Understanding the Role of Neuropsychiatric Symptom in Functional Decline in Alzheimer’s Disease

Abstract

OBJECTIVE:

Neuropsychiatric symptoms (NPS) are major predictors of cognitive impairment and functional decline in dementia. This study explores classes of NPS in Alzheimer’s disease (AD), examines the relationship between resulting NPS classes on rate of functional decline over time, and determines if effects of individual symptoms on functional decline remain after controlling for NPS classes.

METHODS:

Longitudinal analyses of 9,797 participants with Mild Cognitive Impairment or AD at baseline in the National Alzheimer's Coordinating Center Uniform Data Set. Function was measured using the Functional Assessment Questionnaire; NPS by clinician judgment. Latent class analysis (LCA) was used to identify clusters of individuals who share similar NPS profiles. Linear mixed models (LMM) were used to estimate the relationship between NPS classes and individual symptom profiles on functional decline over time.

RESULTS:

LCA revealed 4 distinct classes of NPS: an asymptomatic/mild group (47% of the sample, N=4,721), a second group predominantly with apathy and depression (34.6%, N=3,562), a third group with high rates of multiple symptoms except for hallucination (14.5%, N=264), and a small group with high rates of all symptoms (complex class, 3.1%, N=1,250). NPS classes differentiated baseline function but were not associated with rate of decline. Controlling for NPS classes, more persistent apathy strongly associated with faster rate of decline in function. Effects were observed across NPS classes and disease severity.

CONCLUSIONS:

Specific symptoms rather than classes of symptoms were associated with trajectory of functional decline in AD. In particular, apathy may be useful in tracking longitudinal change in function.

INTRODUCTION

Neuropsychiatric symptoms (NPS) such as apathy, depression, anxiety, irritability, agitation, delusions, and hallucinations are core features of Alzheimer’s disease and related dementia (AD/ADRD) (1, 2). They are associated with worse outcomes for patients and have significant impact on their caregivers and families (3–12). Individual NPS such as apathy and sleep/nighttime behavioral disturbance have been shown to strongly correlate with decline in function in dementia (6, 8, 11, 13–15). The Mayo Clinic Study of Aging examined cognitive decline over time in non-demented individuals and showed annualized change in global cognition differed by individual symptom (16). Participants with depression, apathy, and nighttime behavior had greater decline in all global cognition. In an earlier study, we examined relative contributions of individual NPS on functional decline using a well-characterized cohort of individuals diagnosed with Mild Cognitive Impairment (MCI) or AD enrolled in the National Alzheimer's Coordinating Center Uniform Data Set (NACCUDS) (17). Results showed that different symptoms affected patients’ rate of functional decline differently. In particular, apathy was strongly associated with faster rate of functional decline. Agitation, delusions, and hallucinations also were independently associated with faster functional decline, although the magnitude of the effects was smaller than that from apathy. Findings from these studies suggest that targeting specific behavioral symptoms for interventions may be effective in reducing rate of decline in cognition and function.

Instead of examining individual symptoms, several studies used the person-centered latent class analysis (LCA) to identify qualitatively different subgroups (classes) of patients that are at higher risk of progressing to dementia, which may offer a more holistic view of NPS. An early study from Cache County Dementia Progression Study revealed three classes of patients with dementia based on their NPS profile, a large mild class who were asymptomatic or have one symptom (59%), a second class with predominantly affective symptoms (28%), and a third class with predominantly psychotic symptoms (13%) (18). A later study also from Cache County identified four NPS groups, including asymptomatic, irritable, depressed, and complex (19). A study by Forrester and collaborators using the NACCUDS data found in participants with MCI three groups of patients classified according to NPS (asymptomatic cluster 56%, affective cluster 37%, and severe cluster 7%) (20). Another recent study addressed the impact of NPS in patients diagnosed with MCI, and concluded that the coexistence of certain symptoms, i.e., hyperactivity, affect disturbances, and psychosis, led to higher conversion to dementia (21). Findings from these studies point to the need to identify individuals with different NPS clusters vs. individual symptoms as potential targets for intervention trials.

