Clinical staging in Swedish primary care using the Amsterdam Instrumental Activities of Daily Living Questionnaire

Abstract

INTRODUCTION

We assessed the accuracy of the Amsterdam Instrumental Activities of Daily Living Questionnaire (A‐IADL‐Q) for clinical staging in Swedish primary care.

METHODS

Participants from the Swedish BioFINDER Primary Care study were included. Discriminative performance of the A‐IADL‐Q was evaluated using receiver operating curves. Multinomial and linear regression models assessed associations among A‐IADL‐Q scores, clinical stage, demographics, cognition, and comorbidities.

RESULTS

Among 623 patients, 148 (23.8%) had subjective cognitive decline (SCD), 274 (43.9%) mild cognitive impairment (MCI), and 201 (32.3%) dementia with a mean (standard deviation) age of 76.7 (7.3). The area under the curve (95% confidence interval) for discriminating between SCD versus MCI/dementia was 0.89 (0.86–0.91) and for SCD/MCI versus dementia 0.89 (0.87–0.92). Age (β = −0.25), Mini‐Mental State Examination (β = 0.91) and Montreal Cognitive Assessment (β = 0.57), but no other demographics and comorbidities, were associated with the A‐IADL‐Q.

DISCUSSION

The A‐IADL‐Q may help primary care physicians determine clinical stage and shows promise for use to adequately refer patients to secondary or tertiary care.

Highlights

• First validation study of the Amsterdam‐IADL‐Questionnaire in primary care.

• Demonstrated strong diagnostic accuracy for distinguishing SCD, MCI, and dementia in the context of Alzheimer's disease even when accounting for demographics and comorbidities.

• Provides evidence that the A‐IADL‐Q can be integrated into routine primary care practice to improve timely Alzheimer's disease diagnosis and referral decisions.

1. BACKGROUND

Effective and high‐quality dementia care includes primary and secondary prevention, timely and accurate diagnosis, clear communication of the diagnosis to both the patient and potential caregivers, and comprehensive post‐diagnostic care and treatment. ¹ , ² Primary care typically serves as the initial, and often the only, point of contact with a physician for individuals experiencing cognitive symptoms. With the emergence of new disease‐modifying therapies (DMTs) for Alzheimer's disease (AD), ³ , ⁴ primary care physicians (PCPs) are anticipated to play a crucial role in discriminating between clinical stages for adequate referral to secondary care. This is because only patients with mild cognitive impairment (MCI) and mild dementia, but not moderate to severe dementia and subjective cognitive decline (SCD) currently are eligible for anti‐amyloid therapies. ⁵ , ⁶ Although patients with MCI per definition are still independent in performing activities of daily living (ADL), they often experience subtle impairments in instrumental ADL (IADL). ⁷ Development of accurate blood biomarkers for AD in primary care provides access to biological confirmation of the underlying disease process, ⁸ , ⁹ yet biomarkers need to be complemented with reliable clinical measures to determine clinical staging, particularly for assessing eligibility for disease‐modifying treatments.

The Amsterdam Instrumental Activities of Daily Living Questionnaire (A‐IADL‐Q) was developed and extensively validated to assess everyday functioning in a memory clinic population. ¹⁰ , ¹¹ , ¹² Previous studies on the A‐IADL‐Q demonstrated strong psychometric properties (i.e., measurement invariance, test–retest reliability, construct validity, and interpretability), ¹³ , ¹⁴ , ¹⁵ , ¹⁶ including evidence for the instrument to effectively capture longitudinal functional changes. ¹⁷ , ¹⁸ The A‐IADL‐Q is recommended as part of a core dataset for collecting real‐world data in the clinical context of dementia care. ¹⁹ , ²⁰ However, it remains unclear how the A‐IADL‐Q performs in the primary care setting, with patients often being older and having more comorbidities compared to the memory clinic population. ²¹ It is especially important to validate the tool in this setting as there are results supporting a possible influence of comorbidities on specific IADL abilities. ²²

Therefore, the aims of this study were to: (1) examine the diagnostic accuracy of A‐IADL‐Q for determining clinical stage in a primary care population undergoing clinical evaluation; (2) explore the accuracy of the instrument to discriminate between clinical stages specifically in amyloid‐positive individuals, which is important in the evaluation of eligibility for anti‐amyloid therapies; and (3) evaluate the effect of comorbidities and other possible confounders on the A‐IADL‐Q score.

