SCCM-Weil Research Grant Recipients

Topic: Racialization of ICU Notes and Debiasing Strategies to Lead to Fairer ICU Mortality Prediction

Abstract: Biased data lead to unfair and inequitable algorithms. Artificial intelligence (AI) models in non-medical domains have shown definitive examples of bias, prejudice, and direct consumer harm. Specifically, AI algorithms using natural language processing (NLP) or the quantitative study and classification of language, have also shown bias and prejudicial outputs, such as associating certain occupations with particular genders and associating certain races with stigmatizing words. However, examples of bias in NLP have not been well-studied in healthcare. This is particularly important given the rapid adoption of NLP for clinical outcome prediction. Hence, there is real risk of transmission of disparities, perpetuation of inequities and even possible patient harm if bias in NLP algorithms remains unaddressed.

Here, we seek to understand how text encodes intrinsic race and ethnicity information that if fed unaltered into AI models could lead to unfair and biased algorithms. As a first step to ensure that NLP use in healthcare does not worsen or reinforce existing disparities, the proposed research will focus on whether explicit (e.g. “Black”) and implicit aspects (e.g. “Spanish-speaking”) of notes are predictive of social constructs – specifically race and ethnicity.  Implicit descriptors may also represent how words are contextually related to one another and still encode for race or ethnicity patterns. By recognizing how written language encodes for social constructs and could be racialized, we seek to demonstrate that clinical notes are not immune to human biases and that structural racism exists in language itself. To ensure that NLP-based clinical prediction tools do not contribute to a racist healthcare structure, we seek to identify structurally racist and racialized elements, eliminate them and then construct mortality prediction models. We will determine to what extent mortality predictive performance is affected by debiasing notes. This will directly inform ways we can intervene on the note writing process, use NLP-based tools to promote antiracism in notes and construct fair and equitable note-based algorithms for risk prediction in the ICU.

Topic: Unraveling long COVID pathophysiology: insights from ICU perspectives

Abstract: The COVID-19 pandemic has impacted intensive care medicine significantly and a substantial part of patients develop long COVID syndrome after initial SARS-CoV-2 infection. Symptoms include dyspnea, fatigue and post exertional malaise and can be severely invalidating. Understanding the pathophysiology and long-term effects of long COVID syndrome is crucial for developing targeted treatments which are lacking now. Here we hypothesize a key role for disturbances in microcirculatory function and tissue oxygenation. We investigate how this is caused by an interplay between inflammation, endothelial dysfunction and microvascular dysfunction. This comparative, non-randomized, observational study will include 25 patients who have been admitted to the ICU with COVID-19 and now have long COVID syndrome, from a large cohort at our post-COVID outpatient clinic.  The control group will consist of patients who have been admitted to the ICU with COVID-19, but do not have symptoms of long COVID syndrome. For this study sublingual microcirculatory function and tissue oxygenation measurements will be performed with handheld vital microscopy sublingually before and after exercise. With the same method we are able to visualize leukocyte rolling and adhesion to the blood vessels, to study persistent inflammation. In addition, blood samples will be taken to determine markers of endothelial activation and finally questionnaires will be taken to characterize clinical symptoms. Unraveling the pathophysiology of long COVID from this systemic perspective will provide valuable insights for targeted treatments, which are urgently needed considering the large burden long COVID syndrome has on society.

Topic: Developing a Treatment for Delayed Cerebral Injury After Aneurysmal Subarachnoid Hemorrhage

