Validation of Collett-Lester ’ s Fear of Death Scale in a sample of nursing students

This study aims to evaluate the psychometric properties of Collett-Lester’s Fear of Death Scale. A sample of 349 nursing students answered Fear of Death and Attitude toward death scales. Content validity was checked by expert review; reliability was proven using Cronbach’s alpha; statistical analysis of the items, correlation between items and construct validity were checked by the correlation of the Scale with the Attitude toward death Scale. The multidimensionality of the scale was reviewed through factor analysis with varimax rotation. The Fear of Death Scale possesses good internal consistency and construct validity, confirmed by the significant correlation with the Attitude toward death Scale. Factor analysis partially supports content validity of the subscale items, but presented a modified multidimensional structure that points towards the reconceptualization of the subscales in this sample.


Introduction
Fear of death cannot be directly observed, which is why it should be inferred based on a subject's conduct or self-reported answers.Therefore, there is no simple criterion with which an instrument can be compared to establish its validity.On the opposite, the construct validity of any measure of attitudes towards death is gradually developed over time, as the evolution of Collett-Lester's Fear toward Death Scale (CLFODS) demonstrates.The Scale was created in 1969 to eliminite the content heterogeneity problem of the items in the scales used to measure fear of death.These authors suggested that fear of death is a multidimensional concept with different possible causes, which can make a person react differently to the idea of death as a state and/or as a process.
Likewise, attitudes and emotional reactions can differ for oneself or for others.Thus, four sub-scales were distinguished: fear of one's own death, fear of other persons' death, fear of one's own dying process and fear of other persons' dying process (1)(2) .
The first version of the CLFODS included 36 items, with a different number of items for each subscale.Next, in 1994, a revised version was published that included the same number of items in each subscale (32 items).In 2003, those items in each subscale that did not contribute to the significance of Cronbach's alpha were eliminted, resulting in a final 28-item version (3) .Also, measuring this construct in the health construct is relevant because the meaning of death entails a certain denial and evasion in our society, which includes health professionals (4)(5) .The availability of valid instruments addressing fears of death permit research that helps to visualize one construct and factor at a time, which is considered determinant for people's end-of-life quality of care (6) and quality of life (7)(8) .

Although countless instruments have been used
to measure attitudes towards death, most of them are one-dimensional.Their disadvantage is that they do not permit the identification and distinction of specific elements involved in fears of death.For the same reason, a trend exists to use well-validated and multidimensional scales.In recent years, Collett-Lester's (2,(9)(10)(11) multidimensional fear of death instrument has been validated in several cultural contexts, evidencing acceptable psychometric characteristics.
Recently, experts have translated the CLFODS from the original to a Spanish version (12) .The psychometric characteristics exhibited based on a sample of nursing students and health professionals were as follows: satisfactory reliability, good internal consistency and, moreover, the extracted component offered considerable support for the factor validity of the CLFODS, as opposed to the first versions of the instrument.Moreover, the Spanish version demonstrated convergent and discriminant validity through a positive correlation with anxiety towards death and general anxiety (12)(13) .
In Chile, some instrument have been used, but these do not display adequate reliability and do not measure the death phenomenon as a state and, in turn, as a process, in a multidimensional way (14) .Therefore, in view of positive psychometric results of the CLFODS in other countries, this research aims to assess the psychometric characteristics of reliability and construct validity of the CLFODS -Spanish version in Chilean nursing students, with a view to the availability of a multidimensional measure that permits valuing fear of death not only for use in nursing, but also in other health professionals.

Method
This study was accomplished at two universities in Concepción, involving 349 nursing students in the first to fifth course year, mainly female (80% ), between 17 and 37 years of age (x=21.3s=2.7).The Fear of death Scale and the Attitude toward death Scale were self-applied before the start of a class.Ethical aspects of this research were guaranteed through review and approval by the Institutional Review Board of the School of Medicine and the Nursing Department Head.Next, students were asked for their written informed consent, which guaranteed confidentiality, privacy and anonymity of answers and offered the possibility to withdraw at any time and/or access to help if specifically asked to the researcher.Finally, the researchers decided not to reveal the name of the colleges the students belonged to, due to the private nature of the research question.
To perform the psychometric assessment of the CLFODS, first, the Scale was submitted to expert review to guarantee the understanding of the items, followed by a pilot test of the instrument, in which 30 participants showed good understanding of the scale.
Criterion and construct validity of the instrument was analyzed involving the definitive sample (15) .Moreover, the reliability of the original scale was assessed, followed by exploratory factor analysis.
Reliability was studied through internal consistency, using Cronbach's alpha coefficient.It was considered that the reagents optimally measured between 0.7 and 0.9 (16)(17) .
The validity of the fear of death construct was analyzed through element statistics, inter-element and element-total scale correlation.Moreover, concurrent criterion validity was reviewed through the correlation between the CLFODS and the Attitude toward death Scale (14) .For the latter analysis, a sample of 133 students from the total sample was used.
Finally, exploratory factor analysis was accomplished, which permitted proving whether the factors and variables that constitute the scale are in line with the pre-established theory of multidimensionality (18) .In this study, principal component analysis (PCA) with varimax rotation was used.

