INTRODUCTION
This research used the original form of the Geriatric Depression Scale (GDS-30; Yesavage et al., Reference Yesavage, Brink and Rose1982–1983) for two distinct purposes: as the point of comparison in an examination of the psychometric properties of short forms of the GDS and as a means of determining the prevalence of a broad range of nonsomatic symptoms of depression among patients receiving palliative care.
It is important for palliative care clinicians to be able to identify depression in their patients. Depression affects patients’ physical health (Andersen et al., Reference Andersen, Farrar and Golden-Kreutz2004, Reference Andersen, Farrar and Golden-Kreutz2007), quality of life (Lin et al., Reference Lin, Katon and Von Korff2003; Smith et al., Reference Smith, Gomm and Dickens2003), and mortality (Faller & Bulzebruck, Reference Faller and Bulzebruck2002; Meyer et al., Reference Meyer, Sinnot and Seed2003b). It restricts clinicians’ ability to manage patients’ care by its effect on patient compliance (DiMatteo et al., Reference DiMatteo, Lepper and Croghan2000; Lin et al., Reference Lin, Katon and Von Korff2003), the efficacy of treatments for symptoms (Spiegel, Reference Spiegel1996; Passik et al., Reference Passik, Kirsh and Theobald2002), and patients’ desire for death (Kelly et al., Reference Kelly, Burnett and Pelusi2003; Akechi et al., Reference Akechi, Okamura and Sugawara2004). Patient depression also adds to carer burden (Cassileth et al., Reference Cassileth, Lusk and Strouse1985), affects outcomes for health services by precipitating inpatient admissions (Christakis, Reference Christakis1994; Hinton, Reference Hinton1994) and increases treatment costs beyond those due to illness severity (Unutzer et al., Reference Unutzer, Patrick and Simon1997; Sullivan & Dworkin, Reference Sullivan and Dworkin2003). Depression is a particularly pressing problem in palliative care, where it is both prevalent (Chochinov et al., Reference Chochinov, Wilson and Enns1994; Hotopf et al., Reference Hotopf, Chidgey and Addington-Hall2002; Durkin et al., Reference Durkin, Kearney and O'Siorain2003) and long-lasting (Lloyd-Williams & Riddleston, Reference Lloyd-Williams and Riddleston2002; Meyer et al., Reference Meyer, Sinnot and Seed2003b).
Assessment of depression in palliative patients has taken two forms, which reflect two different aims. First, brief screening instruments (Chochinov et al., Reference Chochinov, Wilson and Enns1997; Robinson & Crawford, Reference Robinson and Crawford2005), often consisting of only one or two items, have been developed to quickly identify patients who warrant further assessment or intervention. Second, longer, multi-item instruments and interviews, originally developed for other populations, have been applied in palliative care (Beck et al., Reference Beck, Ward and Mendelson1961; Razavi et al., Reference Razavi, Delvaux and Farvacques1990; Maher et al., Reference Maher, Mackenzie and Young1996; Kramer, Reference Kramer1999; Le Fevre et al., Reference Le Fevre, Devereux and Smith1999) or adapted for use in palliative care (Cox et al., Reference Cox, Holden and Sagovsky1987; Lloyd-Williams et al., Reference Lloyd-Williams, Friedman and Rudd2000). Their aim is to provide insight into the patient's experience and range of symptoms in addition to identifying patients who warrant further assessment or intervention.
