22 research outputs found
Identification of Bacterial Operons Required During The Plant Immune Response using a RB-TnSeq Approach
No full textToday, the majority of agrochemicals used to increase agricultural output have adverse effects on the environment, and the developments of more sustainable approaches that utilize plant-growth promoting bacteria are needed 1. In order to benefit the plants, these microbes must be able to colonize the plant root in mixed environmental microbial communities that induce immune responses from the plant. Toward understanding the genetic basis of root colonization by plant-growth promoting microbes, a randomly barcoded transposon insertion library sequencing method (Bar-Seq) was utilized to identify genes that protect bacteria from the plant-immune response. Arabidopsis thaliana ecotype Col-0 and two mutants deficient in immune responses were grown for 5 weeks in liquid media and then an immune response was induced for 1 day using flg22 flagellin peptide. Exudates containing anti-microbial chemicals produced by the plants were collected. Bar-Seq libraries of four different plant-associated bacteria were grown in each exudate and the transposon barcodes corresponding to interrupted genes were sequenced. First, the genes are assigned fitness scores by comparing barcode abundance to input. Then, the fitness scores are compared (mutant vs. wild-type, + flg22 vs. – flg22) to identify genes required for growth during the immune response. These results will assist in the efforts to achieve higher crop yields in a more sustainable way, and provide insight into the genetic interactions between bacteria and their hosts.Bachelor of Scienc
Mining for Equitable Health: Assessing the Impact of Missing Data in Electronic Health Records
No full textElectronic health records (EHR) are collected as a routine part of healthcare delivery, and have great potential to be utilized to improve patient health outcomes. They contain multiple years of health information to be leveraged for risk prediction, disease detection, and treatment evaluation. However, they do not have a consistent, standardized format across institutions, particularly in the United States, and can present significant analytical challenges- they contain multi-scale data from heterogeneous domains and include both structured and unstructured data. Data for individual patients are collected at irregular time intervals and with varying frequencies. In addition to the analytical challenges, EHR can reflect inequity- patients belonging to different groups will have differing amounts of data in their health records. Many of these issues can contribute to biased data collection. The consequence is that the data for under-served groups may be less informative partly due to more fragmented care, which can be viewed as a type of missing data problem. For EHR data in this complex form, there is currently no framework for introducing realistic missing values. There has also been little to no work in assessing the impact of missing data in EHR. In this work, we first introduce a terminology to define three levels of EHR data and then propose a novel framework for simulating realistic missing data scenarios in EHR to adequately assess their impact on predictive modeling. We incorporate the use of a medical knowledge graph to capture dependencies between medical events to create a more realistic missing data framework. In an intensive care unit setting, we found that missing data have greater negative impact on the performance of disease prediction models in groups that tend to have less access to healthcare, or seek less healthcare. We also found that the impact of missing data on disease prediction models is stronger when using the knowledge graph framework to introduce realistic missing values as opposed to random event removal
On the Algorithmic Bias of Aligning Large Language Models with RLHF: Preference Collapse and Matching Regularization
No full textAccurately aligning large language models (LLMs) with human preferences is
crucial for informing fair, economically sound, and statistically efficient
decision-making processes. However, we argue that reinforcement learning from
human feedback (RLHF) -- the predominant approach for aligning LLMs with human
preferences through a reward model -- suffers from an inherent algorithmic bias
due to its Kullback--Leibler-based regularization in optimization. In extreme
cases, this bias could lead to a phenomenon we term preference collapse, where
minority preferences are virtually disregarded. To mitigate this algorithmic
bias, we introduce preference matching (PM) RLHF, a novel approach that
provably aligns LLMs with the preference distribution of the reward model under
the Bradley--Terry--Luce/Plackett--Luce model. Central to our approach is a PM
regularizer that takes the form of the negative logarithm of the LLM's policy
probability distribution over responses, which helps the LLM balance response
diversification and reward maximization. Notably, we obtain this regularizer by
solving an ordinary differential equation that is necessary for the PM
property. For practical implementation, we introduce a conditional variant of
PM RLHF that is tailored to natural language generation. Finally, we
empirically validate the effectiveness of conditional PM RLHF through
experiments on the OPT-1.3B and Llama-2-7B models, demonstrating a 29% to 41%
improvement in alignment with human preferences, as measured by a certain
metric, compared to standard RLHF
Informative missingness: What can we learn from patterns in missing laboratory data in the electronic health record?