In this study, we hope to shed light on whether specific symptoms or individuals with clusters of symptoms are associated with functional decline in AD. We used the same NACCUDS data as before to explore latent classes of NPS in patients with MCI or AD. We examined the relationship between resulting NPS classes on rate of functional decline over time, and determined if effects of individual NPS on functional decline remain when NPS classes are accounted for.

METHODS

Data Source and Sample Derivation

Data are drawn from the National Alzheimer's Coordinating Center Uniform Data Set (NACCUDS). Recruitment, participant evaluation, and diagnostic criteria are detailed elsewhere (22). Briefly, beginning in September 2005, participants have been enrolled and followed prospectively at approximately 12-month intervals from 42 National Institute of Aging (NIA) funded Alzheimer's Disease Research Centers (ADRC) located throughout the United States with a standardized protocol. Data from the ADRCs are anonymized and compiled into the NACCUDS, widely recognized as the gold standard in longitudinal, multi-domain neurocognitive and phenotypic data. Recruitment methods vary between ADRC, therefore NACCUDS should not be considered a population-based sample. Written informed consent was provided by all participants and their informants and approved by local Institutional Review Boards (IRB). Research using the NACCUDS database was approved by the University of Washington IRB.

Data used in the current study were drawn from all participants age 50 or older who were enrolled in NACCUDS until the November 2022 data freeze and had at least one follow-up visit. Participants diagnosed as MCI or AD at baseline with a primary etiologic diagnosis of AD, who continued to have the same primary etiologic diagnosis at more than half of their follow-up visits were included in the analysis. Of note, the etiologic diagnoses were primarily made on a clinical basis. Participants who at baseline had Clinical Dementia Rating (CDR) of 0 with their AD diagnosis and those who at baseline were already at CDR 3 were excluded. By year 5, about half of the participants remained in the study. The analysis included data for up to 5 UDS visits. Sample selection steps are described in Supplemental Figure 1.

Measures

Function

Our dependent variable is participants’ function, measured using the Functional Assessment Questionnaire (FAQ) reported from interviews with study partners. The FAQ asks whether the participant had any difficulty or needed help with 10 items in the previous 4 weeks on a scale from 0–3, corresponding to normal, has difficulty but does by oneself, requires assistance, and dependent. Responses to each item were summed to obtain a total FAQ score (range=0–30) and then divided by the number of tasks attempted to obtain a standardized score ranging from 0 to 3. Higher FAQ scores indicate worse functioning. About 1% of participants (N=124) were reported to have not attempted any tasks and therefore had a missing value for all FAQ items were excluded from the analysis.

Neuropsychiatric Symptom (NPS)

At each visit, study clinicians provided a clinical assessment, per NACCUDS protocol, of whether a participant currently manifested the symptom as a meaningful change in behavior (yes=1, no=0) for the following list: apathy-withdrawal, depression, psychosis (visual hallucinations, auditory hallucinations, abnormal/false/delusional beliefs), disinhibition, irritability, and agitation. We classified NPS clusters at baseline using latent class analysis (LCA). Based on how often a symptom was reported by the clinician throughout the participant’s follow-up period, we also categorized each symptom into four mutually exclusive groups: never occurred across all visits, intermittently occurring (more than one but <50% visits, persistently occurring (≥50% visits), and always occurred across all visits.

Demographic and Clinical Characteristics

Participant characteristics included age, sex, race (White, Black, Other), ethnicity (Hispanic/Latino yes/no), years of education, marital status (married/living as married vs. other), and living alone (yes/no) (Form A1). Participant medical history was obtained from Form A5. Medication use was reported using a medication inventory (Form A4) which included all medications (including nonprescription drugs, vitamins, and supplements) taken by the participant within 2 weeks of their visit. Dementia severity was measured by the global CDR (Form B4). Depressive symptoms by the 15-item Geriatric Depression Scale (GDS-15, Form B6) (23). Apolipoprotein E (APOE) genotype was reported by individual ADCs when available and by discretion.

