Faculty Bibliography

In recent years, our understanding of the microbial world within us has been revolutionized by the use of culture-independent techniques. The use of multi-omic approaches can now not only comprehensively characterize the microbial environment but also evaluate its functional aspects and its relationship with the host immune response. Advances in bioinformatics have enabled high throughput and in-depth analyses of transcripts, proteins and metabolites and enormously expanded our understanding of the role of the human microbiome in different conditions. Such investigations of the lower airways have specific challenges but as the field develops, new approaches will be facilitated. In this review, we focus on how integrative multi-omics can advance our understanding of the microbial environment and its effects on the host immune tone in the lungs.

The last decade of research has revolutionized our understanding of respiratory microbiology, revealing that the lungs and airways contain diverse and dynamic microbial communities in health and disease. This "respiratory ecosystem"-a densely interconnected environment of microbial and host interactions-represents a tremendous and under-appreciated source of biological and clinical heterogeneity across patients with acute and chronic lung disease. Unlike other major sources of heterogeneity, such as comorbidities and host genetics, the respiratory microbiome is readily modifiable by clinical interventions, and therefore represents an untapped opportunity for therapeutic manipulation. As a potential "treatable trait" in efforts to subphenotype patients and deliver precision medicine, the respiratory microbiome is a promising therapeutic target. In this Pulmonary Perspective, we identify and discuss multiple challenges, both conceptual and practical, that must be overcome before the respiratory microbiome can be effectively modulated as a therapeutic target. Barriers include: 1) the need to identify specific microbiologic and ecologic "targets" for therapeutic modulation; 2) the need for an improved understanding of the efficacy and persistence of response to respiratory microbiome-modulating interventions; 3) the need for clinicians to be able to access, understand and utilize microbiome data for sub-phenotyping patients, and 4) specific concerns in special populations (including children, patients with chronic lung disease, and critically ill patients). By delineating these barriers, we identify opportunities for prospective research to advance our understanding of the respiratory microbiome, its role in human respiratory disease, and its genuine potential as a therapeutic target.

INTRODUCTION: Hospitals experienced substantial stress during the COVID-19 pandemic-threats to standard operations- but it is not well known how this stress manifested at individual hospitals. We aimed to understand patterns of hospital stress over time, where stress was located within hospitals, and correlations between individual stress measures.
METHOD(S): We conducted a weekly hospital stress survey from November 2020 through May 2021 among site leaders from the SCCM Discovery Severe Acute Respiratory Infection - Preparedness (SARI-PREP) multicenter prospective cohort study. The survey assessed hospital stress ordinally and also assessed ED and ICU stress and deviations from standard operating procedures. Pairwise comparisons of strain measures were calculated by Pearson's correlation coefficients (r).
RESULT(S): Eight hospitals across three health systems in New York, California, and Washington contributed 190 hospital-weeks of data. Sites reported unavailability of some hospital resources resulting in potentially avoidable patient harm during 3.5% of hospital-weeks (with at least one such week at four hospitals); alterations in care processes and/or staffing which were fully compensated for during 57.9% of weeks; and no stress during 38.6% of weeks. During one December 2020 week, hospital stress, ICU stress, and care deviations were all present at 100% of reporting sites. The most common care deviations were increased hospital staffing (39.5%) and cancelling elective surgeries (18.6%). Hospital stress and care deviations were highly correlated (r = 0.81, p < 0.0001). Stress was more common in ICUs (72.4%) than EDs (14.3%), and ICU and ED stress were not correlated (r = 0.19, p = 0.05). While ED stress rose and abated earlier, ICU stress and care deviations persisted (range 2-13 weeks longer) as local case rates declined.
CONCLUSION(S): Hospital stress during the pandemic varied in degree and type both within and among hospitals over time. Care deviations were common but potentially avoidable patient harm was rare. Systematic national assessments of hospital stress, both during and between pandemics, could inform more rapid, proactive public health responses to novel threats. Areas for further study include impacts from persistent low-level stress and longer-term consequences for hospitals and their communities

OBJECTIVES: We designed a prospective cohort study to systematically study patients with severe acute respiratory infection (SARI) and improve hospital preparedness (SARI-PREP). The goal of this project is to evaluate the natural history, prognostic biomarkers, and characteristics, including hospital stress, associated with SARI clinical outcomes and severity.
METHOD(S): In collaboration with the Society of Critical Care Medicine Discovery Research Network and the National Emerging Special Pathogen Training and Education Center (NETEC), SARIPREP is an ongoing, prospective, observational, multi-center cohort study of hospitalized patients with respiratory viral infections. We collected patient demographics, signs, symptoms, and medications; microbiology, imaging, and other diagnostics; mechanical ventilation, hospital procedures, and other interventions; and clinical outcomes. Hospital leadership completed a weekly hospital stress survey. Respiratory, blood, and urine biospecimens were collected from patients on days 0, 3, 7-14 after study enrollment and at hospital discharge. MEASUREMENTS AND MAIN RESULTS: SARI-PREP enrollment began on April 4, 2020 and currently includes 674 patients. Here we report results from the first 400 patients: 216 are from the University of Washington Hospitals, Seattle WA, 142 from New York University, New York NY and 42 from University of Southern California, Los Angeles, CA. Almost all tested positive for SARS-CoV-2 infection (n=397), whereas 3 patients tested positive for an alternative viral pathogen. The mean (+/-SD) age of the patients was 57+/-16 years; 72% were men, 62% were White, 14% were Asian, 12% were Black, and 31% were Hispanic. Most of the patients were admitted to the intensive care unit (96%). The median (interquartile range) hospital length of stay was 22 (9-46) days. Rates of invasive mechanical ventilation (72%) and renal replacement therapy (19%) were common and the rate of hospital mortality was 35%.
CONCLUSION(S): Initial SARI-PREP analysis indicates enrollment of a diverse population of hospitalized patients primarily with SARSCoV-2 infection. The demographics and clinical outcomes of our cohort mirror other large critically ill cohorts of COVID-19 patients. Results of a concomitant, weekly, hospital stress assessment are reported separately

Although high-throughput data allow researchers to interrogate thousands of variables simultaneously, it can also introduce a significant number of spurious results. Here we demonstrate that correlation analysis of large datasets can yield numerous false positives due to the presence of outliers that canonical methods fail to identify. We present Correlations Under The InfluencE (CUTIE), an open-source jackknifing-based method to detect such cases with both parametric and non-parametric correlation measures, and which can also uniquely rescue correlations not originally deemed significant or with incorrect sign. Our approach can additionally be used to identify variables or samples that induce these false correlations in high proportion. A meta-analysis of various omics datasets using CUTIE reveals that this issue is pervasive across different domains, although microbiome data are particularly susceptible to it. Although the significance of a correlation eventually depends on the thresholds used, our approach provides an efficient way to automatically identify those that warrant closer examination in very large datasets.