Self-Report Depression Screening Measures for Older Hispanic/Latin American Adults: A PRISMA Systematic Review

Abstract

Background:

Assessing depression symptoms in Hispanic/Latin American (H/Ls) older adults, a group at high risk for depression, is nuanced due to the influence of cultural characteristics in symptom expression and manifestation. Little is known about the psychometric properties of available measures when used with this population.

Methods:

We conducted a two-stage systematic review of available depression assessment tools. We first identified self-report measures designed for use with adults. We then identified studies where at least one of such measures was used in older H/Ls that reported psychometric properties for the measure(s) used.

Results:

Only 3 measures were identified for use with older H/Ls: the BDI, GDS, and CES-D. However, few data were found to support the validity of the BDI, and the CES-D was not consistently valid across cultural groups. The GDS was found appropriate, though its performance varied based on race/ethnicity, nationality, and cutoff scores. The CES-D and GDS also demonstrated varying psychometric properties based on study setting (research versus clinical) and target population (inpatient psychiatric patients versus community-dwelling individuals).

Limitations:

The number of articles that met criteria for inclusion in our review was small, and there was variation among samples of the few studies included.

Conclusions:

Currently available self-report depression screening measures have acceptable applicability among older H/Ls, but their utility may vary based on their intended use. Modified cutoff scores may be beneficial in maximizing the utility of these measures when given to diverse older adults.

Depression Assessment in Older Hispanic/Latin American Adults: A Systematic Review

The number of individuals that identify as members of a culturally and linguistically diverse (CALD) group has increased from 6.9 million in 2006 to 11.1 million in 2016 and is expected to increase notably by 2030. People that identify as Hispanic or Latin American (H/L) represent the most prevalent CALD group across age groups, including older adults. This population comprised 23% of those age 65 and above in 2016, and by 2030, is projected to comprise 28% of this age group (Administration for Community Living, 2019). Given the expected increase in the older adult population, it is crucial to ensure that the resources to promote the health of these individuals are sufficient and appropriate. Physical and mental health screenings are a fundamental aspect of providing adequate health services, including conditions and disorders for which older individuals are at high risk.

Depression, though less prevalent in older adults than in younger populations, has severe consequences on the well-being of the former. Suicide rates are higher for older than younger individuals and commonly associated with depressive symptoms, which are present in about 15% of community-dwelling older adults (Blazer, 2003). Depression is also associated with symptoms of other cognitive, somatic, and affective disorders in this population, including vascular risk factors and dementia (Hickie et al., 2001; Sweitzer et al., 2002). There is also a high risk for depression for H/L people, particularly among bicultural, immigrant, and less-acculturated individuals (Gerst, 2010; González et al., 2001). Screening for depression is an important prevention strategy to reduce the risk of adverse outcomes related to this disorder in older adults, as it may help detect suicide risk and implement effective treatments that reduce suicidal thoughts (Bruce et al., 2004).

Assessing depression symptoms in the context of aging presents a unique set of challenges, which have been scarcely addressed in research and may bias the clinical evaluation of older adults. Noted issues include limited knowledge of symptom pattern, the inadequacy of tools developed for younger cohorts, the misuse of assessment tools, and deficits in additional assessment (Gonçalves et al., 2009). Depression in older adults is often comorbid with other medical conditions in a complex interplay; this may not always be accounted for in assessment due to factors such as clinical judgement or measurement design. In addition, social and cultural paradigms influence the way in which older adults understand and report their own symptoms. Even with appropriate considerations, the assessment measures used to evaluate older adults should be selected carefully. Although measures like the Geriatric Depression Scale (GDS; Yesavage et al., 1982) are available for this purpose, the studies that have validated them have commonly been conducted with majority non-Hispanic White samples, thus limiting the generalizability and applicability of their findings. A recent systematic review by Balsamo et al. (2018) evaluated the psychometric properties of self-report depression measures among older adults, finding the GDS, the Beck Depression Inventory-II (BDI-II; Beck et al., 1996), and the Center for Epidemiological Studies Depression Scale (CES-D; Radloff, 1977) to be favorable for this purpose; however, little to no demographic data are reported, weakening the generalizability of findings to underrepresented racial and ethnic groups (O’Bryant et al., 2004; Medina et al., 2020).