Statistical Analyses

We carried out data analyses in three steps. First, we used LCA to identify clusters of individuals sharing similar NPS profiles at baseline. LCA posits the existence of underlying unobserved groups of individuals (latent classes) that present different patterns of observed responses on categorical variables, and generates probability of an individual belonging to one of the mutually exclusive latent classes. The optimal number of classes was determined based on bootstrapped likelihood ratio tests and Bayesian Information Criterion (BIC) (24). Second, using the model generated classes of NPS, we estimated a linear mixed model (LMM) to examine the relationship between NPS classes on participants’ functional outcome over time. The main independent variables in this LMM were NPS classes and their interactions with time, measured by years since baseline visit. Estimated coefficients of the NPS classes indicates baseline differences in function by NPS classes. Estimated coefficients of the interaction terms indicate differences in rate of decline in function by NPS classes. Lastly, to examine how NPS classes and individual symptom profiles were independently related to functional decline, we estimated a full LMM model that included both NPS classes and individual symptom profiles together in the LMM model. Estimated coefficients of a variable indicates baseline differences in function by that variable. Estimated coefficients of the interaction terms with time indicate differences in rate of decline in function by that variable.

LMM models controlled for baseline dementia severity (reference group: CDR=0.5) and their interactions with time, and the following demographic and clinical covariates: baseline age, gender, race/ethnicity, years of education, referral source (professionals vs. other), indicators for history of diabetes, hypertension, hypercholesterolemia, ApoE ε4 allele (none, heterozygote, homozygote), and indicators of current use of 20 classes of medications as reported in the NACCUDS. Models also included subjects and ADRCs as random intercepts, assuming subjects were nested within each ADRC. We tested models that included individual random slopes to allow participants to differ in their overall rate of change over time. Likelihood ratio tests suggested that including a random slope did not improve model fit and was subsequently dropped. Initial models also included interaction terms between CDR and NPS indicators. None of the interaction terms were statistically significant and were subsequently dropped. All analyses were performed using Stata 18.0 (25). Statistical significance was set a priori at p<0.05.

Sensitivity Analysis:

The composition of NPS clusters in MCI might be different from those in dementia as disease progresses. We therefore estimated a sensitivity analysis using MCI and dementia patients separately only.

RESULTS

Baseline Sample Characteristics

Average age of the sample was 73.7±9.2 years at baseline, 47% male, 78% non-Hispanic White, 11% non-Hispanic Black, 8% Hispanic (Table 1). Average education was 14.9±3.6 years. Average MMSE was 22.8±5.3. 55% of participants had CDR=0.5, 36% CDR=1, and 9% CDR=2. Of those with APOE genotyping, 52% had 0, 37% had one, and 11% had two APOE4 alleles. The majority of participants had hypertension (53.6%) and hypercholesterolemia (52.8%). Participants were followed for an average of 4±2.2 years. On average, participants were taking 6.3±3.9 medications.

Table 1.

