Identification of Depression: An Examination of the CES-D vs. Single Item
Measures
Linda Nichols, Ph.D.
Abstract
Targeted Journal: Journal of the American Geriatrics Society
Introduction. The identification of depression is an important issue in the care of older persons and a variety of depression scales are used in care and research. In the time constrained environment of ambulatory care, clinicians do not always have time for a multiple-item scale and older patients may fatigue when completing multiple item scales. For quick and simple case finding, ease of use and appropriateness for the population being screened are paramount (Mulrow, et al., 1995; Lyness et al, 1997).
In this paper, using data on 245 Alzheimer’s caregivers, the validity of each of five single-item indicators will be evaluated using clinical cutpoint data for the CES-D as reliable and valid standards. The first will be a Cantril ladder for depression. Cantril self-anchoring scales, consisting of a two sentence inquiry and a nonverbal ladder with 10 points, can be used to measure a respondent’s global perception of an issue, defined by the respondent’s assumptions, perceptions, values and goals, rather than the constructs of the researcher/clinician. In addition, four single items from the General Well-Being Scale (revised) will be examined: I have been in excellent spirits during the past month; During the past month, I felt so depressed that I lost interest in things; Things looked so hopeless this past month that I felt like giving up; Have you felt sad or blue during the past month. These four items are scored using five-point scales.
Background. The two most widely used depression screening instruments are the CES-D and the GDS, with the CES-D being the most widely studied scale (Irwin et al., 1999). In an effort to reduce the burden of administration for both clinicians and patients, several shorter versions of both scales have been developed. Recent articles have looked at the validity and reliability of three different short versions of the CES-D depression scale, including the CESD-10 developed by Andresen et al, 1994 (Boey 1999), and the 11 item Iowa and 10 item Boston versions developed by Kahout, et al., 1993 (Carpenter, et al., 1998; Irwin, et al., 1999). In their evaluation of the CES-D Boston and Iowa forms, Carpenter and colleagues (1998) found that the Boston form which has a yes/no format did not perform as well as the Iowa form which has a 3 item forced-choice format. However, Lyness and colleagues recommended the use of the GDS-Short over the CES-D for older persons due to its simpler yes/no response format. The Cantril ladder may also be simpler for older persons than the four item forced choice version of the CES-D.
Like the CES-D short versions, the 15 item GDS-Short developed by Sheikh, et al., 1986 has been found to have acceptable properties in screening for major depression in older primary care patients (Gerety, et al, 1994; Lyness, et al., 1997). Several researchers have recently explored further reductions in the number of items in the GDS-Short. Hoyl and colleagues (1999) developed a 5 item GDS which was as effective for depression screening as the GDS-S with shorter administration time. To assist primary care practitioners in identifying older persons with depressive symptoms, D’Ath and colleagues (1994) developed three shorter versions, GDS10, GDS4 and GDS1, using a British sample of primary care patients. The GDS1 (Do you feel that your life is empty?) identified 84% of cases identified by the GDS-Short and had sensitivity and specificity of 59% and 75% against a detailed psychiatric interview.
In examining the Boston and Iowa versions of the CES-D, single factor loadings, rather than the four factors originally identified for the CES-D (Radloff, 1977), were found to be the most parsimonious(Kahout, et al., 1993; Carpenter, et al, 1998). This finding suggests that single item screens may correlate well with the CES-D results.
Methods. Baseline data on 245 Alzheimer’s caregivers (58.6% White, 39.8% Black, 78.4% female, 21.6% male) from the Memphis site will be utilized in this study. Measures to be examined are the CES-D and site specific measures of Cantril ladder depression and four single items from the General Well-Being Scale.
We plan to use the point-biserial correlation coefficient, which is expressly designed to measure association between a dichotomy and a polytomous variable. In the study we will use point-biserial correlation coefficients to establish the univariate associations between each of the 5 single-item indicators studied and CES-d data dichotomized at clinical cutpoints. This analysis will provide information about single item indicator validity by documenting the clinical information value of each of the 5 single item scales examined separately, and the square of the parameter estimates will provide analogues to shared or explained variance. We will test for statistical significance between these coefficients to determine if one or more predictors provides more valid information about CES-d cutpoint criterion data.
Frequencies of missing data and distributions of the site specific items will be determined first. Consistency of findings will be confirmed between Blacks and Whites and between males and females.