The assessment of depression symptoms can also represent a challenge in the context of diverse populations. Research has shown that depression is commonly under-recognized among H/Ls in various settings, a tendency partly related to cultural differences in the manifestation and reporting of depressive symptoms listed in currently available depression screening tools (Crockett et al., 2005; MacIntosh & Strickland, 2010). Additional influences include health literacy barriers, somatic presentations, and the use of cultural idioms of distress in depression screening (Lewis-Fernandez et al., 2005). Some studies have examined the psychometric properties of depression measures across multiple ethnic groups, with some finding that specific items (like those related to sleep and appetite disturbances) performed poorly in the detection of depression in these samples (Teresi et al., 2008). Despite this knowledge, little has been done to evaluate the utility of available measures to assess depression symptoms in older H/L adults.

Identifying appropriate assessment tools for older H/L adults is a need yet to be met in research and clinical contexts. The measures that have been validated for use with older adults are not necessarily valid for use with H/Ls, and those that have been validated in H/L populations may be influenced by aging factors when used with older adults. There is a need for critical review of the literature to determine which currently available tests are appropriate for measuring symptoms of depression in this population. Self-report screening measures, in particular, are relevant to evaluate as they represent a crucial aspect of depression assessment (Uher et al., 2012). Addressing this gap may allow for the provision of general recommendations and clinical considerations for use of these tools in the assessment of older H/L adults. Similar work by Wu and Skemp Kelley (2007) provided valuable information for the assessment of depressive symptoms in Chinese older adults, describing variability in the psychometric properties and cutoff scores of available instruments and offering strategies for efficient evaluation of that population, including specific recommended measures. However, to our knowledge, no similar study exists in the context of H/L older adult populations.

The Present Study

We sought to systematically review the literature on depression assessment tools in older Hispanics/Latin Americans. Our search was focused on self-report measures given their widespread use and utility in to assess depressive symptoms in clinical settings (Valente & Saunders, 2005). After identifying commonly used depression measures designed for use with general adult populations, we searched for studies that reported data about the psychometric properties (e.g., reliability, validity) of these materials, using samples that included older H/Ls. Thus, our study had three main aims. We sought to (1) identify depression screening tools available in English and/or Spanish that may be considered appropriate for use with Hispanic/Latin American older adults; (2) examine the psychometric properties of depression screening tools available in English and/or Spanish among older Hispanic/Latin American adults; and (3) provide general guidelines for consideration in the assessment of depression in older H/L adults using screening measures.

Method

We conducted a two-stage systematic review of the available literature following the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines (Moher et al., 2009). Stage 1 consisted of a review of three databases focused on behavioral measurement instruments in health and psychosocial science. The purpose of this search was to compile a list of self-report depression measures available for use with adult populations. Stage 2 consisted of an extended review of three additional databases focused on the reporting of findings related to these identified measures in the psychological and health science literature. The purpose of this search was to identify journal articles in which psychometric data for the self-report depression measures found during Stage 1.

Inclusion/Exclusion Criteria

In Stage 1, we included records for measures that (1) were in English or Spanish, (2) were exclusively self-report, (3) were designed for use with adults, and (4) were designed to measure depressive symptoms. Records for tests that measured constructs similar or related to depression, but not depressive symptoms (e.g., negative affect, worry, rumination), were excluded, as were tests that measured any other clinical disorders in addition to depression (i.e., screening questionnaires for identifying a variety of disorders). Measures that were compilations of two or more existent tests were included if all items in one such measure were designed to measure depressive symptoms. There was no limit in publication date.

In Stage 2, we initially screened studies that (1) were in English or Spanish, (2) were conducted in the United States, Spain, or Latin American countries (3) included Hispanics/Latin Americans (H/Ls), (4) sampled older adult participants, (5) included at least one of the depression screening tools identified in Stage 1, and (6) provided data on the psychometric properties of the depression screening tool used. After conducting an initial search and evaluating results based on these criteria, we identified studies that qualified for secondary screening. We screened these articles and applied an additional set of criteria to detect those with samples that were representative of the H/L older adult population; thus, we selected studies in which (1) at least one fifth of participants were H/L and (2) all participants were adults of age 55 or older. The required minimum of H/L participants in studies was determined as a standard of representativeness, with the consideration that H/Ls comprise over 18% of the general population and over 20% of the older adult population in the United States (United States Census, 2016). As in Stage 1, there was no limit in publication date. We excluded unpublished manuscripts (e.g., theses/dissertations), editorial and opinion pieces, systematic reviews or meta-analyses, books, book chapters, and articles from journals that were not peer-reviewed.

