Abstract
INTRODUCTION
The Informant Questionnaire on Cognitive Decline in the Elderly (IQCODE) is suitable for screening dementia, particularly in populations with limited education. However, its feasibility and psychometrics in South Africa remain unassessed.
METHODS
We analyzed 1309 index participant–informant dyads from the Health and Aging in Africa: Longitudinal Study of an International Network for the Demographic Evaluation of Populations and Their Health Community in South Africa study in rural South Africa, assessing IQCODE completeness, associations with index participant/informant characteristics, single‐factor structure and reliability via factor analysis, and validity through correlations with neuropsychological tests.
RESULTS
Index participants averaged 71.9 years, with > 50% lacking formal education. IQCODE missingness declined across waves and was associated with index participant characteristics. IQCODE scores were correlated with both index participant and informant characteristics. The 16‐item IQCODE demonstrated a single‐factor structure explaining up to 66% of variance, strong internal consistency, and correlations with word and story recall, indicating high convergent validity.
DISCUSSION
IQCODE is a reliable, valid dementia assessment tool in rural South Africa, supporting broader African applications and cross‐national comparisons.
Highlights
Sixteen‐item Informant Questionnaire on Cognitive Decline in the Elderly (IQCODE) responses in rural South Africa support a single‐factor construct, explaining up to 66% of total observed variance.
The single‐factor structure is reliable, as the IQCODE items data show high internal consistency (omega hierarchical = 0.90).
The IQCODE score shows strong convergent validity with significant correlation with word and story recall memory measures.
Index participants’ characteristics are mainly associated with IQCODE completion.
IQCODE contributes to dementia assessment in rural South Africa and other settings with low education populations.
Keywords: dementia assessment, feasibility, Informant Questionnaire on Cognitive Decline in the Elderly, reliability, rural region, South Africa, validity
1. BACKGROUND
Alzheimer's disease and related dementias (ADRDs) are among the leading causes of death worldwide. 1 South Africa is experiencing rapid population aging, with the proportion of adults aged ≥ 60 expected to rise from 9.1% to 15.4% by 2050. 2 Dementia cases in South Africa are projected to increase even more rapidly, from 0.2 million in 2019 to 0.68 million by 2050. 3 However, a substantial number of dementia cases in South Africa remain undiagnosed, 4 largely due to the absence of culturally sensitive cognitive assessment tools that are feasible in settings with low education levels and diverse languages. 5
Population‐based dementia screening in Sub‐Saharan African countries, including South Africa, faces significant challenges. The use of different screening measures on the same population in this region often leads to widely varying diagnoses of dementia. 4 , 6 , 7 Many cognitive tests require skills such as drawing, reading, writing, and calculation—abilities that many older adults, particularly those in rural South Africa, may lack due to limited formal education. 5 , 8 , 9 Additionally, the high prevalence of vision and hearing impairments, as revealed by pilot studies among older adults in resource‐limited areas of South Africa, further complicates the administration of cognitive assessments, as these tests often rely on intact sensory function. 10 , 11
Informant reports from close family members are complementary in dementia assessment and widely used in screening, especially in countries with lower education and literacy levels. 12 , 13 , 14 Among various informant‐based dementia assessment tools, the Informant Questionnaire on Cognitive Decline in the Elderly (IQCODE) 15 is the most commonly used worldwide. 16 The IQCODE is especially suitable for individuals with low education and literacy levels 15 and shows no differences in test accuracy across different languages of administration. 16 A study in Brazil found that the IQCODE outperformed the Mini‐Mental State Examination (MMSE) when directly administered to older adults in accurately screening for dementia cases. 17
IQCODE, which consists of 26 items, measures a single concept—cognitive decline over a 10 year time window by interviewing an informant familiar with the index participant, typically a spouse or child. 15 IQCODE was first implemented in Australia among informants of both the general population and individuals with dementia, demonstrating high internal reliability, test–retest reliability, and discriminant validity. 18 A shorter 16‐item version was later developed, showing strong correlation with the original form and greater sensitivity to early cognitive decline. 19 The 16‐item IQCODE has since been widely adopted in population‐representative aging surveys worldwide, including in the United States, UK, and India, 12 , 14 as well as South Africa. 13 A previous study underscored the strong correlations between IQCODE and objective cognitive function tests for cross‐country comparisons of dementia. 14 However, its performance remains unexamined in South Africa and, more broadly, in Sub‐Saharan Africa. To our knowledge, this is the first study to evaluate the feasibility, inter‐item reliability, and validity of IQCODE in a rural South African context.
RESEARCH IN CONTEXT
Systematic review: The authors reviewed literature on the reliability, validity, and application of the Informant Questionnaire on Cognitive Decline in the Elderly (IQCODE) in African countries using Google Scholar and PubMed. IQCODE is a widely used informant‐based tool for dementia screening, particularly in populations with low education. It facilitates cross‐national comparisons in dementia research, yet its performance in South Africa remains underexplored.
Interpretation: In this rural South African population, IQCODE demonstrated a high completion rate, strong internal consistency, a single‐factor structure, and high convergent validity with memory‐related cognitive measures. These findings support its use in dementia screening for populations with limited formal education.
Future directions: Further research should examine IQCODE's role in cross‐national dementia assessment, including its cultural adaptability and predictive validity. Longitudinal studies can explore its effectiveness in tracking cognitive decline, enhancing its application in diverse global settings.
2. METHODS
2.1. Data
This study uses data from the dementia‐focused sub‐cohort of the Health and Aging in Africa: Longitudinal Study of an International Network for the Demographic Evaluation of Populations and Their Health (INDEPTH) Community in South Africa (HAALSI). HAALSI is an ongoing cohort study since 2014 following 5059 individuals aged ≥ 40 in Agincourt, a rural community in South Africa characterized by high rates of illiteracy and unemployment. 9 HAALSI is a harmonized sister study to the Health and Retirement Study (HRS) in the United States, 20 collecting longitudinal data on social, economic, and biological characteristics related to infectious and chronic health conditions, including cognitive function and dementia. 9
In 2019 and 2020, a new HAALSI study recruited ≈ 700 HAALSI participants aged ≥ 50 ∖ who had taken part in HAALSI Wave 2 (the parent cohort) between 2018 and 2019 to examine the prevalence, incidence, and risk factors of cognitive decline and dementia. 21 In alignment with the Harmonized Cognitive Assessment Protocol (HCAP), 22 the HAALSI dementia sub‐cohort study (HAALSI‐HCAP) implemented a comprehensive approach that included a culturally adaptive cognitive battery, 21 along with informant interviews and neurological evaluations to ensure valid assessment of cognitive impairment and dementia while facilitating cross‐national comparisons.