Baseline Characteristics by NPS Class

Variable	All Sample		class 1		class 2		class 3		class 4
N	9,797		4,721		3,562		1,250		264
Age, mean, SD	73.7	9.2	74.3	9.0	72.9	9.4	74.1	9.0	74.1	9.3
Male, N, %	4,634	47.3	2,294	48.6	1,596	44.8	636	50.9	104	39.4
Non-Hispanic White, N, %	7,622	77.8	3,777	80.0	2,793	78.4	901	72.1	153	58.0
Years of schooling, mean, SD	14.9	3.6	15.2	3.4	14.8	3.6	14.3	3.8	13.3	4.0
Follow up years, mean, SD	4.0	2.2	4.1	2.3	4.0	2.2	3.8	2.1	3.3	1.7
Referred by professionals, N, %	5,780	59.0	2,677	56.7	2,194	61.6	740	59.2	167	63.3
Diabetes, N, %	1,215	12.4	533	11.3	452	12.7	188	15.0	43	16.3
Hypertension, N, %	5,251	53.6	2,464	52.2	1,909	53.6	721	57.7	159	60.1
Hypercholesterolemia, N, %	5,173	52.8	2,483	52.6	1,888	53.0	678	54.2	124	46.8
Apolipoprotein (ApoE) ε4 allele, N, %
None	5,055	51.6	2,521	53.4	1,745	49.0	645	51.6	143	54.1
Homozygote	3,625	37.0	1,690	35.8	1,371	38.5	468	37.4	96	36.4
Heterozygote	1,117	11.4	510	10.8	445	12.5	138	11.0	25	9.5
Mini-Mental State Examination (MMSE), mean, SD	2,234	22.8	1,124	23.8	794	22.3	269	21.5	48	18.0
Functional Assessment Questionnaire (FAQ), mean, SD	11.1	8.6	8.4	7.9	12.7	8.2	15.0	8.6	20.2	7.9
Neuropsychiatric Inventory Questionnaire (NPI-Q), mean, SD	4.0	4.1	2.3	3.0	4.4	3.7	9.4	5.2	7.9	4.9
Total number of medications, mean, SD	6.3	3.9	6.2	3.9	6.5	4.0	6.5	3.9	6.1	3.8
Antihypertensive or blood pressure medications, N, %	5,251	53.6	2,526	53.5	1,899	53.3	688	55.0	141	53.4
Lipid lowering med	4,477	45.7	2,167	45.9	1,642	46.1	568	45.4	101	38.3
Nonsteroidal anti-inflammatory drugs	3,674	37.5	1,898	40.2	1,279	35.9	414	33.1	81	30.7
Anticoagulant or antiplatelet	3,615	36.9	1,846	39.1	1,272	35.7	416	33.3	83	31.4
Antidepressant	3,468	35.4	1,020	21.6	1,749	49.1	574	45.9	122	46.2
Beta-adrenergic blocking agent	1,920	19.6	944	20.0	687	19.3	248	19.8	48	18.2
Angiotensin convert enzyme inh.	1,773	18.1	817	17.3	666	18.7	240	19.2	48	18.2
Calcium channel blocking agent	1,411	14.4	670	14.2	513	14.4	194	15.5	35	13.3
Diuretic	1,313	13.4	614	13.0	477	13.4	181	14.5	37	14.0
Angiotensin II inhibitor	1,038	10.6	519	11.0	363	10.2	131	10.5	26	9.8
Anxiolytic, sedative or hypnotic	1,078	11.0	482	10.2	403	11.3	159	12.7	35	13.3
Diabetes meds	950	9.7	411	8.7	349	9.8	150	12.0	37	14.0
Antiadrenergic agent	843	8.6	406	8.6	285	8.0	125	10.0	22	8.3
Antihypertensive combo therapy	402	4.1	179	3.8	153	4.3	54	4.3	10	3.8
Estrogen hormone therapy	225	2.3	109	2.3	96	2.7	23	1.8	5	1.9
Antiparkinsonian agent	196	2.0	80	1.7	82	2.3	24	1.9	6	2.3
Vasodilator	167	1.7	80	1.7	61	1.7	26	2.1	2	0.8
Antipsychotic agent	343	3.5	47	1.0	121	3.4	110	8.8	59	22.3
Estrogen + progestin therapy	29	0.3	9	0.2	14	0.4	1	0.1	0	0.0

Notes: Between class differences statistically significant at p<0.001 for all variables except for hypercholesterolemia (p=0.172)

NPS Classes

At baseline, clinicians reported NPS ranged from apathy (31.4%), depression (31%), irritability (27.1%), agitation (11.8%), disinhibition (9.4%), delusions (8.1%), to hallucinations (4.3%).