2. METHODS

2.1. Study design and ethics statement

This study used baseline data from the BioFINDER Primary Care study (NCT06120361; https://biofinder.se) with ongoing recruitment of participants from ≈ 25 public and private primary care centers in southern Sweden. After the basic clinical evaluations in primary care, participants were referred to the Memory Clinic of Skåne University Hospital for a full advanced work‐up at a specialist center. Clinicians at the Memory Clinic were blinded to all primary care tests including the A‐IADL‐Q to prevent circular reasoning in the diagnostic process. Data were collected between September 2020 and April 2024. The study design has been extensively described previously. ⁸ , ⁹

All participants signed informed consent, and the study was approved by the Swedish Ethical Review Authority.

2.2. Participants

Participants were eligible for inclusion if they or their informant sought help due to perceived cognitive symptoms, or if their PCP suspected a progressive neurodegenerative disorder. A further criterion was that the PCP confirmed the main symptom to be cognitive in nature, most commonly memory complaints but potentially also executive, visuospatial, language, or attention complaints, rather than explained by another condition. Other inclusion criteria were age ≥ 40 years. Exclusion criteria were already diagnosed dementia, refusal of investigation at the memory clinic, or cognitive impairment due to other diseases such as significant anemia, infection, severe sleep deprivation, psychotic disorder, moderate to severe depressive episode, or alcohol/drug abuse. Participants were selected for the present study if they had a study partner (spouse, relative, or other) who completed the A‐IADL‐Q and their clinical stage was defined. Study partners of participants completed the Swedish language version of A‐IADL‐Q on an iPad or at home on an online form of this questionnaire using Qualtrics, a web‐based survey tool (www.qualtrics.com).

2.3. Diagnostic criteria

After examination in primary care, all participants were referred to the memory clinic for a standard diagnostic work‐up. Clinical staging was determined in a weekly consensus meeting including dementia specialists and neuropsychologists. MCI was defined after consensus decision based on having notable cognitive symptoms and abnormal cognitive test results but no functional impairment using the RBANS (Repeatable Battery for the Assessment of Neuropsychological Status) battery (accounting for premorbid cognitive level), Clinical Dementia Rating (CDR) rating (CDR = 0.5), and Functional Activities Questionnaire (FAQ) score < 9. The MCI definition did not require that a strict threshold in a cognitive domain was met (although all performed < −1 standard deviation [SD] in at least one cognitive domain in the RBANS battery), but was based on the overall clinical assessment. The classification followed the design of the MCI classification of the Mayo Clinic Study of Aging ²³ and was in line with the Diagnostic and Statistical Manual of Mental Disorders Fifth Edition (DSM‐5) criteria for mild neurocognitive disorder. ²⁴ Dementia was defined according to the DSM‐5 criteria for major neurocognitive disorder ²⁴ with further stratification into mild and moderate dementia according the Alzheimer's Association criteria for clinical stages corresponding to stages 4 and 5. ²⁵ SCD was defined as experiencing cognitive symptoms to the level that led the patient to seek help in primary care, but not fulfilling the criteria for MCI or dementia. The CDR score was used as a secondary outcome. ²⁶ Amyloid status was determined using cerebrospinal fluid (CSF) Lumipulse amyloid beta 42/40 ratio (≤ 0.072 for positivity) or visual read of [¹⁸F]flutemetamol positron emission tomography (PET) if there were contraindications for lumbar puncture. ²⁷ , ²⁸

RESEARCH IN CONTEXT

1. Systematic review: With the emergence of new therapeutic interventions for Alzheimer's disease, there is a clear need for early detection and adequate referral in primary care. The Amsterdam‐IADL‐Questionnaire (A‐IADL‐Q) could aid in the diagnostic pathway, due to its user‐friendly design and extensive validation in a memory clinic setting. However, there are no validation studies in primary care.

2. Interpretation: This study provides evidence for the diagnostic accuracy of the A‐IADL‐Q in distinguishing between SCD, MCI and dementia in a primary care population, largely independent from covariates and comorbidities. The A‐IADL‐Q may help primary care physicians determine clinical stage.

3. Future directions: Further research should focus on how these results can be extended to the context of adequate referral of potentially eligible individuals for anti‐amyloid therapies and investigate psychometric properties of the scale and potential bias on an item level.

