The Center for Epidemiologic Studies Depression Scale (CES-D): Is It Suitable for Use with Older Adults?

Abstract With the aim of verifying the suitability of the CES-D scale for use in long-term care institutions for older adults, the CES-D questionnaire was used to collect patient-reported assessments, and two well-known psychometric instruments – the Hospital Anxiety and Depression Scale (HADS) and the Barthel Index of Abilities of Daily Living – were used to collect nurse-reported assessments, based on observations of patients’ behaviours. With regard to possible frequent cases of cognitive impairment and/or insufficient motivation to give sensible responses to CES-D questions, the patient-reported responses were collected from patients during one-on-one sessions with a nurse. The reliability, concurrent validity, and the trustworthiness of the obtained data were supported with proper values of the Cronbach’s alpha coefficient, 0.70 < alpha < 0.85, with significant correlation between CES-D and HADS-Depression, R = 0.50, p < 0.001, and with significant correlation between scores of particular CES-D items vs. final CES-D evaluations of depression, proved by significance p < 0.001 for 18 of 20 CES-D items. These findings supported the effectiveness of the one-on-one session methodology in questionnaire surveys for older adults. The postulation that cases of self-reported depression included somewhat different information about the patient than nurse-reported depression concerning the same patient was supported with the evidence that, in spite of the significant correlation between the Barthel Index and HADS-Depression, R = −0.17, p = 0.016, and in spite of the significant correlation between CES-D and HADS-Depression, the correlation between the Barthel Index and CES-D, equal to R = −0.08 was insignificant at p = 0.244. The findings of this study, considered jointly, support the valuableness of the CES-D scale for use in one-on-one surveys for older adults.


Introduction
The patient is the primary recipient of treatment. Consequently, there is a clear need to recognize and value the patient's satisfaction with all aspects of therapeutic and nursing care. For that reason, physicians and nurses are analyses performed for this population somewhat troublesome, but more interesting from a methodological perspective.