The Geriatric Depression Scale (Yesavage et al., 1982–1983) is a multi-item scale that has many characteristics that are desirable in palliative care. It has excellent sensitivity and specificity in aged community (Olin et al., Reference Olin, Schneider and Eaton1992; Sharp & Lipsky, Reference Sharp and Lipsky2002) and primary-care samples (Lyness et al., Reference Lyness, Noel and Cox1997), was specifically designed for use with aged populations, uses simple and consistent response alternatives, focuses on nonsomatic symptoms in order to minimize overdiagnosis in medically ill populations, is available in a wide range of Asian (Liu et al., Reference Liu, Lu and Yu1998) and European languages (Bach et al., Reference Bach, Nikolaus and Oster1996; Baker & Espino, Reference Baker and Espino1997; Clement et al., Reference Clement, Nassif and Leger1997), identifies patients with suicide ideation without direct questioning (Heisel et al., Reference Heisel, Flett and Duberstein2005), and uses both positively and negatively worded questions to avoid “yea-” or “nay-saying” and establishing expectations. Most other multi-item scales used in palliative care do not share these advantages (Endicott, Reference Endicott1984; Chochinov et al., Reference Chochinov, Wilson and Enns1997; Lloyd-Williams et al., Reference Lloyd-Williams, Friedman and Rudd2001, Reference Lloyd-Williams, Dennis and Taylor2004; Lloyd-Williams & Payne, Reference Lloyd-Williams and Payne2003).
The GDS is also of interest because one of its forms has been recommended for routine use by the Royal College of Physicians and the British Geriatrics Society (Dall & Hopkins, Reference Dall and Hopkins1992), and in many countries (Akamatsu et al., Reference Akamatsu, Saito and Wada2005; Wada et al., Reference Wada, Wada and Ishine2005; Riccio et al., Reference Riccio, Solinas and Astara2007) it is the tool of choice for assessing depression during a Comprehensive Geriatric Assessment (CGA; Osterweil, Reference Osterweil2003). One form of the GDS is also a component in the Abbreviated Comprehensive Geriatric Assessment (Mann et al., Reference Mann, Koller, Mann, van der Cammen and Steurer2004; Overcash et al., Reference Overcash, Beckstead and Extermann2005). CGA has been advocated (Balducci, Reference Balducci2003; Wieland & Hirth, Reference Wieland and Hirth2003) and widely adopted in oncology and hematology (Repetto et al., Reference Repetto, Fratino and Audisio2002; Extermann, Reference Extermann2003; Deschler et al., Reference Deschler, de Witte and Mertelsmann2006). Forms of the GDS are also widely used in cancer research (Duffy et al., Reference Duffy, Terrell and Valenstein2002; Chen et al., Reference Chen, Cantor and Meyer2003).
Because the length of the original form of the GDS (GDS-30; Yesavage et al., 1982–1983) precludes its use in many clinical settings, a variety of “short” forms of the scale have been developed. As a result, palliative clinicians interested in using the GDS are faced with a bewildering array of forms from which to choose. These include three different four-item forms (D'Ath et al., Reference D'Ath, Katona and Mullan1994; van Marwijk et al., Reference van Marwijk, Wallace and De Bock1995; Galaria et al., Reference Galaria, Casten and Rovner2000), two different forms containing one item (D'Ath et al., Reference D'Ath, Katona and Mullan1994; Galaria et al., Reference Galaria, Casten and Rovner2000) and five items (Hoyl et al., Reference Hoyl, Alessi and Harker1999; Molloy et al., Reference Molloy, Standish and Dubois2006), forms containing 10 (van Marwijk et al., 1995), 12 (Sutcliffe et al., Reference Sutcliffe, Cordingley and Burns2000), and 15 items (Sheik & Yesavage, Reference Sheik, Yesavage and Brink1986), and an algorithm based on two of these short forms (Weeks et al., Reference Weeks, McGann and Michaels2003). This study informs choices between these alternatives by comparing the psychometric properties of previously published short forms with those of the original GDS in the same palliative population. There are few previous reports of short forms of the GDS being used in patients with advanced disease (Greenberg et al., Reference Greenberg, Lantz and Likourezos2004; Jerant et al., Reference Jerant, Azari and Nesbitt2004; Murtagh et al., Reference Murtagh, Addington-Hall and Higginson2007), and these reveal little about their psychometric properties.