The parent cohort sample was categorized into five dementia risk groups based on cognitive performance in HAALSI Wave 2: highest, moderate, low, lowest, and those using a proxy for cognitive reporting. A stratified random sampling strategy was used. Within each stratum, a predetermined number of index participants were randomly selected based on a power analysis. In total, 710 individuals from the parent cohort were contacted, and 635 ultimately participated in the HAALSI‐HCAP Wave 1 study. Further details on sampling methodology and the survey can be found in a previous study. 21
In 2022, the HAALSI‐HCAP study conducted its second wave of interviews (HAALSI‐HCAP Wave 2). All participants from HAALSI‐HCAP Wave 1 were invited to continue their participation, and additional participants were recruited from HAALSI Wave 3 as refresh samples using a sampling strategy similar to that of HAALSI‐HCAP Wave 1. In total, 683 index participants participated in HAALSI‐HCAP Wave 2.
HAALSI‐HCAP index participants were asked to designate a reliable informant to complete the HCAP informant interview. Eligible informants had to be at least 18 years old, have contact with the index participant at least two to three times per week, and have known the index participant for a minimum of 5 years. 21 Nearly all HAALSI‐HCAP study index participants had an informant interview completed. Specifically, 630 index participants in Wave 1 (99.2%) and 679 index participants in Wave 2 (99.4%) completed an informant interview. Among them, 504 index participants were followed across both waves, with 125 maintaining the same informant for both survey rounds (24.8%; Figure 1).
FIGURE 1.
HAALSI‐HCAP study informant interview flow chart. HAALSI, Health and Aging in Africa: Longitudinal Study of an International Network for the Demographic Evaluation of Populations and Their Health Community in South Africa; HCAP, Harmonized Cognitive Assessment Protocol.
2.2. Measurements
The HAALSI study fully adapted the HCAP informant battery that incorporated a variety of validated instruments, such as the IQCODE, Blessed Dementia Rating Scale, and 10/66 Dementia Research Group Informant Questionnaire; the instruments were translated into Shangaan, the primary language of the study population, and then back‐translated. A pilot study was conducted among older adults not enrolled in HAALSI, with weekly data quality checks implemented throughout data collection to promptly address any discrepancies. 21
This study focused on the 16‐item IQCODE, a widely used dementia screening tool. 19 The informant was asked to compare the index participant's current performance to their performance 10 years ago for each activity described in the 16 IQCODE items. Each item (Figure 2) initially offered three response options: “improved,” “not much changed,” and “worse.” If informants selected “improved” or “worse,” they were prompted to specify the degree of change with additional options: “much improved,” “a bit improved,” “a bit worse,” or “much worse.” Each IQCODE item was rated on a 5‐point scale: 1 (“much improved”), 2 (“a bit improved”), 3 (“not much changed”), 4 (“a bit worse”), and 5 (“much worse”). The IQCODE score is calculated by averaging the scores across all 16 items, resulting in a range from 1 to 5, in which higher scores indicate greater cognitive decline over the past 10 years. To ensure comparability to the original research, we followed the established practice of allowing up to three missing item values when computing the summary score. 15
FIGURE 2.
Distribution of responses to 16 IQCODE items in HAALSI‐HCAP Wave 1 and 2 (N = 1309). HAALSI, Health and Aging in Africa: Longitudinal Study of an International Network for the Demographic Evaluation of Populations and Their Health Community in South Africa; HCAP, Harmonized Cognitive Assessment Protocol; IQCODE, Informant Questionnaire on Cognitive Decline in the Elderly.
Additionally, the HAALSI‐HCAP study introduced an extra response option—“does not apply/does not do this activity.” This adaptation, also implemented in India, 12 acknowledges that certain items originally developed in Western contexts may not be relevant for older South Africans due to differences in economic resources or educational attainment. Responses of “does not do” were not assigned a rating and were treated as missing. We calculated the total number of “does not do” responses for each index participant.
In addition to informant interviews, HAALSI‐HCAP also adapted a comprehensive dementia cognitive battery directly for index participants, collecting data on general cognitive status, episodic memory, language, executive function and attention, and visuospatial/spatial memory with valid measures. 21 To assess convergent validity, this study uses several cognitive test scores related to episodic memory to examine its’ association with IQCODE score. It includes (1) Immediate and Delayed Word Recall—index participants repeat and later recall 10 words spoken by the interviewer; (2) Immediate and Delayed Story Recall—participants recall and later recall key details from two short, narrated stories. A higher score indicates better cognitive function.
2.3. Analysis strategies
We first described the characteristics of index participants and their informants across HAALSI‐HCAP waves. To evaluate the completeness of IQCODE responses, we analyzed the distribution of responses for all 16 items across both waves. We then examined associations between index participant and informant characteristics and the total number of “does not do” responses using a mixed‐effect linear regression model. This model included random intercepts to account for correlation within index participants across waves and adjusted for wave effects. Covariates include the sex, age, and education level of both index participants and informants; the index participants’ self‐care difficulty; the informants’ years of knowing the index participant; and the informants’ relationship to and frequency of seeing the index participant. The same method was used to assess associations between index participant and informant characteristics and the IQCODE score. Among index participants who participated in both waves, we further investigated whether a change in informant influenced changes in the IQCODE score and the number of “does not do” responses, using linear regression models, controlling for index participant characteristics, including age, sex, education, and changes of difficulty in self‐care (e.g., dressing, feeding, or using the toilet) status across the two waves.
