Preparedness and quality of dementia care in an Italian health authority when introducing anti‐Aβ therapies

Abstract

INTRODUCTION

We aimed to assess dementia care organizations within a southern Italian local health authority (Azienda Sanitaria Locale di Barletta Andria Trani) and to propose an equitable implementation framework for anti‐amyloid therapies.

METHODS

We performed a retrospective analysis of dementia‐related outpatient activity in 2024 patients across 18 district clinics (primary care clinics) and three specialized centers for cognitive disorders and dementia (specialized memory centers), integrating booking, clinical, imaging, and prescribing data.

RESULTS

Among 1158 initial evaluations, specialized memory centers showed substantially higher quality indicators than district clinics: Mini‐Mental State Examination documentation (80.0% vs. 6.6%, p < 0.001), full clinical documentation (95.5% vs. 10.0%), higher use of magnetic resonance imaging and amyloid positron emission tomography, and 5‐fold higher rate of prescribing acetylcholinesterase inhibitors. Validated radiological scoring and biomarker facilities (cerebrospinal fluid, apolipoprotein E genotyping) were largely unavailable.

DISCUSSION

Major organizational gaps limit readiness for disease‐modifying Alzheimer's disease therapies. A four‐tier hub‐and‐spoke model may improve equity, efficiency, and safe implementation.

Highlights

• Organizational fragmentation, not disease prevalence, represents the primary barrier to equitable access to anti‐amyloid therapies in dementia care.

• Specialized diagnostic centers demonstrated superior cognitive assessment rigor and documentation completeness (80% Mini‐Mental State Examination administration, 95% documentation) compared to district clinics (6.6%, 10%), yet were overwhelmed bottlenecks rather than gateways.

• Only 7.6% of patients underwent a comprehensive biomarker assessment (amyloid positron emission tomography) despite 57.7% meeting clinical criteria for mild cognitive impairment or mild dementia.

• Electronic health record fragmentation prevented seamless data flow between organizational levels, fundamentally compromising coordinated care and long‐term outcome tracking.

• A four‐tiered implementation model (Tier I: district‐based screening; Tier II: comprehensive neuropsychology; Tier III: biomarker confirmation; Tier IV: therapeutic hub) preserves accessibility while optimizing specialized resources.

1. BACKGROUND

Alzheimer's disease (AD) represents one of the most urgent public health crises today, affecting tens of millions worldwide with projections indicating significant burden growth by 2050. ¹ , ² , ³ , ⁴ Global health‐care systems spend > 1 trillion dollars annually on direct dementia care costs, including medications, clinical assessments, and facility‐based services, with substantially higher indirect costs from informal caregiving and lost productivity that are not captured in formal health‐care budgets. ³ , ⁵ Italy, like many European countries, faces this crisis amid an aging population and overstretched health‐care infrastructure. ⁶

For >30 years, treatment options remained limited to symptomatic therapies, such as acetylcholinesterase inhibitors and memantine, drugs that modestly slow cognitive decline but cannot alter the disease's neurodegenerative course. ⁷ This reflected a 99.6% failure rate in clinical trials targeting the amyloid cascade hypothesis between 2002 and 2012. ⁸ However, a fundamental shift occurred with the US Food and Drug Administration approvals of aducanumab in 2021, lecanemab in January 2023, ⁹ and of donanemab in July 2024: ¹⁰ all antibodies directed against various forms of amyloid beta (Aβ) in the brain.

Adopting anti‐amyloid therapies (ATTs) require streamlined diagnostic workflows. ¹¹ Eligibility requires confirmation of amyloid pathology through expensive amyloid positron emission tomography (PET) scans or cerebrospinal fluid (CSF) biomarker analysis, with resources unevenly distributed and in short supply. ¹² Close magnetic resonance imaging (MRI) monitoring is mandatory for the timely detection of amyloid‐related imaging abnormalities (ARIAs) in treated patients. ¹³ , ¹⁴ Apolipoprotein E (APOE) genotyping and complex informed consent are required for eligibility evaluation. ¹⁵ , ¹⁶ Multidisciplinary teams including neurologists, neuroradiologists, infusion nurses, and care coordinators ensure safe supervision of ATT infusions. ¹⁷

This infrastructural complexity conflicts with Italy's fragmented regional health‐care organization, in which diagnostic capacity remains dispersed, specialist expertise is concentrated in urban centers, and primary–secondary care integration remains incomplete. A panel of Italian experts recently developed recommendations for the efficient implementation and appropriateness of ATT in Italy, ¹⁸ but at the time of this analysis, the Italian government has not authorized reimbursability of either lecanemab or donanemab. Apart from adopting ATT or other disease‐modifying therapies (DMTs), implementing effective diagnostic and therapeutic workflows is fundamental to improving dementia care worldwide. This study focused on analyzing the quality of dementia care in our territory, with a secondary goal of developing a diagnostic and patient‐care workflow optimized for the local health authority known as Azienda Sanitaria Locale di Barletta Andria Trani (ASL BT) that could potentially be translated to other local sanitary authorities in Italy.

