Dementia and electronic health record phenotypes: a scoping review of available phenotypes and opportunities for future research

Trends in antipsychotic prescribing among community-dwelling older adults with dementia, 2010-2018

Due to an FDA "black box" warning for heightened risk of death, Choosing Wisely (CW) recommends avoiding antipsychotic prescription drugs as first-line treatment for dementia-related agitation. Yet, post-CW trends among community-dwelling patients with dementia remain unknown. In this retrospective cohort study, we used nationally representative Health and Retirement Study survey data linked to Medicare fee-for-service claims (January 1, 2010-December 31, 2018) to analyze prescribing trends during the pre-publication (2010-2012), publication (2013-2015), and post-publication (2016-2018) periods of CW recommendations. We included community-dwelling adults aged ≥65 years with dementia. We utilized multivariable mixed regression models to determine the percentage of patients prescribed any, potentially low-value, and potentially indicated antipsychotics. Among an estimated 2.4-2.7 million patients with dementia, any antipsychotic prescribing increased from 9.4% (95% CI, 6.4%-12.3%) during the pre-publication period (2010-2012) to 15.8% (95% CI, 12.8%-18.8%) (P < 0.001) during the publication period (2013-2015). Potentially low-value and potentially indicated prescriptions also increased. Post-publication period (2016-2018) prescribing of 16.0% (95% CI, 13.0%-19.1%) (P < 0.001) remained higher than pre-publication. Among older Americans with dementia, antipsychotic prescriptions increased after the publication of CW recommendations and held steady in the subsequent post-publication period. Stronger interventions, such as electronic clinical decision support tools and financial incentives, are needed to curb low-value antipsychotic prescribing for this vulnerable population.

Paving the way for Alzheimer's disease blood-based biomarkers in primary care

Blood-based biomarkers (BBBMs) for Alzheimer's disease (AD) have the potential to revolutionize the detection and management of cognitive impairment. AD BBBMs are not currently recommended for use in primary care but may soon be as research demonstrates their clinical utility for differential diagnosis and patient management. To prepare for the incorporation of AD BBBMs into primary care, several practical challenges must be addressed. Here, we describe four immediate challenges: (1) preparing primary care providers to order and disclose AD BBBMs, (2) expanding the dementia-capable workforce, (3) ensuring equitable uptake of AD BBBM testing, and (4) securing access to AD treatment. We conclude by discussing future directions and challenges for use of AD BBBMs in primary care, including screening for preclinical AD and dementia detection algorithms. HIGHLIGHTS: Alzheimer's disease (AD) blood-based biomarkers (BBBMs) may be well suited for primary care. Many changes are needed to prepare the workforce and ensure patient access. Paving the way for AD BBBMs in primary care will require a multi-pronged approach.

How Do Researchers Identify and Recruit Dementia Caregivers? A Scoping Review

BACKGROUND AND OBJECTIVES

Studies involving dementia caregivers are essential to transform care and inform new policies. However, identifying and recruiting this population for research is an ongoing challenge. This scoping review aimed to capture the current methodology for identifying and recruiting dementia caregivers in clinical studies. A focus was placed on methods for underrepresented populations and pragmatic trials to guide pragmatic and equitable clinical studies.

RESEARCH DESIGN AND METHODS

Researchers conducted a literature search using PubMed, PsycINFO, EMBASE, and Web of Science databases. Studies conducted in the US that enrolled at least 10 caregivers and were published within the last 10 years (2013-2023) were included.

RESULTS

Overall, 148 articles were included in the review. The most common method for identification was community outreach, and paper advertisements for recruitment. Caregivers were most often approached in community settings, formal organizations, and/or dementia research centers. Most enrolled caregivers were female, White, and spouses of persons living with dementia. Race and ethnicity were underreported, as were the target recruitment goals. Limited studies were self-reported as pragmatic. Additionally, limited studies reported adaptations for methods of identification and recruitment in underrepresented populations.

DISCUSSION AND IMPLICATIONS

We identified gaps in current practices for the identification and recruitment of dementia caregivers. Future identification and recruitment methodologies should be tailored to the intervention's intent, health care setting, and the research questions that need to be answered, while balancing available resources. Additionally, transparent reporting of identification and recruitment procedures, target recruitment goals, and comprehensive demographic data is warranted.

