The Alzheimer’s Questionnaire: A Proof of Concept Study for a New Informant-Based Dementia Assessment

Abstract

The aim of this pilot study is to determine the feasibility and clinical utility of a brief, informant-based screening questionnaire for Alzheimer’s disease (AD) that can be administered in a primary care setting. The Alzheimer’s Questionnaire (AQ) was administered to the informants of 188 patients in 3 dementia clinics (50 cognitively normal, 69 mild cognitive impairment (MCI), 69 AD). Total score for the AQ is based upon the sum of clinical symptom items in which the informant responds as being present. Clinical symptoms which are known to be highly predictive of the clinical AD diagnosis are given greater weight in the total AQ score. The mean time of administration of the AQ was 2.6 ± 0.6 minutes. Sensitivity and specificity were found to be high for detecting both AD (98.55, 96.00) and MCI (86.96, 94.00) with ROC curves yielding AUC values of 0.99 and 0.95, respectively. This pilot study indicates that the AQ is a brief, sensitive measure for detecting both MCI and AD and could be easily implemented in a primary care setting.

INTRODUCTION

Confidence in making the diagnosis of Alzheimer’s disease (AD) and mild cognitive impairment (MCI) remains elusive. Evidence suggests that physicians, bombarded by demands of care by increasing numbers of medical conditions and available treatments, are not sufficiently sensitive to signs of cognitive impairment or early dementia.

Many physicians do not screen for cognitive problems in their practices unless they receive complaints from either patients or patients’ families [1–3]. This is unfortunate since a majority of patients with a dementing illness do not report cognitive problems to their health care providers and, on average, family members do not seek medical attention for the patient until several years after the onset of symptoms. As a result, recognition of dementia by primary care physicians is poor until it is moderately advanced [3,4]. Providers cite a lack of confidence in diagnosing AD as a primary reason that nearly half of AD patients remain undiagnosed [1,5,6]. Delaying diagnosis results in increased likelihood of disease progression before intervention is attempted [7]. Screening has been proposed to help combat under-diagnosis but validated, structured, interview based instruments are lacking. The desirable characteristics for a clinician-administered screening instrument include high sensitivity, high specificity, short administration time, minimal training requirements for the instrument administrator and simplicity of scoring [7].

We have developed the Alzheimer’s Questionnaire (AQ), a clinician-administered and informant-based screening instrument as a way to quickly and accurately detect cognitive impairment. Scores for some items are weighted based on their ability to accurately predict the clinical AD diagnosis which is made based on the results from other validated instruments. The AQ offers the advantage of asking simple yes/no questions in a weighted format that gives an absolute score without requiring interpretation of individual domains. This will aid clinicians in asking the most pertinent questions when screening for cognitive decline in the primary care setting [2].

METHODS

Development of the AQ

Items for the AQ are based on those from other widely used informant-based assessments [8,10–12], but have been adapted for ease and speed of administration. Items for the AQ were selected and approved by a group of clinicians with extensive experience in dementia assessment. The items were selected based on their face validity to assess each of the AQ domains. Six items were selected to be weighted in the AQ total score as it was agreed by the clinicians that these items would clearly differentiate an impaired individual from a cognitively normal individual.

Study participants

The AD and MCI subjects were drawn from the practices of three physicians (MS, RY, US). The cognitively normal (NC) subjects were administered the AQ as part of their annual assessment for a brain donation program as all are required to provide a collateral informant. Since this is a data gathering project, an IRB exemption was granted.

Included in the study were 188 subjects, 50 of which were designated NC, 69 were MCI cases, and 69 were AD cases. The AD subject met NINCDS-ADRDA [13] criteria for a clinical diagnosis of probable and possible AD. Our NC subjects were defined as having no demonstrable cognitively-based limitations of activities of daily living including employment by informant report. MCI cases were diagnosed as such based on Petersen criteria [14]. Consensus diagnosis with a neurologist, geriatric psychiatrist, and neuropsychologist was used to determine the clinical status of each subject. Rigorous criteria were used to exclude anyone with any type of symptomatic or severe brain related neurological or psychiatric illness. Excluded conditions included mental retardation, epilepsy, cerebral infarction or hemorrhage, multiple sclerosis, brain tumor, major depressive disorder (unipolar or bipolar), schizophrenia, traumatic brain injury, and substance abuse. This was done by prospective interview of the participant and careful scrutiny of the medical records. Each subject was asked to identify an informant to provide additional information on cognitive and functional changes.

