Participants and data collection
The following inclusion criteria were used: aged 18 to 80 years; self-reported diagnosis of diabetes (Type I or II) or cancer (any type); and signed informed consent. For the sake of generalization, we included two separate etiology groups: diabetes and cancer. While both conditions carry a high risk for depressive symptomatology, they differ considerably. For example, cancer is directly related to the risk of death, and its treatment is often intensive, e.g. surgery or chemotherapy . Diabetes, on the other hand, is not directly related to risk of death, and its primary treatment mostly involves management of glucose levels .
An online survey was carried out from September 2010 through November 2011 using the Unipark software package (http://www.unipark.com). Diabetes patients were recruited through the http://www.dvn.nl site (Diabetes Association Netherlands). Cancer patients were recruited through the following sites: http://www.nfk.nl (Dutch Federation of Cancer Patient Organizations); http://www.borstkankertrefpunt.hyves.nl (forum for people dealing with breast cancer issues); http://www.olijf.nl (network for women with gynecological cancer). The patient associations behind these sites were contacted to inform them about the study and request them to host the survey. Once permission was granted, the web page posted a link to the survey, stating the affiliation and the general purpose of the study. Anyone who visited the site could see the posting; those choosing to follow the link could read more about the purpose of the study and eligibility to participate in it. Those who signed the informed consent could go on to complete the survey.
For the sake of simplicity, we refer to the online participants as the Diabetes and the Cancer groups. An additional group of cancer patients completed a brief version of the survey on paper. We refer to the latter participants as the Cancer-paper group. The reason to have an additional group was to cross-check some of our findings in patients that differed in terms of administration (paper vs. computer) and recruitment method (physician instead of internet). Members of the Cancer-paper group were approached by their physicians at the Radiotherapy Department of the University Medical Center Groningen during the period April to November 2011. They were contacted in the context of an ongoing project of the Health Psychology Section, University Medical Center Groningen. Patients giving their informed consent were sent a paper survey by post.
Participants in the Diabetes and Cancer groups first provided information on their gender, age, years since diagnosis, and the existence of other important medical conditions. Next, the McSad classification system was presented as a self-report health check list. The respondents were asked to identify their level of functioning within the previous week. Finally, they completed depression questionnaires and provided information for other scales. Participants in the Cancer-paper group first answered the personal and disease-related questions and then provided the other information, though only on the McSad and a depression scale.
McSad classification system
The McSad was designed to describe Major Unipolar Depression for valuation purposes. It distinguishes six domains of distinct depressive symptoms, in accordance with the DSM-III . Each one (Emotion, Self-appraisal, Cognition, Physiology, Behavior, and Role function) recognizes four levels of dysfunctioning: no (1), mild (2), moderate (3), and severe (4). The Emotion domain combines symptoms of a blue mood and a loss of interest (example of mild dysfunctioning: “Feel more down (or sad, blue, depressed) than usual and don’t enjoy things as usual”). The Self-appraisal domain concerns how one views the self and the world (example of mild dysfunctioning: “Don’t feel very good about myself these days and often see the down-side of everything”). The Cognition domain describes cognitive performance such as concentration, memory, and decision-making (example of mild dysfunctioning: “Have some trouble concentrating and remembering these days, and it seems harder to make decisions”). The Physiology domain refers to physical symptoms of depression such as sleep, energy, and appetite (example of mild dysfunctioning: “Sleep is quite troublesome these days. Don’t have quite the normal get up and go, and have less of an appetite”). The Behavior domain relates to symptoms of psychomotor agitation/retardation and, in the more severe categories, to suicidal ideation (example of mild dysfunctioning: “Things are more of a chore these days and at times I feel sluggish or agitated”). Finally, the Role function domain addresses performance in work, home, or social settings (example of mild dysfunctioning: “Able to function okay at work, home, school, or with friends but often don’t enjoy what I am doing, and/or feel more withdrawn lately”). The full version of this classification system is available in the publication by Bennett et al. . McSad is a self-report instrument. For each of the six domains, respondents are asked to choose the one level that best describes how they had functioned during the past week. The answers generate a single metric representing a descriptive profile (for example, profile 232322 describes mild dysfunctioning in Emotion, Cognition, Behavior, and Role function and moderate dysfunctioning in Self-appraisal and Physiology). All possible combinations of the four levels in the six domains would generate 4096 (46) unique descriptive profiles of depression states.