Information Sources

We used three academic databases in Stage 1 selected for their focus on behavioral measurement instruments in psychological and health sciences: Health and Psychological Instruments (HAPI), APA PsycTests, and Mental Measurements Yearbook with Tests in Print (MMYTP). We used three additional databases in Stage 2: PsycInfo, PubMed, and MedicLatina. These databases were selected given their offering of scholarly articles from peer-reviewed journals, and the availability of materials covering a wide range of subjects in psychology, including psychometrics.

Search

The process in all three Stage 1 databases included four criteria sets: (1) measure subject (i.e., depression-related keywords), (2) test modality (i.e. “self-report” or “self report”), (3) language (“English” or “Spanish”), and (4) age. We identified appropriate depression-related keywords for each database using their respective lists of index terms. Given our interest in identifying measures designed for use with general adult populations, our age criteria did not specifically request results involving older adults; instead, it discarded all results involving children and/or adolescent samples. The search excluded records containing the terms “children,” “adolescents,” “youth,” “child,” “teenager,” “pediatric,” and “kids.”

The Stage 2 process in all three databases included four criteria sets: (1) specific depression measure identified in Stage 1 (e.g., “Beck Depression Inventory, or BDI”), (2) age, (3) language/ethnicity, and (4) type of data reported. Table 1 displays the terms used for search criteria 2 through 4; these terms were consistently used in combination with each depression measure identified in Stage 2.

Table 1.

Index terms used for Stage 2 Search

Age		Language/Ethnicity		Required data
older adults or elderly or geriatric or geriatrics or aging or senior or seniors or older people or age 65 or 65+	AND	Hispanics or latinos or latinas or latinx or chicano or mexican american or spanish speakers	AND	psychometrics or validity or reliability or psychometric properties or sensitivity or specificity

Data Collection Process

For each record identified in Stage 1, we documented the database of origin, assessment measure, citation for the measure, language, and reason for discarding if it did not meet inclusion criteria. Records obtained through the Stage 2 database search were similarly recorded to document the assessment measure, title and citation for the article, and a description of the sample used; these records were then reviewed to assess their initial eligibility. Studies that met inclusion criteria at this point were reviewed further to determine their final eligibility. Studies that met inclusion criteria were reviewed for available psychometric data and diagnostic cutoff scores used by authors to diagnose depression with the measures. Records were originally obtained by the lead author (AOL) and then verified in collaboration with the second author (MNM).

Data Items

The following sample characteristics were collected when available from qualifying studies: number of subjects, clinical or nonclinical categorization, source of recruitment, gender proportion, age range, racial/ethnic group distribution, socioeconomic status, and language(s) spoken. We also recorded the following psychometric data when available: reliability and validity statistics, sensitivity and specificity, positive and negative predictive value, and other psychometric qualities (e.g., results of factor structure analyses, measurement bias).

Risk of Bias in Individual Studies/Across Studies

Outcome reporting bias within eligible studies was reported qualitatively. Table 2 displays study characteristics that may implicate risk for bias in eligible studies, as suggested by PRISMA guidelines (Moher et al., 2009). We did not generate mathematical estimates of risk for bias across study as we did not access study data beyond that which was reported in the reviewed manuscripts.

Table 2.

Reported sources of bias within studies

Article	Race/ethnicity differences	Sex differences	Age differences	Diagnosis differences	Language differences	Location differences
Baker & Espino 1997				No
Baker et al., 1993
Baker et al., 1994					No
Carrete et al., 2001				Yes
Espino et al., 1996	No	No		No
Gatewood-Colwell et al., 1989	No	No
Gonzalez et al., 1995						No
Hahn et al., 2011		Yes		Yes
MacIntosh & Strickland, 2010	Yes
Mahard, 1988
Robinson et al., 2002		Yes		Yes

Note: “Yes” means the study reported significant differences between groups based on a given factor. “No” means the study reported NO significant differences between groups based on a given factor. Empty cells signify a lack of statement in the study about significant differences based on a given factor.