2. METHODS

2.1. Setting and study design

This retrospective, observational cohort study investigated dementia care practices within ASL BT, located in the Apulia region of southern Italy. The study period covered the entire year of 2024 (January 1 to December 31). ASL BT serves ≈ 400,000 residents through a comprehensive network that includes 18 outpatient clinics; 3 specialized memory centers functioning as tertiary referral centers staffed with multidisciplinary teams (neurologists, geriatricians, neuropsychologists); and 15 primary care clinics managed by neurologists, geriatricians, or psychiatrists.

Data were collected from three institutional repositories and cross‐validated to ensure consistency and accuracy: (1) Centralized Booking System (CUP), which identified all outpatient visits coded for cognitive decline, dementia, or related neurocognitive disorders; (2) Picture Archiving and Communication System (PACS), which provided technical details of neuroimaging procedures and allowed for qualitative assessment of radiological reporting quality; and (3) Regional e‐Prescribing Platform, which supplied information on prescriptions for acetylcholinesterase inhibitors and memantine. All datasets were anonymized before aggregation. Two independent data abstractors conducted data extraction, with discrepancies resolved through consensus discussion supervised by a senior neurologist.

2.2. Variables and data collection

For each patient, the following variables were systematically recorded: demographic characteristics (age, sex), visit type, final clinical diagnosis, Mini‐Mental State Examination (MMSE) results, neuroimaging modality and quality metrics, CSF biomarker results when available, pharmacological treatments, APOE genotyping status, referral source, and overall documentation completeness.

2.3. Comparative simulation analysis: status quo versus hub‐and‐spoke model

A comprehensive cost‐efficiency comparison and budget impact projection were conducted to compare two diagnostic models for dementia care within ASL BT, which serves 1158 patients with cognitive impairment. This analysis is a cost‐efficiency projection rather than a full cost‐effectiveness study, as it does not incorporate clinical outcomes such as improvements in diagnostic accuracy or patient health gains. Future implementation research should incorporate outcome metrics and longitudinal follow‐up data to conduct rigorous cost‐effectiveness analysis as the proposed model is operationalized.

The analysis compared: (1) status quo, a theoretical scenario resembling the existent organization in which all patients receive complete diagnostic testing including neuropsychological assessment, structural and functional neuroimaging, biomarker evaluation via lumbar puncture or amyloid PET, and genetic analysis; and (2) the hub‐and‐spoke model, a proposed tiered organizational model with centralized coordination and risk‐stratified resource allocation.

For the hub‐and‐spoke model simulation, the following inputs were used based on published epidemiological data and local prevalence estimates: 50% (579 patients) were considered affected by either mild cognitive impairment (MCI) or mild dementia and advanced to specialist assessment in Tier II; 40% (232 patients) were considered to have biologically proven AD based on amyloid pathology and advanced to Tier III; 10% (23 patients) were considered fulfilling anti‐amyloid monoclonal antibody eligibility criteria in Tier IV. These percentage stratifications were derived from known AD prevalence estimates and anti‐amyloid monoclonal antibody eligibility rates in published implementation studies. ¹⁹

RESEARCH IN CONTEXT

1. Systematic review: Previous literature consistently identifies fragmented diagnostic infrastructure, unequal access to biomarker testing, and inadequate data integration as major barriers to implementing anti‐amyloid monoclonal antibodies in routine practice. While clinical trials demonstrate compelling efficacy, real‐world uptake remains severely limited. This study quantifies organizational dysfunction in a defined Italian territory and proposes an operationalized, evidence‐based implementation framework.

2. Interpretation: Organizational barriers, not disease prevalence, represent the primary obstacle to equitable access. The stark disparity in diagnostic rigor between specialized centers and district clinics reveals systematic disadvantages for primary care patients. Our four‐tiered hub‐and‐spoke model addresses these failures through explicit pathways and standardized referral criteria.

3. Future directions: Blood‐based biomarkers could decentralize biomarker confirmation to primary care, dramatically lowering identification thresholds. Digital health integration and rigorous implementation of scientific methodologies are critical to the equitable deployment of disease‐modifying therapies.