For the purposes of the current study, the McSad was translated into Dutch, using established forward-backward translation guidelines [39, 40], i.e., translated into Dutch and then back to English. The Dutch translation was made by a group of researchers at the Health Psychology Section of the University Medical Center Groningen (one professor, two senior researchers, and one research assistant), all of whom are native Dutch speakers and fluent in English, with extensive experience in using depression scales for research among patients with somatic conditions. Then an expert in English with no prior knowledge of the questionnaire was engaged for the back translation. The translated versions were discussed in order to reach consensus on slight differences in wording. Content validity was assessed by the same group; they assessed whether all depression symptoms are covered by the McSad and how consistently the four levels reflect the levels of dysfunctioning in each domain. Given that the McSad was developed on the basis of DSM criteria, we first made a point-by-point comparison of the DSM and the McSad. Furthermore, we compared the McSad with depression scales commonly used among patients with somatic conditions (PHQ-9, CES-D, and HADS). After assessing the content validity, minor changes were made in levels 3 and 4 of the Behavior domain with regard to suicidal tendency, as the way they had been formulated was considered extremely negative.
In order to test the McSad psychometrically, each domain was assigned a score from 1 (no dysfunctioning) to 4 (severe dysfunctioning).
To test construct validity of the McSad, we examined the correlations of the McSad domains (i.e. domain scores) with depression scales (i.e. total scores), as well as with other scales measuring constructs related to depression, that is self-esteem and extraversion. In light of our conceptual framework, based on theory and previous research, we formulated working hypotheses regarding the construct validity of the McSad. Specifically, we hypothesized higher correlations of the McSad domains with the depression scales (convergent validity) than with self-esteem and extraversion scales (divergent validity).
We hypothesized that the aggregate scores for the Cancer and Diabetes groups on all McSad domains would correlate strongly with their total scores on the two depression scales investigated here, i.e., the Center for Epidemiological Studies Depression Scale (CES-D) [41, 42] and the Patient Health Questionnaire (PHQ-9) . We also looked into the correlation between the McSad domains and the CES-D for the Cancer-paper group. As differences in the administration and recruitment method were not expected to exert any influence on the hypothesized relationships of the McSad with the depression scales, we expected to find strong correlations in this group too.
The Patient Health Questionnaire (PHQ-9) is a depression scale consisting of nine items. These correspond to DSM depression criteria such as blue mood and sleep problems. Items are presented as questions about the frequency of depression symptoms within the past two weeks. Answer categories range from 0 (“not at all”) to 3 (“almost every day”). The item scores are summed to calculate the total depression score, which ranges from 0 (no symptoms) to 27 (highest level of depression). The PHQ-9 has been demonstrated to be a reliable and valid instrument for screening for depression and for assessing its severity, also among populations with a background of medical issues. A mean score of 5.08 on the PHQ-9 questionnaire has been reported for primary care patients . The previously validated Dutch version of the PHQ-9 was used here . In the current samples, internal consistency was good, as indicated by high Chronbach’s alphas of .90 and .83 in the Diabetes and Cancer groups, respectively.
The Center for Epidemiological Studies Depression Scale (CES-D) is a validated self-report scale for assessing depressive symptoms. It consists of 20 items representing symptoms of depression. These concern a depressed mood, feelings of guilt and worthlessness, feelings of helplessness, psychomotor retardation, loss of appetite, and sleep disturbance. The scale addresses the frequency of such symptoms within the last week. Items are scored on a four-point scale ranging from 0 (rarely or none of the time, i.e., <1 day) to 3 (most or all of the time, i.e., 5-7 days). Four of the items are phrased in reverse order and recoded accordingly. Item scores are summed to calculate the scale score, which ranges from 0 to 60, with higher scores representing higher levels of depression. The CES-D has also been validated among patients with chronic medical conditions [46, 47]. An average CES-D score of about 12 has been reported for cancer patients under treatment [48, 49]. The validated Dutch version of the scale was used here . In the current samples, internal consistency was good, as indicated by high Chronbach’s alphas of .93, .89, and .85 in the Diabetes and Cancer and Cancer-paper groups, respectively.
