Gallo et. al.

< - - Back to Measurement matrix         MEGs Search

Quick Links:   Sample   Response Rate   Internal Consistency    DIF Method    DIF Evidence    DIF Impact    Expert Opinion

Gallo et. al.   ( see abstract )

Diagnostic Interview Schedule (DIS)

Name/ Reference

Gallo JJ, Cooper-Patrick L, Lasikar S. Depressive symptoms of whites and African Americans aged 60 years and older. The Journals of Gerontology. (1998) 53B(5):277-285

Source contact info

Joseph J. Gallo, MD, MPH, Department of Mental Hygiene, School of Hygiene and Public Health, The Johns Hopkins University, 624 North Broadway, Room 886, Baltimore, MD 21205.

E-mail: jgallo@welchlink.welch.jhu.edu

Availability (private or public)

public

Conceptual framework

The authors argue that there might be are ethnic differences in the tendency to report symptoms salient for the diagnosis of depression. They hypothesized “that older African Americans might be more likely to report somatic or vegetative symptoms of depression, and less likely to report symptoms of a psychological nature, especially dysphoria, when compared to older Whites”. MIMIC was used to examine DIF in the Major Depression in the Diagnostic Interview Schedule on the bases of self-reported race.

Purpose of measure & application (clinical, research, survey, screening)

Diagnostic Interview Schedule (DIS) (self-report): Screening and diagnostic instrument for assessment of symptoms of major depression.

Sample characteristics

Analysis was performed using household data from respondents aged 60 years and older at the Baltimore , Maryland , and Piedmont , North Carolina , sites (in two independently gathered samples).

“The African Americans in Baltimore tended to include fewer persons not living with a spouse compared to the African Americans in Durham-Piedmont. Among Whites, fewer persons in Baltimore reported less than a high school education and fewer reported not living with a spouse compared to Whites in Durham-Piedmont.”

See Anthony et al., 1985; Eaton & Kessler, 1985; Helzer et al., 1985 below for detailed descriptions of the sampling design, the diagnostic assessments, etc. of the ECA and for reliability and validity of DIS .

Recruitment methods

“The Epidemiologic Catchment Area Program (ECA) was a series of epidemiologic surveys conducted by collaborators between 1980 and 1984 at five sites in the United States: New Haven, Connecticut (Yale University); Baltimore, Maryland (Johns Hopkins University); St. Louis, Missouri (Washington University); Durham-Piedmont, North Carolina (Duke University); and Los Angeles, California (University of California, Los Angeles). At each site, ECA collaborators used multistage probability sampling to select 3,000 to 5,000 adult respondents, and then applied a standardized interview, the Diagnostic Interview Schedule (DIS; Robins, Helzer, Croughan, & Ratcliff, 1981). The standardized questions of the DIS were keyed to individual criteria of the DSM-III case definition for Major Depression.” (pg P278)

“Positive responses to the questions were followed by further questioning to determine whether a threshold for severity had been met and whether the symptom was plausible as a psychiatric symptom; that is, the symptom could not be explained by physical illness, medications, alcohol, or drug use (Robins et al., 1981; Robins & Regier, 1991). Finally, respondents rated the onset and recency of syndromes, that is, clusters of symptoms, so that the timing of disorders could be estimated. To meet the DSM-III diagnostic criteria for Major Depression, an individual must have depressed mood (dysphoria) or loss of interest in things normally enjoyed (anhedonia) for two weeks or more, and at least four of eight additional criteria. However, a diagnosis of Major Depression was not required for inclusion in the analysis.” (pg P278)

In Baltimore and in Durham-Piedmont, information on onset and recency was gathered at the symptom level, so that data on recency of individual symptoms are available for the criteria of DSM-III Major Depression (Von Korff & Anthony, 1982). If the respondent "ever" had the symptom, the respondent was then asked when the last time the symptom was present.” (pg P278)

Data collection method

Epidemiologic surveys conducted as part of the The Epidemiologic Catchment Area Program (ECA) between 1980 and 1984 at five sites in the United States . Only household data from respondents aged 60 years and older at the Baltimore , Maryland , and Piedmont , North Carolina sites were using in this study.

Self reported race was obtained by asking: “Would you please look at this card and give me the letter of the group that best describes your racial background?”: American Indian, Alaskan Native, Asian, Pacific Islander, Black-Not Hispanic, Hispanic, White-Not Hispanic. The study was restricted to persons who identified themselves as either “Black-Not Hispanic” or “White-Not Hispanic”.

