Search

Reservoirs of dense atomic gas (primarily hydrogen), contain approximately 90 percent of the neutral gas at a redshift of 3, and contribute to 2-3 percent of the total baryons in the Universe. These damped Lyman-${\alpha}$ systems (so called because they absorb Lyman-${\alpha}$ photons from within and from background sources) have been studied for decades, but only through absorption lines present in the spectra of background quasars and gamma-ray bursts. Such pencil beams do not constrain the physical extent of the systems. Here, we report integral-field spectroscopy of a bright, gravitationally lensed galaxy at a redshift of 2.7 with two foreground damped Lyman-${\alpha}$ systems. These systems are $>$ 238 $kpc^2$ in extent, with column densities of neutral hydrogen varying by more than an order of magnitude on $ 5.5 \times 10^{8}$ - $1.4 \times 10^{9} M_{\odot}$, showing that they contain the necessary fuel for the next generation of star formation, consistent with relatively massive, low-luminosity primeval galaxies at redshifts $>$ 2., Comment: 31 pages, 4 Figures, 4 Extended Data figures, 2 Extended Data tables, Author's version, Accepted: 4 March 2022, Published online by Nature on May 18, 2022

Background: Behavioural non-pharmaceutical interventions (NPIs) such as social distancing and hygiene (masking, hand hygiene) were implemented widely during the COVID-19 pandemic to control infections. NPI adherence is crucial for effectiveness, but has also been implied in an increase in mental health problems. However, existing studies often have designs unsuited to understanding mechanistic relationships between NPI adherence, mental health symptoms, and health-related cognitions (risk perceptions, control beliefs). Understanding these mechanisms is crucial for the refinement of existing and the designs of effective future NPIs.Purpose: To separate between- and temporal within-person associations between mental health, risk cognitions and NPI adherence.Methods: Six-month ecological momentary assessment study with six 4-day assessment bouts in 397 German adults. Daily measurement of adherence, mental health symptoms and cognitions during bouts. We used dynamic temporal network analysis to estimate between- person, as well as contemporaneous and temporal within-person effects for distancing and hygiene NPIs.Results: Distinct networks of mental health, health cognitions and adherence emerged. Participants with higher control and higher susceptibility were also more adherent. Similar within-person contemporaneous patterns; additionally, distancing and loneliness were associated. Temporally, better adherence to NPIs was associated with better mental health on subsequent days, whereas higher loneliness was associated with better subsequent hygiene adherence.Conclusion: Findings suggest no negative impact of NPI adherence on mental health or vice versa, but instead suggest that adherence might improve mental health symptoms. Control beliefs and risk perceptions are important covariates of adherence - both on between-person and within-person level.

Objectives: Adherence to behavioral non-pharmaceutical interventions (NPIs) is key to control COVID-19 infection rates. However, rates have decreased globally, and the potentially modifiable determinants of ongoing adherence and their interaction with social and physical momentary environments are still poorly understood. Here, we comprehensively examine within-person variations and between-person differences in known behavioral determinants (capability, motivation), as well as the moderating role of situationally variable environmental factors (opportunity) in predicting adherence with hygiene and distance-related NPIs. Methods: Ecological momentary assessment study over six months with monthly assessment bouts (four days each, five daily assessments) in 623 German adults. Repeated daily assessments of Capability-Opportunity-Motivation model of behavior (COM-B) factors. Bayesian multilevel logistic regression models were estimated to examine main effects of COM-B factors and moderating effects of momentary environmental factors. Results: Momentary adherence to NPIs was predicted by within-person changes in COM-B factors (motivation: intentions, goal conflict, control beliefs; opportunities: regulations, norms). Between-person differences in capabilities (habit strength) and motivation (intentions, control beliefs) predicted adherence across situations. Situation-specific environmental factors moderated the motivation-behavior association (regulation measures increased; goal conflict, and non-adherent others decreased the association). Conclusions: Individual momentary (within-person) and stable (between-person) motivation indicators predicted adherence. However, situational environmental factors such as regulations or norms have strong main effects and moderate the motivation-behavior translation. These findings have policy implications, supporting recent claims to not rely on the narrative of “personal responsibility”, but instead combine health education measures to increase individual motivation with consistent regulation.

It has been shown that infants can increase or modify a motorically available behavior such as sucking, kicking, arm waving, etc., in response to a positive visual reinforcement (e.g., DeCasper & Fifer, 1980; Millar, 1990; Rochat & Striano, 1999; Rovee-Collier, 1997; Watson & Ramey, 1972). We tested infants to determine if they would also change their vocal behavior in response to contingent feedback, which lacks the social, emotional, and auditory modeling typical of parent-child interaction. Here, we show that in a single five-minute session infants increase the rate of their vocalizations in order to control the appearance of colorful shapes on an iPad screen. This is the first experimental study to demonstrate that infants can rapidly learn to increase their vocalizations, when given positive reinforcement with no social element. This work sets the foundations for future studies into the causal relationship between the number of early vocalizations and the onset of words. In addition, there are potential clinical applications for reinforcing vocal practice in infant populations who are at risk for poor language skills.

