Overview of the instruments

The majority of studies used existing instruments.13–17 ,20 ,21 ,24 ,26 One of the studies used an instrument developed in the 1960s,26 two studies used instruments developed in the 1970s,15 ,24 two were developed in the 1980s,15 ,17 two in the 1990s14 ,20 and four from 2000 onwards.16 ,21 ,23 ,25 Two studies modified the tool for the purpose of their research either by changing the terminology or the format.25 ,26 Five studies included original measures.16 ,18 ,19 ,22 ,23 The tables give an overview of the research articles (supplementary table 1) and they included measures ranging from more widely used and tested tools (supplementary table 2) to more recent instruments (supplementary table 3).

Indicators

The studies assessed in the review used a variety of indicators. Four of the studies included measures of medical knowledge.13 ,16 ,19 ,23 Attitudes and opinions related to palliative care delivery were also assessed,14 ,16–18 ,22 ,23 ,25 as were perceptions of confidence in dealing with issues related to palliative care delivery20 ,21 and frequency of experience in palliative care delivery.23 A number of studies examined attitudes and emotional reactions to death and dying.14–17 20–22 ,25

Content

Theoretical knowledge was assessed by Arber13 and Kwekkeboom et al16 using the Palliative Care Quiz for Nursing (PCQN).27 Items focused on the philosophy and principles of palliative care, the management of symptoms, and psychosocial and spiritual care of individuals and families. Velayudhan et al19 developed a multiple choice questionnaire for medical students which contained 20 items, and another 15-item version for nursing students on theoretical knowledge, including some psychosocial open-ended questions. Kumar et al23 used the Physical Therapy in Palliative Care—Knowledge, Attitudes, Beliefs and Experiences Scale (PTiPC-KABE Scale)23 to research student physiotherapists. The 37-item self-report measure consisted of quantitative and qualitative data relating to the participants’ perceived knowledge, attitudes, beliefs and experiences of palliative care.

Changes to participants’ confidence in palliative care delivery were measured by Mason and Ellershaw20 ,21 using the Self-Efficacy in Palliative Care (SEPC) Scale.28 The 23-item SEPC contains three distinct subscales (communication, patient management and multi-professional team work). Mavis29 indicated that the self-assessed confidence of medical students correlated with performance on a variety of interventions and competence assessment measures.

Educational studies have demonstrated the importance of positive attitudes as a result of learning.30 In addition to the PCQN,27 Kwekkeboom et al16 included a 12-item scale designed by Bradley et al31 to assess the attitudes of physicians and nurses about care at the end of life. The domains included views about roles and responsibilities of healthcare professionals in caring for terminal patients, the extent to which palliative care provides additional benefits not offered in conventional medical care, views about the role, and importance of clinician–patient communication.

Previous studies demonstrated that students who completed clinical rotations and courses in palliative care expressed more comfort with death and caring for dying patients.32 To address this aspect of effective education, Kwekkeboom et al16 assessed nursing students’ concerns about caring for dying patients using a six-item scale designed by Milton39 representing major areas of concern to nursing students. Participants were asked to rate their degree of comfort when dealing with a dying patient and their family members, their ability to locate resources needed to care for a dying patients, and their ability to handle their own emotions.

Barrere et al,14 Frommelt25 and Mallory17 all used the Frommelt Attitude toward Care of the Dying (FATCOD)33 instrument developed by Frommelt in 1988. Schwartz et al22 explored changes to participants’ attitudes towards death using the Concept of a Good Death Measure.40 The instrument contains 17 descriptive positive statements relating to a ‘good’ death. Based on the research by Walden-Galuszko and associates42 the measure incorporates the concepts of a ‘traditional’ versus a ‘modern’ death, and includes items based on discussions with clinicians and a literature review. The measure assesses three domains: closure, personal control and clinical criteria.

Death anxiety was measured by Mason and Ellershaw20 ,21 using the Thanatophobia Scale (TS)36 to measure attitudes and expected outcomes of providing palliative care. Hegedus et al24 used the Multidimensional Fear of Death Scale (MFODS),43 developed by Neimeyer and Moore in 1994 based on the drawing on Hoelter's34 definition of fear of death: ‘an emotional reaction involving subjective feelings of unpleasantness and concern based on contemplation or anticipation of any several facets related to death’ (Hoelter34 in Hegedus et al,24 page 265). Hurtig and Stewin15 used the Confrontation–Integration of Death Scale (CIDS) developed by Klug.35 CIDS measures two areas of what Hurtig and Stewin15 described as ‘the reconciliation with death construct’: ‘death confrontation’ (contemplation of death), and ‘death integration’ (the positive emotional response to death confrontation)15 (page 31). Personal experience of death was accounted for using an open-ended question.

Schwartz et al22 used the Concerns about Dying (CD) instrument.41 The CD contains 10 descriptive statements designed to assess an individual's comfort level in caring for the dying as well as general concerns about death. The CD is split into three parts: general concerns about death and dying; spirituality; and concerns about working with the dying.

