Search

Graph Neural Networks (GNNs) are a popular approach for predicting graph structured data. As GNNs tightly entangle the input graph into the neural network structure, common explainable AI approaches are not applicable. To a large extent, GNNs have remained black-boxes for the user so far. In this paper, we show that GNNs can in fact be naturally explained using higher-order expansions, i.e. by identifying groups of edges that jointly contribute to the prediction. Practically, we find that such explanations can be extracted using a nested attribution scheme, where existing techniques such as layer-wise relevance propagation (LRP) can be applied at each step. The output is a collection of walks into the input graph that are relevant for the prediction. Our novel explanation method, which we denote by GNN-LRP, is applicable to a broad range of graph neural networks and lets us extract practically relevant insights on sentiment analysis of text data, structure-property relationships in quantum chemistry, and image classification., Comment: 14 pages + 6 pages supplement

Central nervous system lymphoproliferative disorder (CNS-PTLD) after organ transplant is a unique clinicopathological entity and is associated with poor survival rates. When the CNS is involved, intravenous rituximab might not be the treatment of choice, due to its poor CNS penetration. However, intrathecal (IT) administration of rituximab has shown to be safe and efficient in small studies and in case series. We report here the case of a patient with late development of CNS-PTLD after kidney-pancreas transplantation who achieved complete remission after surgical resection and four cycles of IT rituximab and we provide a review of the literature for this treatment option.

Item does not contain fulltext BACKGROUND: Etrolizumab is a gut-targeted anti-β7 integrin monoclonal antibody. In a previous phase 2 induction study, etrolizumab significantly improved clinical remission versus placebo in patients with moderately to severely active ulcerative colitis. We aimed to compare the safety and efficacy of etrolizumab with infliximab in patients with moderately to severely active ulcerative colitis. METHODS: We conducted a randomised, double-blind, double-dummy, parallel-group, phase 3 study (GARDENIA) across 114 treatment centres worldwide. We included adults (age 18-80 years) with moderately to severely active ulcerative colitis (Mayo Clinic total score [MCS] of 6-12 with an endoscopic subscore of ≥2, a rectal bleeding subscore of ≥1, and a stool frequency subscore of ≥1) who were naive to tumour necrosis factor inhibitors. Patients were required to have had an established diagnosis of ulcerative colitis for at least 3 months, corroborated by both clinical and endoscopic evidence, and evidence of disease extending at least 20 cm from the anal verge. Participants were randomly assigned (1:1) to receive subcutaneous etrolizumab 105 mg once every 4 weeks or intravenous infliximab 5 mg/kg at 0, 2, and 6 weeks and every 8 weeks thereafter for 52 weeks. Randomisation was stratified by baseline concomitant treatment with corticosteroids, concomitant treatment with immunosuppressants, and baseline disease activity. All participants and study site personnel were masked to treatment assignment. The primary endpoint was the proportion of patients who had both clinical response at week 10 (MCS ≥3-point decrease and ≥30% reduction from baseline, plus ≥1-point decrease in rectal bleeding subscore or absolute rectal bleeding score of 0 or 1) and clinical remission at week 54 (MCS ≤2, with individual subscores ≤1); efficacy was analysed using a modified intention-to-treat population (all randomised patients who received at least one dose of study drug). GARDENIA was designed to show superiority of etrolizumab over infliximab for the primary endpoint. This trial is registered with ClinicalTrials.gov, NCT02136069, and is now closed to recruitment. FINDINGS: Between Dec 24, 2014, and June 23, 2020, 730 patients were screened for eligibility and 397 were enrolled and randomly assigned to etrolizumab (n=199) or infliximab (n=198). 95 (48%) patients in the etrolizumab group and 103 (52%) in the infliximab group completed the study through week 54. At week 54, 37 (18·6%) of 199 patients in the etrolizumab group and 39 (19·7%) of 198 in the infliximab group met the primary endpoint (adjusted treatment difference -0·9% [95% CI -8·7 to 6·8]; p=0·81). The number of patients reporting one or more adverse events was similar between treatment groups (154 [77%] of 199 in the etrolizumab group and 151 [76%] of 198 in the infliximab group); the most common adverse event in both groups was ulcerative colitis (55 [28%] patients in the etrolizumab group and 43 [22%] in the infliximab group). More patients in the etrolizumab group reported serious adverse events (including serious infections) than did those in the infliximab group (32 [16%] vs 20 [10%]); the most common serious adverse event was ulcerative colitis (12 [6%] and 11 [6%]). There was one death during follow-up, in the infliximab group due to a pulmonary embolism, which was not considered to be related to study treatment. INTERPRETATION: To our knowledge, this trial is the first phase 3 maintenance study in moderately to severely active ulcerative colitis to use infliximab as an active comparator. Although the study did not show statistical superiority for the primary endpoint, etrolizumab performed similarly to infliximab from a clinical viewpoint. FUNDING: F Hoffmann-La Roche.

The new LHCb Vertex Locator (VELO) for LHCb, comprising a new pixel detector and readout electronics, will be installed in 2021 for data taking in Run 3 at the LHC. The electronics centers around the "VeloPix" ASIC at the front-end operating in a trigger-less readout at 40MHz. A custom serializer, called gigabit wireline transmitter (GWT), and associated custom protocol have been designed for the VeloPix. The GWT data are sent from the serializers of the VeloPix at a line rate of 5.12 Gb/s, reaching a total data rate of 2-3 Tb/s for the full VELO detector. Data are sent over 300-m optic-fiber links to the control and readout electronics cards for deserialization and processing in Intel Arria 10 FPGAs. Because of the VeloPix trigger-less design, latency variances up to 12 mu s can occur between adjacent datagrams. It is therefore essential to buffer and synchronize the data in firmware prior to onward propagation or suffer a huge CPU-processing penalty. This article will describe the architecture of the readout firmware in detail with focus given to the resynchronization mechanism and techniques for cauterization. Issues found during readout commissioning, and scaling resource utilization, along with the their solutions, will be illustrated. The latest results of the firmware data-processing chain can be presented as well as the verification procedures employed in simulation. Challenges for the next generation of the detector will also be presented with ideas for a readout processing solution. The new LHCb Vertex Locator (VELO) for LHCb, comprising a new pixel detector and readout electronics, will be installed in 2021 for data taking in Run 3 at the LHC. The electronics centers around the 'VeloPix' ASIC at the front-end operating in a trigger-less readout at 40MHz. A custom serializer, called gigabit wireline transmitter (GWT), and associated custom protocol have been designed for the VeloPix. The GWT data are sent from the serializers of the VeloPix at a line rate of 5.12 Gb/s, reaching a total data rate of 2-3 Tb/s for the full VELO detector. Data are sent over 300-m optic-fiber links to the control and readout electronics cards for deserialization and processing in Intel Arria 10 FPGAs. Because of the VeloPix trigger-less design, latency variances up to 12 $\mu \text{s}$ can occur between adjacent datagrams. It is therefore essential to buffer and synchronize the data in firmware prior to onward propagation or suffer a huge CPU-processing penalty. This article will describe the architecture of the readout firmware in detail with focus given to the resynchronization mechanism and techniques for cauterization. Issues found during readout commissioning, and scaling resource utilization, along with the their solutions, will be illustrated. The latest results of the firmware data-processing chain can be presented as well as the verification procedures employed in simulation. Challenges for the next generation of the detector will also be presented with ideas for a readout processing solution.

