Cognitive impairment and its neurobiological correlates in adults experiencing long-term homelessness: an exploratory observational cross-sectional study protocol.

Hypotheses and research questions

The core hypotheses are as follows:

The research questions which will be addressed through this study are as follows:

Overall design

This study comprises an exploratory cross-sectional observational study of cognitive function and its neuroimaging and biomarker associations in long-term homeless adults (LTHAs). A population-representative cohort of adults in long-term homelessness will be recruited from the community in Dublin, Ireland, to undergo a series of structured biopsychosocial and cognitive assessments and will be invited to provide blood samples and undergo contrast-enhanced brain Magnetic Resonance Imaging (MRI). A control sample of housed adults will be recruited from staff working in homeless services. Controls will be matched on a 1:2 basis to an LTHA that has completed the study protocol based on age, sex and educational attainment. Controls will complete an aligned assessment protocol, including structured biopsychosocial and cognitive assessments, blood sampling and brain MRI and will serve as a housed comparison group.

Public and patient involvement

A study Public and Patient Involvement (PPI) group of five people with lived experience of homelessness was formed in collaboration with Depaul Ireland. The study PPI group have diverse socioeconomic backgrounds, educational attainment and experience of homelessness in Ireland. A programme of regular, subsidised meetings was devised, spaced over the course of the study to ensure ongoing PPI input into the study protocol, the first of which was held early in the initial study design phase. At this meeting, the PPI group provided expertise formed through their lived experience on the study design, informing the recruitment methodology, location of assessments and assessment measures selected. The PPI group also reviewed the study documentation to ensure that it is accessible and acceptable to the target LTHA study population.

Participants and recruitment

Potential participants in homelessness will be identified through gatekeeper organisations, comprising the principal homeless service providers in Ireland: Depaul Ireland, Focus Ireland, the Peter McVerry Trust, Novas and Crosscare. Information on the study, including the eligibility criteria ( Table 1) and assessment protocol will be disseminated to gatekeeper organisation staff and directly to potential participants through presentations at gatekeeper organisation staff meetings and presentations in the service common areas. Staff and potential participants will then be given the opportunity to ask questions about participation in a no-obligation setting, in which the power differential between the study team and potential participants is minimised. Telephone and email contact details will be disseminated to gatekeeper organisation staff and potential participants, facilitating referral of potential participants for assessment.

Following referral, potential participants will be met on site at the referring gatekeeper location in a private location. They will be offered the presence of their keyworker for emotional support. Potential participants will then be screened using the eligibility criteria ( Table 1).

Control participants will also be recruited, comprising housed adults matched based on age, sex and educational attainment to an LTHA that has completed the study assessment protocol. Control participants will be recruited from staff working in the gatekeeper organisations, and thereafter via snowball sampling. Potential control participants will be considered eligible if they are aged over the age of 18 years, with fluent English language ability. Potential control participants will be excluded if they have previous experience of social exclusion (including homelessness and incarceration), a previous concussion or experience of repetitive head impacts or any current or previously identified central neurological disorder (such as stroke or multiple sclerosis). This will be done to ensure that the control group represent a neurologically healthy housed comparison group.

Consent process

Eligible potential participants will be provided with the patient information leaflet (PIL) and consent forms, with the opportunity to have these read out loud offered. Participants’ ability to give informed consent will be assessed using an established teach-back method.³¹ In brief, participants will be provided with the study information (either by reading the study PIL or by having it read out to them). They will then be asked to repeat what they recalled and understood of the provided information, and their understanding will then be clarified and additional tailored information provided as necessary. The cycle will be repeated as needed to ensure that the study information was adequately understood, retained and weighed by the participant to facilitate communication of their informed consent. The rights of the participant to withdraw from the study at any time without giving a reason and without this affecting their future medical care, and to decline to participate in any component of the study of their choosing will be emphasised. The opportunity to ask questions will then be provided, and eligible potential participants will thus be asked to provide informed written consent to participate. A summary of the recruitment and assessment process is provided in Figure 1.

Figure 1. A summary of the recruitment and assessment process for LTHA participants and matched housed controls.⁴⁶

Assessment protocol

LTHA participants will undergo their clinical assessments, including phlebotomy, on site in the homeless service from which they were recruited. They will be offered the presence of their keyworker for emotional support during their assessment. After completion of their clinical assessment, participants will then be asked to attend St James’s Hospital, Dublin, to undergo repeat blood sampling and contrast-enhanced brain MRI. Imaging will be arranged within two weeks of the participant’s clinical assessment when possible. Taxis will be provided for the participant to travel to-and-from their imaging appointment, and they will be offered the presence of their keyworker for emotional support during this appointment.