No full textBackground: In electronic health records, patterns of missing laboratory test results could capture patients’ course of disease as well as reflect clinician's concerns or worries for possible conditions. These patterns are often understudied and overlooked. This study aims to identify informative patterns of missingness among laboratory data collected across 15 healthcare system sites in three countries for COVID-19 inpatients. Methods: We collected and analyzed demographic, diagnosis, and laboratory data for 69,939 patients with positive COVID-19 PCR tests across three countries from 1 January 2020 through 30 September 2021. We analyzed missing laboratory measurements across sites, missingness stratification by demographic variables, temporal trends of missingness, correlations between labs based on missingness indicators over time, and clustering of groups of labs based on their missingness/ordering pattern. Results: With these analyses, we identified mapping issues faced in seven out of 15 sites. We also identified nuances in data collection and variable definition for the various sites. Temporal trend analyses may support the use of laboratory test result missingness patterns in identifying severe COVID-19 patients. Lastly, using missingness patterns, we determined relationships between various labs that reflect clinical behaviors. Conclusion: In this work, we use. computational approaches to relate missingness patterns to hospital treatment capacity and highlight the heterogeneity of looking at COVID-19 over time and at multiple sites, where there might be different phases, policies, etc. Changes in missingness could suggest a change in a patient's condition, and patterns of missingness among laboratory measurements could potentially identify clinical outcomes. This allows sites to consider missing data as informative to analyses and help researchers identify which sites are better poised to study particular questions
Recommended from our members
Crowd-sourced machine learning prediction of long COVID using data from the National COVID Cohort Collaborative.
No full textBACKGROUND: While many patients seem to recover from SARS-CoV-2 infections, many patients report experiencing SARS-CoV-2 symptoms for weeks or months after their acute COVID-19 ends, even developing new symptoms weeks after infection. These long-term effects are called post-acute sequelae of SARS-CoV-2 (PASC) or, more commonly, Long COVID. The overall prevalence of Long COVID is currently unknown, and tools are needed to help identify patients at risk for developing long COVID. METHODS: A working group of the Rapid Acceleration of Diagnostics-radical (RADx-rad) program, comprised of individuals from various NIH institutes and centers, in collaboration with REsearching COVID to Enhance Recovery (RECOVER) developed and organized the Long COVID Computational Challenge (L3C), a community challenge aimed at incentivizing the broader scientific community to develop interpretable and accurate methods for identifying patients at risk of developing Long COVID. From August 2022 to December 2022, participants developed Long COVID risk prediction algorithms using the National COVID Cohort Collaborative (N3C) data enclave, a harmonized data repository from over 75 healthcare institutions from across the United States (U.S.). FINDINGS: Over the course of the challenge, 74 teams designed and built 35 Long COVID prediction models using the N3C data enclave. The top 10 teams all scored above a 0.80 Area Under the Receiver Operator Curve (AUROC) with the highest scoring model achieving a mean AUROC of 0.895. Included in the top submission was a visualization dashboard that built timelines for each patient, updating the risk of a patient developing Long COVID in response to clinical events. INTERPRETATION: As a result of L3C, federal reviewers identified multiple machine learning models that can be used to identify patients at risk for developing Long COVID. Many of the teams used approaches in their submissions which can be applied to future clinical prediction questions. FUNDING: Research reported in this RADx® Rad publication was supported by the National Institutes of Health. Timothy Bergquist, Johanna Loomba, and Emily Pfaff were supported by Axle Subcontract: NCATS-STSS-P00438
Assessing Alternative Financing Methods for the Canadian Health Care System in View of Population Aging