Administration of AQ

The AQ consists of simple yes/no questions in a weighted format pertaining to five domains which are: Memory, Orientation, Functional Ability, Visuospatial and Language (Table 7). Points for each question that are answered “yes” are summed to give a total score. Each subject was accompanied by the informant to a clinic, where the AQ was administered to the informants of consecutive patients.

Appendix 1.

The Alzheimer’s Questionnaire

	Yes	No	Weighted Score
Memory
Does the patient have memory loss?			1
If so, is their memory it worse than a few years ago?			1
Does the patient repeat questions OR statements OR stories in the same day?			2
Have you had to take over tracking events OR appointments? OR Does the patient forget appointments?			1
Does the patient misplace items more than once a month? OR Does the patient misplace objects so that he or she cannot find them?			1
Does the patient suspect others are moving, hiding or stealing items when they cannot find them?			1
Orientation
Does the patient frequently have trouble knowing the day, date, month, year, time? OR Does the patient have to use cues like the newspaper or the calendar to know the day and date more than once a day?			2
Does the patient become disoriented in unfamiliar places?			1
Does the patient become more confused outside the home or when traveling?			1
Functional Ability
Excluding physical limitations (e.g., tremor, hemiparesis, etc.), does the patient have trouble handling money (tips, calculating change?)			1
Excluding physical limitations (e.g., tremor, hemiparesis, etc.), does the patient have trouble paying bills or doing finances OR Are family members taking over finances because of concerns about ability?			2
Does the patient have trouble remembering to take medications or tracking medications taken?			1
Is the patient having difficulty driving? OR Are you concerned about the patient’s driving? OR Has the patient stopped driving for reasons other than physical limitations?			1
Is the patient having trouble using appliances (e.g., microwave, oven, stove, remote control, telephone, alarm clock)?			1
Excluding physical limitations, is the patient having difficulty in completing home repair or other home related tasks (housekeeping)?			1
Excluding physical limitations, has the patient given up or significantly reduced activities such as golfing, dancing, exercising, or crafts?			1
Visuospatial
Is the patient getting lost in familiar surroundings (own neighborhood)?			2
Does the patient have a decreased sense of direction?			1
Language
Does the patient have trouble finding words other than names?			1
Does the patient confuse names of family members or friends?			2
Does the patient have difficulty recognizing people familiar to him/her?			2

Statistical analysis

The data were analyzed by first evaluating the sensitivity and specificity of the AQ with regard to identifying both MCI and AD cases. The accuracy of the AQ was then analyzed by using receiver operating characteristic (ROC) curves and their associated area under the curve (AUC) value. The psychometric properties of the AQ were then analyzed through a principal component factor analysis and by Cronbach’s alpha which assessed the AQ’s internal validity. In addition, correlations of the AQ domain scores were also derived in order to demonstrate internal validity. Analysis of covariance (ANCOVA) was also used to discern statistically significant group differences in AQ scores between the three clinical groups.

RESULTS

The AQ was administered to the informants 188 subjects. Individuals with Mini-Mental Status Examination (MMSE) scores below 20 were excluded in order reduce the amount of overall variability in the data and so that the data better reflected a population that is likely to be seen in a primary care setting for cognitive complaints. The sample consisted of 45.7% (n = 86) females and 54.3% (n = 102) males. Detailed demographic characteristics are displayed in Table 1. The mean time of administration of the AQ was 2.6 ± 0.6 minutes.

Table 1.

Demographic Characteristics of Study Sample

	NC	MCI	AD	Total
N	50	69	69	188
Mean Age (sd)	77.60 (7.33)	74.61 (7.71)	78.68 (7.21)	76.90 (7.61)
Mean Education (sd)	15.48 (2.85)	14.61 (2.60)	14.52 (2.57)	14.81 (2.67)
Mean MMSE (sd)	28.86 (1.31)	27.28 (1.99)	24.09 (2.50)	26.53 (2.83)
Mean AQ Score (sd)	2.12 (2.31)	11.06 (5.12)	17.64 (4.84)	11.10 (7.53)

NC – Normal Control; MCI – Mild Cognitive Impairment; AD – Alzheimer’s Disease.