Biopsychosocial assessment

LTHA participant sociodemographic, medical and life-history self-report data will be collected through clinical interview directly into a study-specific REDCap form. Data on participants’ medical and psychiatric history, smoking history, drug and alcohol use, education and employment history, housing history and forensic history will be collected. Their resting heart rate and blood pressure will be measured using an electronic blood pressure monitor. Their grip strength will be measured and recorded twice in each hand using a hand-held hydraulic dynamometer (JAMAR^®), and their handedness will be documented. Participants will then be asked to complete a battery of validated biopsychosocial and cognitive assessments, as detailed in Table 2. This battery has been selected to test examine participants’ cognitive performance across neurocognitive domains, and is enriched to evaluate their frontal-executive performance in the most depth, as this is where the impairment has previously been identified most often in the published literature. This battery has been refined based on feedback from the study PPI group and expertise provided by a cognitive neurologist and a senior clinical neuropsychologist.

Biobanking

Venous blood will be drawn at the time of the participant’s clinical assessment, and again when they attend for MRI, comprising a short-interval follow-up sample. Three 9 ml Lithium Heparin-coated, one 8 ml Serum Separator Clot Activator, and one 9 ml K3EDTA Vacuette© tubes will be collected and transported immediately for processing and storage of serum, plasma, peripheral blood mononuclear cells (PBMCs) and whole blood. Whole blood stimulation assays will also be performed to examine participants’ innate immune responses to challenge. These feature stimulation of blood with four innate immune agonists (lipopolysaccharide, Resiquimod, polyinosinic:polycytidylic acid and interferon-alpha) with supernatants collected via centrifugation after a 22-hour incubation. All samples will be stored at -80 °C in a secure location for a period not exceeding five years from study completion for future related analyses.

Blood based biomarkers

Plasma samples from participants that provide explicit informed consent for sample transfer will have circulating fluid biomarker levels quantified using NUcleic acid Linked Immuno-Sandwich Assay (NULISA).³² More than 120 biomarkers associated with neurological disease and BBB function, including brain-derived tau (BD-tau) isoforms, amyloid β-40 and -42 (Aβ40 and Aβ42), neurofilament light (NfL), Glial Fibrillary Acidic Protein (GFAP), S100β and ubiquitin carboxy-terminal hydrolase L1 (UCHL1) will be quantified. Plasma samples from the Tallaght University Hospital Institute of Memory and Cognition – Biobank for Research in Ageing and Neurodegeneration (TIMC-BRAiN) biobank, comprising samples from older adults with mild cognitive impairment without Alzheimer’s Disease (AD) pathology, and older adults with AD will be sent for simultaneous analysis. The protocol for banking of these samples has been previously published, and laboratory methodology was consistent between the TIMC-BRAiN biobank and the current study.³³ These samples will serve as additional controls, contextualising the results from the LTHA and matched control participants relative to a population of housed older adults with MCI without AD, and to a population of older adults with a consensus diagnosis of AD.

MRI assessment

All study participants will be invited to attend St James’s Hospital for brain MRI. A 3 Tesla Philips Achieva scanner (Philips, Best, The Netherlands) will be used. The following sequences will be obtained: T1 magnetisation-prepared rapid acquisition with gradient echo (MPRAGE), T2 fluid attenuated inversion recovery (FLAIR), T2*, diffusion tensor imaging (DTI) in 61 directions and dynamic contrast enhanced (DCE) ( Table 3). All neuroimaging will be reported clinically by a consultant neuroradiologist, and clinically significant findings will be followed up as per their recommendation.

Clinical follow-up

Participants will be contacted within two weeks of completion of their assessment to go through their testing results via telephone. Abnormal results will be identified, and follow-up will be arranged through the St James’s Hospital Inclusion Health Service as needed. Participants’ General Practitioners will be informed of their participation in the study and will be copied on any correspondence that results from their participation. Participants will have been informed that this is a component of the study in the Participant Information Leaflet and will have provided explicit consent for this to take place.

