|Home | About | Journals | Submit | Contact Us | Français|
Diagnostic drift characterizes change in diagnosis and diagnostic classification over time. The Clinical Dementia Rating (CDR) is used commonly in dementia diagnosis and staging of dementia severity. Whether increasing efforts to diagnose dementia at earlier symptomatic stages has led to diagnostic drift in the CDR is unknown.
To examine dementia severity as determined by the CDR over time.
Secondary analysis of data from longitudinal studies of aging and dementia.
An Alzheimer’s Disease Research Center (ADRC), where a variety of clinicians contributed CDR ratings over the course of the study.
Adults aged 63 to 83 years with no (CDR 0), very mild (CDR 0.5) or mild (CDR 1) dementia enrolled in the ADRC at any time from August 1979 to May 2007.
Within each CDR group changes in scores on standardized psychometric tests with time were examined using multiple linear regression analyses. These tests included the Mini Mental State Examination, Short Blessed Test, Wechsler Memory Scale Logical Memory IA-Immediate, Blessed Dementia Scale, and a psychometric composite score.
A total of 1768 participants met inclusion criteria. Over time, participants were older, more educated, more likely to be minorities, and less likely to be male. Statistically significant change in psychometric test performance over time occurred only within the CDR 1 group for Logical Memory and the psychometric composite, but the degree of change was minimal.
Despite changes in participant characteristics, the CDR demonstrates general stability for assessment of dementia over almost three decades.
Diagnostic drift describes change in diagnosis and diagnostic classification over time.1 It can influence assessment methods and diagnostic algorithms. Consequently, diagnostic drift is important to consider in longitudinal studies that depend on the accurate and reliable identification of disease and staging of disease severity.2–4 Precise comparison of the prevalence and incidence of dementia over time is particularly important given the projected increase in the number of individuals with dementia of the Alzheimer type (DAT) over the next several decades. The Clinical Dementia Rating (CDR)5,6 has been used for nearly three decades in the evaluation and staging of dementia. The CDR was developed at Washington University for use in longitudinal studies of aging and dementia.5 The validity7 and reliability8 of the CDR have been demonstrated, including in multicenter studies.9 A literature review conducted in November 2008 revealed 708 references since 1982 for the CDR, which has been translated into 60 languages and dialects.
Changes have been made to the CDR and its scoring rules over the years in response to growing knowledge of dementia symptomatology.6 With increasing clinical and research interest in earlier stages of dementia,10,11 a relevant issue is whether improvements in diagnostic precision caused “drift” in the CDR over time such that the average level of cognitive impairment within a particular CDR rating has decreased. The question of “CDR drift” is the focus of the current study.
Data from the initial clinical assessments of newly enrolled participants over the course of our longitudinal studies from August 1979 through May 2007 were used. Participant inclusion criteria were age at the first assessment between 63 and 83 years (the age range of the original sample recruited in 1979) and a CDR indicating no (0), very mild (0.5), or mild (1) dementia at the initial assessment. We excluded participants with moderate (CDR 2) or severe (CDR 3) dementia because we anticipated that any occurrence of “drift” would be toward earlier dementia detection.
Experienced clinicians conducted a clinical assessment to evaluate features indicative of a dementing disorder. Clinicians interviewed the participant and a knowledgeable collateral source, typically a spouse, adult child, or other relative, and completed a detailed neurological examination of each participant. The assessment protocol evaluated intra-individual cognitive change, and the clinician determined whether memory and thinking difficulties consistent with decline from an individual’s baseline level of cognitive function were present. The clinical assessment also included the Blessed Dementia Scale (BDS)12 as a means to assess decline in function in daily activities and two brief cognitive screening instruments: the Mini-Mental State Examination (MMSE)13 and the Short Blessed Test (SBT).14 In determining the CDR and the clinical diagnosis, the clinician was unaware of the results of psychometric test scores.
The CDR quantifies decline in cognitive function from a prior level and ascertains consequent impairment in routine activities. The clinician synthesized information obtained from the clinical assessment to use the CDR to rate cognitive function in six domains: memory, orientation, judgment and problem solving, community affairs, home and hobbies, and personal care. The CDR describes five levels of impairment 0 (none), 0.5 (very mild or questionable), 1 (mild), 2 (moderate), and 3 (severe) for each of these six domains, with the exception of personal care for which scores of 0, 1, 2, and 3 are described. A global CDR score was derived from individual domain scores as delineated by a standard scoring algorithm.6 The clinician also diagnosed the presumed cause of the dementia for all participants with a CDR of 0.5 or greater, in accordance with standard clinical diagnostic criteria as previously described.15 The clinical diagnosis of DAT in our sample, even for individuals at the CDR 0.5 stage, is verified by the neuropathologic diagnosis of Alzheimer’s disease in 92% of cases coming to autopsy.15,16 Demographic characteristics including race and education were obtained per participant and collateral source report.