Sensitivity and specificity of the AQ were found to be high for detecting both MCI and AD. In addition, ROC curve analysis yielded high AUC values. Values for sensitivity, specificity, and AUC are displayed in Table 2. Graphical representations of the ROC analyses are displayed in Figs 1 and 2. Internal validity of the AQ was determined to be high as Cronbach’s alpha was equal to 0.88. Factor analysis was conducted using the principal component analysis method and showed that all 21 items on the AQ loaded strongly onto one factor which accounted for 33.26% of the total variance with an Eigen value of 6.98.

Table 2.

Sensitivity, Specificity, and AUC of the AQ

	Sensitivity (95% CI)	Specificity (95% CI)	AUC (95% CI)
MCI	86.96 (76.70–93.90)	94.00 (83.50–98.7)	0.95 (0.90–0.98)
AD	98.55 (92.20–100.00)	96.00 (86.30–99.50)	0.99 (0.96–1.00)

MCI – Mild Cognitive Impairment;

AD – Alzheimer’s Disease.

Fig. 1.

ROC Curve for MCI (AUC = 0.95).

Fig. 2.

ROC Curve for NC versus AD (AUC = 0.99).

Correlations between the domain scores of the AQ were also evaluated to further demonstrate internal validity and are shown in Table 3. All correlation values are significant at the p < 0.0001 level. Analysis of covariance (ANCOVA) was used to analyze group differences on the AQ. After accounting for the effects of age and education, statistically significant differences on mean AQ score were present between all three clinical groups [F = 177.85 df = (2, 185), p < 0.0001].

Table 3.

Correlation of AQ Domain Scores

Domain	Memory	Orientation	Functional ability	Visuospatial	Language
Memory	———	0.80	0.82	0.55	0.64
Orientation	0.80	———	0.81	0.59	0.63
Functional Ability	0.82	0.81	———	0.59	0.66
Visuospatial	0.55	0.59	0.59	———	0.41
Language	0.64	0.63	0.66	0.41	———

p-value for all correlations is significant at the 0.0001 level.

A separate analysis of the data was conducted with the weights removed from the weighted items. In general, removing the weights did not change sensitivity, specificity, and AUC values (Table 4). Correlations among the AQ domain scores were similar to those found with weighted scores (Table 5). However, the Language domain had notable increases in its correlations with Memory, Orientation, and Functional Ability in the unweighted analysis. In addition, the factor analysis results were almost identical to those of the weighted analysis and Cronbach’s alpha was slightly higher (0.89) for the unweighted analysis.

Table 4.

Sensitivity, Specificity, and AUC of the AQ With Unweighted Items

	Sensitivity (95% CI)	Specificity (95% CI)	AUC (95% CI)
MCI	87.14 (77.00–93.90)	92.73 (82.40–98.00)	0.94 (0.89–0.98)
AD	95.65 (87.80–99.10)	98.18 (90.30–100.00)	0.99 (0.96–1.00)

MCI – Mild Cognitive Impairment; AD – Alzheimer’s Disease.

Table 5.

Correlation of AQ Domain Scores with Unweighted Items

Domain	Memory	Orientation	Functional ability	Visuospatial	Language
Memory	———	0.80	0.81	0.63	0.66
Orientation	0.80	———	0.80	0.65	0.64
Functional Ability	0.81	0.80	———	0.62	0.68
Visuospatial	0.63	0.65	0.62	———	0.44
Language	0.66	0.64	0.68	0.44	———

p-value for all correlations is significant at the 0.0001 level.

In addition, several items on the AQ that appeared to be similar with respect to content and construct were identified and analyzed to determine if any of the items should be eliminated. These consisted of six questions among three of the domains. Each domain contained two questions that were identified for further analysis. Kappa statistics were calculated for each pair of questions to determine the extent to which they were answered similarly.

For the Orientation domain, “Does the patient become disoriented in unfamiliar places?” and “Does the patient become more confused when travelling outside the home?” yielded a Kappa of 0.34 (0.01, 0.67). For the Visuospatial domain, “Is the patient getting lost in familiar surroundings?” and “Does the patient have a decreased sense of direction?” yielded a Kappa of 0.34 (0.05, 0.62). For the Language domain, “Does the patient confuse names of family members or friends?” and “Does the patient have difficulty recognizing people who are familiar to him/her?” yielded a Kappa of 0.34 (0.01, 0.67).

