shown in Table 3. Their mean age was 36.5 years (SD = 8.2) with work experience on average of 10.5 years (SD = 8.4). The majority were junior level doctors (SHOs and Registrars/Residents). Three fourths had received formal training in end-of-life care during medical training. Compared to the undergraduate phase (65.6%), only a quarter (27.4%) had received such training at in-service or postgraduate (PG) level. However, only a third felt that they have had adequate training to feel confident in handling issues at the end-of-life. Small group discussions and role play were cited as the most preferred training methods.

Table 3 Characteristics and formal training in end-of-life care among doctors (N = 390)

Figure 1 shows selected knowledge and attitudes related to end-of-life care among doctors in the study while Table 2 shows the factors associated with each of these aspects after adjusting for confounders using regression analysis.

Breaking bad news

With regards to knowledge on breaking bad news, 75.9% of doctors were aware that the diagnosis and prognosis of advanced, progressive or incurable disease should be discussed with both the patient and family (Fig. 1); 3.8% would inform only the patient; 23.1% would inform only the family; and 1% would decide whom to inform based on the individual circumstances.

With regard to their attitudes on breaking bad news, the majority believed that it would reduce the patient’s anxiety associated with uncertainty (73.6%) and that it would enable the patient to adjust (88.5%). However, 84.6% of doctors also believed that it would make the patient depressed and 10.5% felt that it would be of no benefit to inform the patient. 11% felt that it should be left to the family to break bad news to the patient. Overall, 65.9% of the doctors favoured breaking bad news to patients (Fig. 1).

Postgraduate training (aOR = 1.99) and a shorter work experience (aOR = 2.71) were independently associated with adequate knowledge towards breaking bad news (Table 2).

Advance directives (living wills)

Only 27.7% of doctors were aware of advance directives (Fig. 1). This was independently associated with postgraduate training (aOR = 4.15) as well as with end-of-life care training received at any time during their medical career (aOR = 2.34) (Table 2). When asked whether one would transfuse blood in a patient in vascular shock and a low haemoglobin, even if the patient has made an advance refusal of receiving any blood products, 49.5% responded ‘no’ while 47.4% responded ‘yes’. However, 86.4% knew that attempted suicide (deliberate self-harm) should not be considered as an advance refusal of life-saving treatment.

Withdrawing and withholding life sustaining treatment

Only 14.6% were aware that the time of death is the time of certifying brain death in a patient on a ventilator (Fig. 1); 31.8% thought it was the time of disconnection from the ventilator; 18.7% thought it was the time the heart stops; and 6.9% did not know.

With regards to attitudes on life sustaining treatment, 51.5% of doctors responded that they would not place patients with metastatic carcinoma on a ventilator in order to prolong life for a few weeks, while 85.4% had no reservations in disconnecting a brain-dead patient from the ventilator. Overall, 70.3% responded that they felt more comfortable in withholding than withdrawing life sustaining treatment (Fig. 1).

Doctors with postgraduate training were more likely not to place terminally ill patients on ventilators (aOR = 2.77) and to correctly certify the time of death (aOR = 2.37) (Table 2) while ICU trained doctors were less likely to disconnect brain dead patients from ventilators (aOR = 0.49) (Table 2) and were more comfortable withholding than withdrawing ventilation (aOR = 1.99).

DNACPR decisions

61.3% of doctors responded that they were adequately aware of DNACPR decisions (Fig. 1) while 4.4% had never heard of it. When asked about when it would be appropriate to consider a DNACPR decision, one correct response was given by 42.3% of doctors, two by 19.2% and three by only 2.6%. With regard to who makes the decision of DNACPR, 19.5% of the doctors responded that it should be decided by the medical team in an unconscious patient, while 11% of doctors believed that it was the family’s decision and 61.3% believed it to be a shared decision between the medical team and the family. Only 63.6% of the doctors were aware that DNACPR decisions did not imply withdrawal of life-sustaining treatment (Fig. 1), while 15.1% thought that all life-sustaining therapy should be withdrawn once a DNACPR decision was made and 13.6% were unsure whether to withdraw or not. Although 58.7% of doctors claimed to have been involved with the care of patients having a DNACPR decision, 26.7% felt reluctant to consider a DNACPR decision (Fig. 1). The three most cited themes for the reluctance were that it was ‘not humane’, ‘against our religious beliefs’ and ‘the senior doctor and not junior doctors who should make the decision’.