Materials and Methods
At the beginning of the research, in October 2013, all of the present residents receiving long-term care at the nursing home under investigation were taken into consideration. Then, the course of forming the research group was split into two phases -a phase before the questionnaire survey and a phase during the questionnaire survey. The exclusion criteria used as well as other details concerning participants are presented in Tables 1 and 2. The % was computed with respect to initial number (100%) at each group The % was computed with respect to initial number (100%) at each group During the first phase, N = 45 patients were excluded from the initial group of N = 395 patients, and during the second phase, N = 143 patients were excluded from the remaining N = 350 patients. In both phases, the same exclusion criteria were used, but during the first phase, the needed information was obtained from clinical documentation only, whereas during the second phase, information obtained from patients was utilized also.
Five exclusion criteria were applied, namely, the obvious criterion of "death of participant"; "formal incapacitation" (which makes it impossible to give valid consent); direct "disagreement to take part in the survey"; recognised acute mental or oncological disease; recognised difficulties in communication with the participant. Concerning difficulties in communication, here the indirect disagreement to give honest answers was included if any serious doubts were stated by the reviewer (Chmiel et al., 2012). Moderate mental impartment didn't give a direct reason to exclude a participant, based on the known principle: "Not knowing where I am doesn't mean I don't know what I like" (Mozley et al., 1999). Finally, N = 207 patients were included in the study, N = 135 women and N = 72 men.
With the aim of verifying the suitability of the CES-D scale for use with older adults in long-term care institutions, the questionnaire survey was made with the use of three psychometric instruments: the CES-D scale, the Hospital Anxiety and Depression Scale (HADS), and the Barthel Index of Abilities of Daily Living (Bjelland et al., 2002;Kuźmicz et al., 2008). Consequently, the material for the analyses includes three data sets: the set of demographic and clinical data, extracted from clinical documentation, the set of data obtained with use of the three psychometric instrumentsthe CES-D scale, the HADS, and the Barthel Index -and the set of other notes made by nurses during the one-on-one sessions with patients.
The range of the CES-D scale is equal to 0 ≤ CES-D ≤ 60, and the range of the HADS-Depression subscale is equal to 0 ≤ HADS-D ≤ 21, both with greater values for deeper depressive moods. The range of the HADS-Anxiety subscale is equal to 0 ≤ HADS-A ≤ 21, with greater values for deeper anxiety. The range of the Barthel Index is equal to 0 ≤ Barthel ≤ 100, with greater values signifying lower ability to perform everyday activities.
The resulting scores for the Barthel Index and for the two subscales of HADS, that is HADS-Anxiety and HADS-Depression, were calculated simply as usual by summing all of the scales' scores (Bjelland et al., 2002;Kuźmicz et al., 2008). The scoring of the CES-D is somewhat more complicated. The CES-D depression scale includes twenty simple questions. Sixteen questions are articulated in a negative direction, for example, question 6: I felt depressed, or question 14: I felt lonely. However, four of the items are worded in a positive direction, such as question 12: I was happy, or question 16: I enjoyed life, which makes it possible to control for response bias (Górkiewicz, 2014). Responses to the CES-D are based on the frequency of depressive feelings and behaviours during the week prior to administration of the questionnaire. A respondent is asked to choose and indicate only one of four admissible answers to each question. The frequency "Rarely or none of the time (less than 1 day)" is scored with 0 points; the frequency "Some or a little of the time (1-2 days)" is scored with 1 point; "Occasionally or a moderate amount of the time (3-4 days)" is scored with 2 points; and "Most or all of the time (5-7 days)" is scored with 3 points. Therefore, scoring usually takes no more than about 5-10 minutes. The resulting score is computed as the sum of the all twenty scores, but after converting the reverse scores into the direct shape: 0 → 3; 1 → 2; 2 → 1; and 3→ 0.
The statistical analyses performed in this study included descriptive statistics, the basic tests of statistical significance: the chi-square test for distribution, Student's t-test for mean values, and F-test for standard variances, applied respectively to the random variables under consideration. The relationships between variables were analysed using the Pearson correlation, with partial correlation and linear regression. The value less than 0.05 was chosen as the level of significance.
A Cronbach's alpha was calculated to determine internal consistency reliability, and concurrent validity was assessed by analysing the correlation between the CES-D and the HADS-Depression (HADS-D) subscale. In addition, the trustworthiness of the measurements made with the CES-D scale was supported with estimated Pearson coefficients of correlation between each particular item of the CES-D and the general assessment was computed as the sum of all items.
All of the computations are quite straightforward, so they can be made with any spreadsheet software with a proper set of statistical functions, such as Excel for Windows (Górkiewicz et al., 2001). In case of need, useful statistical calculators with proper instructions are readily available on-line, e.g. (Lowry, 2011).

Results
The distribution of age and duration of care is shown in Tables 3 and 4. The chi-square test showed a significant difference between men and women with respect to the age of participants, p < 0.001, and also with respect to duration of care, p = 0.030.  The descriptive statistics of the CES-D, the Barthel Index and both HADS subscales, HADS-Anxiety and HADS-Depression, are shown in Tables 5 and 6.
The chi-square test showed that the distribution of age in the group of women, presented in Table 3, differed significantly from normal distribution, with a mean value of 77.7 and SD = 11.5 (according to estimates presented in Table 5), p < 0.001. The distribution of age in the group of men, presented in Table 4, did not differ significantly from normal distribution, with a mean of 68.9 and SD = 12.2 (according to estimates presented in Table 6), p = 0.314. Analogously, the distributions of all remaining variables, CES-D, Barthel, HADS-A and HADS-D, did not differ significantly    Table 7. It was observed that the women's group differed significantly from the men's group with respect to age, p < 0.001, and Barthel Index, p = 0.002, but did not differ with respect to CES-D, HADS-Anxiety and HADS-Depression, p > 0.05.