Mooney26 used the revised Collett–Lester Fear of Death Scale,38 originally created in 1969.37 The instrument contains four subscales (seven items each relating to one's own death or the death of others).

Structure

The most common response format was a Likert scale with 12 of the 14 included studies incorporating this format. A 1–5 scale was most frequently used.14 ,16 ,17 ,22–26 Hurtig and Stewin15 included CIDS35 which used a four-point scale in each of two subscales of 10 items (integration factor) and 8 items (confrontation factor) respectively. High scores correspond to a high degree of a factor. Kwekkeboom et al's16 ‘Concerns about caring for dying patients’ questionnaire also recorded responses in a four-point Likert format with 0=‘not at all’ and 4=‘very much so’. Higher scores indicated more concern/worry. The ‘Concept of a Good Death’40 measure used by Schwartz et al22 incorporated a four-point Likert format as well. The measure assessed the perceived essential components at the end of life among a number of dimensions, including spiritual peace, acceptance, freedom from pain, closure, rated along scales ranging from 1 ‘not necessary’ to 4 ‘essential’. The TS37 included in the studies by Mason and Ellershaw20 ,21 recorded responses along a seven-point Likert scale measuring the level of agreement with seven statements of negative attitude towards caring for a dying patient. The ‘Completion Survey’ measure designed by Thompson18 assessed confidence level in dealing with patients who are dying, addressing the level of comfort at the beginning of the educational intervention and at the conclusion of the programme.

The PCQN27 in the studies by Arber13 and Kwekkeboom et al16 used a true/false/don't know format to measure nurses’ knowledge. Knowledge was also assessed by Velayudhan et al19 using a multiple-choice format. Finally, the SEPC28 incorporated by Mason and Ellershaw20 ,21 measured confidence in performing practice-based objectives on a 100 mm visual analogue scale.

Psychometric properties

Of the 14 reviewed articles, two studies omitted validation information for measures and five studies referenced previous validation. The amount of detail in the remaining articles varied. Kumar et al23 reported the test–retest reliability for a pilot of the PTiPC-KABE Scale for 24 participants. In contrast, the SEPC28 and TS37 were rigorously validated in the studies by Mason and Ellershaw.20 ,21 Internal consistency, measured by Cronbach's α, was most often used.14 ,17 ,20–22 The PCQN27 in the Arber13 and Kwekkeboom et al16 studies was assessed for reliability using the Kuder–Richardson (KR-20) formula for dichotomous variables. The CIDS,35 used by Hurtig and Stewin,15 was also assessed using the KR-20. Moderate to good psychometric properties in relation to reliability were reported (coefficients ranging from 0.65 to 0.95). Split-half reliability for the Revised Collett–Lester Fear of Death Scale38 used by Mooney26 ranged from 0.72 to 0.91.38 Test–retest reliability was demonstrated for a majority of the measures incorporated in the studies,13–17 22–26 including the PCQN,27 FATCOD,33 MFODS,38 CIDS,35 Attitudes about Care at the End of Life,31 Revised Collett–Lester Fear of Death Scale,38 PTiPC-KABE Scale,23 Concept of a Good Death40 and CD41 measures. Structural validity was demonstrated through principal component analysis for measures included in three of the studies.16 ,20 ,21 Content validity was assessed for FATCOD and FATCOD Form B,25 ,33 and the ‘Concerns about caring for dying patients’16 measure (Kwekkeboom, personal communication).

Practicality

Of the 14 articles, 13 reported response rates of 60% or greater. The study by Schwartz et al22 reported a response rate of 90% for the inter-clerkship component of the study. In contrast, the response rate for the longitudinal elective component was 53%. Frommelt25 reported the number of people in the control and experimental groups, but not the population. Eight studies incorporated longitudinal designs, each of which were subject to attrition in the post-test component.13 ,14 ,17 ,20–22 ,24 ,26 The explanation for this attrition in the study by Barrere et al14 was the loss of nine students from the traditional programme (13% decrease), five from the accelerated programme (11% decrease) and non-completion of the follow-up questionnaire. Mason and Ellershaw20 ,21 reported a small number of questionnaires returned with incomplete sections, indicative of problems in formulating a response to some items in the 2008 and 2010 studies. No significant demographic differences (eg, gender, previous experience) were reported for this subgroup, although the analyses were not presented. Other issues include a lack of established validity for an instrument as cited by Kumar et al,23 although the research cited good test–retest reliability. Question wording was also an issue in the study by Arber,13 who cited problems with ambiguity in some items found in the PCQN27 within the British context. Mallory17 also noted a limitation of FATCOD33 in its ability to identify all factors (eg, all previous education, all death experiences) that could impact on the participants’ attitudes toward care of the dying. Finally, Barrere et al14 reported limitations in the forced-choice format, preventing any explanations for participant choices.