The Internet of Things (IoT) is rapidly changing the number of connected devices and the way they interact with each other. This increases the need for an automated and secure onboarding procedure for IoT devices, systems and services. Device manufacturers are entering the market with internet connected devices, ranging from small sensors to production devices, which are subject of security threats specific to IoT. The onboarding procedure is required to introduce a new device in a System of Systems (SoS) without compromising the already onboarded devices and the underlying infrastructure. Onboarding is the process of providing access to the network and registering the components for the first time in an IoT/SoS framework, thus creating a chain of trust from the hardware device to its hosted software systems and their provided services. The large number and diversity of device hardware, software systems and running services raises the challenge to establish a generic onboarding procedure. In this paper, we present an automated and secure onboarding procedure for SoS. We have implemented the onboarding procedure in the Eclipse Arrowhead framework. However, it can be easily adapted for other IoT/SoS frameworks that are based on Service-oriented Architecture (SoA) principles. The automated onboarding procedure ensures a secure and trusted communication between the new IoT devices and the Eclipse Arrowhead framework. We show its application in a smart charging use case and perform a security assessment., IEEE Access

Item does not contain fulltext BACKGROUND: Patient-centred care has been recognised as vital for today's healthcare quality. This type of care puts patients at the centre, contributing to positive patient outcomes such as patient autonomy. Empirical research comparing nurses' and patients' perceptions of the support and provision of patient-centred care is limited and focuses solely on nurses and patients working and staying on surgical wards. AIMS AND OBJECTIVES: Comparing patients' and nurses' perceptions of patient-centred care on different types of hospital wards, and exploring if patient empowerment, health literacy, and certain sociodemographic and context-related variables are associated with these perceptions. DESIGN: Cross-sectional design. METHODS: Data were collected in ten Flemish (February-June 2016) and two Dutch (December 2014-May 2015) hospitals using the Individualised Care Scale (ICS). A linear mixed model was fitted. Data from 845 patients and 569 nurses were analysed. As the ICS was used to measure the concept of patient-centred care, it is described using the term 'individualised care.' RESULTS: Nurses perceived that they supported and provided individualised care more compared with patients as they scored significantly higher on the ICS compared with patients. Patients with higher empowerment scores, higher health literacy, a degree lower than bachelor, a longer hospital stay, and patients who were employed and who were admitted to Dutch hospitals scored significantly higher on some of the ICS subscales/subsections. Nurses who were older and more experienced and those working in Dutch hospitals, regional hospitals and maternity wards scored significantly higher on some of the ICS subscales/subsections. CONCLUSION: Nurses perceived that they supported and provided individualised care more compared with patients. RELEVANCE TO CLINICAL PRACTICE: Creating a shared understanding towards the support and provision of individualised care should be a priority as this could generate more effective nursing care that takes into account the individuality of the patient.

BACKGROUND: Collaboration between policymakers, patients and healthcare workers in hospital quality of care policy setting can improve the integration of new initiatives. The aim of this study was to quantify preferences for various characteristics of a future quality policy in a broad group of stakeholders. MATERIALS AND METHODS: 450 policymakers, clinicians, nurses, patient representatives and hospital board members in Flanders (Belgium) participated in five discrete choice experiments (DCE) on quality control, quality improvement, inspection, patient incidents and transparency. For each DCE, various attributes and levels were defined from a literature review and interviews with 12 international quality and patient safety experts. RESULTS: For the attributes with the highest relative importance, participants exhibited a strong preference for quality control by an independent national organization and coordination of quality improvement initiatives at the level of hospital networks. The individual hospital was chosen over the government for setting up an action plan following patient complaints. Respondents also strongly preferred mandatory reporting of severe patient incidents and transparency by publicly reporting quality indicators at the hospital level. CONCLUSIONS: A future quality model should focus on a multicomponent approach with external quality control, improvement actions on hospital network level and public transparency. DCEs provide an opportunity to incorporate the attitudes and views for individual components of a new policy recommendation. ispartof: Health Policy vol:125 issue:12 pages:1565-1573 ispartof: location:Ireland status: published

PurposeThe purpose of authentic assessment can enrich students with a set of skills that can have a significant impact on their employability. The key aim of these actions is to equip students with the practical skills to be work ready. The focus here is to demonstrate effects on student satisfaction and observations from teaching staff accrued over several semesters in an educational institution after introducing authentic assessments.Design/methodology/approachThis paper used a set of scaffolded authentic assessments for students in logistics and supply chain management (LSCM). A combination of literature and expert input was used to design and implement the authentic assessment. A multi-phase systematic approach using cyclical model steps articulated the course learning outcomes (CLOs).FindingsReinforced evidence that authentic assessments create options for divergent learners and provide opportunities for applying practical and higher-order cognitive skills in tertiary education system of an emerging economy. With the focus on student ability in doing things, students with diversified backgrounds and abilities in Asia can be encouraged to take an active role in their own learning. The formats of the newly redesigned assessments allow multidimensional cognitive capabilities such as art development (posters and video tasks) and reflective exercises. Importantly, formative types of authentic assessments help to decrease the level of anxiety by emphasising the aspect of doing and lead to better student satisfaction with the courses. All of these effects are shown to be repeatable in an Asian economy.Practical implicationsAuthentic assessments prepare students for the new world of work through a more focused scaffolding of their learning.Originality/valueStimulation of deeper learning in tertiary education students is relevant for a top performing Asian economy.