Monte Carlo simulation techniques were used to account for uncertainty in cost estimates and patient flow variables, with 10,000 iterations performed to establish 95% confidence intervals around all primary outcome estimates.

Unit costs for diagnostic procedures were derived from the Italian National Health Service (SSN) fee schedules current as of 2024: MMSE administration: €45; standard cranial computed tomography (CT): €85; brain MRI: €320; lumbar puncture with CSF biomarker analysis (Aβ42, Aβ40, phosphorylated tau [p‐tau]181, p‐tau217): €850; amyloid PET imaging: €1200; comprehensive neuropsychological battery: €400; APOE genotyping: €150. All costs represent direct diagnostic pathway expenditures only, with costs for multidisciplinary team consultation time, nursing labor, facility overhead, and administrative functions not included, representing a limitation that may underestimate true system costs.

Costs aggregation in the status quo workflow was represented by the sum of costs of all exams available without using any referral and selection principles, and irrespective from results of previous steps (according with the “all to all” axiom). On the other hand, cost aggregation in the hub‐and‐spoke model represented the sum of all costs incurred by applying the tiered diagnostic pathway which is weighed on inclusion/exclusion criteria and according with known prevalence data and inclusion criteria. ¹⁹

2.4. Statistical analysis

Statistical analysis involved both parametric and non‐parametric methods to evaluate differences between specialized memory centers and primary care clinics. Chi‐squared (χ²) tests served as the main statistical tool for categorical variables, while independent samples t tests compared mean values for continuous variables, with significance level set at p < 0.05 (IBM SPSS Statistics version 28.0).

3. RESULTS

3.1. Study population characteristics

During the 12‐month study period, ASL BT recorded 3174 outpatient visits for cognitive impairment or suspected dementia, including 1158 initial evaluations (36.4%) and 2016 follow‐up visits (63.6%). Of the 1158 initial evaluations, 1082 had retrievable medical records. The distribution across the care network showed that 818 (75.6%) initial evaluations occurred in primary care clinics, whereas 264 (24.3%) occurred in the three specialized memory centers. The remaining 76 evaluations (6.5%) were untraceable, possibly due to lack of alignment between booking systems and actual clinical delivery. Clinical evaluations from primary care clinics were traceable only in 10% of cases. Critical information such as clinical and demographic items was largely unavailable in reports from primary care clinics due to the widespread use of paper records. The average age at initial evaluation was 76.4 ± 7.8 years; 63% of participants were women, consistent with the known female predominance in AD and other dementias. ¹² , ²⁰

3.2. Cognitive testing practices

Cognitive assessment varied significantly across the network (Table 1). All three specialized memory centers used standardized neuropsychological tests including MMSE and domain‐specific batteries. MMSE was administered to 211 out of 264 (80%) patients in specialized memory centers, whereas primary care clinics had a much lower assessment rate, with MMSE available only in 54 out of 818 (6.6%) initial evaluations (p < 0.001). The average MMSE scores differed significantly between sites (specialized memory centers: 20.1 ± 0.4; primary care clinics: 17.12 ± 0.72; p < 0.001). In specialized memory centers, 57.8% of patients had mild or moderate cognitive impairment according to known diagnostic criteria ⁹ , ¹⁰ (Table 1). Functional level, an important factor for interpreting cognitive tests and diagnoses, was recorded in 100% of patients evaluated at specialized memory centers but was unavailable in patients from primary care clinics.

TABLE 1.

Comparison of dementia care quality indicators between specialized centers (specialized memory centers) and district clinics (primary care clinics) reveals significant disparities in cognitive assessment, neuroimaging, pharmacotherapy, and documentation quality (all p < 0.05), with specialized memory centers consistently outperforming primary care clinics across all measured domains.