To identify the latent structure of the 16‐item IQCODE response data, we conducted several exploratory analyses. All the analyses were conducted separately for Wave 1 and 2 data. First, we conducted a parallel analysis by computing the actual eigenvalues from principal axis factoring among the observed data and comparing them to eigenvalues derived from 100 randomly simulated datasets. Factors were retained if their actual eigenvalues exceeded those from the random data. Second, to further explore the optimal number of factors to retain, we used the very simple structure (VSS) procedure. 23 The VSS method evaluates the fitness of simplified factor models by allowing each variable to load on only one or a limited number of factors (i.e., low complexity). It estimates how well these simplified structures approximate the observed correlation matrix, balancing model fit with interpretability. We assessed solutions with one to eight factors and identified the number of factors that maximized the VSS fit index at low complexity levels. To supplement the evaluation of statistical adequacy provided by the VSS results, we conducted the minimum average partial (MAP) test, 24 which identifies the optimal number of factors by computing the average squared partial correlations after successively partialing out factors. The VSS complexity indicator ranges from 0 to 1, with higher values indicating a better‐fitting simple factor solution. The MAP test yields small positive values, and a lower MAP value suggests a better factor solution. The final factor solution will be determined based on the results of the exploratory analysis, considerations of parsimony, and alignment with theoretical expectations.
Once the final factor solution was selected, we reported the factor loadings (i.e., the correlations between each IQCODE item and each factor), communalities (the proportion of variance explained by the factors, calculated as the sum of squared loadings), and uniqueness (the residual variance, computed as one minus the communality). We also reported the proportion of total variance in the observed data explained by the selected factor(s).
To assess internal consistency reliability of the latent structure, we computed McDonald omega coefficients, 25 which are widely used to evaluate the degree of saturation by a general factor. This approach was appropriate given that the IQCODE items are theoretically and empirically supported as measuring a single underlying construct. 15 We specified a hierarchical model with one general factor and three group factors. The analysis was conducted using maximum likelihood (ML) estimation and an oblimin oblique rotation to allow group factors to correlate. We used the Schmid–Leiman transformation to decompose the factor structure into orthogonal general and group components, enabling the calculation of both omega total (proportion of variance that is reliably attributable to all common factors) and omega hierarchical (proportion of total score variance that is reliably attributable to the general factor alone), the two measurements of reliability. Higher values indicate greater reliability, with established cut‐offs: ≥ 0.7 (acceptable), ≥ 0.8 (good), and ≥ 0.9 (excellent). 26 , 27
To assess the convergent validity of the IQCODE, we used a linear regression to examine the association of the IQCODE score with directly assessed cognitive test scores from the same memory domain, including immediate and delayed word recall, immediate and delayed story recall. Analyses were adjusted for index participants’ age, sex, and education, and were conducted separately for Wave 1 and 2 data.
We conducted sensitivity analyses of factor analysis and reliability analysis using imputed datasets to account for the missingness observed in IQCODE item responses (Figures S1 and S2 in supporting information). IQCODE responses had two types of missingness. The first resulted from “does not do” responses, which are not scorable and affected between 0.5% and 17% of responses across the 16 IQCODE items (Figure S1). This pattern is comparable to findings from a study conducted in India, also situated in a context of limited resources and lower education attainment. 12 We assumed these were missing at random (MAR), conditional on index participant and informant characteristics. The second type stemmed from a technical error in some tablets during HAALSI‐HCAP Wave 2, causing informants to inadvertently skip sub‐questions related to the scale of “improved” or “worse” when answering IQCODE item questions (Figure S2 and Table S1 in supporting information), impacting 8% to 16% across items; this was treated as missing completely at random. As missingness may bias results, we used multiple imputation, particularly the multivariate imputation by chained equations (MICE) method, to address it. Further details on the imputation procedure are provided in the supporting information (Description of the imputation method).
Most analyses were conducted using Stata 18 (StataCorp LLC), except for factor analysis and imputation, which were performed in R (R Foundation for Statistical Computing).
3. RESULTS
On average, index participants were 71.9 years old, and > 60% were women. More than half had no formal education. According to informant reports, the majority (> 90%) did not require assistance with dressing, feeding, or using the toilet. Index participants’ characteristics remained consistent between the two waves (Table 1).
TABLE 1.
Descriptive characteristics of index participants and informants.
| Wave 1 | Wave 2 | Total | |
|---|---|---|---|
| (N = 630) | (N = 679) | (N = 1309) | |
| Index participants characteristics | |||
| Index participant age in years, mean (SD) | 70.8 (11.6) | 72.8 (11.5) | 71.9 (11.6) |
| Index participant sex | |||
| Men | 242 (38.4%) | 252 (37.1%) | 494 (37.7%) |
| Women | 388 (61.6%) | 427 (62.9%) | 815 (62.3%) |
| Index participant education level | |||
| No formal education | 347 (55.2%) | 392 (57.8%) | 739 (56.5%) |
| Some primary (1–7 years) or more | 282 (44.8%) | 286 (42.2%) | 568 (43.5%) |
| Informant reports that the index participant needs help in dressing, feeding, or using the toilet | |||
| No | 576 (91.4%) | 629 (92.6%) | 1205 (92.1%) |
| Yes | 54 (8.6%) | 50 (7.4%) | 104 (7.9%) |
| Informant characteristics | |||
| Informant age in years, mean (SD) | 44.0 (18.1) | 44.3 (15.5) | 44.2 (16.8) |
| Informant sex | |||
| Men | 201 (31.9%) | 241 (35.5%) | 442 (33.8%) |
| Women | 429 (68.1%) | 438 (64.5%) | 867 (66.2%) |
| Informant education level | |||
| No education | 104 (16.5%) | 146 (21.5%) | 250 (19.1%) |
| Secondary or less | 294 (46.7%) | 176 (25.9%) | 470 (35.9%) |
| More than secondary | 232 (36.8%) | 357 (52.6%) | 589 (45.0%) |
| Years of knowing the index participant, mean (SD) | 31.9 (14.1) | 32.9 (13.2) | 32.4 (13.6) |
| Knowing the index participant for 10 years or more | |||
| No | 39 (6.2%) | 12 (1.8%) | 51 (3.9%) |
| Yes | 591 (93.8%) | 667 (98.2%) | 1258 (96.1%) |
| Relationship to the index participant | |||
| Spouse/partner | 174 (27.6%) | 140 (20.6%) | 314 (24.0%) |
| Child/in‐law | 289 (45.9%) | 343 (50.5%) | 632 (48.3%) |
| Grandchild/in‐law | 87 (13.8%) | 90 (13.3%) | 177 (13.5%) |
| Brother/sister | 27 (4.3%) | 32 (4.7%) | 59 (4.5%) |
| Parent/in‐law | 10 (1.6%) | 2 (0.3%) | 12 (0.9%) |
| Other | 43 (6.8%) | 72 (10.6%) | 115 (8.8%) |
| Informant's frequency of seeing the index participant | |||
| Lives with index participant | 400 (63.5%) | 516 (76.0%) | 916 (70.0%) |
| Daily | 180 (28.6%) | 120 (17.7%) | 300 (22.9%) |
| Several times a week or less | 50 (7.9%) | 43 (6.3%) | 93 (7.1%) |
Abbreviation: SD, standard deviation.