Objectives To delineate the impact of peripheral musculoskeletal manifestations on stratification of disease phenotype and outcome in new-onset spondyloarthritis (SpA), using a prospective observational nationwide inception cohort, the BelGian Inflammatory Arthritis and spoNdylitis cohorT (Be-Giant). Methods Newly diagnosed adult SpA patients, fulfilling the Assessment of SpondyloArthritis International Society (ASAS) criteria for axial or peripheral SpA, were included in Be-Giant and prospectively followed every six months. Peripheral involvement (defined as arthritis, enthesitis and/or dactylitis) was determined in relation to clinically similar patient subsets at baseline and disease activity patterns during two-year follow-up, identified through K-means cluster analysis and latent class growth analysis. Results From November 2010 to March 2020, 367 patients were enrolled in Be-Giant, of whom 162 (44%) had peripheral manifestations. Two patient clusters [A, axial predominant (n = 248) and B, peripheral predominant (n = 119)] were identified at diagnosis. Longitudinal analysis (n = 115) revealed two trajectories of disease activity in each cluster: one with persistently high disease activity over time (‘High’), the other rapidly evolving to low disease activity (‘Low’). In cluster A patients, peripheral manifestations predisposed to the ‘High’ trajectory [odds ratio (OR) = 2.0, 95% CI: 1.3, 3.1, P = 0.001], despite more rapid initiation of biologics compared with patients without peripheral manifestations (hazard ratio (HR) = 2.1, 95% CI: 1.0, 4.4, P = 0.04 – Cox proportional-hazards model). Conclusion Peripheral musculoskeletal manifestations are major determinants of phenotypical diversity in new-onset SpA. Intriguingly, stratification of axial SpA according to concomitant peripheral involvement identified an endotype with an unfavorable outcome despite more prompt therapeutic intensification with biologics. These observations justify an endotype-tailored approach beyond current ASAS/EULAR management recommendations.

Context Although neuromuscular deficits in people with chronic ankle instability (CAI) have been identified, previous researchers have mostly investigated the activation of multiple muscles in isolation. Investigating muscle synergies in people with CAI would provide information about the coordination and control of neuromuscular activation strategies and could supply important information for understanding and rehabilitating neuromuscular deficits in this population. Objective To assess and compare muscle synergies using nonnegative matrix factorization in people with CAI and healthy control individuals as they performed different landing-cutting tasks. Design Cross-sectional study. Setting Laboratory. Patients or Other Participants A total of 11 people with CAI (5 men, 6 women; age = 22 ± 3 years, height = 1.68 ± 0.11 m, mass = 69.0 ± 19.1 kg) and 11 people without CAI serving as a healthy control group (5 men, 6 women; age = 23 ± 4 years, height = 1.74 ± 0.11 m, mass = 66.8 ± 15.5 kg) participated. Main Outcome Measure(s) Muscle synergies were extracted from electromyography of the lateral gastrocnemius, medial gastrocnemius, fibularis longus, soleus, and tibialis anterior (TA) muscles during anticipated and unanticipated landing-cutting tasks. The number of synergies, activation coefficients, and muscle-specific weighting coefficients were compared between groups and across tasks. Results The number of muscle synergies was the same for each group and task. The CAI group exhibited greater TA weighting coefficients in synergy 1 than the control group (P = .02). In addition, both groups demonstrated greater fibularis longus (P = .03) weighting coefficients in synergy 2 during the unanticipated landing-cutting task than the anticipated landing-cutting task. Conclusions These results suggest that, although both groups used neuromuscular control strategies of similar complexity or dimensionality to perform the landing-cutting tasks, the CAI group displayed different muscle-specific weightings characterized by greater emphasis on TA function in synergy 1, which may reflect an effort to increase joint stability to compensate for ankle instability.

BACKGROUND: Although electroconvulsive therapy (ECT) is a safe and effective treatment for patients with severe late life depression (LLD), transient cognitive impairment can be a reason to discontinue the treatment. The aim of the current study was to evaluate the association between structural brain characteristics and general cognitive function during and after ECT. METHODS: A total of 80 patients with LLD from the prospective naturalistic follow-up Mood Disorders in Elderly treated with Electroconvulsive Therapy study were examined. Magnetic resonance imaging scans were acquired before ECT. Overall brain morphology (white and grey matter) was evaluated using visual rating scales. Cognitive functioning before, during, and after ECT was measured using the Mini Mental State Examination (MMSE). A linear mixed-model analysis was performed to analyze the association between structural brain alterations and cognitive functioning over time. RESULTS: Patients with moderate to severe white matter hyperintensities (WMH) showed significantly lower MMSE scores than patients without severe WMH (F(1,75.54) = 5.42, p = 0.02) before, during, and post-ECT, however their trajectory of cognitive functioning was similar as no time × WMH interaction effect was observed (F(4,65.85) = 1.9, p = 0.25). Transient cognitive impairment was not associated with medial temporal or global cortical atrophy (MTA, GCA). CONCLUSION: All patients showed a significant drop in cognitive functioning during ECT, which however recovered above baseline levels post-ECT and remained stable until at least 6 months post-ECT, independently of severity of WMH, GCA, or MTA. Therefore, clinicians should not be reluctant to start or continue ECT in patients with severe structural brain alterations. ispartof: American Journal Of Geriatric Psychiatry vol:29 issue:11 pages:1117-1128 ispartof: location:England status: accepted

The COVID-19 pandemic, caused by the severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2), has presented itself as one of the most important health concerns of the 2020's, and hit the geriatric population the hardest. The presence of co-morbidities and immune ageing in the elderly lead to an increased susceptibility to COVID-19, as is the case for other influenza-like illnesses (ILI) or acute respiratory tract infections (ARI). However, little is known, about the impact of a previous or current infection on the other in terms of susceptibility, immune response, and clinical course. The aim of the "Prior Infection with SARS-COV-2" (PICOV) study is to compare the time to occurrence of an ILI or ARI between participants with a confirmed past SARS-CoV-2 infection (previously infected) and those without a confirmed past infection (naïve) in residents and staff members of nursing homes. This paper describes the study design and population characteristics at baseline., info:eu-repo/semantics/published

International audience; Business-to-business (B2B) customer retention relies heavily on analytics and predictive modeling to support decision making. Given this, we introduce uplift modeling as a relevant prescriptive analytics tool. In particular, the uplift logit leaf model offers a segmentation-based algorithm that combines predictive performance with interpretability. Applied to a real-world data set of 6432 customers of a European software provider, the uplift logit leaf model achieves superior performance relative to three other popular uplift models in our study. The accessibility of output gained from the uplift logit leaf model also is showcased with a case study, which reveals relevant managerial insights. This new tool thus delivers novel insights in the form of customized, global, and segment-level visualizations that are especially pertinent to industrial marketing settings. Overall, the findings affirm the viability of uplift modeling for improving decisions related to B2B customer retention management.