Primary

Secondary

Sample size and power justification

Target recruitment is for 80 adults in long-term homelessness. As an exploratory observational study, the effect sizes for many of the main outcome measures listed remain unknown. However, it is estimated that recruitment of 80 LTHA would provide sufficient statistical power to identify the prevalence of cognitive impairment and identify its neuropsychiatric characteristics amongst people experiencing homelessness in Ireland for the first time. International estimates for the prevalence of CI amongst PEH experiencing homelessness range from 1.7% to 92.9%, depending on the population studied and test used.^14,15 Assuming a conservative normative population rate of impairment of 0.5%, a sample size of 80 would provide adequate power to detect a minimum prevalence of CI amongst PEH of 3.85%, with an alpha of 0.05.³⁴

Statistical analysis

Statistical analysis will be performed using R version 4.5.2 (2025-10-31 ucrt). Distributions of variables will be visually inspected, and non-normally distributed variables (for example biomarker results) will be log-transformed prior to further analysis.

Prior to between-group analyses, continuous variables will be tested for normality using the Shapiro-Wilk test. Normally distributed continuous variables will be compared using t-tests, and non-normally distributed continuous variables compared using the Wilcoxon rank-sum test. Ordinal variables will be compared using Spearman correlation. Categorical variables will be compared using the chi-square test for independence or Fisher’s exact test (if the expected counts in more than 20% of cells was less than five). The threshold for statistical significance is set at α = 0.05.

Cognitive impairment

The prevalence of cognitive impairment amongst LTHA will be estimated by comparing their MoCA test results to established cut-off scores indicative of mild cognitive impairment and dementia.^35,36 It will be compared to participants’ Subjective Memory Impairment scores to evaluate the accuracy of their self-report, and to provide an estimation of their subjective memory change in the preceding one and ten years.

To determine participants’ domain-specific cognitive performance, their domain-specific cognitive test results will be converted to z-scores using previously published population normative data. The proportion of participants demonstrating impairment in each cognitive domain will be calculated, with impairment defined as performance more than 1.5 standard deviations below the normative mean (z-score < −1.5). These analyses will allow description of the pattern and relative burden of impairment across cognitive domains within the LTHA cohort.

To further contextualise LTHA participants’ cognitive performance, the LTHA and control participants’ domain-specific cognitive test results will also be compared. For these analyses, LTHA participants’ domain-specific cognitive results will be z-scored based on the mean and standard deviation of the control participants’ results. Using the same impairment threshold (z-score < −1.5), this approach will permit comparison of domain-specific cognitive performance relative to a contemporaneously assessed matched housed control sample.

Head injury exposure

The Ohio State University TBI-ID method will be used to estimate the lifetime concussion rate, exposure to repetitive head impacts and traumatic brain injury amongst adults experiencing homelessness. Participants will also be asked to report on frequency, severity and mechanism of all head injuries that they have experienced. The age at the participants’ first injury and their most recent injury will be derived from this data. This data can be further used to explore the relationship between participants’ experience of head injury and other outcome measures, adjusting for age and sex, using multivariable linear regression.

Neuroimaging

Participants’ global, tissue- and area-specific brain volumes will be quantified using an automated convolutional neural network pipeline (AssemblyNet; VolBrain).³⁷ After automated segmentation, region-, area- and tissue-specific volumes will be calculated, and normalised to the total intracranial volume (TIV) to account for differences in head size between participants. Participants’ global-, tissue- and area-specific brain volumes will then compared between LTHA and controls using linear regression models, with the brain volume metric as the outcome and group (LTHA vs control) as the primary predictor. All models will be adjusted for age, sex and TIV.

To explore whether any observed differences in brain volume between LTHA and control participants are attributable to other variables, a structured effect attenuation analysis using multiple linear regression will be performed. A base model, adjusted for age, sex and TIV will be fitted, with its group estimate representative of association between homelessness and brain volume. Variables, selected based on a priori hypotheses, include Charlson Comorbidity Index, mental health problems, educational attainment, smoking, alcohol use and drug use will be added individually into the base model. The adjusted group effect size, the corresponding percentage attenuation of the group effect size relative to the base model, and associated p-values will be reported. Finally, all covariates will be included simultaneously to assess their combined impact on the group association.

Amongst LTHA participants, the associations between homelessness exposure and traumatic brain injury (TBI) history and total and regional brain volumes will be examined using multiple linear regression models, adjusted for age, sex, and TIV. In addition, associations between global cognitive performance, indicated by participants’ total MoCA score, and total and regional brain volumes will be assessed using similarly adjusted regression models.

To account for multiple comparisons across neuroimaging outcomes, p-values will be corrected using the Benjamini–Hochberg false discovery rate (FDR) procedure.