Abstract: Delayed cerebral injury (DCI) is a significant complication of aneurysmal subarachnoid hemorrhage (SAH) that results from a ruptured aneurysm in the cerebrospinal fluid (CSF) space of the brain. DCI leads to long-standing cognitive and memory deficits in patients. Because it occurs 4 to 14 days after the aneurysm ruptures, it is potentially reversible. Our laboratory and others have shown that DCI is the result of inflammation that begins in the coverings of the brain (the meninges) and is directed by an inflammatory cell called a neutrophil. We have found that a neutrophil enzyme, myeloperoxidase (MPO), is critical for the development of DCI. In this proposal, we ask two important questions that will make it possible to develop therapeutic interventions against DCI. First, if the neutrophils accumulate in the meninges, how do they signal brain cells to become inflamed? Our laboratory has found evidence that the cells of one layer of the meninges, the meningothelial cell, may be an important conduit for inflammatory information to the brain. In this proposal, we will investigate how MPO interacts with meningothelial cells through a series of cell culture experiments. Second, can a specific MPO blocker improve DCI in a mouse model of SAH? We will test two MPO blockers by administering them in the CSF space (where the meningothelial cells reside) to attempt to improve the behavioral deficits of the disease. With these two experiments, we will better understand how MPO sends signals from the meninges to the brain and whether administration of MPO blockers into the CSF space is a potential therapeutic intervention for preventing or ameliorating DCI.

Topic: Metabolic Crosstalk in Acute Lung Injury Between Alveolar Macrophages and Epithelium

Abstract: Uncontrolled inflammation is a hallmark of acute lung injury (ALI). Alveolar type II (ATII) cells are key players in ALI. Alveolar macrophages (AMs) are critical to the pathogenesis of ALI, in which local microenvironmental cues shape their inflammatory properties, which can be fluent and amenable to manipulation. The mechanisms of how metabolic intermediates (e.g., lactate, the end product of glycolysis) can affect AM phenotypes are unknown in the setting of ALI. This proposal aims to delineate how lactate produced by alveolar epithelial type II cells is protective in ALI by shifting the AM toward an antiinflammatory phenotype. An innate protective mechanism in ALI involves enhanced glycolysis in ATII cells via induction of phosphofructokinase-2/fructose-2,6-bisphosphatase (PFKFB) 3. Our preliminary data show that macrophages co-cultured with ATII cells display a blunted response to LPS stimulation, which is dependent on lactate produced by the alveolar epithelial cells. Furthermore, locally intratracheally applied lactate can attenuate ALI. Our data suggest that this is associated with an increase in histone lactylation. Histone lactylation is a recently described histone modification that renders promoters of genes accessible to transcription. We hypothesize that lactate released by ATII cells dampens the AMs' excessive generation of proinflammatory cytokines in ALI via a lactate-mediated epigenetic regulation. In this application, we will test whether lactate produced by ATII cells shifts the AM phenotype toward an antiinflammatory, proresolving phenotype and we will determine whether in animals with ATII specific deletion of Pfkfb3 (decreased glycolytic capacity), AMs exhibit a more proinflammatory phenotype (aim 1). We will study how lactate induces the expression of antiinflammatory cytokine transcription by altering histone lactylation (aim 2). Acid aspiration and intratracheal LPS will be used as murine models of ALI. We will use primary human AMs for translational studies. ELISA assays and qPCR will be used to characterize AMs. We will study H3 histone lactylation modifications and transcription factor analysis experiments. This proposal is novel by expanding the scope of understanding of the role of epimetabolic crosstalk between ATII cells and AMs in ALI. It involves state-of-the-art experimental approaches and incorporates a therapeutic approach with potential for translation.