Over the course of our longitudinal studies, individual clinicians have changed. Each new clinician (neurologist, psychiatrist, geriatrician, or clinical nurse specialist) completes extensive training under the supervision of one of us (JCM) which includes observations of clinical assessments and CDR scoring. Subsequently the clinician trainee reviews a series of teaching and reliability videotapes of participants across the full spectrum of CDR stages and with DAT and nonDAT diagnoses. Certification in the CDR is achieved with an 80% agreement or better by the trainee with the “gold standard” CDRs developed by JCM.
At enrollment and every two years thereafter, the collateral source and participant interviews are videotaped for independent, blinded review by a second clinician. Disagreements, if any, in CDR scores or diagnoses between the examining and reviewing clinicians are resolved by discussion during a weekly clinical conference. The clinical conference provides quality assurance, promotes inter-rater reliability, and facilitates uniform administration of the assessment protocol.
A 90-minute psychometric battery was administered by experienced psychometricians within two weeks of the clinical assessment.17 Psychometricians were blinded to the results of the clinical assessment, including CDR score, and to previous psychometric test results. The psychometric composite includes information from tests of episodic memory evaluated by Logical Memory and Associate Learning subtests of the Wechsler Memory Scale (WMS)18 and the Benton Visual Retention Test (Form C, 10-second exposure).19 Measures of semantic memory included the Wechsler Adult Intelligence Scale (WAIS) Information subtest20 and the Boston Naming Test.21 Timed measures of psychomotor and visuospatial ability included WAIS Block Design,20 WAIS Digit Symbol,20 Trail-Making Test Part A and Part B,22 and Crossing-Off.23 Additional psychometric tests were WMS Mental Control,18 word fluency for S and P24 with 60 seconds permitted for each letter, Digit Span (forward and backward),18 and copying Form D of the Benton Visual Retention Test,19 an untimed visuospatial test. Details of the scoring algorithm are described in a previous publication.17 Scores from the tests in the psychometric battery were combined in a composite factor score using the weights obtained from a principal components analysis of 81 nondemented participants that produced a single factor that accounted for 34% of the variance.25
The Washington University Human Studies Committee approved all procedures. Informed consent was obtained from participants and collateral sources after the study was described fully.
Within each CDR stratum, and over the time period that each test was administered, changes in scores with time on a subset of standardized tests from the clinical and psychometric assessments were examined. These tests were the BDS,12 MMSE,13 SBT,14 the Wechsler Memory Scale Logical Memory IA-Immediate19 and the psychometric composite score.26 Logical Memory was selected because it is a commonly used measure of episodic memory. Higher scores on the SBT and BDS and lower scores on the MMSE, Logical Memory, and psychometric factor score indicate greater cognitive impairment. Over the history of the research program, certain tests were administered at some times and not others. The dates over which data for each test were available are as follows: BDS, August 1979 through May 2007; Logical Memory and the psychometric composite, August 1979 through September 2005; SBT, April 1984 through May 2007; and MMSE, October 1996 through May 2007.
Time was treated as a continuous variable using one-day increments with the time scale converted to years after dividing by 365.25. Because inspection of scatterplots indicated that the relationship, if any, between time and scores on each of the tests examined was linear, Pearson product-moment correlations were used to examine the unadjusted association of scores on each test with time.
Multiple linear regression was used to examine whether scores on each test were associated with time of testing after adjustment for age, gender, race (minority vs. non-minority), and education. Although the initial years of the study enrolled only demented individuals thought on clinical grounds to have DAT, the clinical diagnostic criteria were expanded to include nonDAT causes of dementia on August 22, 1990. For participants with CDR 0.5 and 1, the multiple regression analyses were repeated using data collected beginning August 22, 1990, adjusting for receipt of a DAT diagnosis (yes vs. no), as well as the other demographic variables.
To examine change in test scores across time graphically, least-square mean scores on each test per year adjusting for the demographic variables were calculated and plotted for years in which that test was given the entire year.
Changes in participant demographics over time were tested using Pearson product-moment correlations for age and education, and using logistic regression for sex, race, and receipt of a DAT diagnosis. Additional subgroup analyses examined demented participants with a DAT diagnosis only.