Submitted by (omml@ubi.pt) on 2021-07-05T10:47:24Z No. of bitstreams: 1 2021_Bousquet J_A_COVID anaphylaxis.pdf: 12561118 bytes, checksum: 2f801ee76ad2cb3cbdaa02ffabea8e09 (MD5) Approved for entry into archive by Pessoa (pfep@ubi.pt) on 2021-07-05T10:49:11Z (GMT) No. of bitstreams: 1 2021_Bousquet J_A_COVID anaphylaxis.pdf: 12561118 bytes, checksum: 2f801ee76ad2cb3cbdaa02ffabea8e09 (MD5) Rejected by Pessoa (pfep@ubi.pt), reason: Rever os nomes dos autores. Depois da correção é só voltar a submeter. on 2021-07-05T10:54:19Z (GMT) Submitted by (omml@ubi.pt) on 2021-07-05T11:52:24Z No. of bitstreams: 1 2021_Bousquet J_A_COVID anaphylaxis.pdf: 12561118 bytes, checksum: 2f801ee76ad2cb3cbdaa02ffabea8e09 (MD5) Approved for entry into archive by Pessoa (pfep@ubi.pt) on 2021-07-05T13:34:51Z (GMT) No. of bitstreams: 1 2021_Bousquet J_A_COVID anaphylaxis.pdf: 12561118 bytes, checksum: 2f801ee76ad2cb3cbdaa02ffabea8e09 (MD5) Approved for entry into archive by Pessoa (pfep@ubi.pt) on 2021-07-05T13:35:49Z (GMT) No. of bitstreams: 1 2021_Bousquet J_A_COVID anaphylaxis.pdf: 12561118 bytes, checksum: 2f801ee76ad2cb3cbdaa02ffabea8e09 (MD5) Made available in DSpace on 2021-07-05T13:35:49Z (GMT). No. of bitstreams: 1 2021_Bousquet J_A_COVID anaphylaxis.pdf: 12561118 bytes, checksum: 2f801ee76ad2cb3cbdaa02ffabea8e09 (MD5) Previous issue date: 2021-04-02 info:eu-repo/semantics/publishedVersion

Summary Background Etrolizumab is a gut-targeted, anti-β7 integrin, monoclonal antibody. In an earlier phase 2 induction study, etrolizumab significantly improved clinical remission compared with placebo in patients with moderately to severely active ulcerative colitis. We aimed to evaluate the efficacy and safety of etrolizumab in patients with moderately to severely active ulcerative colitis who had been previously treated with anti-tumour necrosis factor (TNF) agents. Methods HICKORY was a multicentre, phase 3, double-blind, placebo-controlled study in adult (18–80 years) patients with moderately to severely active ulcerative colitis (Mayo Clinic total score [MCS] of 6–12 with an endoscopic subscore of ≥2, a rectal bleeding subscore of ≥1, and a stool frequency subscore of ≥1) previously treated with TNF inhibitors. Patients were recruited from 184 treatment centres across 24 countries in North America, South America, Europe, Asia, Oceania, and the Middle East. Patients needed to have an established diagnosis of ulcerative colitis for at least 3 months, corroborated by both clinical and endoscopic evidence, and evidence of disease extending at least 20 cm from the anal verge. In cohort 1, patients received open-label etrolizumab 105 mg every 4 weeks for a 14-week induction period. In cohort 2, patients were randomly assigned (4:1) to receive subcutaneous etrolizumab 105 mg or placebo every 4 weeks for the 14-week induction phase. Patients in either cohort achieving clinical response to etrolizumab induction were eligible for the maintenance phase, in which they were randomly assigned (1:1) to receive subcutaneous etrolizumab 105 mg or placebo every 4 weeks through to week 66. Randomisation was stratified by baseline concomitant treatment with corticosteroids, concomitant treatment with immunosuppressants (induction randomisation only), baseline disease activity, week 14 MCS remission status (maintenance randomisation only), and induction cohort (maintenance randomisation only). All patients and study site personnel were masked to treatment assignment. Primary endpoints were remission (Mayo Clinic total score [MCS] ≤2, with individual subscores of ≤1 and a rectal bleeding subscore of 0) at week 14, and remission at week 66 among patients with a clinical response (MCS with ≥3-point decrease and ≥30% reduction from baseline, plus ≥1 point decrease in rectal bleeding subscore or absolute rectal bleeding score of 0 or 1) at week 14. Efficacy was analysed using a modified intent-to-treat population. Safety analyses included all patients who received at least one dose of study drug during the induction phase. This study is registered at ClinicalTrials.gov , NCT02100696 . Findings HICKORY was conducted from May 21, 2014, to April 16, 2020, during which time 1081 patients were screened, and 609 deemed eligible for inclusion. 130 patients were included in cohort 1. In cohort 2,479 patients were randomly assigned to the induction phase (etrolizumab n=384, placebo n=95). 232 patients were randomly assigned to the maintenance phase (etrolizumab to etrolizumab n=117, etrolizumab to placebo n=115). At week 14, 71 (18·5%) of 384 patients in the etrolizumab group and six (6·3%) of 95 patients in the placebo group achieved the primary induction endpoint of remission (p=0·0033). No significant difference between etrolizumab and placebo was observed for the primary maintenance endpoint of remission at week 66 among patients with a clinical response at week 14 (27 [24·1%] of 112 vs 23 [20·2%] of 114; p=0·50). Four patients in the etrolizumab group reported treatment-related adverse events leading to treatment discontinuation. The proportion of patients reporting at least adverse event was similar between treatment groups for induction (etrolizumab 253 [66%] of 384; placebo 63 [66%] of 95) and maintenance (etrolizumab to etrolizumab 98 [88%] of 112; etrolizumab to placebo 97 [85%] of 114). The most common adverse event in both groups was ulcerative colitis flare. Most adverse events were mild or moderate. During induction, the most common serious adverse event was ulcerative colitis flare (etrolizumab ten [3%] of 384; placebo: two [2%] of 95). During maintenance, the most common serious adverse event in the etrolizumab to etrolizumab group was appendicitis (two [2%] of 112) and the most common serious adverse events in the etrolizumab to placebo group were ulcerative colitis flare (two [2%] of 114) and anaemia (two [2%] of 114). Interpretation HICKORY demonstrated that a significantly higher proportion of patients with moderately to severely active ulcerative colitis who had been previously treated with anti-TNF agent were able to achieve remission at week 14 when treated with etrolizumab compared with placebo; however, there was no significant difference between groups in remission at week 66 among patients with a clinical response at week 14. Funding F Hoffmann-La Roche.