In accordance with the literature, we hypothesized that all six McSad domains would show moderate correlations with self-esteem [51, 52] and weak correlations with extraversion . We then examined the correlations of the McSad domains with Rosenberg’s self-esteem scale  and the Eysenck Personality Questionnaire – Extraversion scale .
Rosenberg’s Self-Esteem scale (RSE) is used to assess one’s level of self-esteem. It includes ten items related to self-esteem as reflected by respondents. For example, one item states, “I feel that I have a number of good qualities”. Items are answered on a four-point scale, ranging from 1 (“strongly agree”) to 4 (“strongly disagree”). Five items (1, 2, 4, 6, 7) are worded positively, the other five (3, 5, 8, 9, 10) negatively. Negative items are reverse-coded. A total scale score is calculated by summing the item scores, thus ranging from 10 to 40, with higher scores indicating higher self-esteem. The reliability and validity of this scale have been demonstrated .
The Eysenck Personality Questionnaire – Extraversion scale (EPQ-e) consists of 12 items to assess extraversion. These are posed as questions; for example, “Are you a talkative person?” The participants are asked to answer on a dichotomous scale (yes: 1; no: 0). The total score for the scale is calculated by summing the item scores; thus, the total can range from 0 to 12, with higher scores indicating higher extraversion. The psychometric properties of the EPQ have been established [57–59].
Comparison of the McSad to the EQ-5D classification system
To assess the capacity of the McSad to reflect the range of depression states experienced by patients with somatic conditions, we examined the degree of variability in depression scores that can be explained by the McSad and compared that outcome to the degree of variability in depression scores that can be explained by the EQ-5D CS . We hypothesized that altogether the six McSad domains would account for a large proportion of variability in the depression scores assessed by the CES-D and the PHQ-9. Further, we expected this proportion to be larger than that explained by all the EQ-5D domains.
The EuroQol – 5 Dimensions (EQ-5DTM) is a generic health status measure. It consists of a classification system and a Visual Analogue Scale. The present study used only its classification system, which comprises five dimensions: mobility, self-care, daily activities, pain/discomfort, and anxiety/depression. Respondents are asked to choose the one level out of three (no, moderate, severe problems) that reflects their current functioning. Their scores on each domain are combined to generate descriptive profiles appropriate for valuation. In order to conduct psychometric testing of the EQ-5D CS, we used domain scores ranging from 1 (no problems) to 3 (severe problems).
All the variables were examined for outliers, missing data, and normality. The Diabetes and Cancer groups were compared with respect to background variables by means of t tests and chi-square tests. Similarly, the Cancer-paper group was compared to the Cancer group.
The McSad was inspected to discern the distribution of levels and the number of different profiles identified among the Diabetes and Cancer groups.
Relationships of the McSad domains with the selected scales were assessed by means of a Spearman’s Rho correlation coefficient, as the assumption of normality of distribution was violated. Correlations higher than .50 were considered strong, those less than .30 weak .
A number of multiple linear regression analyses were performed to compare the McSad to the EQ-5D CS concerning their capacity to reflect the range of depression scores. The first regression analysis examined all six McSad domains, treating them as predictor variables. The answer categories were coded as dummy variables and the Enter method was used, while the total score on the CES-D was used as an outcome variable. The second regression analysis was similar to the first except that the PHQ-9 total score was used as an outcome. The third and fourth regression analyses were almost equivalent to the first two, though not quite; instead of using the McSad domains as predictor variables, the EQ-5D domains were used for that purpose.
All analyses were carried out separately for the Diabetes and Cancer groups. Data from the Cancer-paper group on the McSad and the CES-D were used only to cross-check convergent validity findings. Statistical significance was assumed for p < .05. Data were analyzed using SPSS, version 16 (SPSS Inc., Chicago IL).