Variable	Specialized memory centers	Primary care clinics	Test statistic	p value
Patients with documented MMSE, n (%)	211 (80.0%)	54 (6.6%)	χ2 = 204.9	< 0.001
Mean MMSE	20.1 ± 4.0	17.12 ± 0.72	t = 8.34	< 0.001
Mild‐moderate (MMSE ≥ 20), n (%)	122 (57.8%)	–	–	–
Severe dementia (MMSE < 20), n (%)	89 (42.2%)	–	–	–
Any neuroimaging	189 (71.6%)	493 (60.3%)	11.47	< 0.001
Brain CT	146 (55.3%)	433 (52.9%)	0.44	0.507
Brain MRI	67 (25.4%)	146 (17.8%)	6.74	0.0094
Aβ PET	33 (12.5%)	49 (6.0%)	9.43	0.0021
Acetylcholinesterase inhibitors	137 (52.1%)	80 (9.8%)	204.9	< 0.001
Memantine	29 (11.0%)	51 (6.3%)	4.08	0.043
Combination therapy	14 (5.4%)	10 (1.2%)	13.2	<0.001
Full documentation	252 (95.5%)	82 (10.0%)	822.5	< 0.001
Incomplete documentation	12 (4.5%)	736 (90.0%)	–	–
Standardized templates	264 (100.0%)	0 (0.0%)	818.0	<0.001
Electronic health record use	264 (100.0%)	54 (6.6%)	678.9	<0.001

Abbreviations: Aβ, amyloid beta; CT, computed tomography; MMSE, Mini‐Mental State Examination; MRI, magnetic resonance imaging; PET, positron emission tomography.

3.3. Neuroimaging use and quality

Neuroimaging was performed on 189 (71.6%) of 264 patients from specialized memory centers and in 493 (60.3%) of 818 initial evaluations from primary care clinics (Table 1). CT was obtained in 146 (55.3%) from specialized memory centers and in 433 (52.9%) from primary care clinics. MRI was conducted in 67 (25.4%) from specialized memory centers and in 146 (17.8%) from primary care clinics. Amyloid PET was performed in 33 (12.5%) patients from specialized memory centers and in 49 (6%) from primary care clinics. The use of validated scoring systems was reported in only a small portion of diagnostic reports: Scheltens scoring for medial temporal atrophy in only 3% of reports, the Fazekas scale for vascular lesions in 2% of reports, and the Kipps and Koedam scales for evaluation of parietal and occipital atrophies were never adopted (Table 1). ²⁰

3.4. Fluid biomarkers and CSF analysis

CSF analyses were not conducted for any patient during the whole of 2024, likely due to the absence of a standardized lumbar protocol within the local health authority and the absence of DMTs availability, rendering an invasive examination difficult to justify.

3.5. Drug treatment patterns

Medication use varied significantly across settings (Table 1). Acetylcholinesterase inhibitors were prescribed to 52.1% of specialized memory centers patients compared to only 9.8% in primary care clinics (χ² = 204.9, p < 0.001), likely due to differences in diagnostic accuracy and prescribing protocols. Memantine was prescribed to 11.0% of specialized memory centers patients versus 6.3% in primary care clinics (χ² = 3.2, p = 0.07). Combination therapy (acetylcholinesterase inhibitors plus memantine) was given to 5.4% of specialized memory centers patients and 1.2% in primary care clinics (χ² = 13.2, p = 0.0003). No patients received ATT during 2024, reflecting absent approval in Italy throughout 2024 (Table 1).

3.6. Documentation quality and data systems integration

Out of the 1082 initial traceable evaluations, only 82 patients (7.6%) underwent a comprehensive assessment including Aβ biomarker analysis. Documentation was significantly more complete at specialized memory centers (95% of visits with full, relevant documentation) than at primary care clinics (10%), indicating higher use of software applications for clinical recording, better organizational capacity, and staffing. However, only specialized memory centers adopted standardized documentation templates (Table 1).

3.7. Patient characteristics at specialized memory centers

Of the 264 patients evaluated at specialized memory centers, 211 were identified as having some form of cognitive impairment (Table 2). The diagnostic distribution showed notable diversity: probable AD in 25%, vascular dementia in 17%, mixed dementia in 37%, frontotemporal dementia (behavioral subgroup) in 4.3%, dementia with Lewy bodies in 2.8%, other specified dementias in 6.6%, and cognitive impairment not otherwise specified in 7.3%. The average MMSE score was 20.1 ± 0.4. Approximately 35% of the patients were independent or slightly dependent (Activities of Daily Living [ADL] ≥ 5). Observed comorbidities included ischemic heart disease in 31.1% and hypertension in 8.8%, with the distribution of comorbidities as follows: 3% with no comorbidities, 33.7% with one to two comorbidities, 37.1% with three to four comorbidities, and 26.1% with five or more.

TABLE 2.

Demographic and clinical characteristics of patients with cognitive impairment evaluated at specialized memory centers. Data include cognitive status, functional assessments (ADL and IADL scores), dementia subtypes, common comorbidities, and medication classes. Values are presented as n (%) for categorical variables and mean ± standard error for continuous variables.

