Operating Characteristics of Brief Screens for Dementia in a Multicultural Population
1995; Elsevier BV; Volume: 3; Issue: 2 Linguagem: Inglês
10.1097/00019442-199500320-00002
ISSN1545-7214
AutoresDavid A. Wilder, Peter Cross, Jiming Chen, Barry J. Gurland, Rafael Lantigua, Jeanne A. Teresi, Mabel Bolivar, Priscilla Encarnación,
Tópico(s)Stroke Rehabilitation and Recovery
ResumoOperating characteristics of seven screens for dementia were compared across various groups for 795 persons who had received a criterion diagnostic evaluation. Area under the curve (AUC), based on receiver operating characteristics, was compared between and within scales as an indication of their efficiency. Differences in AUC were only 5% across all the scales for the entire sample, but increased to 11% across sociocultural groups and scales and to 20% across education groups and scales. Two scales (the Mini-Mental State Exam and the Short Portable Mental Status Questionnaire) misclassified most nondementias for the entire sample, and all scales misclassified most nondementias among persons with less than 5 years of education. Findings could support a recommendation that certain shorter scales be used because they perform as well as longer ones, are more consistent across cultural and educational groups, and can be more easily modified to improve performance in culturally diverse populations. Operating characteristics of seven screens for dementia were compared across various groups for 795 persons who had received a criterion diagnostic evaluation. Area under the curve (AUC), based on receiver operating characteristics, was compared between and within scales as an indication of their efficiency. Differences in AUC were only 5% across all the scales for the entire sample, but increased to 11% across sociocultural groups and scales and to 20% across education groups and scales. Two scales (the Mini-Mental State Exam and the Short Portable Mental Status Questionnaire) misclassified most nondementias for the entire sample, and all scales misclassified most nondementias among persons with less than 5 years of education. Findings could support a recommendation that certain shorter scales be used because they perform as well as longer ones, are more consistent across cultural and educational groups, and can be more easily modified to improve performance in culturally diverse populations. There are a number of widely used and well-documented screening techniques for Alzheimer's disease and related dementias (ADRD).1Kane R Kane R Assessing the Elderly: A Practical Guide to Measurement. Lexington Books, Lexington, MA1981Google Scholar, 2Baker FM Screening tests for cognitive impairment.Hosp Community Psychiatry. 1989; 40: 339-340PubMed Google Scholar These screens are currently used in research and clinical settings. Also, the National Advisory Panel on Alzheimer's Disease has recently recommended that the conventional measures of functional deficit (activities of daily living; ADL) used to assess patients to determine their eligibility for admission to and reimbursement for long-term care be modified to detect the unique characteristics of ADRD patients.3Advisory Panel on Alzheimer's Disease, 4th Report of the Advisory Panel on Alzheimer's Disease, 1992. DHHS Publication No. 93-3520. Supt. of Documents, U.S. Government Printing Office, Washington, DC1993Google Scholar Thus investigators, clinicians, and regulators confront the task of choosing among screening techniques for a given application. Most brief screening instruments for ADRD test selected aspects of cognitive status. These brief screens, usually taking not more than 5–10 minutes to administer, are typically systematic, structured questions and tasks and do not require information from a third party. They are intended to be sensitive and specific to the presence of early as well as established dementia, to be reliable and quantitative, to be able to be administered by health professionals or research assistants from a variety of disciplines, not to need more than a few hours of special training to administer, and to be readily scorable at the time of testing. Ideally, they are not substantially biased by such sociocultural influences as ethnicity and education. During recent years, brief screening instruments