Machine learning functional impairment classification with electronic health record data.

BACKGROUND: Poor functional status is a key marker of morbidity, yet is not routinely captured in clinical encounters. We developed and evaluated the accuracy of a machine learning algorithm that leveraged electronic health record (EHR) data to provide a scalable process for identification of functional impairment. METHODS: We identified a cohort of patients with an electronically captured screening measure of functional status (Older Americans Resources and Services ADL/IADL) between 2018 and 2020 (N = 6484). Patients were classified using unsupervised learning K means and t-distributed Stochastic Neighbor Embedding into normal function (NF), mild to moderate functional impairment (MFI), and severe functional impairment (SFI) states. Using 11 EHR clinical variable domains (832 variable input features), we trained an Extreme Gradient Boosting supervised machine learning algorithm to distinguish functional status states, and measured prediction accuracies. Data were randomly split into training (80%) and test (20%) sets. The SHapley Additive Explanations (SHAP) feature importance analysis was used to list the EHR features in rank order of their contribution to the outcome. RESULTS: Median age was 75.3 years, 62% female, 60% White. Patients were classified as 53% NF (n = 3453), 30% MFI (n = 1947), and 17% SFI (n = 1084). Summary of model performance for identifying functional status state (NF, MFI, SFI) was AUROC (area under the receiving operating characteristic curve) 0.92, 0.89, and 0.87, respectively. Age, falls, hospitalization, home health use, labs (e.g., albumin), comorbidities (e.g., dementia, heart failure, chronic kidney disease, chronic pain), and social determinants of health (e.g., alcohol use) were highly ranked features in predicting functional status states. CONCLUSION: A machine learning algorithm run on EHR clinical data has potential utility for differentiating functional status in the clinical setting. Through further validation and refinement, such algorithms can complement traditional screening methods and result in a population-based strategy for identifying patients with poor functional status who need additional health resources.

Sequential implementation of the EQUIPPED geriatric medication safety program as a learning health system.

OBJECTIVES: To present the three-site EQUIPPED academic health system research collaborative, which engaged in sequential implementation of the EQUIPPED medication safety program, as a learning health system; to understand how the organizations worked together to build resources for program scale-up. DESIGN: Following the Replicating Effective Programs framework, we analyzed content from implementation teams' focus groups, local and cross-site meeting minutes and sites' organizational profiles to develop an implementation package. SETTING: Three academic emergency departments that each implemented EQUIPPED over three successive years. PARTICIPANTS: Implementation team members at each site participating in focus groups (n = 18), local meetings during implementation years, and cross-site meetings during all years of the projects. INTERVENTION(S): EQUIPPED provides Emergency Department providers with clinical decision support (education, order sets, and feedback) to reduce prescribing of potentially inappropriate medications to adults aged 65 years and older who received a prescription at time of discharge. MAIN OUTCOME MEASURE(S): Implementation process components assembled through successive implementation. RESULTS: Each site had clinical and environmental characteristics to be addressed in implementing the EQUIPPED program. We identified 10 process elements and describe lessons for each. Lessons guided the compilation of the EQUIPPED intervention package or toolkit, including the EQUIPPED logic model. CONCLUSIONS: Our academic health system research collaborative addressing medication safety through sequential implementation is a learning health system that can serve as a model for other quality improvement projects with multiple sites. The network produced an implementation package that can be vetted, piloted, evaluated, and finalized for large-scale dissemination in community-based settings.

Geriatric Resource Teams: Equipping Primary Care Practices to Meet the Complex Care Needs of Older Adults.

Primary care practices lack the time, expertise, and resources to perform traditional comprehensive geriatric assessment. In particular, they need methods to improve their capacity to identify and care for older adults with complex care needs, such as cognitive impairment. As the US population ages, discovering strategies to address these complex care needs within primary care are urgently needed. This article describes the development of an innovative, team-based model to improve the diagnosis and care of older adults with cognitive impairment in primary care practices. This model was developed through a mentoring process from a team with expertise in geriatrics and quality improvement. Refinement of the existing assessment process performed during routine care allowed patients with cognitive impairment to be identified. The practice team then used a collaborative workflow to connect patients with appropriate community resources. Utilization of these processes led to reduced referrals to the geriatrics specialty clinic, fewer patients presenting in a crisis to the social worker, and greater collaboration and self-efficacy for care of those with cognitive impairment within the practice. Although the model was initially developed to address cognitive impairment, the impact has been applied more broadly to improve the care of older adults with multimorbidity.

Systems-Based Practice to Improve Care Within and Beyond the Emergency Department.

There is evidence that an emergency department (ED) visit signifies a period of vulnerability for older adults. Transition between the ED and community care can be fraught with challenges. There are essential elements for improved care transition from the ED to the community. Starting a new program requires buy-in from leaders, clinical team, and community. Improving care within an ED requires looking beyond the ED. Following implementation science will increase the success of program implementation and dissemination. There are successful alternative approaches that can be learned from when striving to improve care and transitions.

Early induction of oxidative stress in mouse model of Alzheimer disease with reduced mitochondrial superoxide dismutase activity.