A total of 1768 participants met inclusion criteria, including 580 CDR 0, 684 CDR 0.5, and 504 CDR 1 individuals. The majority of CDR 0.5 (72.9%) and CDR 1 (93.8%) individuals were diagnosed with DAT, and the remainder had nonDAT diagnoses, including dementia with Lewy bodies, frontotemporal dementia, and vascular dementia. Figure 1 shows how initial assessments for these participants are distributed across the study years. Fluctuations in the number of initial assessments during the early years of the study are explained by differing emphases for recruiting efforts during those years; enrollment for 2007 is through May of that year.
Demographic information for study participants is presented in Table 1. The demographic characteristics of the sample changed throughout the years. With time, participants enrolling for initial assessments had more years of education (r = .12, p < .0001) and were older (r = .10, p < .0001), even though the age range was restricted by the present study’s inclusion criteria. Additionally, participants were less likely to be male (OR=0.985, 95% CI=0.972–0.998, p=.0229) and more likely to be of minority race (OR=1.111, 95% CI=1.080–1.143, p<.0001) as time increased. For CDR 0.5 and CDR 1 participants with a first assessment from August 1990 forward, the likelihood of having a DAT diagnosis decreased (OR=0.950, 95% CI=0.916-0.985, p=.0051). This trend is probably because enrollment in the initial years of the studies was restricted to individuals with DAT.
Significant unadjusted correlations of test scores with time were found for the MMSE (r=.13, p=.0089) and the psychometric composite (r=.10, p=.0133) within the CDR 0.5 group, and for the BDS (r=.12, p=.0099) and Logical Memory test (r=.13, p=.0069) within the CDR 1 group, such that scores on each test increased with increasing time.
There were no significant changes in test scores with time for the CDR 0 group in the adjusted analyses (Table 2). Within the CDR 0.5 group, there was a statistically significant increase in scores on the psychometric composite (b=.03, p=.0020) with time when adjusted for age, sex, race, and education, but this effect was no longer significant after adjustment for receipt of a DAT diagnosis (Table 2). For the CDR 1 group, scores on the SBT (b=-.13, p=.0447), Logical Memory (b=.04, p=.007), and the psychometric composite (b=.03, p=.004) changed significantly with time in the direction of less cognitive impairment. After adjustment for DAT diagnosis in this group, only Logical Memory and the psychometric composite showed significant effects of small magnitude (Table 2). Figure 2 shows the least-square mean test scores at each year within each CDR group, adjusted for age, sex, race, and education, for time periods that the test was administered the entire year. Analyses conducted for the subset of participants with DAT demonstrated comparable results.
As evidenced by stability of performance on standardized psychometric tests over time, the CDR generally reflects the same level of impairment now as it did nearly 30 years ago. The exceptions were for Logical Memory performance and the psychometric composite score that showed statistically significant effects for less impairment with time, but these effects were limited to the CDR 1 group and were of such slight magnitude that they likely are not clinically meaningful (Figure 2). The CDR has remained stable even as the sample has become older, more highly educated, more diverse, and increasingly female and with turnover in the individual clinicians. The generalizability of the findings to other research groups is suggested by studies that demonstrate the reliability of the CDR in multicenter trials.26
This study demonstrates the general stability of the CDR and emphasizes the utility of ascertainment of cognitive change as operationalized by the CDR.17 The body of literature examining diagnostic drift in dementia assessment is limited. The study has a number of strengths. The participants are well-characterized, and the study is of 28 years duration. As the sample has become increasingly diverse in terms of race, dementia diagnosis, and clinician staff (initially neurologists and psychiatrists and now neurologists, psychiatrists, geriatricians, and clinical nurse specialists) over time, the CDR demonstrated continued stability.
The study has some limitations. Our sample is a convenience sample, which may limit the generalizability of the findings. The stability of the CDR was assessed at the research center that developed the CDR and clinicians were trained at the center. Additional studies conducted at other centers that use the CDR could help to establish further that the CDR demonstrates no substantive evidence of diagnostic drift.
This study was supported by grants P50AG05681 and P01AG03991 from the National Institute on Aging, 1K12RR024994 and the Postdoctoral Program of 1UL1RR024992-01 from the National Center for Research Resources, the Charles and Joanne Knight Alzheimer Research Initiative of Washington University’s Alzheimer’s Disease Research Center (ADRC), and a generous gift from the Alan and Edith Wolff Charitable Trust. Dr. Williams had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis. We thank the ADRC Psychometric Core led by Dr. Martha Storandt for psychometric testing and her guidance in developing this study and Dr. Elizabeth Grant of the Biostatistics Core for data management. We also acknowledge, with gratitude, the ADRC investigators and staff both past and present for clinical assessments, and our participants and their collateral sources.
Disclosure: The authors report no conflicts of interest.