Associations of Alzheimer Disease and Related Dementia Neuropathologies With Timely Diagnosis of Dementia in Healthcare Settings

Abstract

Background and Objectives

A timely diagnosis of dementia may provide valuable time for treatment and planning, yet underdiagnosis is common. This study investigated the relationship between presence of dementia pathologies and timeliness of dementia diagnosis by healthcare providers.

Methods

This was a retrospective study using 5 cohorts at Rush Alzheimer's Disease Center. We included participants who met all of the following criteria: (1) incident dementia based on annual cohort assessments, (2) linkage to Medicare records, and (3) a completed postmortem brain autopsy. Postmortem neuropathologic examinations identified the presence of AD, limbic-predominant age-related TDP-43 encephalopathy neuropathologic change (LATE-NC), vascular pathologies, and neocortical Lewy bodies (LBs). In linked Medicare data, we defined timely diagnosis as the presence of claims with dementia diagnoses within 3 years before or 1 year after the cohort-based dementia onset. We used logistic regressions to quantify associations of neuropathology markers with timely diagnosis vs underdiagnosis.

Results

Of the 500 eligible participants (71% female, 95% non-Latino White, mean [SD] age at cohort dementia onset = 88 [7] years, mean [SD] years from onset to death = 4 [3]), only 54% received a timely diagnosis. After controlling for demographics, time to death, and other neuropathologies, a pathologic diagnosis of AD (OR = 1.91, 95% CI 1.21–3.00) and moderate/severe LATE-NC pathologies (OR = 1.83, 95% CI 1.25–2.68) were independently associated with higher odds of timely diagnosis. Moderate/severe vascular pathologies (OR = 0.94, 95% CI 0.55–1.59) and neocortical LB pathologies (OR = 1.00, 95% CI 0.64–1.55) were not significantly associated with receipt of a timely diagnosis. In a separate multivariable logistic regression, we found that participants with 3 or 4 neuropathologies present had an over 2-fold increase in odds of timely diagnosis (OR = 2.24, 95% CI 1.32–3.82), compared with those with 1 or no neuropathology.

Discussion

In deceased older adults with cohort-determined incident dementia, the healthcare system was twice as likely to capture those with pathologic diagnosis of AD, moderate/severe LATE-NC, and more than 3 copathologies in a timely manner. While findings from this predominantly White and highly educated sample warrant replication in broader population, this study is an important initial step toward understanding biological correlates of timely diagnosis of dementia.

Introduction

A timely diagnosis of dementia may play an important role in efforts to slow the degeneration of cognition, function, or quality of life caused by the development of dementia, by providing valuable time for treatment and/or planning. This requires healthcare providers to recognize the symptoms of dementia and diagnose the clinical syndrome at early stages; however, in clinical settings, only half of dementia cases are diagnosed by healthcare providers in a timely fashion.^1-3 Of note, while the efficacy of current therapeutics for dementia is still being debated, there is overall consensus that any eventual disease-modifying intervention will be more effective if implemented earlier in the course of disease, making timely diagnosis especially important.^4,5

Several demographic and clinical factors have been related to underdiagnosis of dementia in healthcare settings, including race, education, sex, marital status, and level of impairment.^6-8 However, not much is known about how dementia subtype is related to timeliness of diagnosis, in large part because subtyping dementia based on clinical syndrome alone can be difficult. In characterizing patterns of underdiagnosis, insufficient attention has been paid to the biological complexity of dementia, which, rather than being a single disease, is better viewed as a syndrome, where a constellation of symptoms that define the condition can be caused by multiple, often coexisting pathologic processes.^9-11 Determining whether certain underlying pathologies or combinations of pathologies result in dementia that is more likely to be diagnosed by clinicians in a timely fashion compared with others could prove useful in understanding patterns of underdiagnosis. Nevertheless, most studies of dementia lack the data sources required to link underlying pathologies to healthcare diagnoses, including multiple and granular measures of both neurodegenerative and cerebrovascular pathologies, and robust assessment of dementia onset linked to administrative claims records or electronic health records to observe timing of a clinical dementia diagnosis by providers.

Previously, by linking rigorous dementia assessments in the Rush Alzheimer's Disease Center (RADC) cohorts to Medicare claims, we examined demographic and clinical correlates of timely diagnosis.¹ Here, we extend these findings to pathologic correlates of timely diagnosis, by incorporating postmortem pathologic examinations in these cohorts. Specifically, among deceased participants, we investigated how presence of common postmortem dementia pathologies (AD, limbic-predominant age-related TDP-43 encephalopathy neuropathologic change [LATE-NC], vascular pathologies, and Lewy bodies) was related to receipt of a timely diagnosis of dementia from healthcare providers. These findings may offer novel insights into drives of underdiagnosis in clinical practice and inform strategies for more sensitive detection of dementia.