Informants had an average age of 44.2 years, with > 60% being women. Most had at least a secondary education, and the education level among informants in Wave 2 was higher than in Wave 1, with a greater proportion having more than a secondary education. Informants had known the index participants for an average of 32.4 years, with almost all reporting familiarity with the index participant for at least 10 years. Approximately 24% of informants were the index participant's spouse, while approximately half were the index participant's children or children‐in‐law. Approximately 70% of informants lived with the index participant, with this proportion being higher in Wave 2 (Table 1).
The majority of informants reported that index participants’ memory function had not changed significantly compared to 10 years prior (Figure 2). A small proportion (< 10%) indicated improved cognition, while a larger proportion (between 10% and 35%) reported cognitive decline over the same period. The highest proportion of becoming “worse” was reported for the item assessing whether the index participant can “remember address and telephone number” (36.4%), followed by “learn new machines at home” (28.0%), “remember family and friends” (27.5%), “remember day and month” (27.5%), as well as “learn new things in general” (27.2%). The much lower proportion of becoming “worse” was observed in items related to remembering recent things (11.3%) or conversation and making decisions on everyday matters (10.5%). Furthermore, ≈ ≥ 10% of informants responded with “does not do this activity,” particularly for tasks such as following a story from the media, handling everyday arithmetic problems, managing financial matters, and learning to use new household appliances.
Most index participant characteristics were associated with the total number of “does not do” responses reported (Table 2). Index participants who were older, receiving less education, or had difficulties with self‐care had a higher number of “does not do” responses. Only a few informant characteristics showed significant associations (Table 2). Informants who were the index participants’ spouse compared to other relatives (not spouse, children, grandchildren, siblings, or parents) and those who saw the index participant daily rather than living together reported fewer “does not do” responses. Additionally, compared to Wave 1, Wave 2 data showed a significantly lower number of “does not do” responses (−0.674, 95% confidence interval: −0.866, −0.482). In the panel data analysis, changing an informant between waves was not significantly associated with changes in the total number of “does not do” responses (Table S2 in supporting information).
TABLE 2.
Index participant and informant characteristics associated with total number of “does not do” responses, HAALSI‐HCAP Wave 1 and 2.
| Coefficients | p values | 95% confidence intervals | |
|---|---|---|---|
| Index participant characteristics | |||
| Age | 0.029 | *** | (0.019, 0.039) |
| Sex (reference = men) | |||
| Women | −0.094 | (−0.318, 0.130) | |
| Education (reference = no education) | |||
| Some primary or more | −0.268 | * | (−0.472, −0.065) |
| Difficulty in self‐care (reference = no difficulty) | |||
| Have difficulty | 1.532 | *** | (1.187, 1.878) |
| Informant characteristics | |||
| Age | 0.007 | (−0.007, 0.020) | |
| Sex (reference = men) | |||
| Women | 0.014 | (−0.188, 0.216) | |
| Years of knowing the index participant | 0.002 | (−0.008, 0.012) | |
| Education (reference = no education) | |||
| Secondary or less | −0.163 | (−0.448, 0.123) | |
| More than secondary | −0.196 | (−0.524, 0.131) | |
| Relationship with index participant (reference = spouse) | |||
| Child/in‐law | 0.284 | (−0.161, 0.729) | |
| Grandchild/in‐law | 0.513 | (−0.069, 1.094) | |
| Brother/sister | 0.135 | (−0.379, 0.650) | |
| Parent/in‐law | 0.186 | (−0.823, 1.195) | |
| Other | 0.622 | ** | (0.164, 1.080) |
| Frequency of seeing the index participant (reference = living together) | |||
| Daily | −0.322 | * | (−0.565, −0.078) |
| Several times a week or less | −0.331 | (−0.712, 0.051) | |
| Wave (reference = Wave 1) | |||
| Wave 2 | −0.674 | *** | (−0.866, −0.482) |
Note: *p < 0.05, **p < 0.01, ***p < 0.001. A mixed effects linear regression model was used with random intercepts; 800 unique index participants with 1299 observations across two waves were included.
Abbreviations: HAALSI, Health and Aging in Africa: Longitudinal Study of an International Network for the Demographic Evaluation of Populations and Their Health Community in South Africa; HCAP, Harmonized Cognitive Assessment Protocol.
Both index participant and informant characteristics were associated with the IQCODE score (Table 3). Higher IQCODE scores were observed among older index participants, those with no formal education, and those with self‐care difficulties. Informants who knew the index participants for fewer years or were the index participant's child or child‐in‐law, compared to being the spouse, were more likely to give higher IQCODE ratings. In the panel data analysis, there was no significant difference in IQCODE scores at HAALSI‐HCAP Wave 2 between index participants who retained the same informant across both waves and those who changed informants, after adjusting for index participant characteristics and IQCODE score at HAALSI‐HCAP Wave 1 (Table S3 in supporting information).
TABLE 3.
Index participant and informant characteristics associated with IQCODE score, HAALSI‐HCAP Wave 1 and 2 data.