One innovation in the present study is the examination of the psychometric properties of short forms at two points in the trajectory of illness. Patients receiving palliative care often show more marked changes in physical and cognitive function than other medical patients. Thus, the utility of particular symptoms (e.g., fatigue, changes in sleep) in the diagnosis of depression in palliative care patients may change over time. However, few studies have examined the properties of either single- or multi-item screening tools at more than one visit.
The original form of the GDS will also be used to examine the prevalence of nonsomatic symptoms of depression among patients receiving palliative care. Thus far, research on depression in palliative care has given greater attention to identifying effective screening tools (Chochinov et al., Reference Chochinov, Wilson and Enns1997; Urch et al., Reference Urch, Chamberlain and Field1998; Lloyd-Williams et al., Reference Lloyd-Williams, Friedman and Rudd2000; Meyer et al., Reference Meyer, Sinnot and Seed2003a; Robinson & Crawford, Reference Robinson and Crawford2005) and determining the prevalence of depression (Spitzer et al., Reference Spitzer, Endicott and Robins1978; Lynch, Reference Lynch1995; Chochinov et al., Reference Chochinov, Wilson and Enns1997) than to understanding how symptoms associated with depression are expressed in palliative populations. This shortcoming may have important implications for assessment. There is debate about the role that somatic symptoms should play in the diagnosis of depression in palliative patients. Suggestions that diagnosis should focus on items concerning nonsomatic symptoms (Endicott, Reference Endicott1984; Massie & Holland, Reference Massie and Holland1990) assume that these do not show elevated endorsement due to disease processes or treatment. Although little empirical evidence relevant to this assumption is available, this shows that some nonsomatic items are endorsed by most patients receiving palliative care (Sela, Reference Sela2007). This is not surprising because it is well documented that some nonsomatic symptoms (e.g., impaired cognition) may result from disease processes or be side effects of treatment commonly used in palliative care (Bruera et al., Reference Bruera, Miller and McCallion1992; Kamboj et al., Reference Kamboj, Tookman and Jones2005). This research uses the broad coverage of nonsomatic symptoms provided by the GDS-30 to provide descriptive data about the way in which these symptoms are expressed in a palliative population. Although many previous studies have used multi-item scales with patients receiving palliative care, very few (Sela, Reference Sela2007) have reported the frequency of specific symptoms of depression.
In summary, this research had two aims:
1. To compare, at two points in time, the psychometric properties of 10 previously published multi-item short forms of the GDS to those of the GDS-30 in ambulatory patients receiving palliative care.
2. To document the prevalence with which ambulatory patients receiving palliative care report a range of nonsomatic symptoms of depression at two points in time.
METHOD
Participants
One hundred and three patients attending outpatient oncology and palliative care clinics at two teaching hospitals in Adelaide were approached to participate. All patients were fluent in English, over 18 years of age, and judged by their primary medical specialist to be in the palliative phase of their illness, to be sufficiently robust to tolerate a 40-min research interview, and to be free from severe cognitive impairments. Eighty-four patients were recruited (81.6%). Four patients declined to participate (3.9%), and 15 were unable to complete data collection due to physical decline (14.6%).
Usable data were collected on a second clinic visit from 34 of these patients. The mean interval between the first and second data collection points was 35 days and was primarily determined by the patient's clinical needs. In every case, failure to complete the second data collection was due to physical decline or death.
Measures
Data were collected immediately after the patients’ scheduled clinic visit by a research nurse who was not associated with the patients’ care. Three measures were used:
1. Geriatric Depression Scale. The original 30-item form of the GDS (Yesavage et al., 1982–1983) was administered. From this, the 1- (D'Ath et al., Reference D'Ath, Katona and Mullan1994; Almeida & Almeida, Reference Almeida and Almeida1999), 4- (D'Ath et al., Reference D'Ath, Katona and Mullan1994; van Marwijk et al., 1995; Galaria et al., Reference Galaria, Casten and Rovner2000), 5- (Hoyl et al., Reference Hoyl, Alessi and Harker1999; Molloy et al., Reference Molloy, Standish and Dubois2006), 10- (van Marwijk et al., 1995), 12- (Sutcliffe et al., Reference Sutcliffe, Cordingley and Burns2000), and 15-item (Sheik & Yesavage, Reference Sheik, Yesavage and Brink1986) short form scores were calculated (Table 1).