2.4. Materials

2.4.1. The A‐IADL‐Q

Everyday functioning was assessed using the 30‐item version of the study‐partner based A‐IADL‐Q (A‐IADL‐Q‐30), designed to measure IADL impairment in MCI and dementia. ¹² It has been found to correlate well with other IADL measures while showing low correlations with age, education, and depressive symptoms, indicating good construct validity. ¹⁴ Items have been selected cross‐culturally to capture key aspects of IADL, with an emphasis on everyday technology use. ¹² Each item is rated on a 5‐point Likert scale, from “no difficulty” to “no longer able (due to difficulties with their memory, planning, or thinking).” Other validation studies have provided support for its diagnostic accuracy for detecting dementia (area under the curve [AUC] = 0.97; 95% confidence interval [CI: 0.97–0.98]) and sensitivity to change over time. ¹⁶ , ¹⁸

The total A‐IADL‐Q‐30 score is computed based on item response theory (IRT), linking item scores to a latent construct. ²⁹ The resulting T‐score (mean = 50, SD = 10) is normally distributed in memory clinic populations, with lower scores indicating greater impairment. In a mixed‐methods study, the meaningfulness of scores was established, resulting in T‐scores ≥ 60 indicating “no problems,” 50 to 59 “mild problems,” 40 to 49 “moderate problems,” and < 40 “severe problems”. ¹² , ¹⁶ , ³⁰

2.4.2. Comorbidity composites

Medical data were obtained through retrospective chart review by one of the authors (A.F.) and subsequently coded as present (1) or absent (0). We calculated composite scores for three major comorbidities: cardiovascular disease, psychiatric disease, and metabolic disease. Comorbidity composite scores were calculated by averaging binary scores for presence or absence of multiple diagnoses within a certain category. The composite for cardiovascular diseases consisted of history of stroke or transient ischemic attack (TIA), hypertension, ischemic heart disease, atrial fibrillation, congestive heart failure, and hyperlipidemia. The psychiatry composite consists of depression, anxiety, and other major psychiatric disorders. The metabolic composite score is based on diabetes (type 1 or 2) and chronic kidney disease.

2.4.3. Cognitive and functional screening

Global cognition was measured using the Mini‐Mental State Examination (MMSE, ³¹ range 0–30) and the Montreal Cognitive Assessment (MoCA, ³² range 0–30), with higher scores indicating better cognitive functioning, conducted by a specialized dementia care nurse or occupational therapist at the primary care unit. Functional status was rated using the CDR Global ³³ with scores ranging from 0 to 3; 0 reflects no impairment and scores between 0.5 and 3 reflect more impairment from MCI (0.5) to dementia (> 1), and CDR Sum of Boxes with scores ranging from 0 to 18. The CDR was administered by dementia specialists at the memory clinic using a semi‐structured interview. Importantly, CDR rating and clinical staging were performed blinded to the A‐IADL‐Q score. ADL function was also examined using the informant‐based FAQ ³⁴ with a range from 0 (best) to 30 (worst).

2.5. Statistical analysis

Only participants for which the A‐IADL‐Q score was available were included in analyses. Demographic characteristics were analyzed using descriptive statistics. Group differences were calculated using Mann–Whitney (cognitive tests and education), t test (age), or χ² test (binary variables). Unless otherwise specified, the mild and moderate dementia groups were combined into one dementia group due to small subgroup sizes.

Using the optimal cut‑off for distinguishing cognitive impairment (MCI and dementia combined) from normal cognition (SCD) based on the Youden index, receiver operating characteristic (ROC) curve statistics were used to calculate the sensitivity, specificity, and the AUCs for A‐IADL‐Q and FAQ and compare their discriminative ability. Clinical stage (MCI and dementia with SCD as a reference group) was predicted using multinomial regression models, both adjusted and unadjusted for age, sex, years of education, and MMSE. Overall accuracy, predicted probabilities, and odds ratios (ORs) were calculated based on the multinomial regression model. ORs and 95% CIs were calculated both for MCI and dementia with SCD as the reference group. ORs for the A‐IADL‐Q were inversed for interpretation purposes, such that worse A‐IADL‐Q scores gives higher odds for having MCI or dementia.