| Coefficients | p value | 95% confidence interval | |
|---|---|---|---|
| Index participant characteristics | |||
| Age | 0.007 | *** | (0.003, 0.010) |
| Sex (reference = men) | |||
| Women | −0.048 | (−0.122, 0.026) | |
| Education (reference = no education) | |||
| Some primary or more | −0.174 | *** | (−0.241, −0.108) |
| Difficulty in self‐care (reference = no difficulty) | |||
| Have difficulty | 0.240 | ** | (0.103, 0.376) |
| Informant characteristics | |||
| Age | 0.004 | (−0.001, 0.008) | |
| Sex (reference = men) | |||
| Women | −0.027 | (−0.092, 0.039) | |
| Years of knowing the index participant | −0.004 | * | (−0.007, −0.0002) |
| Education (reference = no education) | |||
| Secondary or less | 0.072 | (−0.023, 0.167) | |
| More than secondary | 0.075 | (−0.033, 0.183) | |
| Relationship with index participant (reference = spouse) | |||
| Child/in‐law | 0.199 | ** | (0.051, 0.346) |
| Grandchild/in‐law | 0.148 | (−0.045, 0.340) | |
| Brother/sister | 0.126 | (−0.049, 0.301) | |
| Parent/in‐law | −0.003 | (−0.315, 0.310) | |
| Other | 0.004 | (−0.143, 0.151) | |
| Frequency of seeing the index participant (reference = living together) | |||
| Daily | −0.054 | (−0.131, 0.024) | |
| Several times a week or less | −0.059 | (−0.178, 0.061) | |
| Wave (reference = Wave 1) | |||
| Wave 2 | 0.047 | (−0.015, 0.108) | |
Note: *p < 0.05, **p < 0.01, ***p < 0.001. A mixed effects linear regression model was used with random intercepts; 655 unique index participants with 1062 observations across two waves were included.
Abbreviations: HAALSI, Health and Aging in Africa: Longitudinal Study of an International Network for the Demographic Evaluation of Populations and Their Health Community in South Africa; HCAP, Harmonized Cognitive Assessment Protocol; IQCODE, Informant Questionnaire on Cognitive Decline in the Elderly.
The parallel analysis scree plots (Figures 3 and S3 in supporting information) suggest three factors and six factors among Wave 2 informants and Wave 1 informants, respectively. However, both plots show the first factor accounted for the majority of the common variable, with a steep drop‐off in subsequent factors, indicating strong unidimensionality.
FIGURE 3.
Parallel analysis scree plot for the 16 IQCODE items using Wave 2 informant data. IQCODE, Informant Questionnaire on Cognitive Decline in the Elderly.
For Wave 1 data, VSS complexity 1 achieves a maximum of 0.87 with one factor, and VSS complexity 2 achieves a maximum of 0.9 with two factors. For Wave 2 data, VSS complexity 1 achieves a maximum of 0.98 with one factor, and VSS complexity 2 achieves a maximum of 0.98 with two factors. The differences in VSS fit index are quite small between VSS complexity 1 and VSS complexity 2. In the Wave 1 data, the Velicer MAP test yielded a minimum average squared partial correlation of 0.0196 with two factors and 0.020 with one factor. In Wave 2, the corresponding values were 0.034 for two factors and 0.036 for one factor. Although the MAP values are slightly lower for the two‐factor solution in both waves, the differences compared to the one‐factor solution are small.
Although statistical criteria such as parallel analysis and MAP test suggested multiple factors, the steep drop‐off after the first eigenvalue, minimal differences in VSS and the MAP indices, and theoretical expectations support the selection of a one‐factor solution. In line with the principle of parsimony and theoretical coherence, we therefore retained a one‐factor structure.
Table 4 presents how the one‐factor structure explained the variance of each IQCODE item. Most items exhibit high loadings on the single factor, with loadings generally higher in Wave 2 than in Wave 1 (Table 4). In Wave 1, individual item loadings range from 0.45 to 0.75, whereas in Wave 2, they range from 0.42 to 0.89. Nearly all item loadings increased from Wave 1 to 2, except for Item 11—“following a story in a book or on TV”—which showed a decline. Correspondingly, the communality increased substantially from Wave 1 to 2, and the uniqueness decreased for all items except Item 11. In Wave 2, the proportion of variance explained by the single factor ranged from 0.49 to 0.82 across items, with the exception of Item 11. For more than half of the items, > 70% of their variance was accounted for by the single‐factor model. Overall, the single‐factor model explained 39% of the variance in the observed data at Wave 1 and 66% at Wave 2.
TABLE 4.
Comparison of factor loadings, communality, and uniqueness across 16 IQCODE items, HAALSI‐HCAP Wave 1 and 2 data.
| Item number | Item content | Factor 1 | Communality (h2) | Uniqueness (u2) | |||
|---|---|---|---|---|---|---|---|
| Wave 1 | Wave 2 | Wave 1 | Wave 2 | Wave 1 | Wave 2 | ||
| 1 | Remembering things about family and friends | 0.57 | 0.84 | 0.33 | 0.71 | 0.67 | 0.29 |
| 2 | Remembering things that have happened recently | 0.6 | 0.85 | 0.36 | 0.72 | 0.64 | 0.28 |
| 3 | Recalling conversations a few days later | 0.69 | 0.85 | 0.47 | 0.73 | 0.53 | 0.27 |
| 4 | Remembering address and telephone number | 0.45 | 0.82 | 0.2 | 0.67 | 0.8 | 0.33 |
| 5 | Remembering what day and month it is | 0.56 | 0.84 | 0.32 | 0.71 | 0.68 | 0.29 |
| 6 | Remembering where things are usually kept | 0.7 | 0.85 | 0.49 | 0.72 | 0.51 | 0.28 |
| 7 | Remembering where to find things which have been put in a different place from usual | 0.62 | 0.86 | 0.39 | 0.75 | 0.61 | 0.25 |
| 8 | Knowing how to work familiar machines around the house | 0.69 | 0.77 | 0.47 | 0.6 | 0.53 | 0.4 |
| 9 | Learning to use a new gadget or machine around the house | 0.59 | 0.7 | 0.34 | 0.49 | 0.66 | 0.51 |
| 10 | Learning new things in general | 0.61 | 0.71 | 0.37 | 0.51 | 0.63 | 0.49 |
| 11 | Following a story in a book or on TV | 0.57 | 0.42 | 0.33 | 0.18 | 0.67 | 0.82 |
| 12 | Making decisions on everyday matters | 0.75 | 0.8 | 0.57 | 0.64 | 0.43 | 0.36 |
| 13 | Handling money for shopping | 0.65 | 0.89 | 0.43 | 0.8 | 0.57 | 0.2 |
| 14 | Handling financial matters | 0.52 | 0.86 | 0.27 | 0.74 | 0.73 | 0.26 |
| 15 | Handling other everyday arithmetic problems | 0.61 | 0.9 | 0.37 | 0.82 | 0.63 | 0.18 |
| 16 | Using intelligence to understand what's going on and to reason things through | 0.75 | 0.89 | 0.56 | 0.8 | 0.44 | 0.2 |
Note: Proportion variance explained at Wave 1 is 39% and at Wave 2 is 66%.