have been used in studies with a number of different ethnic groups, including Chinese, Finns, Koreans, Italians, Jews, Latinos, and African-Americans.4Escobar JI Burnam A Karno M et al.Use of the Mini-Mental State Examination (MMSE) in a community population of mixed ethnicity.J Nerv Ment Dis. 1986; 174: 607-614Crossref PubMed Scopus (350) Google Scholar, 5Folstein MF Folstein SE McHugh PR The Mini-Mental State: a practical method of grading the cognitive state of patients for the clinician.J Psychiatr Res. 1975; 12: 189-198Abstract Full Text PDF PubMed Scopus (69806) Google Scholar, 6Kua E The prevalence of dementia in elderly Chinese.Acta Psychiatr Scand. 1991; 83: 350-352Crossref PubMed Scopus (40) Google Scholar, 7Li G Shen YC Chen CH et al.A three-year follow-up study of age-related dementia in an urban area of Beijing.Acta Psychiatr Scand. 1991; 83: 99-104Crossref PubMed Scopus (150) Google Scholar, 8Salmon DP Riekkinen PJ Katzman R et al.Cross-cultural studies of dementia: a comparison of Mini-Mental State Examination performance in Finland and China.Arch Neurol. 1989; 46: 769-772Crossref PubMed Scopus (111) Google Scholar, 9Silverman JM Li G Schear S et al.A cross-cultural family history study of primary progressive dementia in relatives of nondemented elderly Chinese, Italians, Jews, and Puerto Ricans.Acta Psychiatr Scand. 1992; 85: 211-217Crossref PubMed Scopus (18) Google Scholar, 10Yu ES Liu WT Levy P et al.Cognitive impairment among elderly adults in Shanghai, China.J Gerontol. 1989; 44: S97-S106Crossref PubMed Scopus (126) Google Scholar Studies that include African-American or Latino as well as non-Latino whites have generally reported higher rates of cognitive impairment among the minority groups.4Escobar JI Burnam A Karno M et al.Use of the Mini-Mental State Examination (MMSE) in a community population of mixed ethnicity.J Nerv Ment Dis. 1986; 174: 607-614Crossref PubMed Scopus (350) Google Scholar, 5Folstein MF Folstein SE McHugh PR The Mini-Mental State: a practical method of grading the cognitive state of patients for the clinician.J Psychiatr Res. 1975; 12: 189-198Abstract Full Text PDF PubMed Scopus (69806) Google Scholar, 8Salmon DP Riekkinen PJ Katzman R et al.Cross-cultural studies of dementia: a comparison of Mini-Mental State Examination performance in Finland and China.Arch Neurol. 1989; 46: 769-772Crossref PubMed Scopus (111) Google Scholar, 11De la Monte SM Hutchins GM Moore GW Racial differences in the etiology of dementia and frequency of Alzheimer lesions in the brain.Journal of the National Medical Association. 1989; 81: 644-652PubMed Google Scholar, 12Murden RA McRae TD Kaner S et al.Mini-Mental State exam scores vary with education in Blacks and Whites.J Am Geriatr Soc. 1991; 39: 149-155Crossref PubMed Scopus (256) Google Scholar, 13Weissmann M Myers JK Tischler GL et al.Psychiatric disorders (DSM-III) and cognitive impairment among the elderly in a US urban community.Acta Psychiatr Scand. 1985; 71: 366-379Crossref PubMed Scopus (173) Google Scholar Most studies also reported higher rates among persons with less education.14Burvill PW The impact of criteria selection on prevalence rates. Fifth Congress of the International Federation of Psychiatric Epidemiology (Montreal, Canada; 1990).Psychiatric Journal of the University of Ottawa. 1990; 15: 194-199PubMed Google Scholar, 15Fratiglioni L Grut M Forsell Y et al.Prevalence of Alzheimer's disease and other dementias in an elderly urban population: relationship with age, sex, and education.Neurology. 1991; 41: 1886-1892Crossref PubMed Google Scholar, 16Uhlmann RF Larson EB Effect of education on the Mini-Mental State Examination as a screening test for dementia.J Am Geriatr Soc. 1991; 39: 876-880Crossref PubMed Scopus (226) Google Scholar, 17O'Connor DW Pollitt PA Hyde JB et al.Clinical issues relating to the diagnosis of mild dementia in a British community survey.Arch Neurol. 1991; 48: 530-534Crossref PubMed Scopus (11) Google Scholar Some studies have used multiple instruments with the same subjects in an effort to determine whether some work better than others with minority groups or persons with less education.18Davis PB Morris JC Grant E Brief screening tests vs. clinical staging in senile dementia of the Alzheimer type.J Am Geriatr Soc. 1990; 38: 129-135Crossref PubMed Scopus (79) Google Scholar, 19Fillenbaum G Heyman A Williams K et al.Sensitivity and specificity of standardized screens of cognitive impairment and dementia among elderly black and white community residents.J Clin Epidemiol. 