We steered the soil microbiome via applications of organic residues (mix of cover crop residues, sewage sludge + compost, and digestate + compost) to enhance multiple ecosystem services in line with climate-smart agriculture. Our result highlights the potential to reduce greenhouse gases (GHG) emissions from agricultural soils by the application of specific organic amendments (especially digestate + compost). Unexpectedly, also the addition of mineral fertilizer in our mesocosms led to similar combined GHG emissions than one of the specific organic amendments. However, the application of organic amendments has the potential to increase soil C, which is not the case when using mineral fertilizer. While GHG emissions from cover crop residues were significantly higher compared to mineral fertilizer and the other organic amendments, crop growth was promoted. Furthermore, all organic amendments induced a shift in the diversity and abundances of key microbial groups. We show that organic amendments have the potential to not only lower GHG emissions by modifying the microbial community abundance and composition, but also favour crop growth-promoting microorganisms. This modulation of the microbial community by organic amendments bears the potential to turn soils into more climate-smart soils in comparison to the more conventional use of mineral fertilizers.

Internationalisation is a major trend in higher education worldwide. Yet, little evidence is given on the net impact of international students on national economies. This study addresses this gap by estimating the benefits against the costs driven by international students in Belgium and its Flemish region in particular. Using a unique combination of various sources of micro-data, the results show net positive benefits that exceed costs by a factor ranging between 2.4 (lower bound) and 3.1 (upper bound) times. The results vary highly with the level of education, as the ratio is the lowest for doctoral students (1.2–1.6) and highest for master students (5.1–6.3). The effect is mainly driven by a high stay rate of international students, who are likely to work in the country after graduation. When considering indirect effects, our results show that there are no significant peer effects due to the presence of international students in the classroom. ispartof: Higher Education vol:82 issue:3 pages:459-476 status: published

The Wearables for Epilepsy And Research (WEAR) International Study Group identified a set of methodology standards to guide research on wearable devices for seizure detection. We formed an international consortium of experts from clinical research, engineering, computer science, and data analytics at the beginning of 2020. The study protocols and practical experience acquired during the development of wearable research studies were discussed and analyzed during bi-weekly virtual meetings to highlight commonalities, strengths, and weaknesses, and to formulate recommendations. Seven major essential components of the experimental design were identified, and recommendations were formulated about: (1) description of study aims, (2) policies and agreements, (3) study population, (4) data collection and technical infrastructure, (5) devices, (6) reporting results, and (7) data sharing. Introducing a framework of methodology standards promotes optimal, accurate, and consistent data collection. It also guarantees that studies are generalizable and comparable, and that results can be replicated, validated, and shared.

Background Electronic systems for remotely monitoring symptoms during cancer treatment are increasingly being used. Most of them are intended for weekly or periodic symptom monitoring. Objective The aim of this study was to explore the feasibility and usability of a remote system for daily symptom monitoring during systemic anticancer treatment. Methods We offered a remote system for daily symptom monitoring to patients starting their first ever systemic therapy. Patient acceptance was observed as the proportion of patients showing interest in using the system. System users were invited to complete the "Health information Technology Usability Evaluation Scale," and the number of patients' self-reports was used to evaluate patient compliance. Results Of 465 patients, 239 (51.4%) showed interest in using the system; 111 system users reported a fairly good overall usability score (4 of 5), and perceived ease of use scored the highest (4.7 of 5) and perceived usefulness lowest (3.7 of 5). Their daily compliance was 68.8% (±31.1) at 3 weeks and 59.1% (±31.2) at 12 weeks after the start of treatment. Less than half of respondents (49.5%) and only one third (31.2%) perceived that their symptom reports were used by the nurses and doctors, respectively. Conclusion Half of patients starting their first ever systemic treatment showed interest in using a remote system for daily symptom monitoring. For these patients, daily symptom monitoring seems acceptable, and the system is well complied within the first 12 weeks of treatment. Implication for practice A remote system for daily symptom monitoring is feasible and enables quick response to deteriorating symptoms. Use by healthcare professionals is a point of improvement.

Financial education is often considered to be a necessary condition in order to stimulate better consumer choices. We investigate this claim by analyzing the impact of a financial education course on consumer choices made by 688 students of the 8th and 9th grade. We combine a randomized controlled trial with a discrete choice experiment to analyze how financial education and other critical factors of a purchase decision like price, credit availability, information on product quality, and promotion affect consumer decisions at a young age. We find that students favor purchase options with free gifts and positive consumer reviews. They also prefer cash payments over credit. However, we do not find evidence that this behavior is triggered by financial education. Financial education does not automatically result in better consumer choices, even if the financial education course increased students' financial literacy.

This article presents frequency-modulated-continuous-wave (FMCW) radars developed for the detection of vital signs and gestures using two generations of 145-GHz transceivers (TRXs) integrated in 28-nm bulk CMOS. The performance and limitations of high-frequency radars are quantified with a system-level study, and the design and performance of individual circuit blocks are presented in detail. A 145-GHz center frequency and radar operation over an RF bandwidth of 10 GHz yield a displacement responsivity of 2 $\pi $ rad/mm and a windowed range resolution of 30 mm, respectively. Radar operation over a 0.1–7 m range is enabled by an effective-isotropic radiated power of 11.5 dBm and a noise figure of 8 dB. The ICs feature frequency multiplication by 9 in the transmit and receive paths, sub-arrayed dipole antennas, and neutralization of TX–RX leakage via delay control. A single TRX dissipates 500 mW from a 0.9-/1.8-V drive. The use of fast chirps (5–30- $\mu \text{s}$ ) mitigates the effect of 1/ $f$ -noise at the intermediate frequency (IF). Extensive characterization results showcase state-of-the-art performance of the TRXs, while the code-domain multiple-input and multiple-output (MIMO) radars ( $1 \times 4$ and $4 \times 4$ ) built with them demonstrate vital-sign and gesture detections.