Multimodal Attention Network for Dementia Prediction

The early identification of an individual's dementia risk is crucial for disease prevention and the design of insurance products in an aging society. This study aims to accurately predict the future incidence risk of dementia in individuals by leveraging the advantages of neural networks. This is, however, complicated by the high dimensionality and sparsity of the International Classification of Diseases (ICD) codes when utilizing data from Taiwan's National Health Insurance, which includes individual profiles and medical records. Inspired by the click-through rate (CTR) problem in recommendation systems, where future user behavior is predicted based on their past consumption records, we address these challenges with a multimodal attention network for dementia (MAND), which incorporates an ICD code embedding layer and multihead self-attention to encode ICD codes and capture interactions among diseases. Additionally, we investigate the applicability of several CTR methods to the dementia prediction problem. MAND achieves an AUC of 0.9010, surpassing traditional CTR models and demonstrating its effectiveness. The highly flexible pipelined design allows for module replacement to meet specific requirements. Furthermore, the analysis of attention scores reveals diseases highly correlated with dementia, aligning with prior research and emphasizing the interpretability of the model. This research deepens our understanding of the diseases associated with dementia, and the accurate prediction provided can serve as an early warning for dementia occurrence, aiding in its prevention.

Validation of Different Dementia Code-Based Definitions in a Population-Based Study of Rheumatoid Arthritis

OBJECTIVE

Rheumatoid arthritis (RA) has been associated with an elevated dementia risk. This study aimed to examine how different diagnostic dementia definitions perform in patients with RA compared to individuals without RA.

METHODS

The study population included 2050 individuals (1025 with RA) from a retrospective, population-based cohort in southern Minnesota and compared the performance of 3 code-based dementia diagnostic algorithms with medical record review diagnosis of dementia. For the overall comparison, each patient's complete medical history was used, with no time frames. Sensitivity analyses were performed using 1-, 2-, and 5-year windows around the date that dementia was identified in the medical record (reference standard).

RESULTS

Algorithms performed very similarly in persons with and without RA. The algorithms generally had high specificity, negative predictive values, and accuracy, regardless of the time window studied (> 88%). Sensitivity and positive predictive values varied depending on the algorithm and the time window. Sensitivity values ranged 56.5-95.9%, and positive predictive values ranged 55.2-83.1%. Performance measures declined with more restrictive time windows.

CONCLUSION

Routinely collected electronic health record (EHR) data were used to define code-based dementia diagnostic algorithms with good performance (vs diagnosis by medical record review). These results can inform future studies that use retrospective databases, especially in the same or a similar EHR infrastructure, to identify dementia in individuals with RA.

Leveraging multi-site electronic health data for characterization of subtypes: a pilot study of dementia in the N3C Clinical Tenant

Objectives

To provide a foundational methodology for differentiating comorbidity patterns in subphenotypes through investigation of a multi-site dementia patient dataset.

Materials and Methods

Employing the National Clinical Cohort Collaborative Tenant Pilot (N3C Clinical) dataset, our approach integrates machine learning algorithms-logistic regression and eXtreme Gradient Boosting (XGBoost)-with a diagnostic hierarchical model for nuanced classification of dementia subtypes based on comorbidities and gender. The methodology is enhanced by multi-site EHR data, implementing a hybrid sampling strategy combining 65% Synthetic Minority Over-sampling Technique (SMOTE), 35% Random Under-Sampling (RUS), and Tomek Links for class imbalance. The hierarchical model further refines the analysis, allowing for layered understanding of disease patterns.

Results

The study identified significant comorbidity patterns associated with diagnosis of Alzheimer's, Vascular, and Lewy Body dementia subtypes. The classification models achieved accuracies up to 69% for Alzheimer's/Vascular dementia and highlighted challenges in distinguishing Dementia with Lewy Bodies. The hierarchical model elucidates the complexity of diagnosing Dementia with Lewy Bodies and reveals the potential impact of regional clinical practices on dementia classification.

Conclusion

Our methodology underscores the importance of leveraging multi-site datasets and tailored sampling techniques for dementia research. This framework holds promise for extending to other disease subtypes, offering a pathway to more nuanced and generalizable insights into dementia and its complex interplay with comorbid conditions.