Methods

Cohort Studies and Analytical Sample

We included participants in 5 prospective cohorts at RADC with annual cognitive assessments. Since 1994, the Religious Orders Study (ROS) has followed Catholic nuns, priests, and brothers from across the United States (U.S.).¹² The Rush Memory and Aging Project (MAP) began in 1997 and follows older adults from the Chicago metropolitan area.¹² The Minority Aging Research Study (MARS) began in 2004¹³ and the Rush African American Clinical Core (Clinical Core) began in 2008,¹⁰ both following older African American adults recruited from churches, senior buildings, social clubs, and organizations that serve diverse populations in the Chicago metropolitan area. Finally, the Rush Latino Core (LATC) began in 2015 and includes older adults in the Chicago area who self-identify as Latino.¹⁴ Follow-up rate was 85%–90% in these cohorts. Participants in ROS and MAP consent to brain donation after death; brain donation is optional for MARS, Clinical Core, and LATC participants.^10,15 Overall, 82% of cohort participants provided consent for accessing their Medicare records, including 87% in ROS, 95% in MAP, 49% in MARS, 66% in Clinical Core, and 73% in LATC. The request for Medicare claims was approved by the NIH, and research identifiable files were shared through a signed data use agreement. Linked Medicare data were available through 1991–2022; to rule out potential impacts of the COVID-19 pandemic on healthcare practice, we restricted the current analysis to the end of 2019.

From 5,034 RADC cohort participants, we identified 1,015 participants with cohort-determined incident dementia as of December 31, 2019 (Figure 1). By the time of analysis (i.e., September 2024), 890 died, of whom 720 underwent autopsy and had complete neuropathologic data. We further excluded 65 participants who did not consent for Medicare linkage (n = 61) or could not be linked (n = 4). Finally, we excluded 94 participants who were not enrolled in fee-for-service Medicare and thus had no claims for identifying healthcare diagnoses. The final sample included 500 participants.

Figure 1. Flowchart of Sample Inclusion.

(A) Considering disruption of COVID-19 pandemic on seeking and delivering healthcare, we restricted the analyses to December 31, 2019. (B) We excluded 190 participants with cohort-assessed dementia at their baseline evaluation, for whom we did not have a date of incident diagnosis. (C) Pathologies of interest include Alzheimer disease, limbic-predominant age-related TDP-43 encephalopathy neuropathological change, vascular pathologies, and Lewy bodies pathologies. TDP-43 = TAR DNA-binding protein 43.

Neuropathologic Examinations

Brain autopsies followed a standard protocol described previously.¹⁰ The mean postmortem interval from death to brain removal was 9.1 (SD = 8.0) hours. In brief, brain hemispheres were cut coronally using a Plexiglas jig into 1-cm slabs and examined for visible pathology. One hemisphere not designated for freezing was fixed in 4% paraformaldehyde before the macroscopic examination for infarcts. Diagnostic blocks were dissected from a minimum of 10 brain regions: midfrontal, middle temporal, inferior parietal, anterior cingulate, calcarine cortex, entorhinal cortex with amygdala, hippocampus, basal ganglia, thalamus, and midbrain with substantia nigra. Sections (6 µm) from the blocks were stained for assessment of pathology. Neuropathologic diagnoses were made by expert research neuropathologists in AD/Alzheimer's Disease and Related Disorders Association, blinded to age and clinical data.

Pathologic Diagnosis of AD

A pathologic diagnosis of AD was rendered based on the National Institute on Aging-Alzheimer's Association (NIA-AA) criteria, which relies on a combination of neurofibrillary tangles (Braak), neuritic plaque score (Consortium to Establish a Registry for Alzheimer's Disease [CERAD]), and Aβ plaque score (Thal).¹⁶ As described elsewhere, Braak staging and CERAD scores were obtained from manual counts from 5 brain regions (frontal, temporal, parietal, and entorhinal cortices and hippocampus), using a modified Bielschowsky stain.¹⁷ Thal phase was determined using Aβ positivity across 7 brain regions, including neocortical regions, hippocampus, basal ganglia, midbrain, and cerebellum. Those with intermediate or high likelihood of AD according to NIA-AA recommendations received a final pathologic diagnosis of AD.

Moderate/Severe LATE-NC Pathology

Transactive response (TAR) DNA-binding protein 43 (TDP-43), a pathology marker of LATE-NC, was assessed using monoclonal antibodies to phosphorylated TDP-43 (pS409/410; 1:100) in 8 brain regions (amygdala, 3 limbic, and 4 neocortical).¹⁸ Based on the consensus criteria for a LATE disease,¹⁹ presence of TDP-43 cytoplasmic inclusions in each regions classified 4 stages of TDP-43 distribution: stage 0, absent; stage 1, localized to amygdala; stage 2, extension to hippocampus and/or entorhinal cortex, that is, limbic; stage 3, extension to the neocortex. We dichotomized LATE-NC pathology into none/mild (stage 0 or 1) vs moderate/severe (stages 2 or 3). We did not separately examine hippocampal sclerosis, because the 2 are highly correlated: in our cohorts approximately 95% of those with LATE-NC also had co-occurring hippocampal sclerosis.