Results

Study Selection

The Stage 1 search yielded a total of 207 results, of which 53 self-report depression measures were identified (Figure 1). This final list of measures (see Table S1, Supplementary Materials) was examined further in Stage 2.

Figure 1.

PRISMA diagram for Stage 1 Literature Search

The Stage 2 search yielded a total of 383 records, of which 36 articles were eligible for further examination (Figure 2). Only 11 of these records, which collectively employed 3 different measures (Table 3), included H/L older adult samples and were reviewed further for psychometric data.

Figure 2.

PRISMA diagram for Stage 2 Literature Search

Table 3.

Characteristics of depression screening tools used across included studies

Measure	No. of items	Score range	Type of questions	Time to complete	Cut-off score	Score meaning
BDI	21	0–63	0 to 3 points	5 to 10 minutes	10	0 to 9 = no depression, 10 to 15 = mild depression, 16 to 19 = mild to moderate depression, 20 to 29 = moderate to severe depression, 30 to 63 = severe depression
CES-D	20	0–60	0 to 3 points	10 to 15 minutes	16	Higher scores are indicative of higher depression
GDS-15	15	0–30	Yes/No	5 to 10 minutes	11	0 to 4 = no depressive symptoms, 5–8 = mildly depressed, 12–15 = severely depressed
GDS-30	30	0–30	Yes/No	10 to 15 minutes	11	0 to 9 = no depressive symptoms, 10 to 19 = mildly depressed, 20–30 = severely depressed

Note. BDI = Beck Depression Inventory, original version; CES-D = Center for Epidemiological Studies Depression Scale; GDS-15 = Geriatric Depression Scale, 15-item version; GDS-30 = Geriatric Depression Scale, 30-item version.

Study Characteristics

Table 4 summarizes the final list of studies included, as well as relevant descriptive characteristics and psychometric properties of depression measures.

Table 4.

Psychometric properties of depression screening measures reported in studies representative of Hispanic/Latin American older adults