Background: While palliative home care is advocated for people with dementia, evidence of its effectiveness is lacking. Aim: To evaluate the effects of palliative home care on quality and costs of end-of-life care for older people with dementia. Design: Decedent cohort study using linked nationwide administrative databases and propensity score matching. Setting/participants: All home-dwelling older people who died with dementia between 2010 and 2015 in Belgium ( N = 23,670). Exposure: Receiving palliative home care support for the first time between 360 and 15 days before death. Results: Five thousand six hundred and thirty-seven (23.8%) received palliative home care support in the last 2 years of life, of whom 2918 received it for the first time between 360 and 15 days before death. Two thousand eight hundred and thirty-nine people who received support were matched to 2839 people who received usual care. After matching, those using palliative home care support, in the last 14 days of life, had lower risk of hospital admission (17.5% vs 50.5%; relative risk (RR) = 0.21), undergoing diagnostic testing (17.0% vs 53.6%; RR = 0.20) and receiving inappropriate medications, but were more likely to die at home (75.7% vs 32.6%; RR = 6.45) and to have primary care professional contacts (mean 11.7 vs mean 5.2), compared with those who did not. Further, they had lower mean total costs of care in the last 30 days of life (incremental cost:−€2129). Conclusions: Palliative home care use by home-dwelling older people with dementia is associated with improved quality and reduced costs of end-of-life care. Access remains low and should be increased.

This paper provides a systematic literature review of studies investigating the effect of an intervention on the efficiency of a decision-making unit, when efficiency is computed using nonparametric frontier approaches. This paper offers a guide for future research by identifying patterns in (1) the fields of application, (2) applied efficiency models, and (3) analysis of efficiency determinants. Our findings indicate that, despite the prominent role of frontier techniques in the analysis of public sector performances and the importance of the effectiveness and the policy perspective, these two approaches have long been kept separate. Nevertheless, the combination of efficiency and effectiveness is fundamental to evaluate public interventions and to detect inefficiencies at the policy level, especially in key sectors such as education, health, and environment.

Objective: To estimate the prevalence of protein-energy malnutrition in people admitted for a diabetic foot ulcer (DFU) and to assess the relationship between malnutrition and DFU severity and outcome. Methods: This prospective, observational cohort study included individuals consecutively admitted for a DFU between July 2016 and September 2019. The Global Leadership Initiative on Malnutrition (GLIM) criteria determined the prevalence of malnutrition. The SINBAD score reflected DFU severity. Outcome was evaluated at discharge and at 6 months. The independent contribution of nutritional status on DFU severity and outcome was investigated using logistic regression analysis. Results: A total of 110 patients were included. Malnutrition, as defined by the GLIM criteria, was diagnosed in 26 cases; malnutrition was moderate in 9 and severe in 17. DFU severity differed significantly between subjects with malnutrition versus without malnutrition (SINBAD: 3.85 vs. 3.81, p = 0.012). Logistic regression analysis showed that severe malnutrition (p = 0.015) and hemoglobin level (p = 0.003) were independently linked to DFU severity. At 6-month follow-up, 39 DFU were healed, 36 patients had undergone an amputation (32 minor, 4 major) and 8 had died. No differences were noted in outcome at discharge or at 6 months according to nutritional status. Conclusions: In 24% of patients, malnutrition was diagnosed. Severely malnourished individuals presented with more severe ulcers. However, malnutrition had no impact on the short-term outcome of a DFU. (C) 2021 European Society for Clinical Nutrition and Metabolism. Published by Elsevier Ltd. All rights reserved.

Scombrotoxin fish poisoning (SFP) is caused by the ingestion of certain fish species with elevated concentrations of histamine due to decomposition. In fall 2019, the U.S. Food and Drug Administration (FDA) was notified of 51 SFP cases including two hospitalizations from 11 states through the FDA consumer complaint system or directly from state partners. A case patient was defined as an individual who experienced a histamine-type reaction after consumption of tuna imported from Vietnam and an illness onset between 14 August and 24 November 2019. A traceback investigation was initiated at 19 points of service to identify a common tuna source. The FDA and state partners collected 34 product samples throughout the distribution chain, including from a case patient's home, points of service, distributors, and the port of entry. Samples were analyzed for histamine by sensory evaluation and/or chemical testing. Case patients reported exposure to tuna imported from Vietnam. The traceback investigation identified two Vietnamese manufacturers as the sources of the tuna. Twenty-nine samples were confirmed as decomposed by sensory evaluation and/or were positive for elevated histamine concentrations by chemical testing. Both Vietnamese companies were placed on an import alert. Seven U.S. companies and one Vietnamese company initiated voluntary recalls. The FDA released public communication naming the U.S. importers to help suppliers and distributors identify the product and effectuate the foreign company's recall. This SFP outbreak investigation highlights the complexities of the federal outbreak response, specifically related to imported food. Cultural considerations regarding imported foods should be addressed during outbreak responses when timing is critical. Collaboration with countries where confidentiality agreements are not in place can limit information sharing and the speed of public health responses. HIGHLIGHTS

Introduction A need exists to improve the efficiency of clinical trials in burn care. The objective of this study was to validate “Persistent Organ Dysfunction” plus death as endpoint in burn patients and to demonstrate its statistical efficiency. Methods This secondary outcome analysis of a dataset from a prospective international multicenter RCT (RE-ENERGIZE) included patients with burned total body surface area >20% and a 6-month follow-up. Persistent organ dysfunction was defined as persistence of organ dysfunction with life-supportiing technologies and ICU care. Results In the 539 included patients, the prevalence of 0p p+ pdeath was 40% at day 14 and of 27% at day 28. At both timepoints, survivors with POD (vs. survivors without POD) had a higher mortality rate, longer ICU- and hospital-stays, and a reduced quality of life. POD + death as an endpoint could result in reduced sample size requirements for clinical trials. Detecting a 25% relative risk reduction in 28-day mortality would require a sample size of 4492 patients, whereas 1236 patients would be required were 28-day POD + death used. Conclusions POD + death represents a promising composite outcome measure that may reduce the sample size requirements of clinical trials in severe burns patients. Further validation in larger clinical trials is warranted. Study type Prospective cohort study, level of evidence: II