Abbreviations: HAALSI, Health and Aging in Africa: Longitudinal Study of an International Network for the Demographic Evaluation of Populations and Their Health Community in South Africa; HCAP, Harmonized Cognitive Assessment Protocol; IQCODE, Informant Questionnaire on Cognitive Decline in the Elderly.
In Wave 1, the omega total was 0.93 and the omega hierarchical was 0.70. In Wave 2, these values increased to 0.98 and 0.90, respectively. The high omega hierarchical values indicate that the single‐factor model demonstrates acceptable reliability in Wave 1 and excellent reliability in Wave 2.
Regarding convergent validity, the IQCODE score was strongly correlated with cognitive measures (Table S4 in supporting information) directly assessed among index participants in HAALSI‐HCAP Wave 1 and 2. Overall, a greater IQCODE score was associated with lower word recall or story recall scores (Table 5), when controlling for index participants’ age, sex, and education. The strength of the association between the IQCODE score and each recall memory score decreased from Wave 1 to 2, as indicated by reduced absolute values; however, the associations remained statistically significant.
TABLE 5.
Correlation between IQCODE scores with each cognitive measure of memory among all participants in HAALSI‐HCAP Wave 1 and 2 by imputation status.
| Wave 1 data | Wave 2 data | |||
|---|---|---|---|---|
| Non‐imputed | Imputed | Non‐imputed | Imputed | |
| Immediate word recall | −2.925 (−3.803, −2.047) | −3.132 (−3.873, −2.392) | −0.536 (−1.018, −0.053) | −0.970 (−1.383, −0.557) |
| Delayed word recall | −1.311 (−1.722, −0.901) | −1.392 (−1.728, −1.056) | −0.262 (−0.482, −0.043) | −0.460 (−0.647, −0.273) |
| Immediate story recall | −4.612 (−5.854, −3.370) | −5.080 (−6.124, −4.035) | −1.500 (−2.110, −0.890) | −1.583 (−2.092, −1.075) |
| Delayed story recall | −4.995 (−6.393, −3.597) | −5.355 (−6.528, −4.182) | −1.324 (−2.049, −0.600) | −1.885 (−2.494, −1.276) |
Note: 95% confidence intervals are in parentheses. For each cognitive measure of memory, a linear regression model with random intercepts was used, controlling index participants’ sex, age, and education.
Abbreviations: HAALSI, Health and Aging in Africa: Longitudinal Study of an International Network for the Demographic Evaluation of Populations and Their Health Community in South Africa; HCAP, Harmonized Cognitive Assessment Protocol; IQCODE, Informant Questionnaire on Cognitive Decline in the Elderly.
Sensitivity analysis results based on imputed datasets were consistent with the main findings derived from complete‐case analyses. Communality values for each IQCODE item were similar between imputed and non‐imputed datasets at Wave 1 but showed a slight decline in the imputed datasets at Wave 2 compared to non‐imputed data (Tables S5 and S6 in supporting information). Additionally, there were no substantial differences observed across multiple imputations at either wave. Similarly, the proportion of observed variance explained by the single‐factor model remained largely unchanged between imputed and non‐imputed datasets at Wave 1 (Table S7 in supporting information). However, at Wave 2, this proportion decreased from 66% in the non‐imputed dataset to 59% in the imputed datasets. Omega hierarchical values remained consistent between imputed and non‐imputed analyses, indicating high reliability of the single‐factor model. Associations between the IQCODE score and each recall memory cognitive score remained similar between non‐imputed and imputed Wave 1 data, and at Wave 2, these associations increased slightly in absolute value with the use of imputed datasets (Table 5).
4. DISCUSSION
To the best of our knowledge, this study is the first to assess the feasibility and psychometric properties of the IQCODE, administered as part of an informant interview, for dementia and cognitive impairment assessment in rural South African older adults with high levels of illiteracy and low education. Our findings indicate that the 16‐item IQCODE is a feasible, reliable, and valid tool in this context. African countries are experiencing a rapidly growing aging population, yet culturally sensitive cognitive assessments that account for educational and linguistic diversity remain scarce. 5 Given its strong performance, the IQCODE has the potential to become a robust informant interview tool for supplementing cognitive assessments in African countries. Moreover, because the IQCODE has been widely used in other countries, such as the United States, the UK, and India, 14 these findings contribute to global comparisons of cognitive aging and dementia.
This study found a high completion rate of the IQCODE, with high rates of response. Although some items showed relatively high proportions of “does not do” responses, these rates were comparable to, or in some cases lower than, those reported in IQCODE implementations in the United States, UK, and India. 12 , 14 A similar study in India also included “does not do” as a response option for each IQCODE item. 12 In both studies, items related to technology, math, and storytelling had higher proportions of “does not do” responses. The Indian study removed such items and found that the remaining items still supported a single‐factor structure with high reliability and validity. Unlike the Indian study with one‐wave data availability, our two‐wave data allowed us to track changes in “does not do” responses, which declined over time. This observation may suggest improved comprehension among both interviewers and informants. Additionally, factor loadings increased significantly across waves for most items, except for item 11 (following a story in a book or on TV). Despite its lower contribution, including item 11 did not affect the single‐factor structure. Its performance should be further evaluated in future waves, but retaining all items supports cross‐national comparisons.
In this study, informants’ generation (e.g., child or child‐in‐law vs. spouse) showed a strong correlation with the IQCODE score, consistent with previous findings. 14 Other informant characteristics, such as years known by the informant, were found to be strongly associated with the IQCODE score in both this study and other studies. 14 , 15 , 28 Previous studies 14 , 15 , 28 also found an association between the IQCODE score and the quality of the relationship between the informant and index participant, and the informants’ stress, depression, or anxiety, which were not collected in HAALSI‐HCAP surveys to allow comparison. Overall, informants’ characteristics had limited impact on the final IQCODE score, providing reassurance that changes in the IQCODE score primarily reflect cognitive decline in the index participants. Furthermore, compared to other studies, 14 , 15 , 28 this study used panel data of IQCODE for the same index participants in two survey waves and found that changing an informant across waves was not associated with changes in IQCODE scores when holding index participant characteristics constant. Our findings support the feasibility of using different informants in subsequent waves when previous informants are unavailable.