Response rate

“A total of 1,157 subjects aged 60 years and older were interviewed in the household survey in Baltimore , while 1,683 persons aged 60 years and older comprise the household sample in Durham-Piedmont. These figures exclude subjects who did not identify themselves as either African American or White. After excluding persons without complete information on DIS symptom data and all covariates, 968 subjects remained for analysis from the Baltimore site and 1,486 subjects remained from the Durham-Piedmont site.” (pg P279)

Format & design (readability, # of items, time to complete, response categories)

The items were dichotomous, scored one if present within one month of the interview or zero if not present.

Type of measurement (nominal, ordinal, interval, ratio)

Not provided

Scoring (range, direction, rules, missing data)

Only persons with complete data on symptoms along with the covariates of gender, minority status, educational status, marital status, and mental status score were included in this analysis.

Availability of translations & source

Not provided

Psychometric Properties:

Scale construction

Not provided

Basic summary statistics

Not provided

Variability

Not provided

Test-retest reliability

Not provided

Interrater reliability

Not provided

Internal consistency

For studies of DIS reliability and validity see Anthony et al., 1985; Eaton & Kessler, 1985; Helzer et al., 1985 below.

Content validity

See note above

Construct validity

Not provided

Concurrent validity

Not provided

Predictive validity

Not provided

Sensitivity to change

Not provided

Differential Item Functioning:

Variable studied (e.g., groups)

“The MIMIC model, described below, includes dichotomous variables for self-reported race (White = 0; African American = 1), gender (male = 0; female = 1), educational status (12 or more years of schooling = 0; fewer than 12 years = 1), marital status (married or living with someone as though married = 0; not currently married = 1), and mental status score (above a standard criterion value = 0; below the standard criterion value = 1). The Mini-Mental State Examination (MMSE) score (range, 0 to 30) was employed as a measure of mental status). The decisions on how to code the variables were reached after considering what variables might modify the tendency to endorse sadness or loss of interest or pleasure, but before any data analysis was completed.” (pg P279)

Sample size

Participants were individuals aged 60 years and older n=968 in Baltimore, Maryland, and n=1,486 in the Durham-Piedmont region of North Carolina, who identified themselves as African American or White and who had complete data on symptoms of depression in the DIS active in the one month prior to interview.

DIF method used

(e.g., MH, IRT, Logistic regression, MIMIC, other factor analysis)

The MIMIC model (a special application of the latent trait model) was used for DIF detection. Covariates to the MIMIC model were introduced in order to examine differential item functioning by self-reported race, adjusting for differences in the level of the latent trait and for the effect of other covariates such as gender. Adjustment for differences in level of depression due to gender, educational status, MMSE score, and marital status was performed.

DIF is the estimate of the direct effect of the studied variable, e.g., ethnicity on the depression item, after adjusting for other covariates and level of depression.

The authors explain that the MIMIC model components are estimated simultaneously, using the LISCOMP program's limited information generalized least squares estimator for dichotomous items. “In the measurement model, the dichotomous symptoms (symptom present within one month of interview = 1, or not present = 0) are considered to be the indicators of an underlying latent dimension of depression. The estimate of the latent variable for depression is based on the covariation that the symptom items have in common, in contrast to a simple sum of depressive symptoms. In the regression model, the latent variable depression is regressed on the covariates. As in ordinary multiple regression, the estimates of these regression coefficients represent the difference in the value of the outcome variable for one group compared to another. The estimate of the direct effect for ethnicity on a given symptom of depression represents a shift in the measurement characteristics for that symptom comparing African Americans to Whites. A significant negative direct effect implies that, even adjusting for other characteristics and the level of depression, compared to Whites, African Americans endorse the symptom at a lower prevalence than expected; that is, the symptom exhibits differential item functioning. Conversely, a positive direct effect is interpreted to mean that the symptom is more likely to be endorsed by African Americans than by Whites, all other characteristics considered.” (pg P279)

Test of model assumptions

Several model fit indices were examined. The goodness of fit index for the Baltimore sample was .95 and .98 for Durham-Piedmont. Parameter number adjusted indices were .97 and .99 respectively. Unidimensionality was tested by examination of the covariance structure accompanied by a scree test of the eigenvalues. This result is not shown, but the authors claim that it is consistent with a unidimensional item set.