PurposeThe purpose of this paper is to deepen the understanding of adaptive governance, which is advocated for as a manner to deal with dramatic changes in society and/or environment. To re-think the possible contributions of organizations and organization theory, to adaptive governance.Design/methodology/approachBased on social systems theory this study makes a distinction between “governance organizations” and “governance communities.” Organizations are conceptualized as the decision machines which organize and (co-)steer governance. Communities are seen as the social environments against which the governance system orients its operations. This study considers the adaptive mechanisms of organizations and reflect on the roles of organizations to enhance adaptive governance in communities and societies.FindingsDiverse types of organizations can link or couple in different ways to communities in their social environment. Such links can enhance the coordinative capacity of the governance system and can also spur innovation to enable adaptation. Yet, linking with communities can also slow down responses to change and complexify the processes of deliberation in governance. Not all adaptive mechanisms available to organizations can be used in communicating with communities or can be institutionalized, but the continuous innovation in the field of organizations can inspire continuous testing of small-scale adaptive mechanisms at higher levels. Society can thus enhance its adaptive capacity by managing the role of organizations.Originality/valueThe harnessing of insights in organization theory and systems theory for improving understanding of adaptive governance. The finding that both experiment and coordination at societal level are needed, toward adaptive governance, and that organizations can contribute to both.

The waiting time in infants for a cardiac transplant remains high, due to the scarcity of donors. Consequently, waiting list morbidity and mortality are higher than those in other age groups. Therefore, the decision to list a small infant for cardiac transplantation is seen as an ethical dilemma by most physicians. This review aims to describe outcomes, limitations, and ethical considerations in infant heart transplantation. We used Medline and Embase as data sources. We searched for publications on infant (< 1 year) heart transplantation, bridge-to-transplant and long-term outcomes, and waiting list characteristics from January 2009 to March 2021. Outcome after cardiac transplant in infants is better than that in older children (1-year survival 88%), and complications are less frequent (25% CAV, 10% PTLD). The bridge-to-transplant period in infants is associated with increased mortality (32%) and decreased transplantation rate (43%). This is mainly due to MCS complications or the limited MCS options (with 51% mortality in infancy). Outcomes are worse for infants with CHD or in need of ECMO-support.Conclusion: Infants listed for cardiac transplantation have a high morbidity and mortality, especially in the period between diagnosis and transplantation. For those who receive cardiac transplant, the outlook is encouraging. Unfortunately, despite growing experience in VAD, mortality in children < 10 kg and children with CHD remains high. After transplantation, patients carry a psychological burden and there is a probability of re-transplantation later in life, with decreased outcomes compared to primary transplantation. These considerations are seen as an important ethical dilemma in many centers, when considering cardiac transplantation in infants (< 1 year). What is Known: • For infants, waitlist mortality remains high. In the pediatric population, MCS reduces the waiting list mortality. What is New: • Outcomes after infant cardiac transplantation are better than other age groups; however, MCS options remain limited, with persistently high waiting list mortality. • Future developments in MCS and alternative options to reduce waiting list mortality such as ABO-incompatible transplantation and pulmonary artery banding are encouraging and will improve ethical decision-making when an infant is in need of a cardiac transplant.

The latest development in the anterior lumbar interbody fusion (ALIF) procedure is its application in the lateral position to allow for simultaneous posterior percutaneous screw placement. The technical details of the lateral ALIF technique have not yet been described. To describe the surgical anatomy relevant to the lateral ALIF approach we performed a comprehensive anatomical study. In addition, the preoperative imaging, patient positioning, planning of the skin incision, positioning of the C-arm, surgical approach, and surgical technique are discussed in detail. The technique described led to the successful use of the lateral ALIF technique in our clinical cases. No lateral ALIF procedure needed to be aborted during these cases. Our present work gives detailed anatomical background and technical details for the lateral ALIF approach. This teaching article can provide readers with sufficient technical and anatomical knowledge to assist them in performing their first lateral ALIF procedure.

International audience; Marketing messages are most effective if they reach the right customers. Deciding which customers to contact is an important task in campaign planning. The paper focuses on empirical targeting models. We argue that common practices to develop such models do not account sufficiently for business goals. To remedy this, we propose profit-conscious ensemble selection, a modeling framework that integrates statistical learning principles and business objectives in the form of campaign profit maximization. Studying the interplay between data-driven learning methods and their business value in real-world application contexts, the paper contributes to the emerging field of profit analytics and provides original insights how to implement profit analytics in marketing. The paper also estimates the degree to which profit-concious modeling adds to the bottom line. The results of a comprehensive empirical study confirm the business value of the proposed ensemble learning framework in that it recommends substantially more profitable target groups than several benchmarks.

Psychiatric disorders, and especially severe mental illness, are associated with an increased risk of severe acute respiratory syndrome coronavirus 2 infection and COVID-19-related morbidity and mortality. People with severe mental illness should therefore be prioritised in vaccine allocation strategies. Here, we discuss the risk for worse COVID-19 outcomes in this vulnerable group, the effect of severe mental illness and psychotropic medications on vaccination response, the attitudes of people with severe mental illness towards vaccination, and, the potential barriers to, and possible solutions for, an efficient vaccination programme in this population.

Pulsed-laser-induced single-event current measurements on two geometries of waveguide-integrated germanium photodiodes were conducted over a range of operating voltages to examine the impact of photodiode geometry on the transient response. Vertical p-i-n photodiodes exhibit transients with a duration that is relatively independent of the operating voltage, while the transient duration in lateral p-i-n photodiodes depends on the operating voltage. Furthermore, the experimental measurements facilitate identification of device dimensions that impact the transient response. These results can be used to identify potential radiation mitigation strategies for photodiodes operating in a radiation environment. Understanding the implications of design choices is critical for designing integrated photonic systems that balance system performance with tolerance for radiation degradation.

Open Dialogue is a need-adapted approach to mental health care that was originally developed in Finland. Like other need-adapted approaches, Open Dialogue aims to meet consumer’s needs and promote collaborative person-centred dialogue to support recovery. Need-adapted mental health care is distinguished by flexibility and responsiveness. Fidelity, defined from an implementation science perspective as the delivery of distinctive interventions in a high quality and effective fashion is a key consideration in health care. However, flexibility presents challenges for evaluating fidelity, which is much easier to evaluate when manualization and reproducible processes are possible. Hence, it remains unclear whether Open Dialogue and other need-adapted mental health interventions can be meaningfully evaluated for fidelity. The aim of this paper was to critically appraise and advance the evaluation of fidelity in need-adapted mental health care, using Open Dialogue as a case study. The paper opens a discussion about how fidelity should be evaluated in flexible, complex interventions, and identifies key questions that need to be asked by practitioners working in need-adapted mental health care to ensure they deliver these interventions as intended and in an evidence-based fashion.