Moderate/Severe Vascular Pathologies

The presence of moderate/severe vascular pathologies was based on presence and/or severity of any of the following cerebrovascular pathologies: one or more gross chronic infarcts by visual examination; one or more chronic microinfarcts under microscopy using hematoxylin and eosin stain²⁰; moderate-to-severe arteriolosclerosis detected by use of hematoxylin and eosin²¹; moderate-to-severe atherosclerosis detected by visual inspection of vessels in the Circle of Willis²²; and moderate-to-severe cerebral amyloid angiopathy (CAA) detected using amyloid-β immunostaining in 4 cortical regions. Scoring of CAA severity was assessed by semiquantitative ratings based on circumferential staining, number of vessels stained, and percent of the cortex and/or meninges affected; out of 4-level severity rating, we considered a score of 1.5 and above as moderate to severe, consistent with prior publications.²³ Owing to its close relationship with AD neuropathologic change, CAA was separated from vascular pathologies in sensitivity analysis.

Neocortical Lewy Body (LB) Pathology

We used an antiphosphorylated α-synuclein antibody to examine presence of LBs in 7 brain regions (midfrontal, midtemporal, entorhinal, cingulate and inferior parietal cortices, amygdala, and substantia nigra).²⁴ We dichotomized LB pathology into whether LBs present in neocortex (including midtemporal, inferior parietal, or midfrontal cortices), as reported previously.^24,25

Defining Timely Diagnosis and Underdiagnosis

Participants were categorized as having either timely diagnosis or underdiagnosis based on timing of clinical dementia diagnosis in healthcare claims relative to the timing of dementia onset determined by cohort assessment.

Timing of Dementia Onset Determined by Cohort Assessment

At each RADC annual assessment, participants underwent a three-stage process that consisted of a cognitive battery of 21 tests. Using 19 common tests, severity of impairment was rated for 5 cognitive domains. Then based on the impairment rating and other clinical information, a neuropsychologist rendered a clinical judgment regarding the presence of cognitive impairment. A clinician then reviewed all available data including structured medical history, to make a diagnostic classification of dementia based on the National Institute of Neurologic and Communicative Disorders and Stroke and the AD and Related Disorders Association (NINCDS/ADRDA) criteria. Date of cohort-determined dementia onset was specified as the date of cohort evaluation at which a participant was diagnosed with dementia for the first time. This diagnosis is considered a “research diagnosis”; while results of cognitive testing are shared with participants and they are encouraged to follow up with their doctor if they are concerned by the results, the research diagnosis is not routinely disclosed to participants (it is available upon request, although very few participants have requested it).

Timing of Dementia Diagnosis in Healthcare Settings

To identify dementia diagnosis in healthcare settings, we used the validated Bynum-Standard code-list of dementia-relevant diagnostic codes and conditions about number and location of claims.^26,27 The search identified persons with International Classification of Diseases, ninth and tenth revision (ICD-9/10; eTable 1) codes found under 2 conditions: (1) at least 1 claim with dementia diagnosis from the Medicare Provider Analysis and Review (MedPAR; which includes inpatient and skilled nursing facility stays), Home Health, or Hospice files, or (2) at least 2 qualifying claims that were at least 7 days apart, to account for potential rule-out diagnoses. The service date of claims with eligible dementia diagnosis was considered the date of healthcare diagnosis.

Window for Timeliness

A timely diagnosis was defined by the presence of Medicare claims with dementia diagnosis within a 4-year time window, spanning 3 years before or 1 year following the date of cohort-determined dementia onset. Rather than anchoring to a single date, use of a temporal window accounted for the insidious onset of dementia and cohort assessment intervals. Owing to lack of standard operational definition of “timeliness” in the literature, we used reasonable judgment to inform our selection of this window.¹ Specifically, we considered a claims diagnosis made within 1 year after cohort-determined dementia onset to be timely, assuming that once clinical symptoms meet the research-based threshold for dementia, a one-year look-forward period is sufficient for recognition by healthcare providers. Meanwhile, we allowed for a longer 3-year look-back window before the cohort-determined dementia onset because we recognized that some clinicians may diagnose dementia based on milder symptoms that our assessment would classify as mild cognitive impairment (MCI), but this would still mean that dementia onset was captured by the healthcare system in a timely fashion. We chose to define “underdiagnosis” as the absence of a claims diagnosis in this 4-year window; this included participants who received a “late diagnosis” more than 1 year after the date of incident dementia, and who never received a diagnosis through December 31, 2019, or death, that is, “missed diagnosis.”