Assessment type	Specialized memory centers n (%)
MCI/mild dementia, n	124 (57.7%)
ADL score (range 1–6)	5.1 ± 0.08
IADL score (range 0–8)	4.6 ± 0.09
Mixed dementia	37.0%
Probable Alzheimer's disease	25.1%
Vascular dementia	17.1%
Frontotemporal dementia (bvFTD)	4.3%
Dementia with Lewy bodies	2.8%
Other specified dementias (PSP, FTD variants)	6.6%
Comorbidity	Percentage
Ischemic heart disease (IHD)	31.1%
Hypertension (HTN)	28.8%
Hyperlipidemia/dyslipidemia	26.5%
Diabetes mellitus type II	24.2%
Atrial fibrillation (AF)	14.4%
Chronic obstructive pulmonary disease (COPD)	8.3%
Hypothyroidism	10.6%
Previous stroke/TIA	13.3%
Malignancy (various)	7.2%
Chronic kidney disease	6.4%
Obstructive sleep apnea (OSA)	5.3%
Parkinsonism	4.5%
Drug class	Percentage
Antihypertensives	37.1%
Statins/lipid‐lowering agents	28.8%
Antiplatelet agents (ASA, clopidogrel)	24.2%
Antidiabetic agents	20.5%
Anticoagulants (warfarin, DOACs)	11.7%
Beta‐blockers	18.2%
ACE inhibitors/ARBs	19.7%
Calcium channel blockers	14.0%
Antidepressants (SSRI/SNRI)	29.5%
Antipsychotics (quetiapine, olanzapine, haloperidol)	15.9%
Anti‐anxiety agents (lorazepam, buspirone)	11.7%
Sedative‐hypnotics (trittico, zopiclone)	9.8%

Abbreviations: ACE, angiotensin‐converting enzyme; ADL, Activities of Daily Living; ARB, angiotensin II receptor blocker; ASA, acetylsalicylic acid; bvFTD, behavioral variant frontotemporal dementia; DOAC, direct oral anticoagulant; FTD, frontotemporal dementia; IADL, Instrumental Activities of Daily Living; MCI, mild cognitive impairment; PSP, progressive supranuclear palsy; SNRI, serotonin–norepinephrine reuptake inhibitor; SSRI, selective serotonin reuptake inhibitor; TIA, transient ischemic attack.

3.8. Biological diagnosis

One hundred twenty‐four (124) patients (57.7%) evaluated at specialized memory centers had MCI or mild dementia ⁹ , ¹⁰ (Table 3). Thirty‐three of these underwent amyloid PET imaging, with 21 revealing brain amyloid deposition. Due to the absence of reimbursability and unavailability of ATT in Italy in 2024, none of the patients underwent APOE genotyping, which would be required if ATT were approved, and available.

TABLE 3.

Diagnostic workup and biomarker characterization of patients with AD‐associated MCI or mild dementia at specialized memory centers). The table presents sequential steps in patient evaluation, including cognitive screening (MMSE ≥ 20), amyloid PET imaging results showing brain Aβ deposition, and APOE genotyping status. Data are shown as absolute numbers and percentages relative to the population at each screening step. Note: APOE genotyping data was not available for this cohort.

Metric/Step	Number of patients	Percentage (%)	Notes/context
Patients with MCI or mild dementia (MMSE ≥ 20)	124	57.7%	Of total patients evaluated at specialized memory centers
Underwent amyloid PET imaging	33	26.6%	Of patients with MCI or mild dementia (124)
Revealed brain Aβ deposition (amyloid PET positive)	21	63.6%	Of patients who underwent amyloid PET imaging (33)
Underwent APOE genotyping	None	N/A	Required criterion for lecanemab or donanemab eligibility

Abbreviations: Aβ, amyloid beta; AD, Alzheimer's disease; APOE, apolipoprotein E; MCI, mild cognitive impairment; MMSE, Mini‐Mental State Examination; PET, positron emission tomography.

3.9. A proposed four‐tiered implementation model

We suggest a structured four‐tiered model based on international best practices to promote equity while optimizing resources. Upon first screening evaluation by a general practitioner, patients needing to finalize their assessment should enter a multi‐tiered model of evaluation (Figure 1).

FIGURE 1.