This thesis discussed how advanced analytics and real-world evidence can be utilized to improve patient outcomes. In chapter 2 we summarize frequently observed pitfalls clinicians and data scientists should seek to avoid. Furthermore, we describe potential solutions how these issues can be avoided and overcome. In chapter 3 we compared the four publicly available adult intensive care data sets. As one guiding principle, we suggest eICU-CRD be used for studies when large patient numbers are needed, and concerns include a diverse patient cohort and generalizability. HiRID and AmsterdamUMCdb, meanwhile, are better choices for studies requiring granular and frequent measurements. We suggest MIMIC-IV as a good all-round data set given its higher patient count, good data frequencies and well-established concepts. In chapter 4 we applied some of the considerations discussed in chapters 1 and 2 to a machine learning problem, using a gradient boosted model (GBM) to predict re-admission to the ICU. While we fully agreed with the relevance of the research question and merit of the approach taken, we also critically appraised some of the design choices and questioned if the model would be able to accurately predict re-admission in real-world settings. Regardless of potential modelling and data shortcomings, we argued that more studies using machine learning are required to decrease inter-physician differences in treatment policies and improve clinical care through data. In chapter 5 we presented a clinical use case with application of linear models to real-world data. Using the MIMIC-III data set, trends in survival of cancer patients between 2002-2011 admitted to a single center in Boston, MA were studied. A statistically significant decrease in 28-day mortality rates of cancer patients was found (2002: 33.3% vs. 23.6% in 2011). Chapter 6 explored and analyzed factors that differentiate severely ill sepsis patients with normal versus high serum lactate values. Our analyses showed that high serum bicarbonate and serum chloride levels were associated with normal serum lactate while elevated serum sodium levels, AST and presence of liver disease were strongly associated with high serum lactate levels. These findings were consistent across three data sets from the U.S. and Europe. In chapter 7 a machine learning algorithm was built to select features and predict treatment failure in tuberculosis patients. Using various methods, we identified patient characteristics and clinical features with good classification performance to identify patients at high risk of treatment failure. Linear models performed best, which was slightly higher than models previously established in different countries. Clinical application is, however, limited, as the sensitivity of the models is low (30-36%) when tuned to a specificity of approx. 80%. The relatively low number of treatment failures is likely the main limitation on model performance. In chapter 8 we proved machine learning methods can accurately back-predict pre-admission serum creatinine and hemoglobin values. We developed machine learning algorithms that were trained on patients with outpatient laboratory measurements taken before ICU admission, and hemoglobin and serum creatinine measurements available at admission. Gradient boosted models, random forest models and logistic regression models all achieved good classification performance for all subcohorts (AUC > ~0.7). Gradient boosted models performed best for the cohort with a baseline hemoglobin

This thesis discussed how advanced analytics and real-world evidence can be utilized to improve patient outcomes. In chapter 2 we summarize frequently observed pitfalls clinicians and data scientists should seek to avoid. Furthermore, we describe potential solutions how these issues can be avoided and overcome. In chapter 3 we compared the four publicly available adult intensive care data sets. As one guiding principle, we suggest eICU-CRD be used for studies when large patient numbers are needed, and concerns include a diverse patient cohort and generalizability. HiRID and AmsterdamUMCdb, meanwhile, are better choices for studies requiring granular and frequent measurements. We suggest MIMIC-IV as a good all-round data set given its higher patient count, good data frequencies and well-established concepts. In chapter 4 we applied some of the considerations discussed in chapters 1 and 2 to a machine learning problem, using a gradient boosted model (GBM) to predict re-admission to the ICU. While we fully agreed with the relevance of the research question and merit of the approach taken, we also critically appraised some of the design choices and questioned if the model would be able to accurately predict re-admission in real-world settings. Regardless of potential modelling and data shortcomings, we argued that more studies using machine learning are required to decrease inter-physician differences in treatment policies and improve clinical care through data. In chapter 5 we presented a clinical use case with application of linear models to real-world data. Using the MIMIC-III data set, trends in survival of cancer patients between 2002-2011 admitted to a single center in Boston, MA were studied. A statistically significant decrease in 28-day mortality rates of cancer patients was found (2002: 33.3% vs. 23.6% in 2011). Chapter 6 explored and analyzed factors that differentiate severely ill sepsis patients with normal versus high serum lactate values. Our analyses showed that high serum bicarbonate and serum chloride levels were associated with normal serum lactate while elevated serum sodium levels, AST and presence of liver disease were strongly associated with high serum lactate levels. These findings were consistent across three data sets from the U.S. and Europe. In chapter 7 a machine learning algorithm was built to select features and predict treatment failure in tuberculosis patients. Using various methods, we identified patient characteristics and clinical features with good classification performance to identify patients at high risk of treatment failure. Linear models performed best, which was slightly higher than models previously established in different countries. Clinical application is, however, limited, as the sensitivity of the models is low (30-36%) when tuned to a specificity of approx. 80%. The relatively low number of treatment failures is likely the main limitation on model performance. In chapter 8 we proved machine learning methods can accurately back-predict pre-admission serum creatinine and hemoglobin values. We developed machine learning algorithms that were trained on patients with outpatient laboratory measurements taken before ICU admission, and hemoglobin and serum creatinine measurements available at admission. Gradient boosted models, random forest models and logistic regression models all achieved good classification performance for all subcohorts (AUC > ~0.7). Gradient boosted models performed best for the cohort with a baseline hemoglobin