Journal Article	Measure and Setting (Clinical/Research)	Cutoff score	Psychometrics	Descriptive demographics
Baker & Espino, 1997	GDS-15, Clinical setting	0–5 = normal; 6–10 = mildly suggestive of depression; 11–15 = strongly suggestive of depression. Also used a revised cutoff of 4 or above.	Cronbach’s alpha: 0.94. Sensitivity: Using original cut-scores, 39% in patients with MDD and 77% in patients with ODD. Using revised cut-score, 75% for patients with MDD and 85% for patients with ODD. Specificity: Not determined due to lack of true negatives in sample.	41 Mexican American psychiatric patients age 60–99 with a diagnosis of depressive disorders, 43.9% women.
Baker et al., 1993	GDS-15, Research Setting	6	Sensitivity: 35% for African Americans and 49% for Mexican Americans.	55 African American community residents (mean age 77 years) and 41 Mexican American people (age 65+ years) with a diagnosis of affective illness.
Baker et al., 1994	GDS-15, Clinical setting	6 and 4	Sensitivity: 49% using the original cutoff score of 6 or more. With a revised cutoff of 4, sensitivity improved to 75% for the combined sample of Black and Mexican American participants.	55 Black community-dwelling and 57 Mexican American psychiatric patients. The Mexican American sample, age 62–98, was recruited from the psychiatric unit; hospitalized and depressed, 44% women.
Carrete et al., 2001	GDS-30, face-to-face version (GDS-P) and phone version (GDS-T), Research setting	10 and 11	Cronbach’s alpha: 0.85 for GDS-P and 0.88 for GDS-T. Sensitivity: 88% for GDS-P and 84% for GDS-T. Specificity: 82% for GDS-P and 79% for GDS-T.	Random sample of 282 Hispanic elderly individuals, mean age 71.7 years, 56% women, 50% had college or higher education.
Espino et al., 1996	GDS-30, Clinical setting	11	Sensitivity: 0.80 Specificity: 0.61	66 Mexican American participants age 65 or older, 48% women; recruited from private or public community healthcare clinics.
Gatewood-Colwell et al., 1989	BDI, original version, Research setting	16	Cronbach’s alpha: 0.80 Pearson correlation with GDS: 0.79.	51 participants age 60–80 recruited from a community senior service agency. 15 white women, 11 White men, 13 bilingual Mexican American women, & 12 Mexican American men.
Gonzalez et al., 1995	CES-D, Original and Modified versions, Clinical setting	16	Cronbach’s alpha: Original CES-D: 0.90 for the Scaling study, 0.89 for the Replication study, and 0.93 for the Test-Retest study. Modified CES-D: 0.92, 0.91, and 0.92, respectively. Items that did not meet convergent validity criteria: Item 4, “I felt I was just as good as other people;” correlation of −0.11 with the scale. Item 8, “I felt hopeful about the future;” correlation of 0.35 with the scale.	472 older Hispanic/Latin American adult patients from San Francisco, Los Angeles, San Antonio, Miami, New York City, and Caracas, Venezuela. All recruited from rheumatology clinics. Predominantly women, with a mean 8 years of education.
Hahn et al., 2011	CES-D, Research setting	16	Cronbach’s alpha: 0.73 for baseline data and 0.79 for the second wave of participants.	Mexican American, community-dwelling adults age 65 and over, new to caregiving roles, recruited from Texas, New Mexico, Colorado, California, or Arizona. 59.4% female, mean years of education 4.87. Interviewed in English or Spanish.
MacIntosh & Strickland, 2010	CES-D, Research setting	16	Item bias for overall sample was found for one positive affect question in the 10-item subset. Among the Hispanic sub-population, there was item response bias for the positive affect items across time.	4499 adults age 70–85, 58% Hispanic and 42% non-Hispanic White, 56.7% female
Mahard, 1988	CES-D, Clinical and Research settings	16	Cronbach’s alpha: 0.87 in both patient and nonpatient samples.	60 Puerto Ricans age 55–82 living in New York City in the Summer of 1984. Half diagnosed clinically as depressed and receiving treatment in a community mental health clinic, other half community-dwelling adults. 80% female, predominantly low education (median of 4 years) and low income.
Robinson et al., 2002	CES-D, 20-item and 10-item versions, Clinical setting	Original cut points (21 for 20-item and 4 for 10-item) compared to recommended cut points (20 for 20-item and 3 for 10-item).	Sensitivity: 73% for 20-item version with original cut point; 81% with recommended cut point. 76% for 10-item version with original cut point; 84% with recommended cut point. Specificity: 72% for 20-item version with original cut point and 70% with recommended cut point. 70% for 10-item version with original cut point and 64% with recommended cut point.	303 Puerto Rican adults age 55 and over, recruited from primary care clinics in Northeastern U.S. 73% female, urban, low education, low income, interviewed in English or Spanish.
	GDS, Clinical Setting	Original cut points (9 for 30-item, 6 for 15-item) compared to recommended cut points (13 for 30-item, 7 for 15-item).	Sensitivity: 84% for 30-item version with original cut point and 81% with recommended cut point. 76% for 15-item version with original cut point and 73% with recommended cut point. Specificity: 53% for 30-item version with original cut point and 65% with recommended cut point. 64% for 15-item version with original cut point and 71% with recommended cut point.

Note. BDI = Beck Depression Inventory; CES-D = Center for Epidemiological Studies Depression Scale; GDS = Geriatric Depression Scale.

Robinson et al. (2002) reported psychometric data for both the CES-D and the GDS.

Risk of Bias Within Studies

Risk for bias was identified within studies included in the present review (Table 2). Of eleven studies, only three reported on the presence or absence of significant differences between groups based on race/ethnicity (Espino et al., 1996; Gatewood-Colwell et al., 1989; MacIntosh & Strickland, 2010); such significant differences were absent in two studies (Espino et al., 1996; Gatewood-Colwell et al., 1989) and present in one (MacIntosh & Strickland, 2010). Significant sex differences were reported in two studies (Hahn et al., 2011; Robinson et al., 2002) and not supported in another two studies (Espino et al., 1996; Gatewood-Colwell et al., 1989). Considering that many samples included in the present review were predominantly composed of women, who are at higher risk for depression than men, this source of bias is to be expected. Significant differences between clinical or diagnostic groups were the most commonly mentioned; they were reported absent in two studies and present in three. It is reasonable to assume that the inclusion of individuals with a previous diagnosis of depression and/or other mental disorders in study samples may have influenced their findings. Lastly, one study (Baker et al., 1994) reported a lack of significant differences based on language of test administration, and one (Gonzalez et al., 1995) noted no differences based on location of recruitment and testing.