No full textThe cost of the Canadian health care system is approximately 10% of Gross Domestic Product (GDP). Survey-evidence suggests that Canadians do not wish to have additional funds spent on health care but believe that the system should be able to deliver better quality care. Due to low fertility rates and increasing life expectancy, the Canadian population is aging. Over the next 25 years, the dependency ratio will increase, primarily due to the aging of the “baby boom generation” 2. This will place twofold cost pressures on governments responsible for maintaining the health care system: 1) As a consequence of increased life expectancy, on average, Canadians will have a longer period of health care consumption. Although age-specific cost may not increase, with an aging population aggregate annual health care expenditures are expected to increase. 2) The dependency ratio is a proxy for the ability of the population to support itself. The increasing dependency rate may result in a slowdown in GDP growth, given constant technology. In Section I, this paper attempts to quantify these factors. A single measure combining cost and quality is developed to demonstrate the magnitude of the challenge. In Section II, this paper examines a number of different approaches to health care financing including user fees and alternative compensation methods for physicians. The paper highlights documented information from Canada and international experience on the implementation issues involved. The paper evaluates the desirability of implementing these approaches in Canada.Alternative physician reimbursement models, Capitation, DALE, Disability Adjusted Life Expectancy, QAHE, Quality-Adjusted Health Expenditures, QAHE Index, SID, Supplier-Induced Demand
Recommended from our members
On the Value of Biomedical Research and Medical Care
No full textThis dissertation consists of three chapters that investigate the value of biomedical research and medical care.
Chapter 1 studies trends in biomedical research methods and reporting over 25 years and identifies factors associated with biomedical research quality. A comprehensive database on biomedical research quality, combining systematic expert assessment of randomized controlled trial (RCT) methods with bibliometric and funding information in a sample of 20,571 RCTs is assembled using web-scraping and text-mining. Time series describes the proportion of RCTs using adequate methods, inadequate methods and poor reporting. A multinomial logit model tests potential factors associated with methods and reporting, including funding source, first author affiliation, clinical trial registration status, study novelty, team characteristics, technology and geography. 59.3% of RCTs used inadequate methods (N=12 190) and 35.0% were poorly reported (N=7208). The proportion of poorly reported RCTs decreased from 42.5% in 1990 to 30.2% in 2015. The proportion of RCTs using inadequate methods increased from 54.9% in 1990 to 59.5% in 2015. Industry funding, top pharmaceutical company affiliation, trial registration, larger authorship teams, international teams and drug trials were associated with a greater likelihood of using adequate methods. National Institutes of Health funding and university prestige were not. Even though reporting has improved since 1990, the proportion of RCTs using inadequate methods is high (59.3%) and increasing, potentially slowing progress and contributing to the reproducibility crisis. Stronger incentives for the use of adequate methods are needed.
Chapter 2 studies incentives for the use of adequate methods in biomedical research. Specifically, this chapter investigates how the adequacy of methods affects experimental results and publication outcomes of RCTs. RCTs inform medical practice, health care delivery, follow-on research, regulation, and health policy. Yet, many RCTs are inadequately randomized, blinded, and reported. To analyze scientists' and firms' incentives to meet clinical trial standards, I assemble a detailed database on research methods, experimental results, bibliometric information, and funding for 23,321 RCTs published between 1990 and 2018. I estimate the impact of meeting scientific standards on three outcomes: (1) the direction and significance of experimental results; (2) the impact factor of the publishing journal; and (3) the number of citations the publication receives. I find that increasing numbers of inadequacies increase the probability of finding support for product adoption by 7% per inadequacy, but decrease journal prestige and citations. Publication bias and strategic non-disclosure do not appear to drive the results. I conclude that individual scientists benefit marginally from higher quality research, but that pharmaceutical companies lack strong incentives to drive improvement in trial quality.