Clinically meaningful cut‐offs that were previously established in a memory clinic sample ³⁰ were applied to this sample and compared to the consensus diagnosis.

Linear regression analysis was used to identify the proportion of variability in A‐IADL‐Q scores predicted by the demographic (age, sex, years of education), cognitive variables (MMSE, MoCA), and comorbidities. All comorbidity composites were entered simultaneously in the model, allowing estimation of the independent association of each comorbidity composite while controlling for the others. Multicollinearity was assessed using the variance inflation factor (VIF), with values < 5 considered acceptable.

In all analyses, significance testing was two sided, with α set at 0.05 for statistical significance. All analyses were performed using R version 4.3.2 (R Core Team 2023), using among others the “nnet”, ³⁵ “lme4”, ³⁶ and “pROC” packages. ³⁷

3. RESULTS

3.1. Demographics

We included 623 participants with a mean age of 76.7 years (SD 7.3, range 48.5–91.9); 50.8% were women; mean education was 11 years. Of the total sample, 148 (23.8%) participants were classified as SCD, 274 (43.9%) as having MCI, and 201 (32.3%) as having dementia, of which 175 were mild and 26 were moderate dementia. About half of the participants were positive for amyloid pathology (n = 339, 54.4%) using CSF or PET. Of the subsample with available comorbidity data (n = 473), 393 (82.7%) participants had a history of cardiovascular diseases, 200 (42.1%) of metabolic disease, and 241 (50.8%) of psychiatric diseases. Demographic characteristics of the total sample are presented in Table 1. Group differences were found for age and years of education. As expected, all cognitive and functional measures differed across clinical stages. A‐IADL‐Q scores per diagnostic group are visualized in Figure 1A and per CDR subgroup in Figure 1B. Subgroup characteristics of amyloid‐positive individuals with any stage of cognitive impairment are presented in Table S1 in supporting information, including moderate dementia as a separate entity. Among amyloid‐positive individuals, the A‐IADL‐Q scores again showed a stepwise decrease along clinical stages, also when inspecting the mild and moderate dementia groups separately (Figure S1 in supporting information).

TABLE 1.

Demographic and clinical characteristics.

	Total	SCD	MCI	Dementia	p
N	623	148	274	201
Age (mean [SD])	76.7 (7.3)	72.8 (7.8)	76.5 (7.1)	79.8 (5.9)	<0.001
Education, years (mean [SD])	11.1 (3.3)	12.2 (3.2)	11.0 (3.4)	10.3 (3.2)	<0.001
Female sex, n (%)	318 (50.8)	75 (50.7)	135 (49.3)	107 (53.2)	0.671
Amyloid positive, n (%)	340 (54.3)	48 (32.4)	141 (50.9)	151 (75.1)	<0.001
CDR sum of boxes (mean [SD])	2.0 (2.3)	0.1 (0.2)	1.2 (0.9)	4.5 (2.3)	<0.001
CDR global (mean [SD])	0.5 (0.4)	0.0 (0.1)	0.4 (0.2)	0.8 (0.5)	<0.001
MMSE (mean [SD])	25.8 (3.6)	28.4 (1.7)	26.3 (2.6)	23.1 (4.0)	<0.001
MoCA (mean [SD])	20.9 (4.2)	24.3 (2.6)	21.4 (2.9)	17.7 (4.4)	<0.001
FAQ (mean [SD])	8.2 (6.7)	2.5 (3.1)	6.6 (4.6)	14.7 (5.7)	<0.001

Note: Missing data were present for CDR Global (n = 46), CDR Sum of Boxes (n = 46), MMSE (n = 27), MoCA (n = 26), and FAQ (n = 10). Demographic and diagnostic variables, including age, education, sex, cognitive status, and amyloid beta status, were complete.

Abbreviations: CDR, Clinical Dementia Rating (range Global 0–3, Sum of Boxes 0–18); FAQ, Functional Activities Questionnaire (range 0–30); MCI, mild cognitive impairment; MMSE, Mini‐Mental State Examination (range 0–30); MoCA, Montreal Cognitive Assessment (range 0–30); SCD, subjective cognitive decline; SD, standard deviation.

FIGURE 1.

Boxplots for A‐IADL‐Q scores across (A) diagnostic and (B) CDR subgroups.