Blood-brain barrier

Participants’ BBB permeability will be assessed using DCE-MRI. BBB permeability maps will be created for each participant based on the presence of contrast agent in each parenchymal brain voxel over time, as previously described.^38,39 The threshold for increased permeability in each voxel will be set at the 95^th percentile based on a previously imaged healthy control population. The percentage of voxels in which the BBB is impaired relative to the previously imaged healthy control population can thereby be calculated for each participant, reflecting their percentage BBB disruption (BBBD). Percentage BBBD will then be compared between the LTHA and matched housed control groups using linear regression models, with percentage BBBD as the outcome and group (LTHA vs control) as the primary predictor. All models will be adjusted for age and sex. As described above, structured attenuation analysis will be performed to explore whether any observed differences in brain volume between LTHA and control participants are attributable to other variables. Amongst LTHA participants, the associations between homelessness exposure, traumatic brain injury (TBI) history, MoCA score and BBD will also be examined using similarly adjusted linear regression models. To account for multiple comparisons, p-values will be BH-adjusted.

Fluid biomarkers

Blood biomarker levels will be compared between groups. Short interval follow-up samples will be compared using paired analyses to identify biomarker stability over time within the LTHA group. Amongst LTHA participants, the associations between homelessness exposure, traumatic brain injury (TBI) history, cognitive scores, neuroimaging measures and biomarker levels will also be examined using similarly linear regression models adjusted for age and sex. To account for multiple comparisons, p-values will be BH-adjusted.

Data management

All data collection instruments have been imported into Research Electronic Data Capture (REDCap) and will completed directly during the participant assessment. Participant data will be collected through in-person interview into the corresponding REDCap forms. Study data will thereafter be stored in a REDCap electronic database hosted by Trinity College Dublin.^40,41 There is a full audit trail of activities on REDCap and access will be monitored on an ongoing basis. Only pseudonymised data will be input to REDCap thereby reducing the impact of unauthorised, unlawful processing on individuals. All data will be processed in accordance with Article 6 (1)(e) and Article 9 (2)(j) of the General Data Protection Regulations. Data will be stored for a period of five years following the completion of the study to facilitate further analysis as needed and will thereafter be destroyed. A Data Protection Impact Assessment (DPIA) has been completed for this project, and the overall level of risk identified was low.

Limitations

To maximise the population-representativeness of the LTHA population recruited, recruitment will be performed in collaboration with all major homeless service providers in Dublin. Individuals that are rough sleeping will be accessed through recruitment from day centres and drop-in services providing targeted support to this population. However, participation is on an opt-in basis, and there may therefore be self-selection bias introduced in the recruited population, with recruited participants potentially differing from those that decline to take part. Additionally, the neurocognitive assessment components necessitate English language proficiency for participant assessment to ensure valid test administration and interpretation. As such non-fluent English speakers will be excluded, which may limit the generalisability of the findings.

There is a recognised risk that the LTHA population recruited may include some participants with an intellectual disability (ID), a group that is over-represented amongst PEH relative to the general population.⁴² This could result in over-estimation of the prevalence of acquired cognitive impairment in the LTHA cohort. To mitigate against this, efforts will be made to exclude participants with ID based on self-reported diagnosis and reported additional needs identified during the participants’ schooling. Inclusion of a formal measure to identify premorbid functioning, for example the National Adult Reading Test⁴³ was considered. However, following consultation with the study PPI group, this was removed from the study protocol as it was deemed likely to be a significant deterrent to participation amongst the target population living in long-term homelessness.

This study protocol is ambitious and requires sustained participant engagement over two study visits (if they opt in to brain MRI). No study involving a similar in-depth biopsychosocial assessment, neurocognitive assessment, fluid biomarker measurement and brain MRI has been attempted in a community-dwelling population of people experiencing homelessness before.

The decision to split the study assessment over two visits was consciously made, following consultation with the study PPI group. This allows the majority of the participant assessment to take place in the community in an environment in which the participant is familiar and comfortable. This serves to reduce the power imbalance between the study team and participant and may improve the likelihood that they attend their assessment and increase the chance that they complete the assessment. It is acknowledged, however, that as participants’ imaging is therefore scheduled at on a later date, this may reduce the chance that the participant attends for their imaging. Additionally, in contrast with many other studies in this population, no financial remuneration will be provided for participation in this study.

Multiple strategies have however been incorporated in the study protocol to improve participant engagement, including recruitment through gatekeeper organisations to foster trust between the participants and study team, involvement of the participant’s keyworker for emotional support, scheduling of participants’ imaging within two weeks of their initial assessment when possible and the provision of funded taxis to and from their imaging appointment. However, there remains a risk of participant drop-out between appointments.