ObjectivesThe WHO Access, Watch and Reserve (AWaRe) classification has been developed to support countries and hospitals in promoting rational use of antibiotics while improving access to these essential medicines. We aimed to describe patterns of worldwide antibiotic use according to the AWaRe classification in the adult inpatient population.MethodsThe Global Point Prevalence Survey on Antimicrobial Consumption and Resistance (Global-PPS) collects hospital antibiotic use data using a standardized PPS methodology. Global-PPS 2015, 2017 and 2018 data, collected by 664 hospitals in 69 countries, were categorized into AWaRe groups to calculate proportional AWaRe use, Access-to-Watch ratios and the most common indications for treatment with selected Watch antibiotics. Only prescriptions for systemic antibiotics on adult inpatient wards were analysed.ResultsRegional Access use ranged from 28.4% in West and Central Asia to 57.7% in Oceania, whereas Watch use was lowest in Oceania (41.3%) and highest in West and Central Asia (66.1%). Reserve use ranged from 0.03% in sub-Saharan Africa to 4.7% in Latin America. There were large differences in AWaRe prescribing at country level. Watch antibiotics were prescribed for a range of very different indications worldwide, both for therapeutic and prophylactic use.ConclusionsWe observed considerable variations in AWaRe prescribing and high use of Watch antibiotics, particularly in lower- and upper-middle-income countries, followed by high-income countries. The WHO AWaRe classification has an instrumental role to play in local and national stewardship activities to assess prescribing patterns and to inform and evaluate stewardship activities.

The number and type of collaborations between researchers and stakeholders has increased significantly. This responds to the demand from policymakers, funders, and citizens that researchers should help to tackle important social issues (e.g. climate change, healthy aging). However, there is little knowledge about how collaboration processes are experienced, how we can theoretically conceptualize them, and how in this way we can develop efficient collaboration methods that contribute to solving urgent societal problems. In this meta-ethnography, we gathered relevant knowledge from carefully selected qualitative studies. A title/abstract analysis of 3422 articles from Web of Science and ProQuest led to the interpretative analysis of qualitative data from eight publications. The results of this study show that despite the need for a better understanding of complex collaborative interactions, the differences (or asymmetry) between organizations to which various partners belong hinder efficient collaboration. Bridging figures (brokers) can play an essential role if they succeed in drawing diverse stakeholders out of their organizational context into a new dimension that allows creativity and mutual understanding, but that also allows conflict and distortion. In ideal circumstances, this leads to a quasi-automatic transfer of knowledge between partners that takes place naturally and in both directions (knowledge diffusion).

Three-dimensional (3D) complementary metal–oxide–semiconductor (CMOS) integration could enable device scaling beyond the limits of conventional 2D CMOS technology. Such integration requires vertical electrical connections that pass through silicon substrates and interconnect stacked chips. The fabrication of these through-silicon vias (TSVs) creates new challenges in metrology, including the characterization of the thin isolation film deposited on the sidewalls of the TSVs and thickness characterization for wafer thinning. Here, we show that laser-induced terahertz emission microscopy can be used to characterize TSVs. Terahertz emission is observed through the excitation of a transient electric dipole in the depletion field of the TSV using femtosecond laser pulses. The detected terahertz waveform provides information about the local depletion field and thus structural information about the isolation film. By performing a time-of-flight measurement of the terahertz pulse, we can also extract the silicon wafer thickness. Laser-induced terahertz emission, and time-of-flight measurements of the terahertz pulse, can be used to non-invasively characterize through-silicon vias, which are required for three-dimensional CMOS integration.

BACKGROUND: Treatment of patients with epidermal growth factor receptor-mutated (EGFRm) non-small cell lung cancer (NSCLC) continues to evolve expeditiously. OBJECTIVES: This retrospective study investigated real-world treatment patterns and EGFR mutation testing in patients with EGFRm advanced NSCLC in Belgium. METHODS: Data were extracted from medical records of adults diagnosed with EGFRm locally advanced/metastatic NSCLC between 1 September 2015 and 31 December 2017. Patients were followed retrospectively from diagnosis until 1 September 2018, end of clinical activity or death. Data on demographics, patient outcomes and disease characteristics, treatment patterns and EGFR mutation testing at diagnosis and progression were analyzed descriptively. RESULTS: A total of 141 patients were enrolled. At diagnosis, median age was 69 years, 63.1% were female, 88.7% had metastatic disease, 94.3% had adenocarcinoma histology, 76.6% had ECOG 0/1, 70.9% had common EGFR mutations and 29.1% had only rare mutations. Infirst line, 73.8% of patients received first/second-generation EGFR-tyrosine kinase inhibitors (1G/2G EGFR-TKIs), while 21.9% received other systemic treatments. Among 61 patients progressing on and discontinuing a first 1G/2G EGFR-TKI, 45 (73.8%) received subsequent systemic treatment while 16 (26.2%) did not; 20 (32.8%) received osimertinib. Among 65 patients progressing on a first 1G/2G EGFR-TKI, 47 (72.3%) were tested for T790M, of whom 25 (53.2%) were positive. CONCLUSION: These real-world data from Belgium show that a substantial fraction of patients with EGFRm NSCLC do not receive 1G/2G EGFR-TKIs in first line and do not receive subsequent systemic treatment after progression on 1G/2G EGFR-TKIs. Only a third receive osimertinib upon progression on 1G/2G EGFR-TKIs. These observations should be considered in first-line treatment decisions. TRIAL REGISTRATION: ClinicalTrials.gov: NCT03761901-December 3, 2018.

In recent years, the use of machine learning (ML) in computational chemistry has enabled numerous advances previously out of reach due to the computational complexity of traditional electronic-structure methods. One of the most promising applications is the construction of ML-based force fields (FFs), with the aim to narrow the gap between the accuracy of ab initio methods and the efficiency of classical FFs. The key idea is to learn the statistical relation between chemical structure and potential energy without relying on a preconceived notion of fixed chemical bonds or knowledge about the relevant interactions. Such universal ML approximations are in principle only limited by the quality and quantity of the reference data used to train them. This review gives an overview of applications of ML-FFs and the chemical insights that can be obtained from them. The core concepts underlying ML-FFs are described in detail, and a step-by-step guide for constructing and testing them from scratch is given. The text concludes with a discussion of the challenges that remain to be overcome by the next generation of ML-FFs.