ICU training was independently associated with greater awareness of DNACPR decisions (aOR = 1.76) (Table 2). Having had EOL care training (aOR = 2.48) was independently associated with knowing when to consider DNACPR decisions. Doctors with ICU training were more likely to be aware that DNACPR decisions did not entail withdrawal of life-sustaining treatment while being male (aOR = 1.64) and postgraduate training (aOR = 1.74) were associated with less reluctance in making a DNACPR decision.

Good death

On evaluating the knowledge of the principles of a good death, 55.9% of doctors marked either one or both incorrect responses (items 12 and 13 of question 17 in Table 1) as correct. However, doctors scored a mean of 9.2 (SD = 3.9) of a maximum 14 points. In comparison, doctors who had been caregivers of a terminally ill relative scored significantly higher (mean = 11.2; SD = 2.9) (p = 0.001) (Fig. 2).

Fig. 2
figure 2

Good death knowledge score among doctors who have and have not cared for a relative with terminal illness

61% of doctors believed that it should be the caring physician who should take the lead role in ensuring that the patient has a good death; 17.7% thought it should be the family; 2.3% thought it should be a spiritual leader; 1% thought it should be the nursing staff; and 6.7% thought it should be all of the aforementioned categories.

Doctors with shorter service (aOR = 2.25; 95% CI = 1.37–3.7) and undergone postgraduate training (1.97; 95% CI = 1.23–3.16) were more likely to believe that the caring physician was responsible to ensure a good death at the end-of-life while Buddhists (aOR = 1.74; 95% CI = 1.09–2.76) were more knowledgeable about the principles of a good death independent of all other confounders.

Principles of medical ethics and physician aid-in-dying

Only 17.9% of doctors were able to name all four principles of medical ethics (autonomy, beneficence, non-maleficence and justice), while 57.9% could not name a single. ‘Autonomy’ was the most (34.9%) while ‘justice’ was the least (25.9%) named ethical principle. 52.2% of doctors felt that physician aid-in-dying (which includes ‘physician-assisted suicide’ and ‘euthanasia’) should be legalized in Sri Lanka for patients with incurable, progressive and painful disease. None of the socio-demographic or professional characteristics of doctors were associated with favouring physician aid-in-dying. (analysis not shown).


Discussing end-of-life issues and death with patients and families continues to pose an uneasy challenge to most doctors leading to avoidance of such discussions often leaving patients and families unaware and unprepared, thus, losing the opportunity of a dignified end and a good death. The failure of doctors to discuss and to provide adequate guidance and support at the end-of-life has been attributed to multiple factors including prognostic uncertainty, fear of the impact on patients, navigating patient readiness and feeling inadequately trained for, or unaccustomed to, such discussions [1]. It is intuitive to assume that deficiencies in knowledge and attitudes related to end-of-life care, which stems from inadequate training and education in the formative years of doctors would be a significant contributor to this shortfall. This study explored the magnitude of such deficiencies and the factors that contributed to them in order to identify remediable measures. Despite having a low response rate and limited to a single South Asian population, this study provides some important insights highlighting the need and identifying measures likely to improve provision of end-of-life care, which is an area of medical practice that is often neglected.

In our study, knowledge and attitudes about end-of-life care, the concept of ‘good death’ and the principles of medical ethics among doctors practicing in tertiary care clinical settings in Sri Lanka were found to be lacking. Career progression to postgraduate training, having had in-service training in end-of-life care and training in ICU settings, and a shorter service duration were the most frequent independent determinants of better end-of-life care competencies. Doctors who had been caregivers for a terminally ill relative had a better knowledge of the principles of a ‘good death’.

The majority of doctors (n = 296; 75.6%) in our study were aware that it was necessary to disclose prognostic information, even of terminal illness, and felt that it would reduce the patient’s anxiety associated with uncertainty. However, many doctors also believed that disclosure could cause harm to patients by precipitating depression. This ambiguity probably has led to either the non-disclosure of such information or the practice of disclosing information to the family instead of the patient as was found to be preferred in 23.1% of our study participants and more (50%) in a previous study [7]. Although the reluctance to disclose prognostic information is common [1,2,3,4,5], the preference to disclose to the family instead of the patient may reflect the practice of collective decision making in certain cultures, as in ours, in contrast to the greater emphasis on individual autonomy as seen in Western cultures [1,2,3]. In such collective decision-making cultures, withholding information is ethically rationalised on the principle of beneficence based on the belief that disclosure causes more harm than good.

Advance directives (living wills) are legal documents that ensure that patient preferences regarding health, particularly regarding wishes for life-sustaining treatment, are respected when they are no longer able to make decisions for themselves because of illness or incapacity [15]. Advance directives are not legalised in Sri Lanka and this may explain the reason for its lack of awareness among doctors in our study. However, it is striking that almost half of the participants would transfuse blood in a patient when it’s a life saving measure even in the presence of an advance refusal. This may reflect the greater emphasis on beneficence and the lesser emphasis on the principle of autonomy among Sri Lankan doctors.