Demographic and Health Characteristics

Demographics included self-reported age, sex, years of education, and race and ethnicity. Years from cohort incident dementia to death were calculated based on death date collected in cohort studies. We measured all time-varying characteristics at the time of cohort-determined dementia onset. When characterizing cognition, we used the Mini-Mental State Examination ([MMSE], range 0–30) score and a global cognition z-score summarizing performance on the 19 cognitive tests common to all cohorts. Disability in activities of daily livings (ADLs) was measured with the Katz ADL Scale, reflecting the number of ADLs a participant could not perform without help (range 0–6).²⁸ Disability in instrumental activities of daily living (IADLs) was measured as the number of IADLs a participant needed help (range 0–8), adapted from Duke Older Americans Resources and Services project.²⁸ Using Medicare claims, we derived the Elixhauser Index (range 0–31) as a measure of comorbidity burden,²⁹ and number of inpatient and outpatient encounters during 1 year before cohort-determined dementia onset.

Statistical Analysis

We first examined distributions of neuropathologic, demographic, and health variables for participants with timely diagnosis of dementia and for those with underdiagnosis. For continuous variables with normal distributions, t tests were used to compare means across the 2 analytical categories. For skewed continuous variables, Kruskal-Wallis tests were used. χ² tests were used to compare proportions for categorical variables.

To examine the relationship of neuropathology type with timely diagnosis, we conducted a series of multivariable logistic regression models. First, we tested the relationships of each of the 4 neuropathology markers (AD pathologic diagnosis, moderate/severe LATE-NC pathology, moderate/severe vascular pathologies, and neocortical LB pathology) with timely dementia diagnosis in 4 separate models, adjusted for potential confounders including age at dementia onset, sex, race and ethnicity, years of education, and time between cohort-determined dementia onset and autopsy. Then, we included all 4 neuropathology markers (which are not mutually exclusive) and the same covariates in a single model to test associations of each pathology with timely diagnosis independent of other pathologies. For interpretability, we estimated the probability of timely diagnosis for each neuropathology marker and for 2 commonly observed pathology combinations, (1) coexisting AD and LATE-NC, present in 34% of participants and (2) coexisting AD and vascular pathologies, present in 62% of participants, based on parameters from the model with all pathologies and assuming an average profile of covariates (i.e., mean age, years of education, and time from dementia onset to death and mode categories of sex, i.e., female, and race and ethnicity, i.e., non-Latino White).

As sensitivity checks, we conducted separate logistic regression analyses, adding Elixhauser Comorbidity Index and the calendar year of cohort-based dementia onset as additional covariates; these models tested whether associations between neuropathology and timely diagnosis were confounded by comorbidity burden and secular trends in dementia diagnostic practice, respectively. To test for synergistic effects, we added interactions between AD and the 2 commonly co-occurring pathologies described above: LATE-NC and vascular pathologies.

n addition, to understand whether the overall number pathologies present in the brain is related to timely diagnosis, we fitted a separate logistic regression replacing individual neuropathology markers with number of pathologies present, regardless of the specific pathologic type. Since neuropathologies were almost ubiquitously present in this deceased sample, we combined the few participants without any neuropathology with those with 1 neuropathology (n = 85), and compared them with those with 2 neuropathologies present (n = 200) and with 3 or 4 neuropathologies present (n = 215).

Standard Protocol Approvals, Registrations, and Participant Consents

All cohort studies were approved by an institutional review board of Rush University Medical Center; informed consent including for Medicare linkage was obtained as a repository consent.

Data Availability

In accordance with the Data Use Agreement between Rush University Medical Center and the NIH, only named persons on the Data Use Agreement are granted permission to access Medicare records files. All other data for participants of the cohort studies are available on request through the RADC Research Resource Sharing Hub: radc.rush.edu.

Results

Participant Characteristics

Eligible participants (N = 500) were, on average, aged 87.8 (SD = 6.6) years at cohort-determined dementia onset (Table 1) and lived for another 4 years (mean = 3.9, SD = 3.2). The sample was predominately female (70.6%) and non-Latino White (94.8%), with high educational attainment (mean years of education = 16.2, SD = 3.7). At cohort-determined dementia onset, participants on average scored 20.6 (SD = 5.4) in the Mini-Mental State Examination (MMSE), and reported difficulties in 1.8 (SD = 2.1) ADLs and in 4.5 (SD = 2.8) IADLs. The mean Elixhauser Comorbidity Index was 4.9 (SD = 3.4). At autopsy, AD pathologic diagnosis (78.8%) and moderate/severe vascular pathologies (85.6%) were the most commonly observed, followed by moderate/severe LATE-NC pathology (45.2%). Neocortical LB pathology was the least commonly observed neuropathology (22.4%).