Proposed four‐tier diagnostic and therapeutic pathway for Alzheimer's disease within a hub‐and‐spoke organizational framework. Tier I provides broad, accessible specialist evaluations in outpatient clinics with basic cognitive screening and initial investigations, together with initial evaluation of possible contraindications to DMD. Tier II offers detailed neuropsychological assessment to refine diagnostic accuracy and guide referral. Tier III centralizes advanced diagnostic confirmation using biomarkers and high‐resolution MRI, allowing risk stratification for DMTs. Tier IV consists of specialized therapeutic hubs equipped for infusion‐based treatments, APOE genotyping, and structured MRI monitoring to ensure safe and coordinated multidisciplinary care. ARIA, amyloid‐related imaging abnormality; APOE, apolipoprotein E; CSF, cerebrospinal fluid; CT, computed tomography; DMD, disease modifying drugs; MMSE, Mini‐Mental State Examination; MRI, magnetic resonance imaging; PET, positron emission tomography

Tier I is an initial specialist evaluation with enhanced contraindication screening. Initial specialist evaluation may be conducted either at primary care clinics or specialized centers and should include: (1) thorough clinical history; (2) cognitive screening with the MMSE; (3) assessment of daily living activities and functional level; (4) documentation of behavioral symptoms; (5) routine laboratory tests; (6) no‐contrast enhanced cranial CT as an essential initial imaging modality; and (7) systematic screening for contraindications to anti‐amyloid monoclonal antibody therapy, including assessment of frailty status, current anticoagulant use, recent cardiovascular events, uncontrolled hypertension, renal insufficiency, and anticipated poor adherence or inability to comply with follow‐up monitoring.

Patients scoring < 20 on the MMSE or with clear contraindications to anti‐amyloid therapy do not progress through the workflow and are managed at the site of first clinical referral for symptomatic management and community‐supported psychosocial services. Patients scoring ≥ 20 on the MME, with a Clinical Dementia Rating (CDR) global score of 0.5 to 1 and no contraindications, or with atypical cognitive features not identified by standard screening, are referred to Tier II centers for comprehensive neuropsychological assessment. ¹⁸ Of note, patients complaining of a change in cognition (either amnestic or non‐amnestic), but scoring MMSE = 30 will not be labeled “cognitively healthy” but progress to the higher tier to receive a more thorough evaluation to unveil the compromised domain through complete neuropsychological testing. Indeed, according to current criteria for MCI, “there should be evidence of lower performance in one or more cognitive domains that is greater than would be expected for the patient's age and educational background.” ²¹

Tier II, consisting of comprehensive neuropsychological evaluation and clinical reassessment, is finalized at specialized centers. Evaluation includes validated, standardized test batteries capable of detailed characterization of domain‐specific cognitive deficits such as MMSE, Montreal Cognitive Assessment, and Alzheimer's Disease Assessment Scale Cognitive subscale 14. Comprehensive neuropsychological testing at Tier II is primarily intended for: (1) situations of diagnostic uncertainty in which brief testing yields ambiguous results, (2) suspected atypical presentations (such as logopenic primary progressive aphasia or behavioral/dysexecutive variants). ²¹ This evaluation is intended as a clear “add‐on” resource rather than a tool for ruling out patients with a clinical presentation different from amnestic AD. Because psychologists are associated with specialized memory centers and access to their evaluations is restricted only to patients proposed by the neurologists of these centers in a shared manner, waiting times are minimal. Patients meeting criteria for MCI or mild dementia, ¹⁷ , ¹⁸ without contraindications to anti‐amyloid therapy, advance to Tier III for definitive biomarker testing. ²¹ Patients not qualifying for ATT are referred to symptomatic treatment programs.

Tier III, advanced diagnostic evaluation with biomarker confirmation, involves definitive neuropathological testing through CSF analysis of Aβ42, Aβ40, and p‐tau181 and p‐tau217. ¹² , ²² Alternatively, patients unwilling to undergo lumbar puncture may undergo amyloid PET instead. Emerging blood‐based biomarkers, such as plasma p‐tau217, may serve in the future as pre‐screening tools to stratify patients, with negative results potentially excluding them from further invasive testing. Nevertheless, p‐tau217 still requires patient stratification, the establishment of reference intervals, and approval by the Italian Ministry of Health in Italy. ²³ High‐field MRI evaluates brain atrophy, amyloid angiopathy, microhemorrhages, and white matter changes, which are essential for comprehensive risk assessment, especially for potential ARIAs during treatment. ²⁴ The Scheltens, Fazekas, Kipps, and Koedam scales will be adopted to quantify atrophy on MRI for phenotypic characterization and long‐term monitoring of disease progression. Indeed, p‐tau217 demonstrates superior diagnostic accuracy compared to MRI‐based cortical thickness measures for predicting AD. ²⁵