1990; 43: 651-660Abstract Full Text PDF PubMed Scopus (226) Google Scholar The choice of a brief screen for dementia should be guided by criteria that are explicit, logical, and tied to a sound knowledge base. However, the scientific knowledge base about screens for dementia does not provide adequate information for making these decisions (whether they be by health care professionals, researchers, funding agencies, or various committees concerned with practice standards or regulation) within the context of other relevant information, such as functioning. These crucial decisions must often defer to issues of custom, critical mass of usage, collaborative compromise, preferences of funding bodies, or expert consensus. Recent studies have found that some of the items in conventional screens, including those used in this study, are racially or ethnically biased.20Teresi J Golden R Cross P et al.Item bias in cognitive screening measures: comparisons of elderly White, Afro-American, Hispanic, and high and low education groups.J Clin Epidemiol. 1994; Google Scholar However, these studies do not indicate the extent to which these items bias results in actual misclassification of persons as demented or not. Here we assess the relative accuracy of selected brief screens for ADRD in detecting dementia in a multicultural population. The data in this paper were gathered in the course of the North Manhattan Aging Project (NMAP), a registry for ADRD. This registry consists of a Reporting Component and a Survey Component: the data in this paper are from the Reporting Component and from the pilot study21Gurland BJ Wilder DE Cross P et al.Screening scales for dementia: toward reconciliation of conflicting cross-cultural findings.International Journal of Geriatric Psychiatry. 1992; 7: 105-113Crossref Scopus (95) Google Scholar conducted earlier to establish registry methods. Relative accuracy of screens for dementia is determined by comparing sensitivity—the proportion of all diagnosed dementias so identified by the screen, specificity—the proportion of all diagnosed nondementias identified, and receiver operating characteristic (ROC) curves22Metz CE Basic principles of ROC analysis.Seminars in Nuclear Medicine. 1978; 8: 283-298Abstract Full Text PDF PubMed Scopus (4547) Google Scholar for different sociocultural and education groups within the same study sample. ROC curves are plotted with sensitivity on the Y axis and 1 – specificity (the false positive rate) on the X axis; curves are based on sensitivity and specificity at each scale score value. The area under the curve (AUC) is the probability that a randomly chosen diagnosed dementia patient would have a higher screening test score than a randomly chosen diagnosed non-dementia patient;23Hanley JA McNeil BJ The meaning and use of the area under a receiver operating characteristic (ROC) curve.Diagnostic Radiology. 1982; 143: 30-36Google Scholar it is considered a more precise indication of the accuracy of a scale than sensitivity or specificity at any single cut. The NMAP Registry seeks to identify and list all cases of ADRD among persons 65 years and older in a defined geographic area, using 1) a reporting network, or Reporting Component, and 2) a probability sample survey of residents in this area, or Survey Component. The survey is a source of subjects for the Registry studies as well as a check on its completeness. The geographic area chosen for study is in North Manhattan, comprising 13 contiguous census tracts that contain 4,066 older Latinos, predominantly Dominican; 3,511 non-Latino African-Americans; and 1,774 non-Latino white subjects age 65 years and older, according to figures provided by the United States Census, Special Tabulations Branch. Descriptions of the Reporting Component of the Registry, the referral criteria, and methods for obtaining a criterion diagnosis are described in a companion paper.24Gurland B Wilder D Cross P et al.Relative rates of dementia by multiple case definitions, over two prevalence periods, in three sociocultural groups.American Journal of Geriatric Psychiatry. 