Table 1.

Sample Characteristics at Cohort Diagnosis of Incident Dementia

Characteristics Mean (SD) or No. (col. %)	Overall (N = 500)	Timely diagnosis (n = 272)	Under diagnosis (n = 228)	p Value
Age in y	87.8 (6.6)	87.4 (6.1)	88.3 (7.1)	0.13
Years from cohort diagnosis to death	3.9 (3.2)	3.5 (2.9)	4.2 (3.5)	0.01
Female, (%)	353 (70.6)	186 (68.4)	167 (73.3)	0.23
Non-Latino White, (%)	474 (94.8)	260 (95.6)	214 (93.9)	0.39
Years of education	16.2 (3.7)	16.3 (3.8)	16.2 (3.5)	0.83
Presence of neuropathology, (%)
AD pathologic diagnosis (NIA-AA)a	394 (78.8)	225 (82.7)	169 (74.1)	0.02
Moderate/severe LATE-NC pathologyb	226 (45.2)	136 (50.0)	90 (39.5)	0.02
Moderate/severe vascular pathology	428 (85.6)	230 (84.6)	198 (86.8)	0.47
Neocortical Lewy body pathology	112 (22.4)	62 (22.8)	50 (21.9)	0.82
Cognition at cohort diagnosis
MMSE (0–30)c	20.6 (5.4)	19.8 (5.5)	21.6 (5.0)	<0.001
Global cognition (z-score)	−1.3 (0.6)	−1.3 (0.7)	−1.2 (0.6)	0.002
Comorbidity and disability at cohort diagnosis
Elixhauser index (0–31)	4.9 (3.4)	5.3 (3.4)	4.4 (3.4)	0.002
ADL limitations (0–6)d	1.8 (2.1)	1.9 (2.1)	1.7 (2.0)	0.38
IADL limitations (0–8)e	4.5 (2.8)	4.7 (2.8)	4.3 (2.7)	0.13
Contact with healthcare system in the past year
Number of inpatient encounters	0.7 (1.2)	0.8 (1.2)	0.6 (1.3)	0.85
Number of outpatient encounters	8.1 (7.1)	8.3 (7.2)	7.9 (7.0)	0.70

Abbreviation: AD = Alzheimer disease.

^aNIA-AA = National Institute on Aging-Alzheimer's Association.

^bLATE-NC = limbic-predominant age-related TDP-43 encephalopathy neuropathological change.

^cMMSE = Mini-Mental State Examination.

^dADL = activities of daily living.

^eIADL = instrumental activities of daily living.

Of the 500 eligible participants, 272 (54.4%) received a timely dementia diagnosis in Medicare claims within 3 years before or 1 year after cohort-determined dementia onset (Table 1). Participants with timely diagnosis and underdiagnosis were similar in age, sex, education, ADL/IADL disability, and frequency of medical encounters. In this autopsied subset primarily with non-Latino Whites, there was no difference in timely diagnosis status by race and ethnicity, unlike what we have previously reported in the full sample.¹ The timely diagnosis group had slightly shorter survival after cohort-determined dementia onset (mean = 3.5 years, SD = 2.9) compared with the underdiagnosis group (mean = 4.2 years, SD = 3.5; p = 0.01). Consistent with what we previously published,¹ persons with timely diagnosis had more severe cognitive impairment (timely diagnosis MMSE mean = 21.6, SD = 5.0; underdiagnosis MMSE mean = 19.8, SD = 5.5) and higher comorbidity burden at the time of cohort-determined dementia onset (timely diagnosis Elixhauser Index mean = 5.3, SD = 3.4; underdiagnosis Elixhauser Index mean = 4.4, SD = 3.4).

The proportion of persons with timely dementia diagnosis varied by the presence or absence of a given neuropathology type (Figure 2). Fifty-seven percent of persons with a pathologic diagnosis of AD had a timely diagnosis compared with 44% of persons without a pathologic diagnosis of AD. Similarly, those with moderate/severe LATE-NC pathology had a larger portion of timely diagnosed participants (60%) than those without (50%). Timely diagnosis was more similarly distributed in those with moderate/severe vascular pathologies (54%) vs those without (58%), and in those with neocortical LB pathology (55%) vs those without (54%).

Figure 2. Percent With Timely Diagnosis, According to the Presence of Each Neuropathology.

Red bars represented percent with timely diagnosis among participants with each pathology present. Green bars represented percent with timely diagnosis among participants without each pathology present. AD = Alzheimer disease; LATE-NC = Limbic-predominant age-related TDP-43 encephalopathy neuropathologic change; NIA-AA = National Institute on Aging-Alzheimer's Association.