Instruments
The Spanish version of Collett-Lester's Fear of Death Scale comprises four sub-scales that provide multidimensional information on the "Fear of one's own Death", the "Fear of one's own Dying Process", "Fear of www.eerp.usp.br/rlaeRev. Latino-Am.Enfermagem 2011 Sep.-Oct.;19(5):1171-80.other people's Death" and the "Fear of other people's Dying Process".In total, it contains 28 items, grouped in four sub-scales with seven items each.The answers are given on a 1 (nothing) to 5 (much) Likert scale.
Scores are obtained for the total scale and for each subdimension, calculating the average of the respective answers.The highest mean scores indicate greater fear of death or the dying process (18)(19)(20) .
The Attitude Toward Death Scale (ATD) measures the favorable or unfavorable attitude towards death, designed with 43 Likert-type items ranging from 1 to 5.
The highest score corresponds to the most unfavorable attitude towards death (14) .

Results
Univariate analysis of the Fear of death variables reflects a trend near the normal curve.The mean fear of death score among the nursing students was moderatehigh.What the students fear less is their own death.The highest score per sub-scale corresponded to the fear of other people's death (Table 1).

Reliability of the CLFODS
The total internal reliability of the CLFODS corresponded to 0.91, which indicates that 91% of the variability in the obtained scores represents actual differences among people, while 9% reflects random oscillations.Also, for each sub-scale, the obtained Cronbach's alpha coefficients permit guaranteeing that the items or elements are homogeneous and that the scale consistently measures the characteristic for which it was elaborated (Table 2).The correlation matrix, in turn, evidences that, in sub-scale 1: Death of self, item 1 "morir sólo" presents the lowest correlations with items 3, 6, 7. The other correlations of element pairs range between 0.20 and 0.57.Sub-scale 2: Dying of self, shows correlations higher than 0.3, except for item 7, "posibilidad de morir sólo en un hospital lejos de amigos y familiares", which

Construct Validity of the CLFODS
shows a low correlation level with item 1 "degeneración física..", while the rest of the correlations range between 0.20 and 0.67.Sub-scale 3 Death of others displays a low correlation of 0.183 for item 1 "pérdida de una persona querida" with item 6 "sentirse culpable…", while the rest of the correlations vary between 0.20 and 0.57.Sub-scale 4 dying of others shows correlations that vary between more approximate levels, ranging from 0.24 to 0.52.
Going back to Table 2, for the four sub-scales, each of the seven elements is positively correlated inside the respective sub-scale.Element 1 "morir sólo" (0.33), together with element 14 "posibilidad de morir en un hospital…" (0.39), are the elements with the lowest correlation within their respective sub-scales.In turn, it is observed that the reliability coefficient decreases when extracting most elements, except for elements 1, "morir sólo" and 14, "morir en un hospital…", which increase if the element is eliminated.The reliability coefficients continue without significant variations when eliminating these elements though, which is why none of the elements can be eliminated to strengthen the scale.
These results permit guaranteeing that the items or elements are homogeneous and that the four subscales consistently measure the characteristic they were elaborated for.Hence, they are reliable and show construct validity.

Concurrent criterion validity
The concurrent criterion validity between the CLFODS and its sub-scales and the Attitudes toward death Scale, measured through Pearson's correlation coefficient, shows a positive mean correlation (p<0.001).
The correlation between the CLFODS sub-scales and the Attitude toward death Scale ranged between 0.369 and 0.315.A higher correlation coefficient was observed between the total CLFODS and the Attitude toward death Scale (0.431).

Factor analysis
The initial Kaiser-Meyer-Olkin (KMO) test corresponded to 0.903 and the sphericity test showed significance (p>0.000), which guarantees the pertinence of factor analysis.Saturations higher than 0.40 were collected, following the criterion of eigenvalues higher than one.The initial results before the rotation identified five factors, which summarize 54.9% of total data variability.The first factor concentrated 31% of variability.
This first factor revealed ponderations ranging from 0.41 to 0.70 and all in a positive sense, mainly based on the "death of others".
However, to confirm the multidimensionality hypothesis of the scale proposed in theory and to seek the best adjustment, the researchers decided to submit the PCA results to a varimax rotation.
The results produced a five-factor structure.
The first factor (explained variance = 12.6%) can be understood as "fear of other people's death", without item 20, "sentirse sólo por el alivio provocado".The second factor (explained variance = 10.8%) can be understood as "fear of one's own death".The third factor (explained variance = 10.7%) can be understood as "fear of the physical dying process".The fourth factor (explained variance = 10.4%) can be understood as "fear of death and dying according to the psychological meaning".The fifth factor (explained variance = 10,2%) can be understood as "dying of others".
The items that least explain the variability, in turn, within their respective scales, are items 3 (34%) and 16 (32%).The variability of the remaining items ranged between 44% and 70% (Table 4).The reliability and item-element total correlation analysis of these five factors shows that the totalitem correlations exceeded 0.41, except for item 3 (0.37), but this correlation is nevertheless important.
Cronbach's alpha coefficients for each of the five factors investigated all exceeded 0.70, which is considered the minimum accepted reliability level (Table 5).Also, in the five factors, the alpha coefficient decreased if the item was eliminated.

Table 1 -
Descriptive statistics of CLFODS

Table 2 -
Total-element statistics of the 4 CLFODS sub-scales.

mean if the element is eliminated Scale variance if the element is eliminated Corrected element-total correlation Cronbach's Alpha if the element is eliminated
n=349 www.eerp.usp.br/rlae

Table 4 -
Rotated component matrix

Table 5 -
Item-total correlation and Cronbach's alpha reliability of the 5 factors

Factor 1 Factor 2 Factor 3 Factor 4 Factor 5 Item Corr/Elem/total Item Corr/Elem/total Item Corr/Elem/total Item Corr/Elem/total Item Corr/Elem/total
DiscussionThe present study findings demonstrate the adequate psychometric characteristics of the CLFODS in the Chilean population of Nursing students.