Topic: The Impact of a Metabolic Syndrome Phenotype in ARDS

Abstract: Acute respiratory distress syndrome (ARDS), a highly morbid and often fatal form of acute hypoxic respiratory failure, is common, with approximately 200,000 cases per year in the United States. However, the presentation, diagnosis, and clinical course of ARDS remain heterogeneous. The severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) pandemic increased the incidence of ARDS and as a result highlighted predisposing factors not studied before, such as metabolic syndrome (MetS). Similar to ARDS, MetS is a heterogeneous inflammatory state, often referred to as meta-inflammation and characterized by insulin resistance, obesity, dyslipidemia, and hypertension. ARDS develops as a systemic inflammatory response to certain risk factors and, although some metabolic determinants of health (e.g., obesity) are associated with non-COVID-19 ARDS, emerging data on COVID-19 ARDS clearly demonstrate a link with MetS. The need to better classify ARDS is well recognized. The overall objective of this proposal is to determine how a MetS phenotype alters ARDS risk and response to treatment. The central hypothesis is that a MetS phenotype will identify ARDS patients at increased risk for severe disease and worse outcomes. The rationale is based on our own evidence from COVID-19 ARDS and MetS as well as prior data from others on sepsis-related ARDS. To test this hypothesis, we propose two specific aims: 1) identify the effects of a MetS phenotype on ICU outcomes in patients with ARDS, and 2) determine how ARDS patients with a MetS phenotype respond to adjunctive therapies for ARDS (corticosteroids, prone positioning, ventilator strategies, or neuromuscular blockade) compared to patients without MetS. For the first aim, ARDS patients with MetS will be identified and compared to a matched cohort of control ARDS patients without MetS, using a subset of patients from SCCM Discovery’s SAGE trial and VIRUS registry. For the second aim, we will compare how outcomes differ for adjunctive ARDS therapies in patients with and without MetS. This project is innovative because it explores a never-before-tested hypothesis using existing data to discover new mechanisms and treatments for ARDS. The proposal is significant because it will facilitate harmonization of two robust datasets and will provide strong scientific justification for a follow-on R01 application to the NHLBI to study the biologic mechanisms underpinning how metabolic determinants of health contribute to ARDS.

Topic: Geographic Access to Pediatric Healthcare According to Social Determinants of Health

Abstract: Social determinants of health (SDOH) such as socioeconomic status (SES), race, ethnicity, and urbanism profoundly influence child health access and outcomes. Because of health access disparities, diverse or disadvantaged children are more likely to die from critical illness and less likely to undergo lifesaving therapy. SDOH are necessarily shaped by policies and resource distribution. While centralization may achieve economies of scale and promote quality, for critically ill children timely access to care can mean the difference between life or death. Our limited understanding of the complex interplay between SDOH, access, and outcomes represents a barrier to eliminating disparities.

Health resource allocation is susceptible to systemic discrimination. Social gradients and racial gaps in geographic access to high-quality acute pediatric care resulting from resource maldistribution relative to SDOH is a potential explanation for observed disparities. With researchers challenged to elucidate mechanistic drivers of inequity, this avenue deserves investigation. Our overarching hypothesis is that a disproportionate number of vulnerable children, as measured by SDOH, have low geographic access to care. Preliminary analyses in Massachusetts revealed that 88.3% of the pediatric population reside within a 60-minute drive of hospitals that admit children. Wealth indices (100 is the U.S. average) were significantly lower in areas more than 60 minutes from a hospital (mean difference, 48; 95% CI, 19-77, P = 0.004). Massachusetts is rich in health resources so observed disparities justify the study of geographic access in the United States and quality critical care resource allocation across other SDOH.

To meet this need we will perform cross-sectional analyses of pediatric healthcare distribution using sophisticated geospatial approaches that incorporate SES, diversity, and urbanism. First, we will quantify children’s geographic access to hospitals defined by 30-, 60-, and 120-minute driving radii (Aim 1). Next, we will characterize SDOH of families living increasing distance (less geographic access) from hospitals (Aim 2). Finally, we will evaluate geographic access according to SDOH stratified by hospital features, critical service availability, and quality indicators (Aim 3).

This research will deliver a novel framework to evaluate geographic distribution of pediatric health infrastructure according to SDOH, providing insights into geographic access as a contributor to inequity and improved hospital-level risk adjustment, with public health implications to analyze current service allocation and inform expansion.