Results of Individual Studies

The type of psychometric data reported varied across the eleven studies included. Six provided internal consistency as measured by Cronbach’s alpha, six provided sensitivity and/or specificity data, one provided item-level test bias results as measured by differential item analysis (DIF), and one provided convergent validity data as measured by correlation between depression measures. Additionally, one study (Gonzalez et al., 1995) provided information about two items of the original, 20-item CES-D that did not meet convergent validity (see CES-D below).

Beck Depression Inventory (BDI)

Only one study that included psychometric characteristics of the BDI met criteria to be considered representative of older H/Ls. Gatewood-Connell and colleagues (1989) analyzed a sample of 51 older adults, in which about half were older H/L men and women, to evaluate the validity and reliability of the original BDI (Beck et al., 1961). Results indicated a Cronbach’s alpha of 0.80 for this measure, which is appropriate for research settings but lower than the recommended cutoff for clinical purposes (0.90; Nunnally and Bernstein, 1994). The BDI also had a correlation of 0.79 with the Geriatric Depression Scale (GDS), which was included by the authors as evidence for the convergent validity of the BDI. No data were found for the updated (second) version of the BDI – the BDI-II (Beck et al, 1996).

Geriatric Depression Scale (GDS)

The psychometric characteristics of the GDS differed between study samples; notably, some studies identified in this review evaluated the original, 30-item GDS (Yesavage & Brink, 1983) while others used the 15-item GDS (Sheikh & Yesavage, 1986). Among community-dwelling older H/L adults, a Spanish version of the 30-item GDS translated by Carrete and colleagues (2001) yielded internal consistency values ranging from α =0.85 to 0.88; these properties are considered adequate for research purposes, but potentially insufficient in clinical contexts (Nunnally and Bernstein, 1994). The authors also identified sensitivity values ranging from 84% to 88%. Among psychiatric patients in an inpatient clinical facility, this same measure had high internal consistency (α = 0.94); however, its sensitivity was inadequate in such clinical groups when using the original cutoff score of 5; using a revised score of 4, sensitivity improved significantly (from 39% to 75% for patients with major depression disorder and from 77% to 85% for those with other depressive disorders).

Studies that recruited participants without a history of depression seemed to yield more consistent and adequate psychometric values, even if the participants were recruited from health care clinics. Using a cutoff of 11 on the GDS-30, Espino and colleagues (1996) reported sensitivity of 80% and specificity of 61% among Mexican American participants recruited from such clinics. In a sample of over 300 community-dwelling Puerto Rican participants, sensitivity on the GDS-30 ranged from 81% to 84% and specificity from 53% to 65%, depending on the cutoff score used. In addition, in the latter sample, sensitivity and specificity for the GDS-15 ranged from 73% to 76% and 64% to 71%, respectively, depending on the cutoff score used. Two other studies similarly reported that using recommended cutoffs prevented the GDS-15 from demonstrating acceptable sensitivity values and recommended modified cutoff scores. A lower cutoff score of 4 increased sensitivity from 39% to 75% in patients with major depression (Baker & Espino, 1997). Baker and colleagues (1994) reported similar findings using a cutoff score of 4 compared to a sensitivity of 49% when using a cutoff of 6 in a sample of community-dwelling Mexican Americans with a diagnosis of affective illness.