While oxidative stress has been linked to Alzheimer's disease, the underlying pathophysiological relationship is unclear. To examine this relationship, we induced oxidative stress through the genetic ablation of one copy of mitochondrial antioxidant superoxide dismutase 2 (Sod2) allele in mutant human amyloid precursor protein (hAPP) transgenic mice. The brains of young (5-7 months of age) and old (25-30 months of age) mice with the four genotypes, wild-type (Sod2(+/+)), hemizygous Sod2 (Sod2(+/-)), hAPP/wild-type (Sod2(+/+)), and hAPP/hemizygous (Sod2(+/-)) were examined to assess levels of oxidative stress markers 4-hydroxy-2-nonenal and heme oxygenase-1. Sod2 reduction in young hAPP mice resulted in significantly increased oxidative stress in the pyramidal neurons of the hippocampus. Interestingly, while differences resulting from hAPP expression or Sod2 reduction were not apparent in the neurons in old mice, oxidative stress was increased in astrocytes in old, but not young hAPP mice with either Sod2(+/+) or Sod2(+/-). Our study shows the specific changes in oxidative stress and the causal relationship with the pathological progression of these mice. These results suggest that the early neuronal susceptibility to oxidative stress in the hAPP/Sod2(+/-) mice may contribute to the pathological and behavioral changes seen in this animal model.

Increased expression of p130 in Alzheimer disease.

A number of recent findings support the notion of mechanistic parallels between Alzheimer disease (AD) and oncogenic processes, specifically, that neurons in AD, like cancer cells, display aberrant mitotic cell cycle re-entry. However, the mechanism that drives postmitotic neurons to reenter cell cycle remains elusive. In this study, we focused on the retinoblastoma-related protein p130 in AD. p130 is a transcriptional regulator that complexes with E2F4/5 in the nucleus and suppresses genes that regulate entry into the cell cycle. Interestingly, our results show that there are increases in p130 in cytoplasm of susceptible pyramidal neurons as well as neuroglia, often surrounding senile plaques, and within Hirano bodies in AD. By marked contrast, p130 is found at background levels in non-diseased, age-matched controls. Our data suggest that, despite its upregulation, the aberrant localization of p130 to the neuronal cytoplasm facilitates neuronal cell cycle re-entry in AD.

Distribution, levels and phosphorylation of Raf-1 in Alzheimer's disease.

Extracellular signal-regulated kinase (ERK), a member of the mitogen-activated protein kinase pathway, has been increasingly implicated in the pathogenesis of Alzheimer's disease due to its critical role in brain function. While we previously demonstrated that ERK is activated in Alzheimer's disease, the upstream cascade leading to its activation had not been fully examined. In this study, we focused on Raf-1, one of the physiological activators of the ERK pathway. Raf-1 is activated by phosphorylation at Ser338 and Tyr340/341 and inhibited by phosphorylation at Ser259. Interestingly, phosphorylation at all three sites on Raf-1 was increased as evidenced by both immunocytochemistry and immunoblot analysis in Alzheimer's disease brains compared to age-matched controls. Both phospho-Raf-1 (Ser259) and phospho-Raf-1 (Ser338) were localized to intracytoplasmic granular structures, whereas phospho-Raf-1 (Tyr340/341) was localized to neurofibrillary tangles and granules in pyramidal neurons in Alzheimer's disease hippocampus. There is extensive overlap between phospho-Raf-1 (Ser338) and phospho-Mek1/2, the downstream effector of Raf-1, suggestive of a mechanistic link. Additionally, increased levels of Raf-1 are associated with Ras and MEK1 in Alzheimer's disease as evidenced by its coimmunoprecipitation with Ras and Mek1, respectively. Based on these findings, we speculate that Raf-1 is activated to effectively mediate Ras-dependent signals in Alzheimer's disease.

Steroidogenic acute regulatory protein (StAR): evidence of gonadotropin-induced steroidogenesis in Alzheimer disease.

BACKGROUND: Alzheimer disease (AD) is clinically characterized by progressive memory loss, impairments in behavior, language and visual-spatial skills and ultimately, death. Epidemiological data reporting the predisposition of women to AD has led to a number of lines of evidence suggesting that age-related changes in hormones of the hypothalamic-pituitary-gonadal (HPG) axis following reproductive senescence, may contribute to the etiology of AD. Recent studies from our group and others have reported not only increases in circulating gonadotropins, namely luteinizing hormone (LH) in individuals with AD compared with control individuals, but also significant elevations of LH in vulnerable neuronal populations in individuals with AD compared to control cases as well as the highest density of gonadotropin receptors in the brain are found within the hippocampus, a region devastated in AD. However, while LH is higher in AD patients, the downstream consequences of this are incompletely understood. To begin to examine this issue, here, we examined the expression levels of steroidogenic acute regulatory (StAR) protein, which regulates the first key event in steroidogenesis, namely, the transport of cholesterol into the mitochondria, and is regulated by LH through the cyclic AMP second messenger pathway, in AD and control brain tissue. RESULTS: Our data revealed that StAR protein was markedly increased in both the cytoplasm of hippocampal pyramidal neurons as well as in the cytoplasm of other non-neuronal cell types from AD brains when compared with age-matched controls. Importantly, and suggestive of a direct mechanistic link, StAR protein expression in AD brains colocalized with LH receptor expression. CONCLUSION: Therefore, our findings suggest that LH is not only able to bind to its receptor and induce potentially pathogenic signaling in AD, but also that steroidogenic pathways regulated by LH may play a role in AD.