This study further found that the 16‐item IQCODE data collected in rural South Africa supports a single‐factor structure representing cognitive decline, consistent with the original development of the IQCODE. 15 At Wave 2, this single factor explained ≈ 66% of the total variance in IQCODE scores among rural South African older adults, comparable to findings from India 12 and Lebanon, 8 both of which also had larger populations of lower education. Although the explained variance was lower at Wave 1 (39%), it is closer to the range (42%–61%) reported in a review of IQCODE performance. 15 The declining trend in “does not do” responses may contribute to the substantial increase in total variance explained by the single factor, though further investigation is needed. Additionally, the 16‐item IQCODE exhibited high internal consistency and reliability, with the omega hierarchical value reaching 0.9, aligning with previous research. 8 , 15 , 17
The IQCODE score was also closely correlated with other cognitive measures of memory, consistent with previous studies in high‐income settings 29 and low income and low educated settings 8 , 12 , 14 , 30 as well as review studies. 16
Several limitations should be noted. First, the HAALSI‐HCAP data collection experienced technical issues in Wave 2 due to tablet errors, which introduced missingness in responses for IQCODE items. Along with the missingness due to “does not do activity” responses, we addressed them by imputation. While the main findings remained consistent after imputation, this approach introduces some uncertainty in individual‐level dementia assessments. This emphasized the importance of implementing robust data collection procedures in future studies to minimize such challenges. Second, while the IQCODE demonstrated strong performance, the high proportion of “does not do” responses for specific items suggests contextual limitations. However, the decreasing trend of these responses in the second wave may reflect improved understanding of the IQCODE items by informants over time, possibly due to greater familiarity with the instrument between interviews. These findings highlight the need for further adaptation of the tool to better suit the activities and experiences of rural South African populations. Third, local cultural values may also influence the measurement properties of the IQCODE, though these were not captured in the current dataset. For example, cultural norms emphasizing respect for elders—and concerns about dishonor or stigma—may lead informants to underreport declines in the everyday cognitive functioning of older adults. This underreporting could, in turn, result in an underestimation of both the prevalence of cognitive decline and its correlation with other cognitive or dementia assessments. Accounting for these cultural influences may reveal even stronger associations between IQCODE scores and other cognitive measures than observed in our current findings. Future studies may need to incorporate measures of local cultural values when collecting informant‐based data. Fourth, the IQCODE score was incorporated into the dementia assessment process in Wave 1, 13 limiting our ability to independently evaluate its criterion validity. Finally, while our study demonstrates the feasibility and utility of IQCODE in this specific cohort, and findings can be generalizable to other settings with similar low education and literacy populations, its broader generalizability to other Sub‐Saharan African contexts with higher socio‐economic levels requires further investigation.
In conclusion, this study demonstrates that IQCODE is a reliable and valid instrument for assessing cognitive decline in rural South African settings and addressing the growing dementia burden in low‐resource contexts. However, ongoing efforts to refine the tool and address contextual limitations are essential to optimize its effectiveness.
CONFLICT OF INTEREST STATEMENT
The authors report no conflicts of interest. Author disclosures are available in the supporting information.
CONSENT STATEMENT
Written or verbally informed consent was obtained from all study participants.
Supporting information
Supporting Information
Supporting Information
ACKNOWLEDGMENTS
The HAALSI study is nested within the SAMRC/Wits University Rural Public Health and Health Transitions Research Unit (Agincourt Health and Socio‐Demographic Surveillance System), which is supported by the University of the Witwatersrand, Medical Research Council, and Dept of Science and Innovation, South Africa. The Health and Ageing in Africa: Longitudinal Studies in South Africa study was supported by the U.S. National Institute on Aging of the National Institutes of Health (NIH; grant number P01 AG041710). The HAALSI‐HCAP study was supported by the National Institute on Aging (grant number R01 AG054066). The funders had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.
Guo M, Taporoski TP, Farrell MT, et al. Validity and reliability of the Informant Questionnaire on Cognitive Decline in the Elderly (IQCODE) for dementia assessment in rural South Africa. Alzheimer's Dement. 2025;21:e70584. 10.1002/alz.70584
REFERENCES
- 1. GBD 2021 Causes of Death Collaborators . Global burden of 288 causes of death and life expectancy decomposition in 204 countries and territories and 811 subnational locations, 1990‐2021: a systematic analysis for the Global Burden of Disease Study 2021. Lancet 2024;403:2100‐2132. doi: 10.1016/S0140-6736(24)00367-2 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2. Solanki G, Kelly G, Cornell J, Daviaud E, Geffen L. Population ageing in South Africa: trends, impact, and challenges for the health sector. S Afr Health Rev. 2019;2019:175‐182. [PMC free article] [PubMed] [Google Scholar]