Note: Both clinical staging (SCD/MCI/dementia) and CDR assessment was performed blinded to A‐IADL‐Q scores. A‐IADL‐Q, Amsterdam Instrumental Activities of Daily Living Questionnaire; CDR, Clinical Dementia Rating; MCI, mild cognitive impairment; SCD, subjective cognitive decline.

3.2. ROC curves

The ROC curves with corresponding sensitivities and specificities for distinguishing between all clinical stages using the A‐IADL‐Q are depicted in Figure 2. Based on the highest Youden index, the optimal cut‐off for discriminating between SCD versus cognitive impairment, that is, MCI or dementia, was 58.25, with an AUC of 0.89 (95% CI 0.86–0.91; Figure 2A), a sensitivity of 76%, and a specificity of 91%. The AUC for discriminating SCD and MCI versus dementia was 0.89 (95% CI 0.87–0.92; sensitivity/specificity = 76%/90%; Figure 2B), for SCD versus dementia 0.96 (95% CI 0.95–0.98; sensitivity/specificity = 93%/91%; Figure 2C), for MCI versus dementia 0.86 (95% CI 0.82–0.89; sensitivity/specificity = 70%/90%; Figure 2D), and for SCD versus MCI 0.83 (95% CI 0.79–0.87; sensitivity/specificity = 70%/83%; Figure 2E). The A‐IADL‐Q and FAQ showed comparable accuracy across all diagnostic entities, including dementia versus non‐dementia (A‐IADL‐Q AUC [95% CI] = 0.89 [0.86–0.92]; sensitivity/specificity = 76%/90%); the FAQ AUC [95% CI] = 0.90 [0.88–0.93]; sensitivity/specificity = 81%/83%). (See also Figure S2 in supporting information.) Importantly, the FAQ was used in the diagnostic classification, while the A‐IADL‐Q was administered independently (blinded to the diagnostic process).

FIGURE 2.

Receiver operating curves for comparing (A) SCD (n = 148) versus MCI + dementia (n = 478), (B) SCD + MCI (n = 425) versus dementia (n = 201), (C) SCD (n = 148) versus dementia (n = 201), (D) MCI (n = 277) versus dementia (n = 201), and (E) SCD (n = 148) versus MCI (n = 277). AUC, area under the curve; CI, confidence interval; MCI, mild cognitive impairment; SCD, subjective cognitive decline.

3.3. Multinomial regression models for identifying clinical stages

To identify the potential diagnostic value of A‐IADL‐Q for differentiating among all clinical stages (SCD, MCI, and dementia), multinomial regression analyses were performed adjusted for age, sex, and years of education. For the ORs, as shown in Table 2, we found an association, with lower scores on the A‐IADL‐Q reflecting higher odds of having MCI or dementia, for which each 1‐point decrease in A‐IADL‐Q was associated with 17% higher odds of MCI versus SCD and 32% higher odds of dementia versus SCD. Overall accuracy of the model was 70% (95% CI [66–73], κ = 0.52, p < 0.0001). Classification of subgroups is depicted in Table S2 in supporting information. Subgroup analysis in amyloid‐positive participants only did not substantially change the results, as depicted in Table S3 in supporting information.

TABLE 2.

Multinomial regression model for differentiating MCI and dementia from SCD using the A‐IADL‐Q.

	MCI			Dementia
	B	SE	OR [95% CI]	B	SE	OR [95% CI]
(Intercept)	9.76 **	1.95		15.16 **	2.62
A‐IADL‐Q	−0.19 **	0.02	0.83 [0.79–0.86]	−0.38 **	0.03	0.68 [0.64–0.72]
Age	0.05 *	0.02	1.05 [1.02–1.09]	0.11 *	0.03	1.11 [1.06–1.19]
Female sex	−0.09	0.25	0.91 [0.56–1.47]	−0.26	0.33	0.77 [0.40–1.46]
Years of education	−0.08	0.04	0.92 [0.85–0.99]	−0.13	0.05	0.88 [0.79–0.97]

Abbreviations: A‐IADL‐Q, Amsterdam Instrumental Activities of Daily Living Questionnaire; CI, confidence interval; MCI, mild cognitive impairment; OR, odds ratio; SCD, subjective cognitive decline; SE, standard error.

^* p < 0.01.

^** p < 0.001.