Center for Epidemiological Studies Depression Scale (CES-D)

The CES-D performed appropriately among older, H/L patients recruited from medical clinics, with reported internal consistencies ranging from α=.89 to 0.93. However, Gonzalez and colleagues (1995) identified two items of the original CES-D that did not support convergent validity. In their analyses, item 4 (“I felt I was just as good as other people”) and item 8 (“I felt hopeful about the future”) not only had low or negative correlations (−0.11 and 0.35, respectively) with the scale overall, but removing them improved the measure’s internal consistency alpha. Regarding item 4, the authors suggested that the concept of being “as good as others” may not be a culturally appropriate comparison for Spanish speakers, for whom endorsing this may be equivalent to bragging, while providing a negative response may be a “culturally sanctioned avoidance of self-aggrandizement.” This idea is consistent with the fact that item 4 also correlated similarly with the CES-D and a self-efficacy measure. For item 8, the authors noted that the direct translation of this item to Spanish had a meaning closer to “I felt optimistic about the future,” which participants had difficulty understanding. The authors also reported that these two items also had a higher missing value rate than the other 18 items in the original CES-D evaluation. These findings suggest that older H/L adults may not interact with all questions of the CES-D in the same way as other sociodemographic groups. Alternatively, the vocabulary used in the translation of these items may not be culturally appropriate, as it may be interpreted differently by older H/L adults than was intended in the original measure.

The CES-D showed varying performance among community-dwelling participants from different nationalities. In a study of community-dwelling Mexican Americans, internal consistency ranged from α= 0.73 to 0.79, reflecting internal consistency below recommended cutoffs (Nunnally & Bergstein, 1994). In contrast, two studies where the measure was administered to Puerto Rican older adults yielded internal consistency of α=0.87 as well as a sensitivity of 64% and specificity of 64% using traditional cutoff score of 4; notably, using the authors’ recommended cutoff score of 3 caused an increase in sensitivity to 81% and specificity to 72% (Mahard, 1988; Robinson, 2002). In addition to variation between participants from different nationalities, Hispanic participants overall appeared to express positive affect differently than non-Hispanic Whites based on the results of differential item functioning (DIF) analyses conducted by MacIntosh and Strickland (2010); in other words, item-level bias significantly contributed to differences in total scores depending on ethnoracial group. This difference was greater than the variation expected to occur by chance, and thus, revealed measurement bias in CES-D items for older H/Ls.

Discussion

Summary of Evidence

Although previous research has summarized psychometric characteristics of tools used in late-life depression assessment (Balsamo et al., 2018), a failure of these studies to provide information about the sociodemographic characteristics of the samples included brings into question the generalizability of their findings. Therefore, the goal of this study was to examine available self-report depression screening tools for use with older Hispanics/Latin Americans. We sought to pursue this goal through a systematic review and established three main aims. For our first aim, which involved the identification of potentially useful depression screening tools, we found five measures available in English and Spanish that may be deemed appropriate for use with older adults. Of these, however, our search only yielded studies to support the validity of four such measures for use with older H/Ls, of which two were different versions of the same screening tool (GDS-15 and GDS-30). Notably, despite our broad search, all these are measures created in the United States and validated among Spanish speakers of H/L origin. Our second aim was to examine the psychometric properties of depression screening tools available in English and/or Spanish among older Hispanic/Latin American adults. Overall, we found some information reported on the psychometric characteristics of these measures from studies that included older H/Ls; while these studies are scarce, their samples appear representative and their findings may therefore be generalizable, with considerations described below. However, it is important to delineate the advantages and disadvantages of these measures as inferred from the characteristics of the studies in which their properties have been reported. For this purpose, the performance of these depression measures among different type of samples must be distinguished.

Based on a single study that met criteria for inclusion in this review, the Beck Depression Inventory appeared appropriate for use in research but did not necessarily meet minimal criteria for clinical use. While there were more studies with evidence regarding the Geriatric Depression Scale, these were divided as some reported data on the original 30-item version and others on the 15-item version. Based on internal consistency, the 30-item version was only well-supported among psychiatric patients. The authors of two studies recommended lower cutoff scores for the GDS-15 to improve sensitivity in both clinical and community samples, suggesting that older H/L participants may address the measure differently and therefore demonstrate a different score distribution than the original validation sample. Overall, both versions of the measure performed better among community-residing participants without a history of depression diagnosis in terms of sensitivity and specificity; even then, these values were higher among Puerto Rican than Mexican American participants when using the 30-item GDS, showing a certain variability based on cultural background.