2. The single item relating to depression from the Edmonton Symptom Assessment System (ESAS; Bruera et al., Reference Bruera, Kuehn and Miller1991). This self-report item uses an 11-point numerical analogue format.
3. Custom-designed, single items for self-reported rating of will to live and hope using an 11-point numerical analogue format.
Table 1. Items included in full and short forms of the Geriatric Depression Scale
RESULTS
Psychometric Properties of Short Forms
Six conventional psychometric properties were assessed using the following criteria: a correlation above .75 with the GDS-30 (Table 2); internal consistency above .75 for scales with 10 or more items, above .65 for scales with 5 items, and above .60 for scales with 4 items (Table 3); test–retest reliability similar to that for the GDS-30 (Table 4); convergent validity similar to that shown by the GDS-30 for two related but distinct constructs, will to live and hope (Table 5); and concurrent validity similar to that shown by the GDS-30 for patient ratings of depression on the ESAS (Table 6). Short forms of the GDS containing 10 or more items showed good psychometric properties according to most criteria. In general, scales containing 1, 4, and 5 items showed different patterns of results for different criteria. However, the 4-item scale by D'Ath et al. (Reference D'Ath, Katona and Mullan1994) and the 5-item scale by Molloy et al. (Reference Molloy, Standish and Dubois2006) met as many criteria as the longer forms.
Table 2. Correlations between the full and short forms of the Geriatric Depression Scale at Visits 1 and 2
ar(22).
Table 3. Internal consistency (Cronbach α) for multi-item forms of the Geriatric Depression at Visits 1 and 2
Table 4. Test–retest reliability of forms of the Geriatric Depression Scale
ar(26). br(24). cr(22).
Table 5. Correlation between forms of the Geriatric Depression Scale and ratings of will to live, hope, and depression
*p ≤ 0.5. **p ≤ 0.01. ***p ≤ 0.001.
Table 6. Relative frequency of items from the Geriatric Depression Scale differentially endorsed by patients likely to have severe depression and other patients
*p ≤ .5; **p ≤ .01; ***p ≤ .001.
In addition, the distribution was examined for the short forms with fewer than 10 items because these necessarily yield a restricted range of scores. Ideally, the distribution of scores on brief screening instruments follow a reversed J-curve (i.e., bottom-heavy with a long positive tail) that allows a range of different cut scores to be used in different clinical contexts. Such distributions were produced by the GDS-4 by D'Ath et al. (Reference D'Ath, Katona and Mullan1994) and the GDS-5 by Molloy et al. (Reference Molloy, Standish and Dubois2006) (Table 7).
Table 7. Distribution of scores on GDS short forms with fewer than 10 items
Prevalence of Nonsomatic Symptoms of Depression
Patients did not equally endorse all symptoms of depression included in the GDS-30. The only somatic symptom included in the GDS-30, fatigue (not “full of energy”), was reported by more than half the patients at both visits. However, this was also true for nonsomatic symptoms related to anhedonia (“Dropped many activities and interests”; “Hard to get started on new projects”; “Prefer to stay at home rather than going out and doing new things”) at both visits, and for psychomotor agitation (“Often get restless and fidgety”) at Visit 1. Patients also commonly reported helplessness, hopelessness, that their lives were not exciting, and that their thinking was not clear. In contrast, none of the items relating to depressed affect (“Often feel downhearted and blue”; “Frequently get upset over little things”; “Frequently feel like crying”) were endorsed by more than one third of patients at either visit. Indeed, at both visits more than 80% of patients endorsed items that reflected positive affect (“Basically satisfied with life”; “In good spirits most of the time”; “Feel happy most of the time”; “Think it is wonderful to be alive now”). Worrying about the past was also uncommon.