Purification

The necessity for purification in MIMIC models has not been established or investigated.

Evidence of uniform DIF

At the Baltimore site, examination of the estimates for the direct effect of ethnicity on the symptoms of depression indicate several items with differential item functioning according to self-reported ethnicity. Sleep disturbance was less likely to be endorsed by African Americans than Whites, given level of depression and the other potentially influential covariates. Difficulty with concentrating and thoughts of death were more likely to be reported by older African Americans one month prior to interview, when compared with older Whites. Older African Americans were less likely to report other symptoms, but the difference did not reach generally accepted standards for statistical significance. The key symptom of sadness was noted to be less likely to be endorsed by African Americans at both sites (the point estimate for the direct effect is negative).

In the Durham-Piedmont sample, three symptoms demonstrated statistically significant differences in African Americans compared with Whites at the Durham-Piedmont site; namely, sadness, loss of interest, and thoughts of death.

While older African Americans were more likely to endorse the item bundle of “thinking about death or suicide" than were Whites in the MIMIC model, the main difference in endorsement arose from the item about thoughts of death.

Evidence of non-uniform DIF

MIMIC models do not permit examination of non-uniform DIF.

Magnitude of DIF

The magnitude of the direct effects are presented in Table 5 but are not discussed.

Re the item bundle of "thinking about death or suicide", Whites in Baltimore and Durham-Piedmont endorsed thoughts of death at similar rates (6.04% and 6.32%, respectively), while African Americans in the two samples endorsed thoughts of death at higher rates than Whites (11.21% in Baltimore and 9.27% in Durham-Piedmont). Endorsement of wanting to die was similar in Whites and African Americans (1.61% among Whites and 1.40% among African Americans in Baltimore , and 1.35% among Whites and 1.35% among African Americans in Durham-Piedmont). No African Americans at either site endorsed thoughts of committing suicide (compared to 0.13% of Whites in Baltimore and 0.62% in Durham-Piedmont).

Impact of DIF

The one month prevalence estimates for symptoms were higher among African Americans than among Whites in Baltimore , but tended to be lower among African Americans compared with Whites in Durham-Piedmont.

According to the authors, “The interpretation of the significantly higher regression estimate (of depression on the covariates of race, gender, education, MMSE score, and marital status) for African Americans at the Baltimore site was that older African Americans are higher on the latent trait of depression when compared with older Whites.” (pg P280) At the Durham-Piedmont site, mean levels of depression were lower for older African Americans than for Whites. Women, those with lower education and unmarried persons had lower levels of depression.

The authors did not discuss the magnitude of the impact of DIF on the measure.

Strengths according to authors:

1. The authors noted that the method used to compare symptoms of depression according to self-reported race, i.e., the MIMIC model, has several advantages over other methods that might have been used to compare item response.

2. The authors mentioned that the present study is an extension of a line of research delineating the salience of the standard criteria for depression across subjects defined by age, race, settings of care, and other characteristics, and that this study focused on the endorsement of symptoms by older African Americans and Whites.

Limitations according to authors:

1. “It was assumed that participants understand and respond consistently to questions about race and/or ethnicity”. (pg 283) Thus, the validity and assessment of the concept of race can be questioned.

2. “Labels employed for ethnicity imply a homogeneity within groups; differences ascribed to "race" most likely reflect social more than genetic differences. On the other hand, failure to account for heterogeneity in measurement of depressive symptoms might be misleading in comparing prevalence of symptoms or disorders according to ethnicity”. (pg 283)

3. “Somatic complaints of respondents was not explicitly included in the model. Thus, the extent to which African Americans ascribed their symptoms to a physical illness is unknown”. (pg 283)

4. “Chronic medical conditions may also affect the pattern of depressive symptoms endorsed and are not included in the model. Instead, estimates for other characteristics that can be assessed more reliably than self-report of medical diagnoses (such as level of educational attainment) but which are themselves associated with functional and medical status were adjusted”. (pg 283)

5. “The tendency to report symptoms could have been affected in unknown ways by the ethnic or racial background of the interviewer”. (pg 283)

6. “The MIMIC model itself present limitations” [not discussed in the manuscript]. (pg 283)

7. “Differences between sites likely reflect characteristics of the samples that were not included in the model, such as rural-urban contrasts, or may simply reflect sampling fluctuation. Older African American adults at the rural site may be generally more reluctant to report symptoms”. (pg 283)

Additional strengths and limitations by expert reviewer:

This is a well executed study examining DIF using the MIMIC model. The only limitations related to the method include inability to model non-uniform DIF by examining differences in the slope parameter. A more detailed description of the tests of model assumptions and the impact of DIF on the total score would have been useful. However, the discussion section is an excellent presentation of other related findings and of the possible differences in meaning and cultural world view that may affect findings.