Figure 3 shows the predicted probabilities of SCD, MCI, and dementia based on the A‐IADL‐Q score. The model estimates that individuals with higher A‐IADL‐Q scores have a greater predicted probability of being diagnosed with SCD, whereas those with lower A‐IADL‐Q scores are more likely to be classified as MCI or dementia. The predicted probability shifts from SCD to MCI being the most likely diagnosis at an A‐IADL‐Q score of ≈ 65 and from MCI to dementia at an A‐IADL‐Q score of ≈ 49.

FIGURE 3.

Predicted probabilities for the A‐IADL‐Q by diagnostic group. Note: At an A‐IADL‐Q score of ≈ 65 the predicted probability shifts from SCD to MCI being the most likely diagnosis and at an A‐IADL‐Q score of ≈ 49 it shifts from MCI to dementia. A‐IADL‐Q, Amsterdam Instrumental Activities of Daily Living Questionnaire; MCI, mild cognitive impairment; SCD, subjective cognitive decline

3.4. Applying clinically meaningful cutoffs

To provide a qualitative interpretation of scores, we applied previously published cut‐offs for clinical meaningfulness (Table S2), ³⁰ resulting in n = 220 classified as little to no IADL problems (T‐score ≥ 60), of which the majority had SCD (n = 124; 56.4%), 86 MCI (39.1%), and 10 dementia (4.5%). Of the participants classified as having mild IADL problems (n = 204; T‐score 50–59), the majority had MCI (n = 142; 69.6%), 23 SCD (11.3%), and 39 dementia (19.1%) and of those classified as having moderate to severe IADL problems (n = 199; T‐score ≤ 49), the majority had dementia (n = 152; 76.4%), 1 had SCD (0.5%), and 46 MCI (23%).

3.5. Linear regression models for examining the influence from demographics and comorbidities

To examine the influence of age, sex, education, cognitive screeners, and comorbidity composites we computed a multivariable linear regression model, as depicted in Table 3. VIF ranged from 1.0 (sex) to 1.7 (MoCA), indicating minimal multicollinearity. We found an effect for age, to the extent that 1 year older was associated with .25 lower A‐IADL‐Q score, and an effect for the cognitive screening tests where a 1‐point increase in MMSE and MoCA led to a 0.91 and 0.57 higher A‐IADL‐Q score, respectively. We did not find an effect for other demographics or the comorbidity composite scores on the A‐IADL‐Q.

TABLE 3.

Linear regression model.

	B	SE	T	95% CI
(Intercept)	40.45 *	6.63	6.10	27.42 to 53.48
Age	−0.25 *	0.07	−3.68	−0.38 to −0.11
Sex	0.27	0.89	0.30	−1.47 to 2.01
Education	−0.08	0.14	−0.58	−0.36 to 0.20
MMSE	0.91 *	0.15	6.13	0.62 to 1.20
MoCA	0.57 *	0.12	4.67	0.33 to 0.81
Cardiovascular	−2.62	1.70	−1.54	−5.96 to 0.72
Psychiatric	−2.31	1.48	−1.56	−5.22 to 0.60
Metabolic	−1.63	2.19	−0.74	−5.93 to 2.67

Note: Education in years, comorbidities concern predefined composite scores.

Abbreviations: CI, confidence interval; MMSE, Mini‐Mental State Examination; MoCA, Montreal Cognitive Assessment; SE, standard error.

^* p < 0.001.

4. DISCUSSION

In this study, we evaluated the diagnostic accuracy of the A‐IADL‐Q in a primary care setting. We found a high discriminative ability to distinguish individuals with MCI or dementia from SCD (AUC = 0.89) and SCD or MCI from dementia (AUC = 0.89). The ability to discriminate between SCD and MCI was slightly lower (AUC = 0.83). Our results suggest that the A‐IADL‐Q not only differentiates between clinical stages, but also reflects clinically meaningful levels of IADL impairment that are relevant for clinical assessment and care. Inspecting possible confounders of the A‐IADL‐Q, we found a small effect for age and the cognitive screening tests but no effect for other demographics or the comorbidity composite scores.