The results also show the importance of the wording of questions about nonsomatic symptoms. Although the majority of patients endorsed anhedonia items relating to behaviors (e.g., “Dropped many activities and interests”), only a minority endorsed anhedonia items relating to perceptions. That is, only about one third indicated that they were often bored and less than one quarter felt that their lives were empty. Similarly, patients did not respond in the same way to the four items assessing impaired cognition. In particular, more patients rejected the idea that their mind was “as clear as it used to be” than the idea that it was “easy to make decisions.”
Despite the high frequency with which many items were endorsed, few individuals endorsed a large number of symptoms, and therefore most patients did not meet the standard cut scores for the GDS-30 (Yesavage et al., 1982–1983). At Visit 1, 42.3% and 5.1% of patients were identified as likely to be experiencing mild and severe depression, respectively. At Visit 2, 25.9% and 14.8% of patients were identified as likely to be experiencing mild and severe depression, respectively.
Most items on the GDS-30 were endorsed by at least one third of patients at Visit 1 and/or Visit 2. Despite this, 23 of the 30 items discriminated between patients identified as likely to have severe depression and other patients at Visit 1 and/or Visit 2. Ten of these items were the same at Visits 1 and 2 (Table 6). Included among these 10 are the GDS-1 by D'Ath et al. (Reference D'Ath, Katona and Mullan1994), 3 items from the GDS-4 by D'Ath et al. (Reference D'Ath, Katona and Mullan1994), and 4 items from the GDS-5 by Molloy et al. (Reference Molloy, Standish and Dubois2006). The larger number of items included in the GDS-10, GDS-12, and GDS-15 did not significantly improve their overlap with these 10 items.
DISCUSSION
Patient-centered care requires that clinicians have insight into patients’ experience of psychological distress. For this purpose, multi-item scales are superior to single-item and algorithm-based screening tools. A subset of five previously published short forms of the GDS showed good psychometric properties in the current sample (Sheik & Yesavage, Reference Sheik, Yesavage and Brink1986; D'Ath et al., Reference D'Ath, Katona and Mullan1994; van Marwijk et al., 1995; Sutcliffe et al., Reference Sutcliffe, Cordingley and Burns2000; Molloy et al., Reference Molloy, Standish and Dubois2006). Overall, these captured most of the information gained from asking patients the full 30 items, had good internal consistency and adequate test–retest reliability, showed concurrent and convergent validity similar to that of the full scale, and produced distributions of scores that may be of use in clinical contexts. In general, the psychometric properties of the 4-item short form by D'Ath et al. (Reference D'Ath, Katona and Mullan1994) and the 5-item short-form by Molloy et al. (Reference Molloy, Standish and Dubois2006) were similar to those of the three short forms containing 10 or more items (Sheik & Yesavage, Reference Sheik, Yesavage and Brink1986; van Marwijk et al., 1995; Sutcliffe et al., Reference Sutcliffe, Cordingley and Burns2000). The 5-item tool may be of particular interest in palliative contexts. In other settings, it shows good psychometric properties in patients with cognitive impairment (Molloy et al., Reference Molloy, Standish and Dubois2006).
In summary, five short forms of the GDS (D'Ath et al., Reference D'Ath, Katona and Mullan1994; van Marwijk et al., 1995; Shah et al., Reference Shah, Phongsathorn and Bielawska1996; Sutcliffe et al., Reference Sutcliffe, Cordingley and Burns2000; Molloy et al., Reference Molloy, Standish and Dubois2006) hold promise as clinically useful tools in palliative care because they have many psychometric strengths and use simpler and more consistent response alternatives than most other multi-item scales (Zigmond & Snaith, Reference Zigmond and Snaith1983; Cox et al., Reference Cox, Holden and Sagovsky1987). Preferences between these short forms will be influenced by the relative importance assigned to minimizing burden versus understanding patients’ experience, the relevance of their content to the clinical situation, and the results of subsequent investigations examining their validity against a gold standard.