Discussion

This study underscores the critical role of multi-site data analyzes in understanding the relationship between comorbidities and disease subtypes. By utilizing diverse healthcare data, we emphasize the need to consider site-specific differences in clinical practices and patient demographics. Despite challenges like class imbalance and variability in EHR data, our findings highlight the essential contribution of multi-site data to developing accurate and generalizable models for disease classification.

Large language models and generative AI in telehealth: a responsible use lens

OBJECTIVE

This scoping review aims to assess the current research landscape of the application and use of large language models (LLMs) and generative Artificial Intelligence (AI), through tools such as ChatGPT in telehealth. Additionally, the review seeks to identify key areas for future research, with a particular focus on AI ethics considerations for responsible use and ensuring trustworthy AI.

MATERIALS AND METHODS

Following the scoping review methodological framework, a search strategy was conducted across 6 databases. To structure our review, we employed AI ethics guidelines and principles, constructing a concept matrix for investigating the responsible use of AI in telehealth. Using the concept matrix in our review enabled the identification of gaps in the literature and informed future research directions.

RESULTS

Twenty studies were included in the review. Among the included studies, 5 were empirical, and 15 were reviews and perspectives focusing on different telehealth applications and healthcare contexts. Benefit and reliability concepts were frequently discussed in these studies. Privacy, security, and accountability were peripheral themes, with transparency, explainability, human agency, and contestability lacking conceptual or empirical exploration.

CONCLUSION

The findings emphasized the potential of LLMs, especially ChatGPT, in telehealth. They provide insights into understanding the use of LLMs, enhancing telehealth services, and taking ethical considerations into account. By proposing three future research directions with a focus on responsible use, this review further contributes to the advancement of this emerging phenomenon of healthcare AI.

Validity and reliability of the Turkish version of the Australian National University-Alzheimer's Disease Risk Index (ANU-ADRI)

INTRODUCTION

There is still a requirement for concise, practical scales that can be readily incorporated into everyday schedules and predict the likelihood of dementia onset in individuals without dementia. This study aimed to assess the reliability of the ANU-ADRI (Australian National University Alzheimer's Disease Risk Index)-Short Form in Turkish geriatric patients.

METHODS

This methodological study involved 339 elderly patients attending the geriatric outpatient clinic for various reasons. The known-group validity and divergent validity were assessed. The ANU-ADRI was administered during the baseline test and again within one week for retest purposes. Alongside the ANU-ADRI, all participants underwent a comprehensive geriatric assessment, including Activities of Daily Living (ADL), mobility assessment (Performance-Oriented Mobility Assessment (POMA) and Timed Up and Go Test), nutritional assessment (Mini Nutritional Assessment (MNA)), and global cognition evaluation (Mini-Mental State Examination (MMSE)).

RESULTS

The scale demonstrated satisfactory linguistic validity. A correlation was observed between the mean scores of the ANU-ADRI test and retest (r = 0.997, p < 0.001). Additionally, there existed a moderate negative linear association between the ANU-ADRI and MMSE scores (r = -0.310, p < 0.001), POMA (r = -0.406, p < 0.001), Basic ADL (r = -0.359, p < 0.001), and Instrumental ADL (r = -0.294, p < 0.001). Moreover, a moderate positive linear association was found between the ANU-ADRI and the Timed Up and Go Test duration (r = 0.538, p < 0.001).

CONCLUSION

The ANU-ADRI-Short Form was proved as a valuable tool for clinical practice, facilitating the assessment of Alzheimer's disease risk within the Turkish geriatric population.

Leveraging electronic health records and knowledge networks for Alzheimer's disease prediction and sex-specific biological insights

Identification of Alzheimer's disease (AD) onset risk can facilitate interventions before irreversible disease progression. We demonstrate that electronic health records from the University of California, San Francisco, followed by knowledge networks (for example, SPOKE) allow for (1) prediction of AD onset and (2) prioritization of biological hypotheses, and (3) contextualization of sex dimorphism. We trained random forest models and predicted AD onset on a cohort of 749 individuals with AD and 250,545 controls with a mean area under the receiver operating characteristic of 0.72 (7 years prior) to 0.81 (1 day prior). We further harnessed matched cohort models to identify conditions with predictive power before AD onset. Knowledge networks highlight shared genes between multiple top predictors and AD (for example, APOE, ACTB, IL6 and INS). Genetic colocalization analysis supports AD association with hyperlipidemia at the APOE locus, as well as a stronger female AD association with osteoporosis at a locus near MS4A6A. We therefore show how clinical data can be utilized for early AD prediction and identification of personalized biological hypotheses.