We explored LCA with one to five classes and tested model fit using bootstrapped likelihood ratio tests and Bayesian Information Criteria (BIC) and found that a 4-class model was the best fit (Table 2, Figure 1). Class 1 (47% of the sample, N=4,721) was constituted by participants who had few NPS at baseline. Most symptoms were endorsed at <5% except for depression (10.2%), irritability (9.1%), and apathy (6.4%). Class 2 (34.6% of the sample, N=3,562) was constituted by participants who had higher rates of apathy (51.9%), depression (49.9%), and irritability (24.2%) but lower rates (<10%) of other symptoms. Class 3 (14.5% of the sample, N=1,250) was constituted by participants who had high rates of many symptoms (more than 50% were endorsed for apathy, depression, irritability, or agitation) except for hallucination (5.6%). Class 4 (3.1% of the sample, labeled as “complex”, N=264) had high rates of all symptoms. Baseline characteristics differed significantly by NPS. (Table 1).

Table 2.

Latent Class Analysis Model Statistics for One- to Four-Class Solutions.

Model	ll (model)	df	AIC	BIC
One class	−28957.04	7	57928.07	57978.39
Two class	−26798.18	15	53626.36	53734.19
Three class	−26619.43	23	53284.86	53450.21
Four class	−26476.53	31	53015.07	53237.93

Notes: ll= Log-likelihood; df= Degree of freedom; AIC=Akaike Information Criterion; BIC=Bayesian Information Criterion

Figure 1.

Estimated conditional probabilities (y-axis) observed in the latent class analysis (LCA) for NPS domains (x-axis).

dep=depression; hall=hallucinations; del=delusion; disin=disinhibition; beirrit=irritability; agit=agitation.

Looking at individual symptoms throughout the entire follow-up period, apathy remained the most common NPS over time, endorsed by the clinicians in 61.5% of all visits, followed by irritability (55.3%), depressed mood (51.2%), agitation (38.5%), disinhibition (29.7%), delusions (25.9%), and hallucinations (17.1%).

Estimated Relationships between NPS Classes and Individual Symptoms and Functional Decline

To explore their independent effects of NPS classes and individual symptom profiles on FAQ over time, a full LMM model was estimated including both NPS Classes and individual symptom profiles as independent variables. Detailed estimation results are in Supplemental Table 1.

Figure 2 plots predicted FAQ scores generated from the LMM estimating the relationship between NPS classes on functional decline after controlling for individual symptom profiles. Results showed that controlling for NPS symptom profiles, the mild NPS class (class 1) had lower FAQ scores (better function) than other classes at baseline. Function declined over time for all NPS classes, but there were no differences in the rate of decline between NPS classes.

Figure 2.

LMM predicted FAQ scores over time by NPS classes.

AD=Alzheimer’s disease; FAQ=Functional Assessment Questionnaire; LMM=Linear mixed model; Cl=Confidence Interval

Results from linear mixed models (LMM). Outcome=FAQ score at each visit. Model predictions based on values of the covariates as observed.

Vertical bars indicate 95% confidence interval around coefficient estimates.

Figure 3 plots predicted FAQ scores generated from the model for each individual symptom profiles after controlling for NPS classes. Results also showed that controlling for NPS classes and other symptom profiles, baseline FAQ scores were higher in those with intermittent or persistent/always apathy, hallucinations, and agitation compared to those who never had the symptom. Over time, rate of decline in function was faster in those with intermittent and persistent/always apathy, agitation, and delusions compared to those without the symptom. There were no differences baseline function and in the rate of decline in FAQ for depression, disinhibition, or irritability.

Figure 3.

LMM predicted FAQ scores over time by neuropsychiatric symptoms.

AD=Alzheimer’s disease; FAQ=Functional Assessment Questionnaire; LMM=Linear mixed model; Cl=Confidence Interval

Results from linear mixed models (LMM). Outcome=FAQ score at each visit. Model predictions based on values of the covariates as observed.

Vertical bars indicate 95% confidence interval around coefficient estimates.

Sensitivity Analysis

Since the composition of NPS clusters in dementia might be different from MCI as disease progresses, we estimated a sensitivity analysis for MCI and dementia patients separately. Data revealed a 3-class structure in MCI and a 4-class structure in dementia (Supplemental Figure 2a and 3a). The three milder classes closely resembled each other in MCI and dementia patients, and the complex class was only seen in dementia patients.