1995; 3: 6-20Abstract Full Text Full Text PDF Scopus (35) Google Scholar A compendium instrument was compiled from five widely used dementia screening scales: the Kahn-Goldfarb Mental Status Questionnaire (MSQ);25Kahn RL Goldfarb Al Pollack M et al.Brief objective measure for the determination of mental status in the aged.Am J Psychiatry. 1960; 117: 326-328Crossref PubMed Scopus (655) Google Scholar the Short Portable Mental Status Questionnaire (SPMSQ);26Pfeiffer E A short portable mental status questionnaire for the assessment of organic brain deficit in elderly patients.J Am Geriatr Soc. 1975; 22: 433-444Crossref Scopus (4130) Google Scholar the Comprehensive Assessment and Referral Interview (CARE) Cognitive Scale, Dementia Version (in a previous paper this scale was called the CARE Diagnostic Scale; the term diagnostic has been dropped to avoid confusion with the application of formal diagnostic criteria); the Blessed Memory-Information-Concentration (MIC) test;27Gurland B Golden R Challop I Unidimensional and multidimensional approaches to the differentiation of depression and dementia in the elderly.in: Corkin S Davis K Growden J Alzheimer's Disease: A Review of Progress (Aging). Volume 19. Raven, New York1982: 119-125Google Scholar and the Mini-Mental State Examination (MMSE).28Katzman R Brown T Fuld P et al.Validation of a short orientation-memory-concentration test of cognitive impairment.Am J Psychiatry. 1983; 140: 734-738Crossref PubMed Scopus (1627) Google Scholar A sixth previously published scale, the CARE Cognitive Scale, Homogeneous Version,29Folstein MF Folstein SE McHugh PR Mini-Mental State: a practical method for grading the cognitive state of patients for the clinician.J Psychiatr Res. 1975; 12: 189-198Abstract Full Text PDF PubMed Scopus (41442) Google Scholar based on latent structure analysis of cognitive items in the CARE, and a "culture-fair" scale containing 23 of the 53 items from the other scales in the compendium instrument derived from analysis of NMAP data using mathematical mixture models (see Note at the end of this article) are also included in the scale comparisons in this paper. The compendium instrument was organized to minimize redundancy between constituent measures and allow a smooth flow of the interview. The interview averaged 20 minutes in length. A Spanish translation of the compendium instrument was prepared by a bilingual interviewer trained in the administration of the English version. The translation was reviewed by two bilingual professionals with extensive experience in mental testing; adjustments were made until the original and translated versions coincided. In a few items, an equivalent rather than a literal translation was chosen. Interviewers fluent in both Spanish and English (Dominican, Puerto Rican, Panamanian, or Colombian) were assigned to subjects according to the subject's expressed preference for language. This was based on specified criteria applied to information collected in an evaluation carried out by the Clinical Core team of neurologists, physicians, and neuropsychologists who were blind to the screen scores. The Clinical Core team procedures, described elsewhere,30Golden RB Teresi JA Gurland BJ Development of indicator scales for the Comprehensive Assessment and Referral Evaluation (CARE) interview schedule.J Gerontol. 1984; 39: 138-146Crossref PubMed Scopus (130) Google Scholar, 31Stern Y Andrews H Pittman J et al.Diagnosis of dementia in a heterogeneous population: development of a neuropsychological, paradigm-based diagnosis of dementia and quantified correction for the effects of education.Arch Neurol. 1992; 49: 453-460Crossref PubMed Scopus (358) Google Scholar replicate the diagnostic evaluation that a person might undergo at an Alzheimer's disease research center. A criterion diagnosis of dementia means the subject met DSM-III-R criteria for dementia, covering all subtypes. In the Transcultural Pilot Study (TPS) for the NMAP,21Gurland BJ Wilder DE Cross P et al.Screening scales for dementia: toward reconciliation of conflicting cross-cultural findings.International Journal of Geriatric Psychiatry. 