A similar pattern emerged for the CES-D when used in community-dwelling samples as it did for the GDS when used in samples without a history of depression diagnosis; that is, the CES-D showed better internal consistency among Puerto Rican than Mexican participants, though even for the former sample, sensitivity and specificity values were only moderately acceptable and varied notably based on cutoff score. The internal consistency of the CES-D among patients from medical clinics was very good, but the measure had problematic items that threatened the overall validity of the measure when used with older H/Ls. Clinicians evaluating participants from this demographic group in clinical contexts using the CES-D should take the cultural inadequacy of items 4 and 8 into account, whether that implicates reconsidering their relevance in the overall assessment of depression symptoms or using an abbreviated version of the CES-D without these items (notably, this latter option is limited in that the psychometrics of these abbreviated versions in this population may also be unverified).

The small number of studies that met inclusion criteria for this systematic review preclude our ability to make firm recommendations. Therefore, we cautiously provide preliminary recommendations based on the reported results. The data summarized in the present study suggest that the utility of the measures identified may vary depending on their intended purpose. The BDI seems to be an appropriate instrument for the assessment of depression among older H/Ls in research contexts, although this assumption relies on a single data point obtained after a comprehensive systematic review. The best use of the GDS may be a subject of debate; based on internal consistency, it appears appropriate for use with patients who are hospitalized in psychiatric clinics, but various studies have urged to reconsider the use of Cronbach’s alpha values as a single determinant of validity (Dunn et al., 2014; Peters, 2014). When accounting for sensitivity and specificity, the GDS may be better suited for mentally healthy participants; and regardless of the population, professionals evaluating older H/Ls may consider using a different cutoff score to obtain a more desirable balance of sensitivity and specificity, as recommended by Baker and Espino (1997), Baker and colleagues (1994), and Robinson and colleagues (2002). Finally, the CES-D demonstrates weaknesses in regard to its cultural appropriateness, at least among older H/L adults. The acceptability of its psychometric properties is highly variable depending on demographic characteristics, including race/ethnicity and nationality, and it has been found through different methodologies that individuals from diverse backgrounds interact with certain items of this measure inconsistently. These issues may intervene with research findings in studies in which older H/L adults represent a relevant proportion of the sample, and for clinical purposes, providers may do well to select a more robust screening tool when evaluating older H/Ls. If used regardless, clinicians may consider using different cutoffs based on findings by Robinson and colleagues (2002) to maximize the utility of this measure.

Limitations

Although the present study provides a summary of depression measure properties based on a thorough systematic review, there are methodological issues to consider. The number of articles that met criteria for inclusion in our review was small, and there was variation among samples of the few studies included, greatly limiting interpretation and the recommendations that can be made based off these findings. Sample size varied from 44 to over 4000, and there was little homogeneity in terms of gender, education, and nationality of participants. Studies with diverse racial/ethnic samples often reported including specific percentages of “Hispanic” participants without clarifying their country of origin or their level of bilingualism or acculturation, key factors often excluded from psychological assessment literature (O’Bryant et al., 2004; Medina et al., 2020). Overall, as seen in Table 2, there was inconsistency in the reporting of significant differences at the level of relevant sociodemographic characteristics. These limitations increment the risk of bias at the study and outcome level. At the outcome level, it is also worth mentioning that the studies identified do not provide information on the positive or negative predictive values of the depression screening measures we describe, even though these values may be more accurate descriptors of the utility of a measure (Trevethan, 2017).

Conclusions

Overall, the data on the psychometric properties of depression screening measures for use with older H/L adults is variable and reveals a limited awareness among researchers and clinicians of the nuances unique to evaluating depression in this population. Of the three measures considered in this review, the Geriatric Depression Scale shows the most promise as an adequate test for use with diverse older adults, although not without caveats described above. Our findings suggest that using existing depression screening measures when working with these populations involves additional considerations for both clinicians and researchers in order to compensate for the fact that such measures were designed without accounting for cultural differences in the experience and manifestation of depressive symptoms. Future research should place greater emphasis on the development and validation of depression measures created with consideration of the needs of diverse clients and patients, or ideally, designed through an emic approach to aim for items written in culturally relevant terms.

Highlights.

• We reviewed depression scales for use with older Hispanic/Latin American adults

• Only three scales were found to contain psychometric data in this population

• The validity of these measures may vary based on their intended use

• Sensitivity of these measures may also be improved with modified cutoff scores