HADS-
A series of four linear regression models Y = b 0 + b 1 · Age + b 2 · Gender, where the outcome variable was Y = CES-D, the Barthel Index, and both of the HADS subscales, HADS-Anxiety and HADS-Depression, were estimated. The distributions of the regression errors were evaluated visually, based on the Y = b 0 + b 1 · Age charts. Then, for each model separately, the two null hypotheses were considered -the first hypothesis being that b 1 = 0 and the second hypothesis that b 2 = 0. It was concluded that only at the model: Barthel Index = b 0 + b 1 · Age + b 2 · Gender, should the null hypothesis b 2 = 0 be rejected, p = 0.003. For each of all of the seven remaining hypotheses under consideration, a significance of p > 0.05 was found. The internal consistency of the Barthel Index and the CES-D scale was supported with estimated values of Cronbach's alpha, equal to alpha = 0.848 and alpha = 0.816 for the CES-D, in women's and men's subgroups respectively, and alpha = 0.726 and alpha = 0.741 for the Barthel Index, in women's and men's subgroups respectively.
The trustworthiness of the obtained data was supported with the estimated Pearson coefficients of correlation between scores of each particular item of the CES-D and the general score computed as the sum of all items, for number of pairs N = 207 (Table 8). The concurrent validity of the CES-D and HADS-Depression (HADS-D) subscale was supported with the Pearson coefficient of correlation, R = 0.50, p < 0.001 for N = 207, and also with the coefficient of partial correlation between these scales (with eliminated influence of the Barthel Index), R = 0.50, p = 0.024 (Table 9). The main results were associated with two ascertained relationships between variables considered in this study. The first relationship consisted of a positive correlation between the positively directed items of the CES-D scale and the negatively directed general CES-D score. This relationship is shown in Table 8. The second relationship consisted of a significant correlation between a nurse-reported HADS-D score of a patient's depression and nurse-reported Barthel Index score, in comparison with a lack of correlation between a patient-reported CES-D score of his/her depression and a nurse-reported Barthel Index score. This relationship is shown in Table 9.