Topic: The role of ARG-1 in Microglia During Efferocytosis After Neonatal Brain Hypoxia-Ischemia 

Abstract: Hypoxic-ischemic brain injury (HI) remains the major cause of long-term disabilities in children because therapies are limited to therapeutic hypothermia, which provides only partial neuroprotection. A critical determinant of neuronal survival is neuroinflammation and, most importantly, biological processes that help to ameliorate it. In pursuit of new pathways that may serve as new therapies by minimizing the initial inflammatory response, this proposal focuses on the relatively underexplored arginase (ARG-1) pathway in microglia/macrophages (mi/ma). ARG-1 is a key regulatory enzyme of inflammatory and tissue repair pathways in the body that become activated after HI, such as macrophage polarization, oxidant/nitrosative stress, apoptosis, clearance of dead cells (efferocytosis), and tissue repair. While recent studies associate increased ARG-1 expression with better outcomes in various neuropathologies, and ARG-1 has become widely used as a major marker of pro-repair microglia (mi/ma), the exact role of ARG-1 expressed by ARG-1+mi/ma in neuroprotection remains unknown. We are the first to describe the spatiotemporal changes in ARG-1 activity and expression during neurodevelopment and after HI. After HI, the subset of ARG-1+mi/ma accumulate in the injured hemisphere within 4 hours of injury, where they engulf dead and dying neurons, suggesting the involvement of ARG-1+mi/ma in efferocytosis. We also showed that ARG-1+mi/ma remain at the injury site for a long time; however, HI suppresses ARG-1 expression at 8 days, possibly worsening the outcome. 

Our proposal is built on the strong scientific premise that multiple regulatory molecules of efferocytosis upregulate ARG-1, possibly leading to synthesis of polyamines necessary for building a cytoskeleton for phagocytosis, thereby enhancing clearance of dead cells. This is how efferocytosis prevents stimulation of proinflammatory cytokines and contributes to the resolution of inflammation. In the brain, the role of ARG-1+mi/ma in the regulation of efferocytosis has not been explored. Our work will precisely elucidate the extent that brain efferocytosis depends on ARG-1 (aim 1) and whether inducing the ARG-1 pathway will accelerate efferocytosis and limit the inflammatory environment (aim 2), thereby improving histologic and neurologic outcomes (aim 3). The hypothesis to be tested is that ARG-1 is pivotal for regulation of efferocytosis functions of ARG-1+mi/ma, contributing to neuroprotection.

Topic: Move to Music Video Intervention: A Music and Video Guided Exercise Intervention for ICU Survivors

Abstract: Each year, more than 5.7 million people are admitted to intensive care units (ICUs). Because of inactivity during their ICU stay and following discharge, ICU survivors experience more profound and persistent physical impairments compared with other hospitalized patients. To address this problem, we have developed an innovative and engaging music-guided exercise intervention called Move to Music (M2M), which targets inactivity as a reversible risk factor to promote better patient outcomes. Our preliminary M2M study provides promising findings for improving physical strength and subsequent overall functioning; however, in response to feedback from M2M participants, we propose adding a video component to create an M2M video intervention (M2M-V) to further promote ICU survivors’ motivation and ability to self-manage exercise. 

The M2M-V intervention uses rhythmic music and verbal cues to guide targeted upper and lower extremity exercises, while the video provides ongoing guidance and feedback. This low-cost, tailored, and easily implemented intervention has not been tested among ICU survivors. Our specific aim is to determine the preliminary efficacy of the M2M-V intervention on physical outcomes by randomly assigning ICU survivors to two groups (M2M-V and M2M). We anticipate that ICU survivors who receive the M2M-V intervention will demonstrate improved physical health, physical activity, and handgrip strength, as well as higher adherence to their exercise program compared with those assigned to M2M. Findings from this preliminary study will provide important data to support a larger clinical trial designed to assess long-term rehabilitation and recovery for ICU survivors. Our premise is that a tailored, self-managed music and video guided exercise intervention (M2M-V) will reduce physical inactivity and improve physical outcomes among ICU survivors. The proposal is directly aligned with SCCM’s research priority areas to advance and improve patient experience and outcomes for ICU survivors.