Multivariable Analysis

In separate, adjusted models for each pathology type (Table 2, Models 1–4), a pathologic diagnosis of AD and moderate/severe LATE-NC were both associated with increased odds of timely diagnosis, whereas moderate/severe vascular pathologies and neocortical LB pathology were not. These patterns persisted in a model with all pathologies included at the same time (Table 2, Model 5): pathologic diagnosis of AD (OR = 1.91, 95% CI 1.21–3.00) and moderate/severe LATE-NC pathology (OR = 1.83, 95% CI 1.25–2.68) were associated with an almost two-fold increase in the odds of timely diagnosis, independent of other pathologies, while moderate/severe vascular pathologies (OR = 0.94, 95% CI 0.55–1.59) and neocortical LB pathology (OR = 1.00, 95% CI 0.64–1.55) were not significantly associated with timely diagnosis. All results were robust to additional adjustment for Elixhauser Comorbidity Index and calendar year of cohort-based dementia onset, as well as exclusion of CAA from vascular pathologies (results not shown).

Table 2.

Adjusted Odds Ratios (95% CIs) for Timely Diagnosis (N = 500)

Variable	Model 1	Model 2	Model 3	Model 4	Model 5
Age at cohort diagnosis of incident dementia (per y)	0.95 (0.92–0.98)	0.95 (0.92–0.98)	0.96 (0.93–0.99)	0.96 (0.93–0.99)	0.94 (0.91–0.97)
Male (vs female)	1.15 (0.76–1.76)	1.09 (0.72–1.64)	1.11 (0.74–1.67)	1.11 (0.74–1.68)	1.12 (0.74–1.70)
Non-Latino White (vs all other)	1.76 (0.77–4.03)	1.64 (0.72–3.74)	1.62 (0.71–3.68)	1.62 (0.71–3.69)	1.78 (0.77–4.09)
Years of education (per y)	0.99 (0.94–1.04)	1.00 (0.95–1.05)	0.99 (0.94–1.04)	0.99 (0.94–1.04)	1.00 (0.95–1.05)
Time from cohort diagnosis of incident dementia to death (per y)	0.89 (0.84–0.95)	0.88 (0.83–0.94)	0.90 (0.85–0.96)	0.90 (0.85–0.96)	0.88 (0.82–0.94)
AD pathologic diagnosis (NIA-AA, vs not)	1.93 (1.23–3.02)				1.91 (1.21–3.00)
Moderate/severe LATE-NC pathology (vs not)		1.85 (1.27–2.70)			1.83 (1.25–2.68)
Moderate/severe vascular pathology (vs not)			0.92 (0.55–1.55)		0.94 (0.55–1.59)
Neocortical Lewy body pathology (vs not)				1.05 (0.69–1.62)	1.00 (0.64–1.55)

Abbreviations: AD = Alzheimer disease; LATE-NC = limbic-predominant age-related TDP-43 encephalopathy neuropathologic change; NIA-AA = National Institute on Aging-Alzheimer's Association.

Note: Estimates were obtained from multivariate logistic models, adjusted for all covariates listed above.

Based on parameter estimates derived from Model 5, we predicted the probability of timely diagnosis associated with each pathology for average participants in Figure 3. The highest probability of receiving a timely diagnosis was observed among those with pathologic diagnosis of AD (53%) or moderate/severe LATE-NC pathology (52%). Those with moderate/severe vascular pathologies and neocortical Lewy body pathology had less than 40% probability of receiving a timely diagnosis.

Figure 3. Predicted Probability of Timely Diagnosis.

Predicted probabilities were based on parameters in the main model, that is, Table 2 Model 5, for average participants with mean age/years of education/time from dementia onset to death, female sex, and non-Latino White. AD = Alzheimer disease; LATE-NC = Limbic-predominant age-related TDP-43 encephalopathy neuropathologic change; NIA-AA = National Institute on Aging-Alzheimer's Association.

Next, using the same parameters from Model 5, we estimated the probability of timely diagnosis for 2 commonly observed mixed pathology combinations, including (1) coexisting AD and LATE-NC pathologies and (2) coexisting AD and vascular pathologies. In Figure 4A, compared with when either AD or LATE-NC pathology was present, having both present raised the probability of timely diagnosis from 53% to 67%. However, in secondary analysis fitting a separate multivariable logistic regression that included the interaction between AD and LATE-NC pathologies, the interaction term was not significant (p value = 0.38; eTable 2), suggesting a potential additive rather than synergistic effect of the 2 pathologies. In Figure 4B, presence of moderate/severe vascular pathologies on top of AD did not modify the probability of timely diagnosis (51% probability for co-existing AD and vascular pathologies, vs 53% probability for AD alone). The interaction term in multivariable-adjusted model also was not statistically significant (p value = 0.10; eTable 3).

Figure 4. Predicted Probability of Timely Diagnosis, by Presence of Mixed Neuropathology.