To our knowledge, this is the first study to validate the A‐IADL‐Q in a primary care population, adding evidence for its use as a practical and valid tool for dementia assessment in primary care. Previous studies in memory clinic settings have shown that traditional IADL instruments often exhibit limited validity and reporting on their psychometric properties is scarce, while the A‐IADL‐Q has been extensively validated, showing strong psychometric quality. ³⁸ Moreover, where some traditional instruments fail to distinguish early‐stage cognitive impairment, ³⁹ our findings provide support for the predictive ability of A‐IADL‐Q across different stages of cognitive impairment as well as different stages of AD (biomarker confirmed). Notably, the A‐IADL‐Q demonstrated comparable sensitivity and specificity to the FAQ across all diagnostic contrasts, despite being administered completely independently from the diagnostic process. The FAQ, by contrast, was part of the clinical workup and contributed to the consensus diagnoses against which both instruments are here evaluated, introducing incorporation bias that inherently favors FAQ performance. The fact that the A‐IADL‐Q achieves comparable accuracy under these conditions represents a notable strength and underscores its potential as an unbiased functional measure in primary care. This robustness strengthens the relevance for early clinical detection of neurodegenerative changes in AD, specifically in combination with biomarker confirmation.

The overall accuracy of the A‐IADL‐Q is promising, especially in distinguishing dementia from non‐dementia, but the ability to accurately distinguish between MCI and dementia remains a challenge and needs further refinement, including optimized pen and paper cognitive tests or potentially novel digital tests. ⁴⁰ Even though the clinical staging indicates there is no lack of independence in ADL function in the early disease stages, there are often slight and/or fluctuating impairments in complex activities resulting in compensatory mechanisms that can vary on an individual basis (such as writing more notes, double checking, depending on GPS to navigate). ¹⁸ These subtle deficits remain difficult to detect with traditional questionnaires. Consistent with this heterogeneity, a small subset of patients with dementia in our sample showed relatively high A‐IADL‐Q scores, suggesting incorrect report from the informant or limited IADL impairment. This may reflect variability in disease stage or clinical presentation, as functional difficulties can differ across dementia subtypes, with some atypical presentations showing relatively preserved IADL functioning. ⁴¹

In previous validation studies, the A‐IADL‐Q remained robust across different age groups and minor item bias related to age was reported. ¹³ , ¹⁵ Although we did find that age was a significant predictor of the A‐IADL‐Q score in this study, the effect size was small (β = −0.25), indicating limited clinical impact. Given that a previously established threshold for clinically meaningful cognitive decline was set at −2.4, ¹⁵ , ³⁰ this would correspond to an ≈ 10 years’ age difference to reach a clinically meaningful level of decline. The cognitive screening instruments MMSE and MoCA were both significantly associated with the A‐IADL‐Q score, although with minimal effect sizes. This finding is in line with previous literature, in which studies reported that global cognitive functioning explained between 21% and 42% of the variance in IADL. ⁴² , ⁴³ In clinical practice, application of previously published normative scores could aid in correcting for age and education level, ¹⁶ and the combination of both cognitive and functional measures would be recommended to obtain a reliable diagnosis. ⁴³

We did not find an effect for comorbidities on the A‐IADL‐Q score. This confirms previous reporting in a systematic review that comorbidity measures are poor predictors of IADL performance among older individuals presented in clinical settings. ⁴⁴ Yet, the associations of comorbidity measures on the A‐IADL‐Q were not previously studied in primary care, while these patients are generally older and have more comorbidities compared to the memory clinic. ²¹ Our findings provide support for the use of the instrument in the context of cognitive changes with minimum bias introduced by comorbidities. Still, the results should be interpreted with caution, as the data on comorbidities was collected manually and is binary (present or absent) and thereafter combined in composite scores, potentially leading to a less nuanced picture.

The emergence of anti‐amyloid therapies has highlighted the need of timely and accurate identification of AD in primary care. ⁴⁵ As PCPs generally tend to hold a reserved position in diagnosing and providing care, that is, “watchful waiting,” the recent therapeutic developments require a more active position and timely referral to secondary care. ⁴⁵ , ⁴⁶ Given its minimum sensitivity to demographic and comorbidity variables, the A‐IADL‐Q can be effectively used in assessing clinical stage at early stages as well as in separating individuals with moderate to severe dementia who would not be eligible for treatment. Particularly in combination with cognitive screening tests and blood‐based biomarkers the A‐IADL‐Q has the potential to be integrated in efficient, standardized care pathways in primary care. ⁴⁷ However, to reliably use the instrument in the context of eligibility for anti‐amyloid treatments and ensure generalizability across clinical settings, it should be examined in a larger and more diverse sample. In addition, future studies in primary care settings should investigate the diagnostic value of self‐reported A‐IADL‐Q scores and the potential divergence between self‐ and informant reports, particularly in individuals in the earliest stages of cognitive decline who may not yet have an available study partner.