Key references:

1. Robins, L. N., Helzer, J. E., Croughan, J., & Ratcliff, K. S. (1981). National Institute of Mental Health Diagnostic Interview Schedule: Its history, characteristics, and validity. Archives of General Psychiatry, 38, 381-389.

2. Robins, L. N., & Regier, D. A. (1991). Psychiatric disorders in America: The Epidemiologic Catchment Area Study. New York: Free Press.

3. Von Korff, M., & Anthony, J. C. (1982). The NIMH Diagnostic Interview Schedule modified to record current mental status. Journal of Affective Disorders, 4, 365-371.

4. West, S. G., Finch, J. F., & Curran, P. J. (1995). Structural equation models with nonnormal variables: Problems and remedies. In R. H. Hoyle (Ed.), Structural equation modeling: Concepts, issues, and applications (pp. 56-75). Thousand Oaks, CA: Sage.

5. Anthony, J. C., & Petronis, K. R. (1991). Suspected risk factors for depression among adults 1844 years old. Epidemiology, 2, 123-132. Baker, F. M. (1991). A contrast: Geriatric depression versus depression in younger age groups. Journal of the National Medical Association, 83, 340-344.

6. Eaton, W. W., & Kessler, L. G. (1985). Epidemiologic field methods in psychiatry: The NIMH Epidemiologic Catchment Area Program. New York: Academic Press.

7. Helzer, J. E., Robins, L. N., McEvoy, L. T., Spitznagel, E. L., Stolzman, R. K., Farmer, A., & Brockington, I. F. (1985). A comparison of clinical and diagnostic interview schedule diagnoses: Physician examination of lay-interviewed cases in the general population. Archives of General Psychiatry, 42, 657-666.

( see abstract )