Background: White matter lesions (WMLs) on brain magnetic resonance imaging (MRI) in multiple sclerosis (MS) may contribute to misdiagnosis. In chronic active lesions, peripheral iron-laden macrophages appear as paramagnetic rim lesions (PRLs). Objective: To evaluate the sensitivity and specificity of PRLs in differentiating MS from mimics using clinical 3T MRI scanners. Method: This retrospective international study reviewed MRI scans of patients with MS ( n = 254), MS mimics ( n = 91) and older healthy controls ( n = 217). WMLs, detected using fluid-attenuated inversion recovery MRI, were analysed with phase-sensitive imaging. Sensitivity and specificity were assessed for PRLs. Results: At least one PRL was found in 22.9% of MS and 26.1% of clinically isolated syndrome (CIS) patients. Only one PRL was found elsewhere. The identification of ⩾1 PRL was the optimal cut-off and had high specificity (99.7%, confidence interval (CI) = 98.20%–99.99%) when distinguishing MS and CIS from mimics and healthy controls, but lower sensitivity (24.0%, CI = 18.9%–36.6%). All patients with a PRL showing a central vein sign (CVS) in the same lesion ( n = 54) had MS or CIS, giving a specificity of 100% (CI = 98.8%–100.0%) but equally low sensitivity (21.3%, CI = 16.4%–26.81%) Conclusion: PRLs may reduce diagnostic uncertainty in MS by being a highly specific imaging diagnostic biomarker, especially when used in conjunction with the CVS.

We study the star formation history (SFH) of the ultra-diffuse galaxy (UDG) Dragonfly 44 (DF44) based on the simultaneous fit to near-ultraviolet to near-infrared photometry and high signal-to-noise optical spectroscopy. In fitting the observations we adopt an advanced physical model with a flexible SFH, and we discuss the results in the context of the degeneracies between stellar population parameters. Through reconstructing the mass-assembly history with a prior for extended star formation (akin to methods in the literature) we find that DF44 formed 90 per cent of its stellar mass by $z\sim 0.9$ ($\sim 7.2$ Gyr ago). In comparison, using a prior that prefers concentrated star formation (as informed by previous studies of DF44's stellar populations) suggests that DF44 formed as early as $z\sim 8$ ($\sim 12.9$ Gyr ago). Regardless of whether DF44 is old or very old, the SFHs imply early star formation and rapid quenching. This result, together with DF44's large size and evidence that it is on its first infall into the Coma cluster, challenges UDG formation scenarios from simulations that treat all UDGs as contiguous with the canonical dwarf population. While our results cannot confirm any particular formation scenario, we can conclude from this that DF44 experienced a rare quenching event., 25 pages, 15 figures. Accepted for publication in MNRAS

Randomised controlled trials (RCTs) play an important role in multiple sclerosis (MS) research, ensuring that new interventions are safe and efficacious before their introduction into clinical practice. Trials have been evolving to improve the robustness of their designs and the efficiency of their conduct. Advances in digital and mobile technologies in recent years have facilitated this process and the first RCTs with decentralised elements became possible. Decentralised clinical trials (DCTs) are conducted remotely, enabling participation of a more heterogeneous population who can participate in research activities from different locations and at their convenience. DCTs also rely on digital and mobile technologies which allows for more flexible and frequent assessments. While hospitals quickly adapted to e-health and telehealth assessments during the COVID-19 pandemic, the conduct of conventional RCTs was profoundly disrupted. In this paper, we review the existing evidence and gaps in knowledge in the design and conduct of DCTs in MS.