Panel A: By presence of AD and LATE-NC pathology. Panel B: By the presence of AD and vascular pathologies. Predicted probabilities were based on parameters in the main model, that is, Table 2 Model 5, for average participants with mean age/years of education/time from dementia onset to death, female sex and non-Latino White. AD = Alzheimer disease; LATE-NC = limbic-predominant age-related TDP-43 encephalopathy neuropathologic change; TDP-43 = TAR DNA-binding protein 43.

Finally, we examined the probability of timely diagnosis by number of neuropathologies present (Figure 5; see model estimates in eTable 4). Those with zero or 1 neuropathology had 43% probability of timely diagnosis; having 2 neuropathologies present increased the probability to 51%, which did not reach statistical significance (p value = 0.25). Participants with 3 or 4 neuropathologies present had significantly higher probability of timely diagnosis at 63%, compared with those with 0 or 1 neuropathology (p value<0.01).

Figure 5. Predicted Probability of Timely Diagnosis, by the Number of Neuropathologies Present.

Predicted probabilities were based on parameters in eTable 4, for average participants with mean age/years of education/time from dementia onset to death, female sex, and non-Latino White.

Discussion

Leveraging the linkage of Medicare claims to rigorous annual dementia assessments and postmortem pathologic examinations from cohort studies, we found that the presence of specific dementia-causing pathologies was related to different probabilities of a timely diagnosis by healthcare providers. Persons with pathologic diagnosis of AD and moderate/severe LATE-NC at autopsy were most likely to have had a timely dementia diagnosis. Furthermore, participants with mixed pathologies, in particular those with more than 3 copathologies, were more likely to have received a timely diagnosis, compared with those with one or no pathology. This study is an important initial step to understanding biological correlates of timely diagnosis of dementia, supplementing what we previously reported on clinical correlates.¹

One likely explanation for these findings is the difference in clinical manifestations of dementia across pathology types and combinations. The healthcare system may on average be more sensitive to impairment in certain cognitive domains than others. AD is the most common and well-known cause of dementia,^10,11 and its key clinical feature, loss of memory, may be more readily recognized by healthcare providers as dementia. This may be in part due to the fact that existing dementia screening tools, such as MMSE, focus heavily on memory deficits, making impairment in other cognitive domains less easily detected.^30,31 The other pathology associated with timely recognition, LATE-NC, is known to mimic AD clinically,¹⁹ which may explain earlier identification of dementia cases with LATE-NC pathology. This suggests a potential for improving timely recognition of dementia in clinical practice, by incorporating tools that more effectively capture impairment in nonmemory domains, such as executive function or visuospatial ability. Alternatively, healthcare providers may be more likely to recognize persons who decline more rapidly and present with more severe cognitive impairment overall. Several clinical-pathological studies have reported distinctive cognitive trajectories associated with specific neuropathologies, and patterns in general support this possibility: AD and LATE-NC are more robustly related to cognitive deterioration; in contrast, LB pathology plays a larger role in decline proximate to death, and vascular pathologies tend to have relatively weaker associations.^32,33 Our findings regarding mixed pathologies may also support this explanation: co-occurring AD and LATE-NC pathologies is associated with accelerated cognitive decline compared with each pathology alone, consistent with the signal here of coexisting AD and LATE-NC raising the probability of timely diagnosis.^34,35 We further found that participants with the greatest number of copathologies had the highest probability of a timely diagnosis, consistent with evidence that additional pathologies exacerbate deterioration.^10,36-39

Other explanations may also play a role. For example, if different pathologies are differentially related to the occurrence of certain comorbidities, healthcare encounters intended for these comorbidities may provide opportunities for clinicians to detect dementia. This possibility is consistent with our previous work that higher comorbidity burden is associated with greater likelihood of timely diagnosis.¹ Alternately, competing demands of caring for these comorbidities may distract from a dementia diagnosis. In other neurologic diseases, such as multiple sclerosis, comorbidities are linked to an elevated risk of diagnostic delay,^40,41 suggesting that similar diagnostic overshadowing may exist in the context of dementia. Differential risks of death may also play a role, such that certain pathologies are associated with shorter or longer survival, thereby reducing/increasing opportunities for diagnosis.