LMM estimation results showed substantively similar results on the relationship between NPS classes and individual symptom profiles in both MCI and dementia subsamples. Specifically, while different NPS classes were associated with differences in baseline levels of function, they were not associated with rate of decline in function (Supplemental Figure 2b and 3b). Controlling for NPS classes, more persistent apathy was strongly associated with faster rate of decline in function than those with intermittent or without apathy (Supplemental Figure 2c and 3c).

DISCUSSION

As neuropsychiatric symptoms (NPS) are major predictors of functional decline in dementia, it is important to determine if there are specific symptoms or clusters of symptoms that may be suitable targets for interventions for preventing and delaying functional decline. In this study, we explored whether specific NPS or specific individuals who manifest clusters of symptoms were associated with functional decline in a large cohort of extensively characterized patients with AD across the spectrum of disease severity. From a specific NPS perspective, we simultaneously examined profiles of seven common behavioral symptoms (apathy-withdrawal, depressed mood, visual or auditory hallucinations, delusions, disinhibition, irritability, and agitation) based on ADRC research clinicians’ reports.

We used LCA to identify individuals with clusters of NPS. Our data revealed four distinct classes of NPS: a large class of participants who had few NPS (47%, asymptomatic/mild class), a class of participants who predominantly had apathy and depression (34.6%), a class of participants who had high rates of multiple symptoms except for hallucinations (14.5%), and a small class of participants who were had high rates of all symptoms (complex class, 3.1%). These different classes of NPS are consistent with those from the literature that showed large mild class who were asymptomatic or mild symptom and small group of participants with complex NPS. We extended the current literature by examining the relationship between NPS classes and patients’ functional decline over time. Results showed that while these NPS classes differentiated baseline function, they were not associated with differential rates of decline in function, suggesting that the NPS classes may be limited in helping clinician, caregiver or family’s predict the likely rate of functional deterioration and need for services.

In an earlier study we showed individual NPS, particularly apathy, was strongly correlated with rate of functional decline (17). Here we further showed that after controlling for NPS classes, apathy remained strongly associated with rate of functional decline. There are several reasons why apathy may be a particularly strong predictor of functional outcomes. First, adequate functioning relies on executive or cognitive control, which involves planning, task selection, and switching, all heavily dependent on the dorsolateral and ventrolateral prefrontal cortices and cognitive territory of the basal ganglia, particularly the dorsal caudate nucleus (26, 27). These areas have been shown to be impaired in executive or cognitive domain of apathy (28). Second, adequate functioning also requires a certain level of engagement and motivation. Inability to respond to emotional rewards makes it challenging to engage in activities that are typically rewarding. These functions, heavily dependent on the orbital and ventromedial prefrontal cortices as well as the limbic territories of the basal ganglia, particularly the ventral striatum, have been shown to be impaired in emotional apathy (28, 29). Further, the reduced participation in social activities, also referred to as social dimension of apathy by some researchers (30), further impairs personal and social functioning (31). These aspects of apathy may contribute to a negative cycle that accelerates functional impairment.

It is worth mentioning that apathy may serve not only as a symptom, but also as a potential marker for disease progression and apathetic patients may be inherently at higher risk of decline. Some studies suggest that development and progression of apathy is influenced by the Aβ pathology starting at very early stages of AD independent of cognitive changes (32). Several imaging (33) and CSF studies (34) have also shown relationships between apathy and higher level of Tau.

The relationship between apathy and accelerated functional decline is complex and multifaceted. This complexity is reflected in the ongoing efforts to define distinct domains of apathy (26, 30). We emphasize that relationships reported in the current observational study should be interpreted as associations and no causal pathways are implied. Discussions above highlights the critical importance of early detection and ongoing monitoring of apathy as a key marker for neurodegenerative disease and faster decline, and an important treatment target with potential to significantly improve quality of life of those affected. Identifying therapeutic interventions that directly affect these pathways may reduce apathy and improving function.