Topic: Using Video Games to Reduce Diagnostic Errors in Acute Respiratory Failure

Abstract: Each year, two million Americans are hospitalized with acute respiratory failure, and one out of five will not survive their hospitalizations. Early diagnosis and treatment can prevent death. Yet, because several different conditions can cause acute respiratory failure, accurate diagnosis is difficult. As a result, diagnostic errors remain a critical, yet potentially avoidable, problem.
 
This proposal seeks to address two key knowledge gaps. First, the current structure of medical training requires direct one-on-one learning from expert physicians, which is time-consuming, unstructured, and limited by the availability of experts. Second, improvement in diagnostic ability requires a controlled and repetitive training environment followed by immediate feedback, which is often impossible to achieve in routine practice.
 
To address these problems, we propose to use video game technology to reduce diagnostic errors through two pathways: 1) widely disseminating the mental models used by physicians to make diagnoses, and 2) using an enjoyable yet realistic virtual training environment to hone diagnostic skills.
 
Our aims are to: 1) identify the most relevant clinical characteristics used by physicians to diagnose the cause of acute respiratory failure, and 2) develop a video game to reduce diagnostic errors in acute respiratory failure.
 
This proposal is directly aligned with SCCM’s research priority areas to incorporate cognitive psychology, systems engineering, social science, and simulations into critical care education and training. The findings from this study will form the foundation for future R-series proposals to pilot and conduct a multicenter randomized trial of a video game to reduce diagnostic errors in acute respiratory failure.

Topic: Nano-Scale Drug Carriers Targeted to Alveolar Marginated Neutrophils for the Treatment of ARDS

Abstract: Acute respiratory distress syndrome (ARDS) is an acute inflammation of the alveoli (air sacs) of the lungs that afflicts 190,000 Americans each year and has a mortality rate of 40%. There are no FDA-approved drugs for ARDS due to three pharmacologic challenges. First, the patients have multisystem organ failure and therefore tolerate off-target drug side effects poorly. Second, the alveoli are covered with viscous liquid, preventing inhaled drug delivery. Third, ARDS has many mechanisms, so targeting a single molecular pathway is unlikely to work.
 
To overcome these challenges, we have created a nanotechnology to deliver unprecedented amounts of multiple drugs to inflamed alveoli at very high concentrations, thus lowering off-target side effects. Our technology, called D20-tagged liposomes (DLs), consists of 100-nanometer-diameter lipid spheres that are loaded with small-molecule drugs and are covalently coated with our newly discovered D20 tag. On IV injection, DLs massively accumulate in the alveolar marginated neutrophils, which are markedly increased in number in human ARDS and animal models of acute lung injury.
 
In this proposal, our aims are to further translational development of DLs in two ways: 1) assess the therapeutic efficacy of DL-loaded drugs in two ARDS-like mouse models, and 2) test DLs in ex vivo human lungs. Pursuing these two aims will provide fundamental insights into mechanisms of drug delivery and potentially create a new therapy for ARDS.

Topic: PTSD in ICU Family Caregivers: Trajectories, Risk, and Association With Patient Health Outcomes

Abstract: Emotional distress is common among family caregivers of patients admitted to the intensive care unit (ICU), with nearly two-thirds of caregivers affected by posttraumatic stress disorder (PTSD). PTSD is associated with negative outcomes including reduced physical health, unemployment, and suicide attempts. Furthermore, research in non-ICU populations has shown that heightened caregiver emotional distress is associated with worse patient health outcomes. Interventions to reduce ICU family caregiver PTSD have been met with limited success, which may indicate an incomplete understanding of the underlying mechanisms. I am proposing a prospective cohort study that will provide a comprehensive understanding of ICU family caregiver PTSD. The specific aims are to: 1) define six-month PTSD trajectories for ICU family caregivers and identify patient and caregiver characteristics that predict trajectory membership, 2) identify stressors and preferences for emotional support among caregivers with significant PTSD symptoms, and 3) measure the association between caregiver PTSD trajectory and patient health outcomes.