Velandai Srikanth

Background and ObjectivesEnlarged perivascular spaces (EPVSs) in the basal ganglia (BG-EPVS) are an important marker of cerebral small vessel disease (cSVD), and EPVS in the centrum semiovale (CSO-...

Objectives: Leveraging routine electronic health records (EHR) for dementia detection is a growing field, but quality and clinical utility of existing models are unclear. This systematic review aimed to evaluate performance, methodological quality, and risk of bias of EHR-based dementia prediction models. Materials and Methods: We systematically searched Medline, EMBASE, Scopus, IEEE Xplore, and ACM from conception until July 2024. All studies and grey literature describing development or validation of probabilistic prediction models using EHR data for dementia detection were included. Risk of bias was assessed using PROBAST. Results: Fifty-six studies (434 prediction models, 155 external validations) were included. Most models were prognostic (66%), used US data (71%), relied solely on structured data, and 47 (11%) were externally validated. Modeled outcomes were extremely heterogeneous: gold-standard clinical criteria were used in 17 models (4%), with others reliant on diagnostic codes for case ascertainment. Discriminative metrics were frequently reported (82% of models), but calibration was rarely assessed (16%). All models were judged high risk of bias, driven by poor outcome definition, inadequate handling of missing data, and potential overfitting. Discussion: Our review highlights significant issues with methodological rigor and reporting transparency in existing EHR dementia prediction models. Ambiguous outcomes, flawed case ascertainment, and incomplete performance reporting, all limit clinical usefulness. Overall, model performance was difficult to assess and compare across studies due to incomplete reporting. Conclusion: Electronic health record-based dementia prediction is still in its infancy. Methodological rigor and interdisciplinary collaboration are essential to meet clinical needs and achieve real-world impact.

The most common cause of convexity subarachnoid hemorrhage (cSAH) in younger patients (younger than 60 years) is reversible cerebral vasoconstriction syndrome (RCVS). Evidence on the long-term outcomes of future vascular events and functional outcome after cSAH due to RCVS is limited. We aimed to assess the rates and baseline predictors of our primary outcomes (cSAH, intracerebral hemorrhage (ICH), and ischemic stroke), functional outcome, and mortality after cSAH attributed to RCVS. Individual patient data pooled analysis in patients with cSAH attributed to RCVS. A systematic literature search was conducted in PubMed and EMBASE. Two independent reviewers screened studies and extracted data. Quality assessment was assessed using the Newcastle-Ottawa Scale. Early events during the accepted time frame of an RCVS episode (<3 months) were classified as progression rather than recurrence. Follow-up was truncated at 5 years. Primary outcomes were recurrent cSAH, ICH, and ischemic stroke. Secondary outcomes were mortality and functional status measured by modified Rankin Score (mRS). We identified 21 eligible cohorts finally including 138 patients from 9 collaborative centers, which provided individual patient data. The mean age was 49.3 ± 12.1 years, and 110 (79.7%) were female. During a mean follow-up of 1.8 years, annual rates were cSAH recurrence 0.81% (95% CI 0.1-2.91), ICH 0.81% (95% CI 0.1-2.91), and ischemic stroke 0.81% (95% CI 0.1-2.91). Progression during the initial episode (shortly after admission) occurred in 10 patients for cSAH, 2 for ICH, and 8 for ischemic stroke, respectively. Of 106 patients (76.8%) with available outcome data, 100 (94.3%) achieve a mRS of ≤1 at follow-up, indicating no significant disability. Two patients died (annual rate 0.5%, 95% CI 0.09-2.9): one within 10 days and the other within 2 months of the cSAH. Our data suggest a favorable prognosis for most patients after RCVS-associated cSAH with low rates of recurrent events and a high proportion achieving functional independence. Major limitations include retrospective data collection and potential selection bias from centers providing individual patient data. Nevertheless, these findings provide prognostic information to inform clinical practice.

Background: Hospital electronic health record (EHR) systems contain extensive data of varying quality and suitability for research. Identifying research suitable data from hospital systems is rarely performed using a structured approach. Objective: To define a core set of research suitable data, from multiple hospital datasets and systems for the National Centre for Healthy Ageing Data Platform. Method: Candidate data were identified based on: published literature, researcher survey and quality audit. We recruited researchers and clinicians to participate in a modified Delphi method to achieve consensus on the core data. Live online surveys were conducted across 11 datasets. Participants rated the relevance of each data item using a Likert scale and provide additional feedback. Acceptance criteria were set at >70% agreement. Items meeting this threshold were included in the core dataset. Results: The end user group contained 16 university and 17 clinician researchers. An initial 144 data items were presented for consideration. Experts provided additional details for specialised items. Consensus to include an item was reached after one ( n = 127) or two ( n = 3) rounds with an additional three items identified by end-users. The final core dataset included 133 items across all 11 datasets. Conclusion: Our approach proved effective in reaching consensus among diverse researchers, ensuring that the selected data items are relevant for a range of purposes. Implications for health information management practice: Having a core dataset will support targeted activities that ensure these items are accurate, readily accessible and used appropriately to support patient care, administrative tasks and research.