Although there is consensus in the medical, ethical, and legal communities that the withholding and withdrawing of life-sustaining treatment at the end-of-life are morally equivalent, there is a growing body of arguments challenging this assertion [16, 17]. However, withdrawal remains more emotionally challenging than withholding life-sustaining treatment as was also seen in our study. Notably, many doctors were not aware of determining the correct time of death when certifying brain death, which precludes justified withdrawal of ventilatory support.

Do-not-resuscitate orders were first reported in 1976 and are made when a patient has declined resuscitation, has a terminal prognosis, when cardiopulmonary resuscitation (CPR) is likely to be non-beneficial or if it is considered that the patient will not survive with sufficient quality of life [18]. In 2005 the term DNACPR replaced DNR to indicate that the order was limited to CPR and did not include pre-arrest life-sustaining interventions [19]. The DNACPR order is made by the caring physician in consultation with other members of the health care team and preferably after discussion with the patient, but the final decision remains the responsibility of the physician [20]. Although there is yet no specific legislature or guidelines on DNACPR in Sri Lanka, it is widely practiced applying the ethical principle of beneficence, when it is considered that administering CPR would not be in the best interest of the patient. In this context, it is significant that a notable proportion of our study participants were not aware of the DNACPR decision (38.7%) and its indications (35.9%).

Thus far, there is no universal definition of a good death. Although twelve guiding principles that define a ‘good death’ have been repeatedly identified in many studies, it is accepted that they should be applied respecting individual diversity in goals of care, perspectives, and preferences [13, 21, 22]. The relatively high knowledge score of the principles of a good death among the study participants is likely to be an overestimate due to a positive marking bias as reflected by over 50% marking the two incorrect responses as correct. As expected, doctors who had been caregivers for a relative with a terminal illness were better informed of the concept of a ‘good death’ [23]. In Sri Lanka, physician aid-in-dying is illegal. However, many doctors felt that it should be legalised for patients with chronic, incurable and progressive diseases resulting in unendurable pain.

In this study, postgraduate training, in-service training in end-of-life care and training in ICU settings were found to be independent determinants of better end-of-life care competencies. However, only about a quarter of our study population had received any form of postgraduate or in-service training in end-of-life care, thus reflecting the deficiencies in knowledge and attitudes related to end-of-life care. Inconsistency of structured end-of-life care curricula as an integral part of medical training, variability in the exposure to care of the dying and lack of opportunities to practice skills under supervision have been identified as universal causes of inadequacy of competencies in end-of-life care [24,25,26].

Studies on the influence of training interventions on patient and family outcomes are scarce [1]. Our study retrospectively identified education and training as independent determinants of better end-of-life care competencies. Future research should prospectively evaluate the impact of interventions such as in-service training and real-life encounters under supervision including palliative care consultations, on competencies in end-of-life care.

Some limitations should be considered when interpreting this study: the study population was limited to tertiary care doctors and the majority were at junior level, there was a wide variability of clinical disciplines from which the doctors were recruited that have variable exposure to end-of-life issues, the response rate to the survey was low and the inherent bias of postal surveys with only the inclined responding. Surveys do not necessarily reflect actual practice, but remains an acceptable approximate measurement. Although the findings may not be generalisable, this study adds important data to the literature of end-of-life care from a predominantly Buddhist Asian population. Furthermore, this study has succeeded in identifying the deficiencies in end-of-life care and through identification of associations, suggest remedial measures that are likely to be effective. The ones who did not return the Google form were either non-eligible doctors or non-responders. Thus, a non-response rate could not be ascertained in this study. Nevertheless, the minimum sample required of 322 was achieved in the study, with 450 returning completed Google forms out of the 3400 who were invited to the study. Previous studies have reported improved response rates with multimodality (postal plus electronic) surveys as replicated in our study and shown that the conclusions drawn from such surveys could be interpreted meaningfully with due recognition of the relatively low response rates [27, 28].


Our study has identified significant deficiencies in knowledge and attitudes related to end-of-life care in Sri Lanka and highlights the need for mandatory training with simulations and real-life encounters under supervision in both undergraduate and in-service or postgraduate medical curricula. Application of the principles of medical ethics and good death should be adapted according to the social and cultural norms unique to the population. Normalising end-of-life discussions by embedding patient preferences at the end-of-life into routine history taking is likely to dissipate the anxiety that surround such discussions and enhance competent provision of end-of-life care.