While this study shows promise for accurate clinical stage classification in primary care using the A‐IADL‐Q total score, future studies may benefit from a more in‐depth assessment of psychometric properties of A‐IADL‐Q in this specific population and inspect potential bias on an item level. Recruitment of participants restricted to Sweden limits the generalizability of our study and results need external validation in independent diverse international primary care cohorts. Although primary care is usually the first point of contact in health care, clinical practice can vary greatly across countries, ⁴⁸ warranting cross‐country replication of our findings. The optimal T‐score cut‐off to distinguish dementia from non‐dementia in our primary care cohort was 58.3, notably higher than the 51.4 reported in the original memory clinic validation study. ¹⁴ This likely reflects differences in population characteristics, with primary care participants exhibiting a broader cognitive spectrum and higher baseline functioning than those referred to memory clinics, emphasizing the need for context‐specific thresholds across clinical settings.

In conclusion, this study adds to previous studies demonstrating the diagnostic accuracy of the A‐IADL‐Q in assessing clinical stage in cognitively impaired older individuals, particularly those with a high prevalence of comorbidities, within a primary care environment. This makes the A‐IADL‐Q a promising instrument to aid PCPs in identifying MCI and dementia, providing clinical management and support, and identifying suitable candidates for anti‐amyloid treatment.

CONFLICT OF INTEREST STATEMENT

Ayesha Fawad, Sophie M. van der Landen, Pontus Tideman, Angela van der Putten‐Toorenburg, Elke Butterbrod, Danielle van Westen, Susanna Calling, Patrik Midlöv, Niklas Mattsson‐Carlgren, Beata Borgström Bolmsjö and Maria H. Nilsson report no conflicts of interest. Ruben Smith has received consultancy/speaker fees from Eli Lilly, Novo Nordisk, Roche, and Triolab. Erik Stomrud has acquired research support (for the institution) from Beckman Coulter, Bristol Myers Squibb, C2N Diagnostics, Eisai, Fujirebio, GE Healthcare and Roche Diagnostics. Oskar Hansson is an employee of Eli Lilly. Sietske A. M. Sikkes is a recipient of the IHI‐AD‐RIDDLE project (grant agreement No. 101132933). AD‐RIDDLE is supported by the Innovative Health Initiative Joint Undertaking (IHI JU). The JU receives support from the European Union's Horizon Europe research and innovation programme and COCIR, EFPIA, EuropaBio, MedTech Europe and Vaccines Europe, with Davos Alzheimer's Collaborative, Combinostics OY., Cambridge Cognition Ltd., C2N Diagnostics LLC, and neotiv GmbH. Sietske A. M. Sikkes has acquired research support (for the institution) from Health∼Holland, Topsector Life Sciences & Health (PPP allowance: DEFEAT‐AD, LSHM20084; Remote‐DEM, LSHM22026), Alzheimer Nederland (SPREAD+ # WE.32‐2022‐01), and Ministry of Health, Welfare and Sports (#90001586), ZonMw in the context of Onderzoeksprogramma Dementia, part of the Dutch National Dementia Strategy (TAP‐dementia, #10510032120003), ZonMW (VIMP, #7330502051 and #73305095008, NWO (YOD‐MOLECULAR, #KICH1.GZ02.20.004) as part of the NWO Research Program KIC 2020‐2023 MISSION—Living with dementia. YOD‐MOLECULAR receives co‐financing from Winterlight Labs, ALLEO Labs, and Hersenstichting. Team Alzheimer also contributes to YOD‐MOLECULAR. Sebastian Palmqvist has acquired research support (for the institution) from Avid Radiopharmaceuticals and ki elements through ADDF. In the past 3 years, he has received consultancy/speaker fees from BioArtic, Danaher, Eisai, Eli Lilly, Novo Nordisk, and Roche.

CONSENT STATEMENT

All participants provided informed consent.

Supporting information

Supporting Information: dad270344‐sup‐0001‐SuppMat.docx

Supporting Information: dad270344‐sup‐0002‐SuppMat.pdf