Objective: To examine general practitioners’ (GPs) referral patterns to allied health services for people with dementia compared with those without dementia across two large Australian Primary Health Networks (PHNs). Design: A retrospective cohort study using routinely collected general practice data. Logistic regression was used to compare odds of allied health referrals, adjusting for age, sex and socioeconomic status. Setting: De-identified patient and episode activity data from 537 GP practices across two PHNs in Australia between 2018 and 2023. Participants: Data from 1 153 304 patients and 28 667 517 GP episodes of care were analysed. After merging records, 693 328 unique patients were identified, including 16 610 patients with dementia. Subcohorts included patients with dementia, stroke, Parkinson’s disease and combinations of these conditions. Results: The dementia cohort (n=16 610) had a similar overall allied health referral rate (36.1%) to the control cohort (n=48 977) (35.4%). Patients with dementia only were significantly less likely to receive any allied health referral compared with those with stroke (adjusted OR (aOR) 0.76, 95% CI 0.72 to 0.80; p<0.001) or Parkinson’s disease (aOR 0.72, 95% CI 0.66 to 0.78; p<0.001). Those with dementia and stroke were also less likely to receive referrals than those with stroke only (aOR 0.71, 95% CI 0.61 to 0.82; p<0.001). No significant difference was found between dementia with Parkinson’s and Parkinson’s only groups (p=0.48). Patients with dementia were consistently less likely to be referred to key allied health services (p<0.05). Conclusion: Despite strong evidence supporting allied health interventions for dementia, referral rates remain comparatively low. Enhancing GP referral resources and education, integrating dementia-specific care pathways and implementing supportive policy changes are needed to improve access and equity in dementia care.

INTRODUCTION: Timely diagnosis is crucial for managing neurodegenerative conditions. This study investigated whether time from symptom onset to diagnosis differs by clinical syndrome and sex.METHODS: This retrospective, cross‐sectional study included 591 participants with Alzheimer's disease (AD), frontotemporal dementia (FTD) subtypes (behavioral variant FTD [bvFTD], semantic dementia [SD], and progressive non‐fluent aphasia), logopenic progressive aphasia (LPA), and syndromes associated with movement disorders (corticobasal syndrome, FTD with motor neuron disease [FTD‐MND], and progressive supranuclear palsy). Bayesian regression models were used to compute diagnostic timelines.RESULTS: Compared to AD (3.35 years; 95% credible interval [CrI]: 3.03–3.72), SD and bvFTD had additional delays of 9.7 (95% CrI: 1.96–20.64) and 14.82 months (95% CrI: 6.94–25.42), respectively, while FTD‐MND was shorter by 11.62 months (95% CrI: −15.7 to −4.68). Men with bvFTD had 23.64 month longer delays than women (95% CrI: 10.35–44.33).DISCUSSION: Diagnostic delays may reflect syndrome‐specific clinical features, diagnostic complexity, and sociocultural factors. Findings highlight the need for improved diagnostic pathways and pre‐clinical biomarkers to facilitate earlier identification.Highlights: Bayesian analyses revealed that diagnostic delays differ by syndrome and sex. Alzheimer's disease (AD) was diagnosed on average 3.35 years after symptom onset. Diagnoses were delayed in semantic and behavioral variant frontotemporal dementia (bvFTD) compared to AD. Men with bvFTD had longer delays than women. Findings support need for improved diagnostic pathways and pre‐clinical biomarkers.

Purpose: Although chronic disease management (CDM) has been reported to improve medication adherence after stroke or transient ischaemic attack (TIA), the impact on specific patterns of medication adherence is unclear. We aimed to evaluate the population effect of receiving a CDM claim on trajectories of medication adherence in long‐term survivors of stroke/TIA.Methods: A cohort study was undertaken using observational data from PRECISE (42 Australian Stroke Clinical Registry hospitals [Victoria and Queensland; 2012–2015] linked with medication dispensing and primary care claims). Community‐dwelling adults with ≥ 1 primary care visit were included. The exposure was a CDM claim (versus no claim) in primary care within 7–18 months post‐stroke/TIA. Medication adherence (antihypertensive, antithrombotic, lipid‐lowering) was assessed between 19 and 30 months post‐stroke/TIA, using group‐based trajectory models. Average treatment effects were estimated using multi‐level logistic regression with inverse probability treatment weights.Results: Among 11 580 survivors of stroke/TIA (median age 70 years, 42% female; 45% with CDM claim), four distinct adherence patterns were identified: near‐perfect adherence, high adherence, declining adherence, and non‐use. After adjustment, having a CDM claim (vs no claim) promoted near‐perfect adherence (odds ratio [OR]: 1.16 [95% CI 1.08–1.25]) for antithrombotic medications. Whereas, having a CDM claim (vs no CDM claim) promoted high adherence for antihypertensive (OR: 1.33 [95% CI 1.24–1.44]) or lipid‐lowering (OR: 1.26 [95% CI 1.16–1.37]) medications. The odds of non‐use were also reduced by 17%–23% in those with (vs without) a CDM claim.Conclusions: CDM claims were associated with favourable trajectories of medication adherence in long‐term survivors of stroke/TIA.