To put the magnitude of the estimated relationship between apathy and functional decline into context, a retrospective analysis using data from the NACCUDS reported average annual change in FAQ of 1.248 points (35). In the current study, we found that compared to individuals who never had apathy, individuals with intermittent and persistently/always apathy had worse function by 0.603±0.098 points at baseline and 0.857±0.094 points per year faster rate of decline. The magnitude of these estimates suggest that reducing apathy may have clinically meaningful impact on patients’ function. Compared to measures of underlying biologies, apathy may be more readily assessed clinically, suggesting that it may serve as a potential target for tracking longitudinal change in function.

In addition to apathy, results also showed that individuals with intermittent and persistent/always delusions had faster rate of decline compared to those who were never endorsed by the ADRC research clinicians as having delusions, but the magnitude of the effects were much smaller.

Clinically, these results highlight the importance of characterizing specific behavioral symptoms, particularly apathy, in the clinical management of dementia and as they can add valuable insight to prognosis and expected rate of decline. There is also some evidence that precision in defining the nature of the symptom may provide additional benefit beyond acknowledging the number or extent of symptoms. Identifying specific NPS symptoms in patients will allow clinicians to implement more personalized treatment and management plans that take into account patients’ current and expected future needs. In addition, the pattern of NPS can provide insight into the underlying biological mechanisms and the circuits involved at different stages of neurodegeneration. It should be noted that the characterization in this study was done by clinicians working in NIA-funded ADRCs with expertise in dementia evaluation that may be beyond primary care practitioners who typically see the most patients with cognitive impairment and dementia. It is possible that enhancing medical education and training that increases awareness and detection of specific NPS in dementia, could enhance patient evaluation and provide greater impact on detection and describing prognosis.

It is perhaps unsurprising that apathy, which is characterized by reduced motivation and independent goal-directed activity, should be so strongly associated with functional decline over time. However, its impact on function, which appears to be more significant than other more dramatic NPS that often dominate clinicians’ time and attention, is likely under-appreciated in most clinical settings. As such, it may be especially important to detect and quantify these symptoms early in the course of illness, and to develop interventions specifically aimed at reducing apathy. Given how common apathy is in AD, interventions focused on reducing frequency and severity of the symptoms may have important implications on functional stability.

Analysis of MCI and dementia separately showed different NPS classes were not associated with differential rate of decline in function. Controlling for NPS classes, more persistent apathy was strongly associated with faster rate of decline in function than those with intermittent or without apathy in both MCI and dementia. These results suggest that the effects of apathy on the rate of functional decline are consistent across dementia severity.

This study has several limitations. First, the NACCUDS study from which the sample is derived from is a highly educated, less ethnic-racially diverse cohort of research participants who have fewer medical and psychiatric comorbidities than the general population of the same age, and is not representative of the general population (36). The cohort is primarily recruited and selected for its cognitive status and may not fully reflect the profiles of these symptoms in a broader aging population. Second, NPS was defined based on clinician judgement, although clinician judgement should take into account all available information per study protocol, it may not be based on recently updated diagnostic criteria that provide a more standardized definition (37). Lastly, although diagnostic protocols are uniform across ADRCs, variations in practices remain.

The study also has several strengths including its large sample size, long duration of follow up, broad range of dementia severity, and extensive clinical characterization of participants and case ascertainment of behavioral symptoms by ADRC-based dementia expert clinicians from multiple centers utilizing informant input. The large sample size allows for more precise estimation and inclusion of symptoms that are less common. Potential biases that may arise from less common symptoms of potential biases that may arise from different practices across ADCs may be reduced, further enhancing robustness of study results.

CONCLUSIONS

Our study suggests that characterizing individual neuropsychiatric symptoms, particularly apathy, could provide valuable insight to expected rate of decline and may be sensitive targets for clinical trials tracking longitudinal change in function. Enhancing medical education for primary care providers who serve most patients with dementia, to improve awareness of NPS in those with MCI and dementia may be a worthwhile strategy to providing patient centered care with better prediction of prognosis and targeted approaches for treatment and management.