1992; 7: 105-113Crossref Scopus (95) Google Scholar the compendium cognitive screening instrument was administered to 550 subjects 65 years of age and older, including approximately equal-sized groups of Latinos, African-Americans, and non-Latino whites. Recruitment was designed to produce a rich but unknown mix of dementia and non-dementia patients. Using the published scoring systems for the five screening scales, sharply contrasting results for absolute and relative rates of cognitive impairment for the three cultural groups were found. These differences were reduced considerably by adjusting the cutpoints. However, when adjusted to the high sensitivity level required for the NMAP Registry, some scales had lower specificity than others. At this 90%-or-better level of sensitivity, the CARE Cognitive Dementia Scale produced the fewest false positives. It was therefore selected for determining referrals of all screen-positives and a proportion of screen-negatives to the Clinical Core in the reporting and survey phases of the study. Some subjects from the TPS came from the targeted NMAP area and were therefore included in the Registry from the beginning. However, all pilot study subjects have been excluded from the NMAP sample analysis for this paper so as to permit comparisons of the results from two independent samples. The subjects for this paper are 795 persons not also in the pilot study who were screened in the Reporting Component of the Registry and given a diagnosis of Dementia or No Dementia by the criterion assessment team. Demographic characteristics of this group of 795 persons are shown in Table 1.TABLE 1Age and education distributions (%) for three sociocultural groupsNon-Latino WhiteAfrican-AmericanLatinoTotalAge, years(n = 136)(n = 299)(n = 355)(n = 790) 65–7417.625.141.731.3 75–8430.946.236.639.2 85+51.528.821.729.5Education, years(n = 132)(n = 290)(n = 356)(n = 778) <511.417.652.832.6 5–1140.960.037.146.3 ≥1247.722.410.121.1Note: NMAP, Reporting Component, Extended Prevalence Period, Transcultural Pilot Study subjects excluded.N = 795 diagnosed subjects from three sociocultural groups. Open table in a new tab Note: NMAP, Reporting Component, Extended Prevalence Period, Transcultural Pilot Study subjects excluded. N = 795 diagnosed subjects from three sociocultural groups. The sample reflects the differences among the three ethnic groups comprising the NMAP target area. Latinos are the largest group, and they have the youngest age distribution and the least education. Non-Latino whites are the smallest group, and they are proportionately older and have more education than the other two groups. The African-American group is somewhere between the other two on all demographic parameters. There is considerable overlap of items among the seven screening scales contained in the compendium instrument. Correlations between pairs of scales ranged from 0.772 to 0.940 for the 550 persons screened in the TPS, from 0.849 to 0.954 for the 1,468 additional persons screened by the Reporting Component of the Registry during the extended prevalence period in the first 2 years of the Reporting Component, and from 0.829 to 0.947 for the 795 persons in the Reporting Component who were also clinically examined and diagnosed. Correlations in all samples were somewhat lower for the SPMSQ and for the MMSE with the other five scales than for the other scales with each other. A major concern of the Registry was to be able to maintain a high level of sensitivity required for thorough case-finding, without allowing specificity to become so low that the numbers of false positives needing a full clinical examination exceeded project resources. Analysis of NMAP data has therefore concentrated on comparisons across scales, with cuts set at high sensitivity. Some studies might want to be certain that most persons captured by a screen truly have dementia, and set their standard at 90% specificity. However, this results in low sensitivity and identifies only relatively advanced cases of dementia with all the screens we have tested. For the group of 795 persons from the Reporting Component of the Registry who were not part of the pilot sample, specificities ranged from 0.61 (CARE Homogeneous) to 0.45 (MMSE) when they were cut to provide 90%-or-better sensitivity (see Table 2). In the pilot sample, 162 persons were evaluated by the clinical diagnostic team; with