Tier IV, a therapeutic hub for DMT administration, consists of tertiary cognitive centers equipped with infusion facilities and multidisciplinary care teams that coordinate the administration of ATT in settings that ensure comprehensive safety monitoring and emergency medical support. Pre‐treatment APOE genotyping is a required part of patient assessment, as APOE ε4 carriership is the strongest genetic predictor of the incidence and severity of ARIAs. ²⁶ MRI scans are scheduled to identify vasogenic edema (ARIA‐E) and hemorrhages or superficial siderosis (ARIA‐H) in patients receiving ATT in a timely way. ²⁷ , ²⁸ , ²⁹ Tier IV centers must have immediate access to corticosteroids, neurocritical care resources, and emergency imaging to manage acute adverse events.

3.10. Total diagnostic costs and budget impact

For the status quo model, all diagnostic procedures were applied to all 1158 patients: status quo total per‐patient cost = €3050 (rounded €3100; Table 4).

TABLE 4.

Cost comparison and operational efficiency comparison between traditional centralized diagnostic pathways (status quo) and a hub‐and‐spoke model for AD‐associated cognitive impairment screening in specialized memory clinics. The table presents annual diagnostic costs, per‐patient costs, savings stratified by DMT eligibility, and modality‐specific cost reductions (amyloid PET, CSF biomarker analysis, MRI, and CT). Additionally, it demonstrates improvements in diagnostic efficiency, including reductions in unnecessary testing rates and total diagnostic tests performed, alongside accelerated diagnostic pathway completion times. Data represent projections for a cohort of 1158 patients evaluated over 5‐ and 10‐year periods.

Metric	Status quo (€)	Hub‐and‐spoke (€)	Savings (%)
Total annual diagnostic costs	3,675,670	2,046,260	45.4%
Cost per patient	3100	1692	45.4%
Cost per treatable patient (DMT eligible)	156,078	85,234	45.4%
Annual amyloid PET scanning costs	348,000	69,600	80.0%
Annual CSF biomarker analysis costs	464,000	46,400	90.0%
Annual MRI imaging costs	578,900	144,725	75.0%
Annual CT screening costs	404,800	202,400	50.0%
5‐year total savings	—	—	45.4%
10‐year total savings	—	—	45.4%
Total diagnostic tests (1158 patients)	13,896	8172	−41.2%
Unnecessary testing rate	60%–65%	15%–20%	−73%
MRI scans performed	1158	290	75%
Complete diagnostic pathway	16–24 w	6–10 w	62% faster

Abbreviations: AD, Alzheimer's disease; CSF, cerebrospinal fluid; CT, computed tomography; DMT, disease‐modifying treatment; MRI, magnetic resonance imaging; PET, positron emission tomography.

For the hub‐and‐spoke model, procedures were applied selectively based on tier assignment and progression criteria: Tier I (all 1158 patients): MMSE + CT = €130 per patient; Tier II (579 patients): neuropsychological testing = €400 per patient; Tier III (232 patients): CSF biomarkers = €850 or amyloid PET = €1200; MRI = €320 per patient; Tier IV (23 anti‐amyloid eligible patients): APOE genotyping = €150 per patient. hub‐and‐spoke per‐patient cost = €1692.

The cost per treatable patient (anti‐amyloid monoclonal antibody‐eligible): for the status quo it is €156,078 per treatable patient and for the hub‐and‐spoke model it is €85,324 per treatable patient.

The hub‐and‐spoke diagnostic model demonstrated substantial cost savings compared to the status quo approach (Table 4). Total annual diagnostic costs decreased from €3,675,670 to €2,046,260, representing a 45.4% reduction in overall spending. Per‐patient diagnostic costs fell from €3100 to €1692, maintaining the same proportional savings. Most notably, the unnecessary testing rate decreased dramatically from 60% to 65% in the status quo model to 15% to 20% in the hub‐and‐spoke model, representing a 73% reduction in unnecessary procedures (Table 4).