Chapter 3 (with David Cutler and Thomas Getzen) studies the contribution of medical care to life expectancy gains. Using two hundred years of national and Massachusetts data on medical care and health, we examine how central medical care is to life expectancy gains. While common theories about medical care cost growth stress growing demand, our analysis highlights the importance of supply side factors, including the major public investments in research, workforce training and hospital construction that fueled a surge in spending over the 1955-1975 span. There is a stronger case that personal medicine affected health in the second half of the twentieth century than in the preceding 150 years. Finally, we consider whether medical care productivity decreases over time, and find that spending increased faster than life expectancy, although the ratio stabilized in the past two decades.Health Polic
Recommended from our members
Validation of Biomechanical Computed Tomography for Fracture Risk Classification in Metastatic Hormone-sensitive Prostate Cancer.
No full textBACKGROUND: Guidelines recommend dual-energy x-ray absorptiometry (DXA) screening to assess fracture risk and benefit from antiresorptive therapy in men with metastatic hormone-sensitive prostate cancer (mHSPC) on androgen deprivation therapy (ADT). However, <30% of eligible patients undergo DXA screening. Biomechanical computed tomography (BCT) is a radiomic technique that measures bone mineral density (BMD) and bone strength from computed tomography (CT) scans. OBJECTIVE: To evaluate the (1) correlations between BCT- and DXA-assessed BMD, and (2) associations between BCT-assessed metrics and subsequent fracture. DESIGN, SETTING, AND PARTICIPANTS: A multicenter retrospective cohort study was conducted among patients with mHSPC between 2013 and 2020 who received CT abdomen/pelvis or positron emission tomography/CT within 48 wk before ADT initiation and during follow-up (48-96 wk after ADT initiation). OUTCOME MEASUREMENTS AND STATISTICAL ANALYSIS: We used univariate logistic regression to assess the associations between BCT measurements and the primary outcomes of subsequent pathologic and nonpathologic fractures. RESULTS AND LIMITATIONS: Among 91 eligible patients, the median ([interquartile range) age was 67 yr (62-75), 44 (48.4%) were White, and 41 (45.1%) were Black. During the median follow-up of 82 wk, 17 men (18.6%) developed a pathologic and 15 (16.5%) a nonpathologic fracture. BCT- and DXA-assessed femoral-neck BMD T scores were strongly correlated (R2 = 0.93). On baseline CT, lower BCT-assessed BMD (odds ratio [OR] 1.80, 95% confidence interval or CI [1.10, 3.25], p = 0.03) was associated with an increased risk of a pathologic fracture. Lower femoral strength (OR 1.63, 95% CI [0.99, 2.71], p = 0.06) was marginally associated with an increased risk of a pathologic fracture. Neither BMD (OR 1.52, 95% CI [0.95, 2.63], p = 0.11) nor strength (OR 1.14, 95% CI [0.75, 1.80], p = 0.57) was associated with a nonpathologic fracture. BCT identified nine (9.9%) men eligible for antiresorptive therapy, of whom four (44%) were not treated. Limitations include low fracture numbers resulting in lower power to detect fracture associations. CONCLUSIONS: Among men diagnosed with mHSPC, BCT assessments were strongly correlated with DXA, predicted subsequent pathologic fracture, and identified additional men indicated for antiresorptive therapy. PATIENT SUMMARY: We assess whether biomechanical computer tomography (BCT) from routine computer tomography (CT) scans can identify fracture risk among patients recently diagnosed with metastatic prostate cancer. We find that BCT and dual-energy x-ray absorptiometry-derived bone mineral density are strongly correlated and that BCT accurately identifies the risk for future fracture. BCT may enable broader fracture risk assessment and facilitate timely interventions to reduce fracture risk in metastatic prostate cancer patients
Übertragbarkeit der gesundheitsökonomischen Evaluationen : Methoden für die multinationalen Studien
No full textFilonenko A. Generalizability of health economic evaluations : methods for multinational patient-level studies. Bielefeld (Germany): Bielefeld University; 2010.Background: Cost-effectiveness outcomes collected in multinational health economic studies may vary across countries and geographic areas due to numerous clinical and socioeconomic factors. The between-country variability poses the question of generalizability of the cost-effectiveness estimates, i.e. applicability of the trial-wide results for the decision-making in particular jurisdiction.