cuts set at 90%-or-better sensitivity, specificity ranged from 0.78 (CARE Dementia) to 0.56 (MMSE and Kahn-Goldfarb MSQ). Lower specificities for all scales in the Reporting Component of the Registry than in the pilot sample reflect the fact mat a majority of reported persons were considered to be at risk, and many had cognitive impairments, even though they did not have dementia, whereas the pilot sample included many normal volunteers who were more easily distinguished from persons diagnosed with dementia. In the Registry sample of 795, the numbers of false positives at the 90%-or-better sensitivity cuts ranged from 139 (CARE Homogeneous) to 199 (MMSE). In the pilot sample with 162 patients examined clinically, false positives ranged from 27 (CARE Dementia) to 54 (MMSE and Kahn-Goldfarb MSQ).TABLE 2Specificity levels and numbers of false positives for seven screening scales, with cuts made at ≥ 0.90 sensitivity level for criterion diagnosis of dementia in two culturally mixed samplesTranscultural Pilot Study (n = 162)NMAP Reporting Registry (N = 795)Scales and Cuts (Number of Errors)SpecificityFalse PositivesSpecificityFalse PositivesBlessed ≥ 7; ≥ 8*Highest number of errors at which sensitivity was ≥ 0.90 was the same for both study samples for all scales except the Blessed, which maintained the high sensitivity standard with one additional error in the NMAP Registry Reporting Component sample.0.65430.56157CARE Dementia ≥ 30.78270.59146Kahn-Goldfarb MSQ ≥ 10.56540.53165MMSE ≥ 70.56540.45199SPMSQ ≥ 20.70370.48194CARE Homogeneous ≥ 2**The CARE Homogeneous Scale was not used in the TPS analysis, and the "culture-fair" scale was derived later, using NMAP data.**The CARE Homogeneous Scale was not used in the TPS analysis, and the "culture-fair" scale was derived later, using NMAP data.0.61139"Culture-fair" ≥ 4**The CARE Homogeneous Scale was not used in the TPS analysis, and the "culture-fair" scale was derived later, using NMAP data.**The CARE Homogeneous Scale was not used in the TPS analysis, and the "culture-fair" scale was derived later, using NMAP data.0.49181Note: (Kahn-Goldfarb) MSQ = Mental Status Questionnaire; MMSE = Mini-Mental State Exam; SPMSQ = Short Portable Mental Status Questionnaire; CARE = Comprehensive Assessment and Referral Interview.* Highest number of errors at which sensitivity was ≥ 0.90 was the same for both study samples for all scales except the Blessed, which maintained the high sensitivity standard with one additional error in the NMAP Registry Reporting Component sample.** The CARE Homogeneous Scale was not used in the TPS analysis, and the "culture-fair" scale was derived later, using NMAP data. Open table in a new tab Note: (Kahn-Goldfarb) MSQ = Mental Status Questionnaire; MMSE = Mini-Mental State Exam; SPMSQ = Short Portable Mental Status Questionnaire; CARE = Comprehensive Assessment and Referral Interview. In accordance with the interest in the use of brief cognitive screening scales in a multicultural population, ROC curves22Metz CE Basic principles of ROC analysis.Seminars in Nuclear Medicine. 1978; 8: 283-298Abstract Full Text PDF PubMed Scopus (4547) Google Scholar and the AUC23Hanley JA McNeil BJ The meaning and use of the area under a receiver operating characteristic (ROC) curve.Diagnostic Radiology. 1982; 143: 30-36Google Scholar were calculated for each of the seven screening scales for the Registry Reporting Component sample as a whole, for three ethnic groups: Latino, non-Latino white, and African-American; and for three educational groups: less than 5 years, 5–11 years, and 12 or more years of education. Better performance of a scale is indicated by a steeper vertical arm and a higher AUC. This is illustrated in Figure 1, which shows ROC curves for the CARE Homogeneous Scale for the three educational groups. A low number of errors on the screening scale results in high sensitivity and low specificity (upper right-hand quadrant), and a high number of errors results in high specificity and low sensitivity (lower left-hand quadrant), for all three educational groups. Starting with no errors, specificity increases more with each error for the group with ≥12 years of education than for the group with 5–11 years, and increases more for the latter than for the group with < 5 years of