Discussion
In this study, self-reported depression, as measured using the CES-D scale with the residents of a nursing home, was confronted with nursereported depression, as measured by the HADS-Depression (HADS-D) scale with the same group of patients. However, it should be noted once again that in general about 20% of those who achieve a high CES-D score in fact do not meet the full psychiatric criteria for major or clinical depression (Counselling Resource Research Staff, 2012). Accordingly, in this study, the CES-D scores, collected using the ground rule: "proper instrument + proper procedure + proper attitude", were considered as an aggregate indicator of patients' well-being. In addition, the Barthel Index was used with the same research group as usual, in a nurse-reported mode.
With regard to possible frequent cases of cognitive impairment and/or insufficient motivation to give sensible responses to CES-D questions, the patient-reported responses were collected from patients during one-on-one sessions with a nurse, who read a single CES-D item step-by-step, in case of need giving some necessary explanation and encouragement, and then listened to the patient's response and recorded it on the CES-D form. It should be emphasized that during a session, the nurse should properly document the course of interaction with a patient (Dijkstra, 2002;Shinkfield et al., 2015). The applied procedure created the best circumstances possible for the respondents to reflect and formulate proper answers to the questionnaire questions (Collins, 2003).
Because one-on-one sessions are very time-consuming and must be carried out by skilled interviewers, the following fundamental questions arose: Is all of the effort made in applying the CES-D superfluous? Does the patientreported CES-D measurement of depression add any valid information to the evaluation made by a nurse with the use of the HADS-D scale?
In this study, the possible valuableness of the CES-D measurements was supported with the main finding that the correlation between the Barthel Index and HADS-D, equal to R = -0.17, was significant at p = 0.016 < 0.05 for N = 207 pairs. However, the correlation between the Barthel Index and CES-D, equal to R = -0.08, was insignificant at p = 0.244 > 0.05 for N = 207 pairs. This finding supported the postulation that, in spite of the significant correlation between the CES-D and HADS-D, R = 0.50, p < 0.001 for N = 207 pairs, self-reported depression included somewhat different information about the patient than nurse-reported depression of the same patient. The partial correlation method made this contrast more expressive, transforming R = -0.17, p = 0.016 into R = -0.145, p = 0.04 and R = -0.08, p = 0.244 into R = -0.002, p = 0.990 (Table 9).
The results obtained by this study correspond with known findings concerning differences in patient and care-provider perceptions (Berlowitz et al., 1995;Chmiel et al., 2010;McCormack, 2004;Ready et al., 2004;Rothwell et al., 1997). Thus, disagreements in patient and care-provider assessments are usual, but they cannot be interpreted on every occasion only in terms of either a patient's trustworthiness or of a care-provider's competency. However, some weighty doubts often arise with respect to patients' competence (Gerrie et al., 2006;Trummer et al., 2006), as well as with respect to the style of communication with patients employed by health care workers (Berry, 2009;Botek, 2015).
Unfortunately, the best possibilities to give proper assessments do not guarantee consistency, validity, and trustworthiness of the obtained data, especially with respect to data obtained in institutional long-term care settings for older adults, who can have cognitive impairment or dementia. The reliability and concurrent validity of the obtained data were supported with proper values of Cronbach's alpha > 0.70 and with a significant correlation between CES-D and HADS-Depression. Concerning trustworthiness of data obtained with the CES-D questionnaire, it is known that in practice, the standard way of confirming an individual's credibility consists of a repeated examination method in which individuals receive the same set of questions on two different occasions (Górkiewicz et al., 2005). Nevertheless, with respect to the CES-D scale, it should be noted that this scale can be used to confirm a patient's credibility on the base of a non-repeated examination, owing to the proper share of opposite questions among the twenty items of the questionnaire (Górkiewicz, 2014). In this study, it was found that among the 20 items of the CES-D scale, the patient-reported scores of 18 items of the CES-D scale (including 3 of the 4 opposite items) were highly significantly correlated with the general score of the CES-D, with a significance level of p < 0.001. This gives clear support for credibility of the obtained data. Only two items of the CES-D scale occurred somewhat confusing for respondents, the details of which are shown in Table 8. Both of these mind-bending items can be wrongly interpreted by respondents in terms of aloneness, isolation, and separation from other people. For example, item 4: "I felt that I was just as good as other people" can give the wrong impression and be interpreted as "I felt like an outsider, like a haughty judge assessing others", and item 19: "I felt that people disliked me" can be misinterpreted to mean "I felt like an outcast, like a recluse". Thus, all of the wavering with these items seems to be quite natural and spontaneous because of older patients' very strong emotional orientation concerning their social environment (Fiori et al., 2006). There were two main limitations of this study. First, the study participants were recruited from a single nursing home only. In addition, only universal psychometric instruments were used. In consequence, specific problems of the residents and of the care providers at the nursing home were omitted.
There are clear prospects for overcoming the above limitations in further research, due to possible cooperation with other long-term care institutions, and with possible use of other psychometric instruments, such as the Polish version of the Newcastle Satisfaction with Nursing Scale (Gutysz-Wojnicka et al., 2007) and/or a Polish version of a scale to measure staff satisfaction with work in elderly care (Donahue et al., 2008;Engström et al., 2006;Peña-Sánchez et al., 2011). Moreover, in our opinion, larger-scale studies on efficient assessment methods are still urgently required to overcome the complexity of the issues present in nursing homes offering long-term care.

Conclusions
In this paper, it was demonstrated that the Center for Epidemiologic Studies Depression Scale (CES-D) can be successfully used to collect patient-reported assessments from older adults in long-term care institutions, during one-on-one sessions with a nurse.
It was proven that patient-reported CES-D measurements of depression add some additional information to evaluations made by nurses with use of the observational scale known as the Hospital Anxiety and Depression Scale (HADS).
The reliability, concurrent validity, and trustworthiness of the obtained data were supported with proper values of Cronbach's alpha coefficient, with significant correlations between CES-D and HADS assessments of depression, and with significant correlations between scores of particular CES-D items vs. final CES-D evaluation of depression.