There are several limitations to note. Beyond the 2 common mixed pathologies examined here, we were not able to separately examine the large number of possible combinations of pathologies, due to our modest sample size. For the same reason, in understanding the association between number of copathologies with timely diagnoses, we cannot further distinguish whether this is due to (1) an additive pathologic burden regardless of specific markers or (2) a higher probability of including “diagnosis-promoting” markers, for example, AD and LATE-NC, as the total count of pathologies increases. In addition, this analysis assumes pathologies assessed via postmortem autopsy were already present or at a similar level of severity at the time of cohort-determined dementia onset. This is likely, as AD pathology is known to develop and persist decades before onset of clinical symptoms,^42,43 and the preclinical periods of vascular and LB pathologies may last for about a decade, longer than the average lag of 4 years between dementia and pathology identification. That being said, our findings require replication using in vivo biomarkers in future studies when validated biomarkers are available for all pathologies included in this analysis. Lastly, our findings may not be generalizable to the broader population. Participants in this study reside either in the Chicago area or in religious communities around the United States and consented to brain donation; they are predominantly non-Latino White and highly educated, with an advanced average age at dementia onset. Therefore, care-seeking behaviors, provider diagnostic practices, and distribution of neuropathologies observed here may differ from those in other regions, contexts, or demographic groups. In addition, autopsy rates, consent rates for Medicare linkage, and fee-for-service enrollment were all lower in the Black and Latino cohorts resulting in a sample that was almost entirely made up of non-Latino White participants, which may further limit generalizability beyond White older adults. In the future, these findings need to be replicated in more representative samples of older adults, and—when available for all neuropathologies examined here—using in vivo biomarkers.

Nevertheless, the study has important strengths. We used a unique data set combining pathologic assessment with healthcare claims to assess underlying pathology in relation to timeliness of healthcare identification of dementia. Moreover, the design of RADC cohort studies ensured high internal validity of our findings, including but not limited to (1) uniform pathologic and clinical assessment across all participants, (2) pathologic assessment conducted blinded to cohort assessment, and (3) high follow-up and autopsy rate. As an initial look into the pathologic correlates of a timely dementia diagnosis, this study revealed systematic patterns linked to biological processes not directly observable to healthcare providers. These findings may motivate future research into dementia diagnostic practice and inform measures to reduce underdiagnosis.

Glossary

ADL

activities of daily living

CAA

cerebral amyloid angiopathy

IADL

instrumental ADL

LATC

Latino Core

LATE-NC

limbic-predominant age-related TDP-43 encephalopathy neuropathologic change

LB

Lewy body

MAP

Memory and Aging Project

MARS

Minority Aging Research Study

MMSE

Mini-Mental State Examination

NIA-AA

National Institute on Aging-Alzheimer's Association

RADC

Rush Alzheimer's Disease Center

ROS

Religious Orders Study

Author Contributions

Y. Chen: drafting/revision of the manuscript for content, including medical writing for content; study concept or design; analysis or interpretation of data. A. Chen: drafting/revision of the manuscript for content, including medical writing for content; analysis or interpretation of data. M.C. Power: drafting/revision of the manuscript for content, including medical writing for content; major role in the acquisition of data; study concept or design; analysis or interpretation of data. F. Grodstein: drafting/revision of the manuscript for content, including medical writing for content; study concept or design; analysis or interpretation of data. A. Kapasi: drafting/revision of the manuscript for content, including medical writing for content; analysis or interpretation of data. A.W. Capuano: drafting/revision of the manuscript for content, including medical writing for content; analysis or interpretation of data. B.S. Lange-Maia: drafting/revision of the manuscript for content, including medical writing for content. A. Moghtaderi: drafting/revision of the manuscript for content, including medical writing for content. E.K. Stapp: drafting/revision of the manuscript for content, including medical writing for content. J. Bhattacharyya: drafting/revision of the manuscript for content, including medical writing for content. R.C. Shah: drafting/revision of the manuscript for content, including medical writing for content; analysis or interpretation of data. L.L. Barnes: drafting/revision of the manuscript for content, including medical writing for content. D.A. Bennett: drafting/revision of the manuscript for content, including medical writing for content; analysis or interpretation of data. B.D. James: drafting/revision of the manuscript for content, including medical writing for content; major role in the acquisition of data; study concept or design; analysis or interpretation of data.

Study Funding

National Institute on Aging (R01AG072559, R01AG079226, R01AG015819, R01AG022018, R01AG017917, Rush AD Research Center grant P30AG10161, P30AG072975).

Disclosure

Y. Chen, and A. Chen reports no disclosures relevant to the manuscript. M.C. Power reports research grants from the United States (U.S.) NIH and U.S. Department of Defense, and the Prince Georges County Department of Family Services. F. Grodstein, A. Kapasi, A.W. Capuano, B.S. Lange-Maia, A. Moghtaderi, E.K. Stapp, and J. Bhattacharyya reports no disclosures relevant to the manuscript. R.C. Shah reports being the site principal investigator or sub-investigator for AD clinical trials for which his institution (Rush University Medical Center) has been compensated in the last 24 months (Athira Pharma, Inc., Annovis Bio, Inc., Edgewater NEXT, Eisai, Inc., and Genentech, Inc.), being a noncompensated scientific advisor to Novo Nordisk, and being a compensated scientific advisory board member of Lundbeck on the diagnosis and management of dementia. L.L. Barnes, D.A. Bennett, and B.D. James reports no disclosures relevant to the manuscript. Go to Neurology.org/N for full disclosures.