Objectives: The purpose of this research is to illustrate how selected methods can improve generalizability of the results of patient-level health economic studies by a) exploring between-country variability in incremental costs, effectiveness and resource use, b) calculation of trial-wide and country-level incremental cost, effectiveness and resource use while accounting for patient- and country-level covariates, and c) assessing the potential of covariates to predict cost and effectiveness estimates for the settings outside the study.
Methods: Review of published health-economic evidence and international HTA guidelines was conducted to evaluate applied or recommended methods and analytical strategies to improve generalizability of the health economic outcomes. In the case study, qualitative and quantitative homogeneity test by Simon and Gail is used to explore between-country heterogeneity of the treatment effects measured by incremental costs, effectiveness, resource use (length of hospitalization) and incremental net monetary benefit. Hierarchical modelling is applied to calculate trial-wide and country-level estimates, while accounting for clustering and incorporating country- and patient-level covariates.
Results: Simon and Gail test indicated qualitative homogeneity for incremental effectiveness, costs, length of hospitalization and net monetary benefit between treatment and control arms. Trial-wide and country-level mean incremental costs and effectiveness were estimated using hierarchical models with and without covariates. Results of hierarchical modelling suggested, that new treatment was more efficacious, saved costs and resource use in majority of the countries.
Conclusions: Homogeneity test and hierarchical models are complementary methods to explore heterogeneity and to estimate trial-wide and country-level parameters for cost-effectiveness analysis. Hierarchical models allow for country-level estimates and adjustment for covariates. In the case study the patient-level covariates showed effect on incremental cost and effectiveness; country-level covariates had very small impact on estimates. The use of country-level covariates as predictors for incremental cost and resource use to improve generalizability of health economic studies beyond the study setting merits further investigation
Crowd-sourced machine learning prediction of long COVID using data from the National COVID Cohort CollaborativeResearch in context
No full textSummary: Background: While many patients seem to recover from SARS-CoV-2 infections, many patients report experiencing SARS-CoV-2 symptoms for weeks or months after their acute COVID-19 ends, even developing new symptoms weeks after infection. These long-term effects are called post-acute sequelae of SARS-CoV-2 (PASC) or, more commonly, Long COVID. The overall prevalence of Long COVID is currently unknown, and tools are needed to help identify patients at risk for developing long COVID. Methods: A working group of the Rapid Acceleration of Diagnostics-radical (RADx-rad) program, comprised of individuals from various NIH institutes and centers, in collaboration with REsearching COVID to Enhance Recovery (RECOVER) developed and organized the Long COVID Computational Challenge (L3C), a community challenge aimed at incentivizing the broader scientific community to develop interpretable and accurate methods for identifying patients at risk of developing Long COVID. From August 2022 to December 2022, participants developed Long COVID risk prediction algorithms using the National COVID Cohort Collaborative (N3C) data enclave, a harmonized data repository from over 75 healthcare institutions from across the United States (U.S.). Findings: Over the course of the challenge, 74 teams designed and built 35 Long COVID prediction models using the N3C data enclave. The top 10 teams all scored above a 0.80 Area Under the Receiver Operator Curve (AUROC) with the highest scoring model achieving a mean AUROC of 0.895. Included in the top submission was a visualization dashboard that built timelines for each patient, updating the risk of a patient developing Long COVID in response to clinical events. Interpretation: As a result of L3C, federal reviewers identified multiple machine learning models that can be used to identify patients at risk for developing Long COVID. Many of the teams used approaches in their submissions which can be applied to future clinical prediction questions. Funding: Research reported in this RADx® Rad publication was supported by the National Institutes of Health. Timothy Bergquist, Johanna Loomba, and Emily Pfaff were supported by Axle Subcontract: NCATS-STSS-P00438