This study also used the GDS-30 to gain insight into the ways in which nonsomatic symptoms associated with depression are expressed in palliative populations. Some nonsomatic symptoms of depression were endorsed by more than half the sample. These primarily related to behavioral indices of anhedonia that were also likely to be influenced by disease progression (e.g., “Dropped many activities and interests”). A similar pattern was reported by Sela (Reference Sela2007), who used a different screening instrument. The high frequency with which patients at both visits endorsed items relating to positive affect and satisfaction with life is noteworthy, given these patients’ prognosis and their symptom burden, and helps to balance the high frequency with which these patients endorsed items relating to loss of activities and interests and lack of energy.
Despite the high level of endorsement of many nonsomatic symptoms, there was little evidence that advanced illness and side effects of treatment led to overdiagnosis of severe depression by the GDS-30. Moreover, most of the items on the GDS-30 helped to discriminate between patients likely to have severe depression and other patients at Visit 1 and/or Visit 2. Thus, those ambulatory palliative patients identified by the GDS-30 as likely to have severe depression report most of the nonsomatic symptoms that are seen in the other geriatric populations on which the GDS was based.
The current research had several strengths, including assessment of the psychometric properties of the scales at more than one point during the trajectory of illness. However, the findings should be interpreted with caution because the sample size was limited, especially at Visit 2, and included only patients well enough to attend an ambulatory out-patient clinic. For obvious practical reasons, scores for the nine short forms of the GDS were derived from the GDS-30 rather than from the independent administration of these scales. The psychometric properties reported here therefore need to be verified when the scales are administered independently. In addition, the effectiveness that these short forms of the GDS (van Marwijk et al., Reference van Marwijk, Arnold and Bonnema1993; D’Ath et al., 1994; Lyness et al., Reference Lyness, Noel and Cox1997; Arthur et al., Reference Arthur, Jagger and Lindesay1999; Molloy et al., Reference Molloy, Standish and Dubois2006) have shown in screening for depression in other populations needs to be verified in palliative contexts by validating them against a gold standard diagnostic test. It will also be important to verify that the high acceptability of these short forms in other populations is also observed among patients receiving palliative care (D'Ath et al., Reference D'Ath, Katona and Mullan1994).
The current study yielded two outcomes. It identified five existing short forms of the GDS (Sheik & Yesavage, Reference Sheik, Yesavage and Brink1986; D'Ath et al., Reference D'Ath, Katona and Mullan1994; van Marwijk et al., 1995; Sutcliffe et al., Reference Sutcliffe, Cordingley and Burns2000; Molloy et al., Reference Molloy, Standish and Dubois2006) that hold the promise of providing insight into patients’ experience of depression while limiting burden on patients and staff. Two of these are sufficiently brief to be included in routine screening. It also showed that items concerning some nonsomatic symptoms of depression are endorsed by many patients receiving palliative care. Despite this, most of the nonsomatic symptoms assessed in this study were helpful in identifying patients with severe depression.
ACKNOWLEDGMENTS
This research was funded by the National Health and Medical Research Council, Australia. We wish to acknowledge the other chief investigators for the “Trajectory of Illness Study,” Prof. Neil Piller, Dr. Roger Hunt, and Prof. Adrian Esterman. This research would not have been possible without the extraordinary contribution to recruitment and data collection of our dedicated research nurse, Ms. Libby Heinsohn-Deer. We also thank Dr. Tim Price and Dr. Roger Hunt for their assistance in recruitment and data collection. Finally, we thank the patients, carers, and staff of Lyell McEwin and The Queen Elizabeth Hospitals, South Australia, for the generous contribution of their precious time and energy.