Prevalence and trend of central nervous system-active medication polypharmacy among US commercially insured adults with vs without early-onset dementia: a multi-year cross-sectional study

BACKGROUND

Limited data exist on the prevalence and trend of central nervous system (CNS)-active medication polypharmacy among adults with early-onset dementia (EOD) and whether these estimates differ for adults without EOD but with chronic pain, depression, or epilepsy, conditions managed by CNS-active medications.

METHODS

A multi-year, cross-sectional study using 2012-2021 MarketScan Commercial Claims data was conducted among adults aged 30 to 64 years with EOD and those without EOD but having a diagnosis of chronic pain, depression, or epilepsy as comparison groups. For each disease cohort, the primary outcome was CNS-active medication polypharmacy defined as concurrent use of ≥ 3 CNS-active medications on the US Beers Criteria list that overlapped for > 30 consecutive days during 12 months following a randomly selected medical encounter with the disease diagnosis. A separate multivariate modified Poisson regression model was used to estimate time trends in CNS polypharmacy in each disease cohort. Differences in trend estimates between EOD and non-EOD disease cohorts were examined by an interaction between EOD status and yearly time.

RESULTS

From 2013 to 2020, the annual crude prevalence of CNS polypharmacy was higher among adults with EOD (21.2%-25.0%) than adults with chronic pain (5.1%-5.9%), depression (14.8%-21.7%), or epilepsy (20.0%-22.3%). The adjusted annual prevalence of CNS polypharmacy among patients with EOD did not significantly change between 2013 and 2020 (adjusted prevalence rate ratio [aPRR], 0.94; 95% CI, 0.88-1.01), whereas a significant decreasing trend was observed among non-EOD cohorts with chronic pain (aPRR, 0.66; 95% CI, 0.63-0.69), depression (aPRR, 0.81; 95% CI, 0.77-0.85), and epilepsy (aPRR, 0.86; 95% CI, 0.83-0.89). The interaction analysis indicated that patients with epilepsy and depression (vs with EOD) had a decreasing probability of CNS-active medication polypharmacy over time (aPRR, 0.98 [95% CI, 0.98-0.99]; P < .001 for interaction for both conditions).

CONCLUSIONS

The prevalence of CNS polypharmacy among US commercially insured adults with EOD (vs without) was higher and remained unchanged from 2013 to 2021. Medication reviews of adults with EOD and CNS polypharmacy are needed to ensure that benefits outweigh risks associated with combined use of these treatments.

Explainable machine learning for predicting conversion to neurological disease: Results from 52,939 medical records

Objective

This study assesses the application of interpretable machine learning modeling using electronic medical record data for the prediction of conversion to neurological disease.

Methods

A retrospective dataset of Cleveland Clinic patients diagnosed with Alzheimer's disease, amyotrophic lateral sclerosis, multiple sclerosis, or Parkinson's disease, and matched controls based on age, sex, race, and ethnicity was compiled. Individualized risk prediction models were created using eXtreme Gradient Boosting for each neurological disease at four timepoints in patient history. The prediction models were assessed for transparency and fairness.

Results

At timepoints 0-months, 12-months, 24-months, and 60-months prior to diagnosis, Alzheimer's disease models achieved the area under the receiver operating characteristic curve on a holdout test dataset of 0.794, 0.742, 0.709, and 0.645; amyotrophic lateral sclerosis of 0.883, 0.710, 0.658, and 0.620; multiple sclerosis of 0.922, 0.877, 0.849, and 0.781; and Parkinson's disease of 0.809, 0.738, 0.700, and 0.651, respectively.

Conclusions

The results demonstrate that electronic medical records contain latent information that can be used for risk stratification for neurological disorders. In particular, patient-reported outcomes, sleep assessments, falls data, additional disease diagnoses, and longitudinal changes in patient health, such as weight change, are important predictors.