Back to   TOP    Measurement matrix         MEGs Search

Diagnostic Interview Schedule (DIS)
Name/ Reference	Gallo JJ, Cooper-Patrick L, Lasikar S. Depressive symptoms of whites and African Americans aged 60 years and older. The Journals of Gerontology. (1998) 53B(5):277-285
Source contact info	Joseph J. Gallo, MD, MPH, Department of Mental Hygiene, School of Hygiene and Public Health, The Johns Hopkins University, 624 North Broadway, Room 886, Baltimore, MD 21205. E-mail: jgallo@welchlink.welch.jhu.edu
Availability (private or public)	public
Conceptual framework	The authors argue that there might be are ethnic differences in the tendency to report symptoms salient for the diagnosis of depression. They hypothesized “that older African Americans might be more likely to report somatic or vegetative symptoms of depression, and less likely to report symptoms of a psychological nature, especially dysphoria, when compared to older Whites”. MIMIC was used to examine DIF in the Major Depression in the Diagnostic Interview Schedule on the bases of self-reported race.
Purpose of measure & application (clinical, research, survey, screening)	Diagnostic Interview Schedule (DIS) (self-report): Screening and diagnostic instrument for assessment of symptoms of major depression.
Sample characteristics	Analysis was performed using household data from respondents aged 60 years and older at the Baltimore , Maryland , and Piedmont , North Carolina , sites (in two independently gathered samples). “The African Americans in Baltimore tended to include fewer persons not living with a spouse compared to the African Americans in Durham-Piedmont. Among Whites, fewer persons in Baltimore reported less than a high school education and fewer reported not living with a spouse compared to Whites in Durham-Piedmont.” See Anthony et al., 1985; Eaton & Kessler, 1985; Helzer et al., 1985 below for detailed descriptions of the sampling design, the diagnostic assessments, etc. of the ECA and for reliability and validity of DIS .
Recruitment methods	“The Epidemiologic Catchment Area Program (ECA) was a series of epidemiologic surveys conducted by collaborators between 1980 and 1984 at five sites in the United States: New Haven, Connecticut (Yale University); Baltimore, Maryland (Johns Hopkins University); St. Louis, Missouri (Washington University); Durham-Piedmont, North Carolina (Duke University); and Los Angeles, California (University of California, Los Angeles). At each site, ECA collaborators used multistage probability sampling to select 3,000 to 5,000 adult respondents, and then applied a standardized interview, the Diagnostic Interview Schedule (DIS; Robins, Helzer, Croughan, & Ratcliff, 1981). The standardized questions of the DIS were keyed to individual criteria of the DSM-III case definition for Major Depression.” (pg P278) “Positive responses to the questions were followed by further questioning to determine whether a threshold for severity had been met and whether the symptom was plausible as a psychiatric symptom; that is, the symptom could not be explained by physical illness, medications, alcohol, or drug use (Robins et al., 1981; Robins & Regier, 1991). Finally, respondents rated the onset and recency of syndromes, that is, clusters of symptoms, so that the timing of disorders could be estimated. To meet the DSM-III diagnostic criteria for Major Depression, an individual must have depressed mood (dysphoria) or loss of interest in things normally enjoyed (anhedonia) for two weeks or more, and at least four of eight additional criteria. However, a diagnosis of Major Depression was not required for inclusion in the analysis.” (pg P278) In Baltimore and in Durham-Piedmont, information on onset and recency was gathered at the symptom level, so that data on recency of individual symptoms are available for the criteria of DSM-III Major Depression (Von Korff & Anthony, 1982). If the respondent "ever" had the symptom, the respondent was then asked when the last time the symptom was present.” (pg P278)
Data collection method	Epidemiologic surveys conducted as part of the The Epidemiologic Catchment Area Program (ECA) between 1980 and 1984 at five sites in the United States . Only household data from respondents aged 60 years and older at the Baltimore , Maryland , and Piedmont , North Carolina sites were using in this study. Self reported race was obtained by asking: “Would you please look at this card and give me the letter of the group that best describes your racial background?”: American Indian, Alaskan Native, Asian, Pacific Islander, Black-Not Hispanic, Hispanic, White-Not Hispanic. The study was restricted to persons who identified themselves as either “Black-Not Hispanic” or “White-Not Hispanic”.
Response rate	“A total of 1,157 subjects aged 60 years and older were interviewed in the household survey in Baltimore , while 1,683 persons aged 60 years and older comprise the household sample in Durham-Piedmont. These figures exclude subjects who did not identify themselves as either African American or White. After excluding persons without complete information on DIS symptom data and all covariates, 968 subjects remained for analysis from the Baltimore site and 1,486 subjects remained from the Durham-Piedmont site.” (pg P279)
Format & design (readability, # of items, time to complete, response categories)	The items were dichotomous, scored one if present within one month of the interview or zero if not present.
Type of measurement (nominal, ordinal, interval, ratio)	Not provided
Scoring (range, direction, rules, missing data)	Only persons with complete data on symptoms along with the covariates of gender, minority status, educational status, marital status, and mental status score were included in this analysis.
Availability of translations & source	Not provided
Psychometric Properties:
Scale construction	Not provided
Basic summary statistics	Not provided
Variability	Not provided
Test-retest reliability	Not provided
Interrater reliability	Not provided
Internal consistency	For studies of DIS reliability and validity see Anthony et al., 1985; Eaton & Kessler, 1985; Helzer et al., 1985 below.
Content validity	See note above
Construct validity	Not provided
Concurrent validity	Not provided
Predictive validity	Not provided
Sensitivity to change	Not provided
Differential Item Functioning:
Variable studied (e.g., groups)	“The MIMIC model, described below, includes dichotomous variables for self-reported race (White = 0; African American = 1), gender (male = 0; female = 1), educational status (12 or more years of schooling = 0; fewer than 12 years = 1), marital status (married or living with someone as though married = 0; not currently married = 1), and mental status score (above a standard criterion value = 0; below the standard criterion value = 1). The Mini-Mental State Examination (MMSE) score (range, 0 to 30) was employed as a measure of mental status). The decisions on how to code the variables were reached after considering what variables might modify the tendency to endorse sadness or loss of interest or pleasure, but before any data analysis was completed.” (pg P279)
Sample size	Participants were individuals aged 60 years and older n=968 in Baltimore, Maryland, and n=1,486 in the Durham-Piedmont region of North Carolina, who identified themselves as African American or White and who had complete data on symptoms of depression in the DIS active in the one month prior to interview.
DIF method used (e.g., MH, IRT, Logistic regression, MIMIC, other factor analysis)	The MIMIC model (a special application of the latent trait model) was used for DIF detection. Covariates to the MIMIC model were introduced in order to examine differential item functioning by self-reported race, adjusting for differences in the level of the latent trait and for the effect of other covariates such as gender. Adjustment for differences in level of depression due to gender, educational status, MMSE score, and marital status was performed. DIF is the estimate of the direct effect of the studied variable, e.g., ethnicity on the depression item, after adjusting for other covariates and level of depression. The authors explain that the MIMIC model components are estimated simultaneously, using the LISCOMP program's limited information generalized least squares estimator for dichotomous items. “In the measurement model, the dichotomous symptoms (symptom present within one month of interview = 1, or not present = 0) are considered to be the indicators of an underlying latent dimension of depression. The estimate of the latent variable for depression is based on the covariation that the symptom items have in common, in contrast to a simple sum of depressive symptoms. In the regression model, the latent variable depression is regressed on the covariates. As in ordinary multiple regression, the estimates of these regression coefficients represent the difference in the value of the outcome variable for one group compared to another. The estimate of the direct effect for ethnicity on a given symptom of depression represents a shift in the measurement characteristics for that symptom comparing African Americans to Whites. A significant negative direct effect implies that, even adjusting for other characteristics and the level of depression, compared to Whites, African Americans endorse the symptom at a lower prevalence than expected; that is, the symptom exhibits differential item functioning. Conversely, a positive direct effect is interpreted to mean that the symptom is more likely to be endorsed by African Americans than by Whites, all other characteristics considered.” (pg P279)
Test of model assumptions	Several model fit indices were examined. The goodness of fit index for the Baltimore sample was .95 and .98 for Durham-Piedmont. Parameter number adjusted indices were .97 and .99 respectively. Unidimensionality was tested by examination of the covariance structure accompanied by a scree test of the eigenvalues. This result is not shown, but the authors claim that it is consistent with a unidimensional item set.
Purification	The necessity for purification in MIMIC models has not been established or investigated.
Evidence of uniform DIF	At the Baltimore site, examination of the estimates for the direct effect of ethnicity on the symptoms of depression indicate several items with differential item functioning according to self-reported ethnicity. Sleep disturbance was less likely to be endorsed by African Americans than Whites, given level of depression and the other potentially influential covariates. Difficulty with concentrating and thoughts of death were more likely to be reported by older African Americans one month prior to interview, when compared with older Whites. Older African Americans were less likely to report other symptoms, but the difference did not reach generally accepted standards for statistical significance. The key symptom of sadness was noted to be less likely to be endorsed by African Americans at both sites (the point estimate for the direct effect is negative). In the Durham-Piedmont sample, three symptoms demonstrated statistically significant differences in African Americans compared with Whites at the Durham-Piedmont site; namely, sadness, loss of interest, and thoughts of death. While older African Americans were more likely to endorse the item bundle of “thinking about death or suicide" than were Whites in the MIMIC model, the main difference in endorsement arose from the item about thoughts of death.
Evidence of non-uniform DIF	MIMIC models do not permit examination of non-uniform DIF.
Magnitude of DIF	The magnitude of the direct effects are presented in Table 5 but are not discussed. Re the item bundle of "thinking about death or suicide", Whites in Baltimore and Durham-Piedmont endorsed thoughts of death at similar rates (6.04% and 6.32%, respectively), while African Americans in the two samples endorsed thoughts of death at higher rates than Whites (11.21% in Baltimore and 9.27% in Durham-Piedmont). Endorsement of wanting to die was similar in Whites and African Americans (1.61% among Whites and 1.40% among African Americans in Baltimore , and 1.35% among Whites and 1.35% among African Americans in Durham-Piedmont). No African Americans at either site endorsed thoughts of committing suicide (compared to 0.13% of Whites in Baltimore and 0.62% in Durham-Piedmont).
Impact of DIF	The one month prevalence estimates for symptoms were higher among African Americans than among Whites in Baltimore , but tended to be lower among African Americans compared with Whites in Durham-Piedmont. According to the authors, “The interpretation of the significantly higher regression estimate (of depression on the covariates of race, gender, education, MMSE score, and marital status) for African Americans at the Baltimore site was that older African Americans are higher on the latent trait of depression when compared with older Whites.” (pg P280) At the Durham-Piedmont site, mean levels of depression were lower for older African Americans than for Whites. Women, those with lower education and unmarried persons had lower levels of depression. The authors did not discuss the magnitude of the impact of DIF on the measure.