education. A horizontal line has been drawn at the 90% sensitivity level to show how the ROC curves for the three groups intersect at different specificity levels. Forty-nine different ROC curves cannot be shown, but some important differences and similarities among the curves can be reported. For the sample as a whole, AUC ranged only 4.9%, from 86.9% with the culture-fair scale to 82.0% with the SPMSQ, over the seven scales (Table 3). To avoid tautology in the analyses of this paper, none of the screen scales have been adjusted for ethnic or educational group. However, it is worth noting that the AUC for the SPMSQ only increased to 0.837 when the recommended "corrections" were applied. Some subjects did not provide adequate information for scoring any conventional screening scale. The majority of these persons were in nursing homes or other long-term home care; and when they were seen by the clinical assessment team, 90% were diagnosed with dementia. As a result, persons providing inadequate data for scoring the scales were considered screen-positives in the NMAP study, and they were treated as scoring at the highest levels of errors on every scale when computing ROCs. This group of 795 from the NMAP Reporting Component includes 70 such persons; 63 were diagnosed with dementia, and 7 without dementia. AUC for every scale was lower when these 70 were excluded, but rank order of AUC for the scales was not affected. AUC ranged nearly 11% when ethnicity was introduced, from 90.0% with the Blessed Scale among non-Latino whites to 79.1% for the SPMSQ among Latinos; and AUC varied over 20% with education, from 92.7% with the culture-fair test among the high school graduates to 72.4% for the SPMSQ among those with <5 years of schooling. Differences in AUC for individual scales across ethnic groups ranged from 8.0% for the MMSE to 4.0% for the CARE Dementia Scale. AUC differences for individual scales across education groups ranged from 17.6% for the SPMSQ to 9.4% for the CARE Dementia Scale. All scales had their highest AUC with the non-Latino whites, and all had their lowest AUC with the Latino subjects.TABLE 3Area under the curve (AUC) for seven cognitive screening scales and range of differences in AUC across three sociocultural and three educational groups (N = 795 diagnosed subjects)ScreeningOverallDifferences in AUC by:ScaleAUCSocio-CultureEducationBlessed0.8670.0570.127CARE Dementia0.8520.0400.094Kahn-Goldfarb MSQ0.8560.0440.112MMSE0.8390.0800.131SPMSQ0.8200.0680.176CARE Homogeneous0.8640.0430.115"Culture-fair"0.8690.0480.118Note: NMAP, Reporting Component, Extended Prevalence Period, Transcultural Pilot Study subjects excluded. Three ethnic groups: non-Latino white, African-American, and Latino; and three education groups: < 5 years of school, 5–11 years of school, and ≥ 12 years of school completed.(Kahn-Goldfarb) MSQ = Mental Status Questionnaire; MMSE = Mini-Mental State Exam; SPMSQ = Short Portable Mental Status Questionnaire; CARE = Comprehensive Assessment and Referral Interview. Open table in a new tab Note: NMAP, Reporting Component, Extended Prevalence Period, Transcultural Pilot Study subjects excluded. Three ethnic groups: non-Latino white, African-American, and Latino; and three education groups: < 5 years of school, 5–11 years of school, and ≥ 12 years of school completed. (Kahn-Goldfarb) MSQ = Mental Status Questionnaire; MMSE = Mini-Mental State Exam; SPMSQ = Short Portable Mental Status Questionnaire; CARE = Comprehensive Assessment and Referral Interview. Not every screening scale generated scores that exactly equaled 90% sensitivity when cut-scores were used. Therefore, a different procedure was adopted to provide a more equal basis for comparisons. For each scale, specificity at the point of intersection of the ROC curve with the horizontal line indicating the 90% sensitivity level was interpolated from a straight line joining the closest points above and below 90% produced by actual scale score values. This was done separately for the sample as a whole and for the different ethnic and education groups. With sensitivity for the criterion diagnosis set at 90% using ROC curves, all seven scales had their highest specificities among the non-Latino whites; and all had their lowest specificities among the La
Referência(s)