- 3. Nichols E, Steinmetz JD, Vollset SE, et al. Estimation of the global prevalence of dementia in 2019 and forecasted prevalence in 2050: an analysis for the Global Burden of Disease Study 2019. The Lancet Public Health. 2022;7:e105‐25. doi: 10.1016/S2468-2667(21)00249-8 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4. Farina N, Jacobs R, Turana Y, et al. Comprehensive measurement of the prevalence of dementia in low‐ and middle‐income countries: sTRiDE methodology and its application in Indonesia and South Africa. BJPsych Open. 2023;9:e102. doi: 10.1192/bjo.2023.76 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5. Akinyemi RO, Yaria J, Ojagbemi A, et al. Dementia in Africa: current evidence, knowledge gaps, and future directions. Alzheimer's & Dementia. 2022;18:790‐809. doi: 10.1002/alz.12432 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6. Paddick S‐M, Longdon AR, Kisoli A, et al. Dementia prevalence estimates in sub‐Saharan Africa: comparison of two diagnostic criteria. Global Health Action. 2013;6:19646. doi: 10.3402/gha.v6i0.19646 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7. Farrell MT, Bassil DT, Guo M, et al. Estimating dementia prevalence using remote diagnoses and algorithmic modelling: a population‐based study of a rural region in South Africa. Lancet Glob Health. 2024;12:2003‐2011. doi: 10.1016/S2214-109X(24)00325-5 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8. Phung TKT, Chaaya M, Asmar K, et al. Performance of the 16‐item Informant Questionnaire on Cognitive Decline for the Elderly in an Arabic speaking older population. Dement Geriatr Cogn Disord. 2015;40:276‐289. doi: 10.1159/000437092 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9. Gómez‐Olivé FX, Montana L, Wagner RG, et al. Cohort profile: Health and Ageing in Africa: a longitudinal study of an INDEPTH community in South Africa (HAALSI). Int J Epidemiol. 2018;47:689‐690j. doi: 10.1093/ije/dyx247 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10. Louw C, Swanepoel DW, Eikelboom RH, Hugo J. Prevalence of hearing loss at primary health care clinics in South Africa. Afr Health Sci. 2018;18:313‐320. doi: 10.4314/ahs.v18i2.16 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11. Farrell MT, Jia Y, Berkman LF, Wagner RG. Do you see what eye see? Measurement, correlates, and functional associations of objective and self‐reported vision impairment in aging South Africans. J Aging Health. 2021;33:803‐816. doi: 10.1177/08982643211012839 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12. Khobragade P, Nichols E, Meijer E, et al. Performance of the Informant Questionnaire on Cognitive Decline for the Elderly (IQCODE) in a nationally representative study in India: the LASI‐DAD study. Int Psychogeriatr. 2024;36:177‐187. doi: 10.1017/S1041610222000606 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13. Bassil DT, Farrell MT, Weerman A, et al. Feasibility of an online consensus approach for the diagnosis of cognitive impairment and dementia in rural South Africa. Alzheimer's & Dementia: Diagnosis, Assessment & Disease Monitoring. 2023;15:e12420. doi: 10.1002/dad2.12420 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14. Nichols E, Gross AL, Zhang YS, et al. Considerations for the use of the Informant Questionnaire on Cognitive Decline in the Elderly (IQCODE) in cross‐country comparisons of cognitive aging and dementia. Alzheimer's & Dementia. 2024;20:4635‐4648. doi: 10.1002/alz.13895 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15. Jorm AF. The Informant Questionnaire on Cognitive Decline in the Elderly (IQCODE): a review. Int Psychogeriatr. 2004;16:275‐293. doi: 10.1017/s1041610204000390 [DOI] [PubMed] [Google Scholar]
- 16. Quinn TJ, Fearon P, Noel‐Storr AH, Young C, McShane R, Stott DJ. Informant Questionnaire on Cognitive Decline in the Elderly (IQCODE) for the diagnosis of dementia within community dwelling populations. Cochrane Database of Systematic Reviews. 2014. doi: 10.1002/14651858.CD010079.pub2 [DOI] [PubMed] [Google Scholar]
- 17. Perroco TR, Bustamante SEZ, Moreno Mdel P, et al. Performance of Brazilian long and short IQCODE on the screening of dementia in elderly people with low education. Int Psychogeriatr. 2009;21:531‐538. doi: 10.1017/S1041610209008849 [DOI] [PubMed] [Google Scholar]
- 18. Jorm AF, Jacomb PA. The Informant Questionnaire on Cognitive Decline in the Elderly (IQCODE): socio‐demographic correlates, reliability, validity and some norms. Psychol Med. 1989;19:1015‐1022. doi: 10.1017/s0033291700005742 [DOI] [PubMed] [Google Scholar]
- 19. Jorm AF. A short form of the Informant Questionnaire on Cognitive Decline in the Elderly (IQCODE): development and cross‐validation. Psychol Med. 1994;24:145‐153. doi: 10.1017/s003329170002691x [DOI] [PubMed] [Google Scholar]
- 20. Sonnega A, Faul JD, Ofstedal MB, Langa KM, Phillips JW, Weir DR. Cohort profile: the Health and Retirement Study (HRS). Int J Epidemiol. 2014;43:576‐585. doi: 10.1093/ije/dyu067 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21. Bassil DT, Farrell MT, Wagner RG, et al. Cohort profile update: cognition and dementia in the Health and Aging in Africa Longitudinal Study of an INDEPTH community in South Africa (HAALSI dementia). Int J Epidemiol. 2022;51:e217‐26. doi: 10.1093/ije/dyab250 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22. Langa KM, Ryan LH, McCammon RJ, et al. The Health and Retirement Study Harmonized Cognitive Assessment Protocol project: study design and methods. Neuroepidemiology. 2020;54:64‐74. doi: 10.1159/000503004 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23. Revelle W, Rocklin T. Very Simple Structure: an alternative procedure for estimating the optimal number of interpretable factors. Multivariate Behav Res. 1979;14:403‐414. doi: 10.1207/s15327906mbr1404_2 [DOI] [PubMed] [Google Scholar]
- 24. Velicer WF. Determining the number of components from the matrix of partial correlations. Psychometrika. 1976;41:321‐327. doi: 10.1007/BF02293557 [DOI] [Google Scholar]
- 25. Zinbarg RE, Yovel I, Revelle W, McDonald RP. Estimating generalizability to a latent variable common to all of a scale's indicators: a comparison of estimators for ωh. Applied Psychological Measurement. 2006;30:121‐144. doi: 10.1177/0146621605278814 [DOI] [Google Scholar]
- 26. Nunnally J, Bernstein L. Psychometric Theory. New York: McGraw‐Hill Higher INC; 1994. [Google Scholar]
- 27. Revelle W, Zinbarg RE. Coefficients alpha, beta, omega, and the glb: comments on Sijtsma. Psychometrika. 2009;74:145‐154. doi: 10.1007/s11336-008-9102-z [DOI] [Google Scholar]
- 28. Nygaard HA, Naik M, Geitung JT. The Informant Questionnaire on Cognitive Decline in the Elderly (IQCODE) is associated with informant stress. Int J Geriatr Psychiatry. 2009;24:1185‐1191. doi: 10.1002/gps.2243 [DOI] [PubMed] [Google Scholar]
- 29. Potter GG, Plassman BL, Burke JR, et al. Cognitive performance and informant reports in the diagnosis of cognitive impairment and dementia in African Americans and whites. Alzheimers Dement. 2009;5:445‐453. doi: 10.1016/j.jalz.2009.04.1234 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30. Goudsmit M, van Campen J, Franzen S, van den Berg E, Schilt T, Schmand B. Dementia detection with a combination of informant‐based and performance‐based measures in low‐educated and illiterate elderly migrants. Clin Neuropsychol. 2021;35:660‐678. doi: 10.1080/13854046.2020.1711967 [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Supporting Information
Supporting Information