4. DISCUSSION

4.1. Quality of dementia care

A systematic review of dementia care organizations within ASL BT identifies the absence of coordinated diagnostic pathways, biomarker infrastructure, and specialized referral networks, which critically lowered the quality of dementia care in 2024. The health‐care system in our territory is divided into two semi‐autonomous branches: primary care clinics, which are widely distributed across the area and manage 82% of cognitive concerns, and three specialized memory centers that make use of standardized neuropsychological batteries, advanced imaging protocols, and rigorous diagnostic procedures aligned with research standards. However, specialized memory centers are overwhelmed and function more as bottlenecks than gateways. ³⁰ This organizational structure results in predictable outcomes: patients with subtle cognitive complaints often receive superficial evaluations without formal neuropsychological testing; referral pathways lack clarity and are inconsistently applied; and essential patient data do not flow smoothly between clinics. ³⁰

Indeed, in our system we found a significant difference in MMSE score between specialized memory centers, where the mean score was 20.1 ± 0.4; and primary care clinics, where the mean score was 17.12 ± 0.72. While variations in assessment rigor between primary and specialty care settings are expected and reflect normal organizational divisions of labor, the magnitude of disparity observed, with specialized centers achieving 80% MMSE documentation and 95% overall documentation completeness compared to 6.6% and 10% in primary care clinics, respectively, suggests systematic organizational fragmentation that impedes patient care coordination and equitable access to diagnostic pathways.

4.2. Efficiency and sustainability of dementia health‐care models

Simulations show that the hub‐and‐spoke model achieves efficiency through intelligent risk stratification, not patient loss. Universal screening conceived by the status quo model proved organizationally impossible (diagnostic saturation), while the tiered pathways (hub‐and‐spoke model) leverage Bayesian logic: pre‐test probability increases at each tier, making expensive tests more selective and cost efficient. The hub‐and‐spoke model improves clinical quality through centralized expertise, standardized protocols, and geographic equity via distributed Tier 1 screening. ²³ It systematically reduces unnecessary and redundant investigations, freeing up clinical and infrastructural resources to focus on targeted diagnostic pathways and therapeutic outcomes. ²³ Sufficient MRI capacity is crucial to support the intensive monitoring needed for patients on ATT, especially for timely detection of ARIAs. ²⁷ , ²⁸ , ²⁹

4.3. International challenges and dementia care proposals

Translating lecanemab and donanemab from clinical trials to real‐world practice shows that the main barriers are organizational. In Italy, the health system is organized at the regional level. A recent survey indicated that out of 450 specialized memory centers in Italy, only a small percentage (10.4%) meet the criteria for effective DMT prescription and administration, ³⁰ with 72.3% concentrated in northern Italy. Internationally, the United Kingdom estimates 30,200 annually eligible patients, with wait times projected at 56 to 129 months without intervention, while blood‐based biomarker triage could reduce this to 17 to 25 months. ³¹ Sweden's specialist consultation bottleneck requires 25% more specialists for 6‐month diagnostic timelines. ³² Germany demonstrates insurance‐based inequity with 76‐month wait times for public insurance versus 40 months for private insurance. ³³ Canada has only 2% of its system capacity for access to DMTs, with 382,000 Canadians awaiting eligibility assessment. ³⁴ Ireland implements a structured hub‐and‐spoke model through regional specialist memory clinics and regional memory assessment and support services. ²³ , ³⁵ A systematic review of 53 practice guidelines identified a lack of consensus regarding diagnostic tools and routine biomarker testing, underscoring the need for clinician guidance on appropriate therapy use and benefit‐to‐risk evaluation. ³⁶ Our proposed hub‐and‐spoke model aims to maintain equity while optimizing resource allocation and focuses on organizing access fairly and sustainably rather than restricting it.

5. CONCLUSION

Whatever the mechanism of action and the cost–benefit ratio of DMTs in treating AD, health‐care systems require substantial investment in a specialist workforce, imaging capacity, biomarker testing, and monitoring infrastructure. Success depends not on dramatic innovation but on sustained commitment to systematic organization, interdisciplinary collaboration, and unwavering dedication to equitable health‐care delivery. Implementation of a successful workflow would enable improvement of quality and sustainability of care in dementia, which has been too long dismissed due to a lack of successful treatments in fighting the disease course. Our data suggest that without organizational reform, the introduction of DMTs risks widening health‐care inequities rather than improving outcomes.

CONFLICTS OF INTEREST STATEMENT

Claudia Ada Di Paola and Rosangela Vallarelli had salary funding by Regional Counsil of Regione Puglia (DGR 1424/2025). All other authors declare no conflicts of interest. Author disclosures are available in the Supporting Information.

CONSENT STATEMENT

Data were anonymized before analysis. Patients were never contacted. Informed consent not required under Italian law.

DECLARATION OF GENERATIVE AI AND AI‐ASSISTED TECHNOLOGIES IN THE WRITING PROCESS

AI was not used in any phase of the preparation of this paper.

Supporting information

Supporting Information: dad270403‐sup‐0001‐ICMJE.pdf

DATA AVAILABILITY STATEMENT

The data supporting the findings of this study are available upon reasonable request.