DISCUSSION

Two important and conclusive findings are highlighted within the present study. First, the AQ is a sensitive measure for detecting both AD and MCI. Second, the AQ is a time-efficient and easily administered tool with a simple scoring system. As the time taken to administer AQ is less than 3 minutes, making it easy to implement in a primary care setting to screen for cognitive problems. The simplicity of the AQ is reflected in that the total score is easily calculated by summing the number of items that have a “yes” response.

The rationale for weighting certain items on the AQ is that they reflect the presence of cognitive symptoms which are known to be highly predictive of the clinical AD diagnosis, such as disorientation to time (e.g., day of the week, month) and repeating statements and questions within a short period of time [15]. This differentiates the AQ from other informant-based instruments that give equal weight to all of their items as it is then problematic to accurately differentiate cognitive symptoms that are related to AD versus normal aging. The result of utilizing weighted scores for those items that are highly predictive of clinical AD is that high diagnostic accuracy, as demonstrated by the sensitivity, specificity, and ROC curves, is achieved which strongly supports the clinical validity of AQ. In addition, this study also demonstrated high internal validity of the AQ through factor analysis and also with a high Cronbach’s alpha. Specifically, the factor analysis shows that the items of the AQ accurately assess memory and other cognitive components that are indicative of MCI and AD.

Analyses of the data without the weights showed no significant differences among the statistical measures; however the inclusion of weights on certain items appears to optimize sensitivity and overall diagnostic accuracy for AD. The unweighted analysis also showed an increase in correlation values among certain domains. Specifically, the Language domain showed increased correlations with Memory, Orientation, and Functional Ability. The reason for this is unclear, but it is possible that removing the weights simply made the data fit a more linear pattern. In addition, questions that appeared to be overlapping in construct measurement did not overlap as shown by the low rate of agreement within the question pairs in each domain. Although these items appear to be similar, they are measuring distinct phenomena.

Although several other informant-based dementia questionnaires have been developed, they have not been validated as accurate instruments in detecting individuals with MCI. This is important as identifying individuals in the earliest stages of cognitive decline will be necessary as the development of disease-modifying therapies become available. Currently-used instruments that are clinician administered such as the MMSE [16, 17], the neurobehavioral cognitive examination [18], the 7 minute screen [19], the time and change test [20], the memory impairment screen [21], the clock drawing test [22], and the mini-cog [23] have demonstrated relatively good diagnostic ability in AD patients. However, the ability of these instruments to identify individuals with MCI is questionable.

In addition, currently used informant-based instruments have not been shown to accurately identify individuals with MCI. The most common clinician-administered [16–23] and informant-based [8–12,24–26] instruments have demonstrated specificities and sensitivities exceeding 80% in identifying AD cases and all take less than 10 minutes to administer. Relative to the most widely used of these instruments, the AQ has higher sensitivity and specificity with regard to identifying AD cases (Table 6), but also high sensitivity and specificity in identifying MCI. In addition, its administration time is comparable and in many cases takes less time to administer.

Table 6.

Comparison of AQ Performance with AD8 and IQCODE in AD

Instrument	Sensitivity	Specificity	AUC	Cronbach’s alpha
AQ	98.55	96.00	0.99	0.88
AD8	85.00 [8]	86.00 [8]	0.83 [8]	0.86 [25]
IQCODE	79.00 [26]	82.00 [26]	0.85 [26]	0.93–0.97 [9]

It is important to note that the AQ is not intended to replace a full diagnostic work-up that is typically done when assessing individuals with memory problems. It should also be noted that the AQ was not used in a general practice setting so it is unclear whether the results of this study represent that of the general geriatric population. This study utilized patients who were seen by dementia specialists and as a result the sample used is biased to a certain extent. Although the ultimate goal is to employ this instrument in general practice, it was employed in specialty practices during this pilot study. In spite of these shortcomings, the AQ may be an extremely useful tool to clinicians who require the use of a brief and accurate assessment of cognition in order to determine if a patient might require further evaluation. Given its diagnostic accuracy, ease of scoring, ease of administration, and short length of time needed for administration the AQ would be of great value to many clinicians who have an extremely limited amount of time in order to assess individuals with memory and cognitive problems.